Browse the corpus

Current pneumonia treatment and prevention strategies are based mainly on data obtained from large clinical studies carried out in the 1980s. One such study, sponsored by the Board of Science and Technology for International Development (BOSTID), National Academy of Sciences, yielded valuable information on the pathogens present during acute respiratory infections (ARIs) in children <5 years old from resource-limited countries [1]. However, interpretation of the wide range of reported ARI incidence rates was complicated in part by the lack of a standardized case definition at the 10 participating study sites [2]. A subsequent literature review of pneumonia etiology studies, conducted between 2000 and 2010 on children aged <5 years, revealed wide disparity in case definitions, specimen collection techniques, and laboratory methods, which increased the complexity of data collation and analysis [3]. Other studies have demonstrated substantial interclinician variation in the interpretation of clinical signs of severe disease in children and young infants [4–7]. Standardization of the clinical [8], radiological [9], laboratory [10], and data management methods [11] at all PERCH sites has been prioritized since inception, as we wished to ensure that any observed variation in pneumonia etiology between sites was not attributable to methodological differences. The objectives of the clinical standardization program were to ensure that study staff (1) adhered strictly to the clinical case definitions; (2) were consistent in their assessment, recognition, and interpretation of clinical signs; (3) used standardized equipment and techniques for obtaining clinical measurements; and (4) used standardized methods for obtaining key clinical samples for laboratory testing. This paper describes the PERCH clinical standardization program of clinical training, retraining, and staff assessment that ran throughout the study.

ed standardized equipment and techniques for obtaining clinical measurements; and (4) used standardized methods for obtaining key clinical samples for laboratory testing. This paper describes the PERCH clinical standardization program of clinical training, retraining, and staff assessment that ran throughout the study. METHODS Study Sites At all sites (Table 1), clinical assessment and enrollment of PERCH cases and controls were carried out by doctors, nurses, and clinical officers (health workers with at least 3 years of formal clinical training). Nurses and field workers or research assistants took anthropometric measurements, assisted clinical staff with procedures, and identified and located PERCH community controls. Table 1. Profile of Pneumonia Etiology Research for Child Health (PERCH) Study Sites

METHODS Study Sites At all sites (Table 1), clinical assessment and enrollment of PERCH cases and controls were carried out by doctors, nurses, and clinical officers (health workers with at least 3 years of formal clinical training). Nurses and field workers or research assistants took anthropometric measurements, assisted clinical staff with procedures, and identified and located PERCH community controls. Table 1. Profile of Pneumonia Etiology Research for Child Health (PERCH) Study Sites Country Training Language Study Site Setting Start of Enrollment Staff Responsible for Enrollment of PERCH Cases and/or Controls Cadre No. (%) Total Kenya English Kilifi Rural August 2011 Doctor 3 (13) 23 COa 18 (78) Nurse 2 (9) South Africa English Johannesburg Urban August 2011 Doctor 1 (8) 13 Nurse 12 (92) Zambia English Lusaka Urban October 2011 Doctor 4 (23) 17 COa 3 (18) Nurse 10 (59) The Gambia English Basse Rural November 2011 Doctor 7 (23) 31 Nurse 24 (77) Mali Frenchc Bamako Urban January 2012 Doctor 12 (67) 18 Nurse 6 (33) Bangladesh Banglac & English Dhaka Urban January 2012 Doctorb 37 (100) 37 Matlab Rural January 2012 Thailand Thaic & English Sa Kaeo Mixed January 2012 Doctor Nurse 2 (11) 17 (89) 19 Nakhon Phanom Mixed February 2012 Total 158 Abbreviations: CO, clinical officer; PERCH, Pneumonia Etiology Research for Child Health. aClinical officers are health workers with at least 3 years of formal clinical training. bIn Bangladesh, all enrollment decisions were made by doctors, although nurses helped to identify potential cases and controls.

Country Training Language Study Site Setting Start of Enrollment Staff Responsible for Enrollment of PERCH Cases and/or Controls Cadre No. (%) Total Kenya English Kilifi Rural August 2011 Doctor 3 (13) 23 COa 18 (78) Nurse 2 (9) South Africa English Johannesburg Urban August 2011 Doctor 1 (8) 13 Nurse 12 (92) Zambia English Lusaka Urban October 2011 Doctor 4 (23) 17 COa 3 (18) Nurse 10 (59) The Gambia English Basse Rural November 2011 Doctor 7 (23) 31 Nurse 24 (77) Mali Frenchc Bamako Urban January 2012 Doctor 12 (67) 18 Nurse 6 (33) Bangladesh Banglac & English Dhaka Urban January 2012 Doctorb 37 (100) 37 Matlab Rural January 2012 Thailand Thaic & English Sa Kaeo Mixed January 2012 Doctor Nurse 2 (11) 17 (89) 19 Nakhon Phanom Mixed February 2012 Total 158 Abbreviations: CO, clinical officer; PERCH, Pneumonia Etiology Research for Child Health. aClinical officers are health workers with at least 3 years of formal clinical training. bIn Bangladesh, all enrollment decisions were made by doctors, although nurses helped to identify potential cases and controls. cMultiple-choice questions (MCQs) were translated into French (Mali) or Thai (Thailand); staff in Bangladesh took MCQs in English.

aClinical officers are health workers with at least 3 years of formal clinical training. bIn Bangladesh, all enrollment decisions were made by doctors, although nurses helped to identify potential cases and controls. cMultiple-choice questions (MCQs) were translated into French (Mali) or Thai (Thailand); staff in Bangladesh took MCQs in English. Preparatory Phase The PERCH case definition (Table 2) was based on the 2005 World Health Organization (WHO) clinical definition of severe and very severe pneumonia [8]. The definition relies on the presence of prespecified clinical signs, without information from chest radiograph (CXR) or pulse oximetry. The PERCH enrollment period predated the 2013 reclassification of severe and very severe pneumonia by the WHO [12]. Table 2. Pneumonia Etiology Research for Child Health (PERCH) Clinical Case Definition of Severe and Very Severe Pneumoniaa

Preparatory Phase The PERCH case definition (Table 2) was based on the 2005 World Health Organization (WHO) clinical definition of severe and very severe pneumonia [8]. The definition relies on the presence of prespecified clinical signs, without information from chest radiograph (CXR) or pulse oximetry. The PERCH enrollment period predated the 2013 reclassification of severe and very severe pneumonia by the WHO [12]. Table 2. Pneumonia Etiology Research for Child Health (PERCH) Clinical Case Definition of Severe and Very Severe Pneumoniaa Case Sign or Symptom Detailed Definition Pneumonia (nonsevere) Cough or difficulty breathing plus fast breathing Cough On history and/or examination Difficulty breathing Fast, labored, deep, irregular, or noisy breathing Fast breathing Respiratory rate (breaths/min): ≥60 (<2 mo); ≥50 (2–11 mo); ≥40 (1–5 y) Severe pneumonia Cough or difficulty breathing plus lower chest wall indrawing Lower chest wall indrawing Inward movement of the lower bony chest wall on inspiration; child must be calm and not crying Very severe pneumonia Cough or difficulty breathing plus any of the following signs or symptomsb: Central cyanosis Blue discoloration of lips, gums, and tongue; should be assessed under good lighting conditions Head nodding Flexion of the head with inspiration; more commonly seen in young children and infants. Most easily seen if child is upright Unable to drink or breastfeed This must be observed in the clinical environment, by study staff: <2 mo: feeding poorly (eg, poor attachment to breast, weak suck) ≥2 mo: inability to take anything (fluids or solids) by mouth Vomiting everything This must be observed in the clinical environment, by study staff: Child is given a drink: if child has not vomited by the end of the clinical assessment, and before study procedures are carried out, then s/he is not “vomiting everything” Lethargy or unconsciousness AVPU scorec = V, P, or U Convulsions this illness Based on detailed description by parent or guardian. For inclusion in PERCH, convulsions must be prolonged (≥15 min) or multiple (≥2 within a 24-h period during the current illness)d Abbreviation: PERCH, Pneumonia Etiology Research for Child Health.

y or unconsciousness AVPU scorec = V, P, or U Convulsions this illness Based on detailed description by parent or guardian. For inclusion in PERCH, convulsions must be prolonged (≥15 min) or multiple (≥2 within a 24-h period during the current illness)d Abbreviation: PERCH, Pneumonia Etiology Research for Child Health. aBased on World Health Organization (2005) clinical case definition of severe and very severe pneumonia (Pocket Book of Hospital Care for Children). bLower chest wall indrawing is not a defining sign of very severe pneumonia as it may disappear if the child becomes exhausted. cAVPU score: (1) clinician first assesses whether the child is alert; A = alert (child takes an age-appropriate interest in their environment); if child not alert, clinician tests, in sequence, V, P, and U, stopping when the child gives a positive response; (2) clinician calls the child’s name without simultaneously touching him or her; V = response to voice (any consistent visual, verbal, or motor response to voice); (3) clinician presses on the base of the child’s fingernail using a pencil or pen; P = response to pain (child withdraws digit); (4) U = unresponsive or unconscious (no response to pain). dDefinition of complex febrile seizure used by American Academy of Pediatrics (Pediatrics 2011; 127: 389–94); PERCH adopted a stringent definition of “convulsions this illness” to avoid enrolling large numbers of children with cough and simple febrile seizures.

cAVPU score: (1) clinician first assesses whether the child is alert; A = alert (child takes an age-appropriate interest in their environment); if child not alert, clinician tests, in sequence, V, P, and U, stopping when the child gives a positive response; (2) clinician calls the child’s name without simultaneously touching him or her; V = response to voice (any consistent visual, verbal, or motor response to voice); (3) clinician presses on the base of the child’s fingernail using a pencil or pen; P = response to pain (child withdraws digit); (4) U = unresponsive or unconscious (no response to pain). dDefinition of complex febrile seizure used by American Academy of Pediatrics (Pediatrics 2011; 127: 389–94); PERCH adopted a stringent definition of “convulsions this illness” to avoid enrolling large numbers of children with cough and simple febrile seizures. Through a series of teleconferences and 2 face-to-face meetings between all PERCH principal investigators (PIs), consensus was achieved on how to elicit, recognize, and interpret each of the signs and symptoms comprising the PERCH clinical case definition (Table 2), and on the choice of methods and equipment for obtaining key clinical measurements (pulse oximetry, anthropometry, respiratory rate) and clinical samples (nasopharyngeal [NP] and oropharyngeal [OP] swabs, induced sputum [IS], lung aspirates, blood, urine).

mptoms comprising the PERCH clinical case definition (Table 2), and on the choice of methods and equipment for obtaining key clinical measurements (pulse oximetry, anthropometry, respiratory rate) and clinical samples (nasopharyngeal [NP] and oropharyngeal [OP] swabs, induced sputum [IS], lung aspirates, blood, urine). Training materials and advice were sought from a wide variety of sources (see Acknowledgments). Many of the clinical video clips, audio recordings and photographs were recorded at PERCH sites by the principal trainer (J. C.), with written informed consent from the patient’s parents or guardians. Training Courses Initial clinical standardization training occurred at all sites immediately prior to a period of pilot enrollment. All sites enrolled to the main study for 24 months, with refresher training carried out in the first and second year. The initial 3-day training and subsequent 2-day refresher training courses were conducted at all sites by the principal trainer, with support from site project leaders. All cadres of PERCH staff (doctors, nurses, clinical officers, research assistants, and field workers) were trained together; interested local non-PERCH clinicians were invited to participate.

2-day refresher training courses were conducted at all sites by the principal trainer, with support from site project leaders. All cadres of PERCH staff (doctors, nurses, clinical officers, research assistants, and field workers) were trained together; interested local non-PERCH clinicians were invited to participate. Training courses comprised lectures, discussion of case scenarios in small groups, practical sessions, and ward-based clinical teaching. The initial training lectures covered the background to the PERCH study, rationale for clinical standardization, recognition of the critically ill child, clinical assessment of the child with cough or difficulty breathing, vital signs, pulse oximetry, techniques for collection of NP/OP swabs, and anthropometry. Discussion of PERCH case scenarios, designed to test the trainees’ ability to identify signs and symptoms that constitute study inclusion and exclusion criteria, took place in groups of 8–10 people, each group being led by the principal trainer and/or a local facilitator. Trainees were divided into groups of 5–8 for hands-on instruction in clinical assessment. Staff members were asked to conduct clinical assessments on children, and were assessed on their ability to elicit and correctly interpret clinical signs.

Evaluation Time-Point MCQ % Score, Median (IQR) MCQ % Score, Median (IQR) Doctor Clinical Officer Nurse Nurses Sees Cases and Controls Nurses Sees Controls Only No. No. No. P Valuea No. No. P Valuea Postbaseline training 64 100 (90–100) 16 100 (90–100) 64 90 (80–100) .07 49 90 (90–100) 15 90 (70–100) .10 Prerefresher training 1 42 85 (75–90) 13 75 (60–85) 40 75 (62.5–85) .02 31 80 (65–95) 9 65 (55–65) .02 Postrefresher training 1 44 95 (90–100) 13 90 (80–95) 42 90 (75–100) .05 32 90 (80–100) 10 77.5 (65–90) .08 Online MCQ 1 46 100 (90–100) 17 90 (80–90) 40 90 (80–100) .03 30 90 (80–100) 10 90 (80–100) .91 Prerefresher training 2 39 90 (80–100) 15 85 (75–85) 42 70 (60–85) <.001 31 75 (65–85) 11 65 (55–75) .13 Postrefresher training 2 36 100 (95–100) 13 85 (85–95) 44 82.5 (72.5–95) <.001 31 90 (75–95) 13 80 (70–90) .46 Online MCQ 2 42 100 (95–100) 14 65 (60–80) 43 85 (65–95) <.001 33 85 (60–95) 10 87.5 (80–95) .59 Abbreviations: IQR, interquartile range; MCQ, multiple-choice question; PERCH, Pneumonia Etiology Research for Child Health. a P value obtained by Kruskal-Wallis test. Figure 1. Distribution of multiple choice question (MCQ) scores by site and training time-point: postbaseline training (A), pre- and postrefresher training 1 (B), and refresher training 2 (C). Boxplots display the distribution of MCQ scores. The number beneath each boxplot indicates the number of Pneumonia Etiology Research for Child Health (PERCH) clinicians and nurses who took the MCQ at each site. The diamond and horizontal line within the boxes represent the mean and median, respectively. The box reflects the interquartile range (IQR) and the whiskers extend to 1.5 multiplied by the IQR in either direction, or maximum and minimum values (if no outliers). The circle indicates outliers (values lying outside 1.5 multiplied by the IQR). Abbreviations: BAN, Bangladesh; GAM, The Gambia; KEN, Kenya; MAL, Mali; MCQ, multiple-choice question; pre, pre-course MCQ; post, post-course MCQ; RF1, refresher training 1; RF2, refresher training 2; SAF, South Africa; THA, Thailand; ZAM, Zambia.

, each group being led by the principal trainer and/or a local facilitator. Trainees were divided into groups of 5–8 for hands-on instruction in clinical assessment. Staff members were asked to conduct clinical assessments on children, and were assessed on their ability to elicit and correctly interpret clinical signs. Practical skills were taught through training videos, demonstrations, and hands-on practice in small groups, with key points highlighted in summary lectures. Clinically stable children acted as subjects for the anthropometry training. Staff learned NP/OP swab collection by practicing on each other. Clinicians from sites where IS samples were routinely collected from children (Kenya, The Gambia, South Africa) trained staff from the other 4 sites. The clinical team in Kenya reviewed video recordings of the collection procedures in The Gambia and South Africa, to ensure that they were consistent with procedures in Kenya. IS training was included in the refresher courses, as was guidance on reducing blood culture contamination rates through improved phlebotomy technique. Ethical approval to perform diagnostic percutaneous needle lung aspiration among PERCH cases was only obtained in The Gambia, Mali, South Africa, and Bangladesh. Clinicians from The Gambia (where lung aspiration is performed frequently on children with focal consolidation on CXR [13]) trained PERCH staff from the other 3 countries. Pleural aspirates and gastric aspirates were not included in the training as they were not designated PERCH procedures, but were carried out as routine hospital procedures if clinically indicated.

n is performed frequently on children with focal consolidation on CXR [13]) trained PERCH staff from the other 3 countries. Pleural aspirates and gastric aspirates were not included in the training as they were not designated PERCH procedures, but were carried out as routine hospital procedures if clinically indicated. All courses finished with a multiple-choice question (MCQ) examination, presentation of certificates, prizes for those achieving top scores, and a group photograph. All participants were invited to provide feedback, using a Likert scale to grade the quality of different course components. Clinical Standardization Guidelines Guidelines summarizing key information from the training program were distributed to all staff at the time of refresher training, with an electronic version made available on the internal PERCH study website. Training Website A training website (www.perchtraining.org), developed in association with a company specializing in digital healthcare (see Acknowledgments), had the following objectives: (1) to act as a repository for the clinical standardization training materials, thereby supporting the training of any new staff who had missed their initial site training course; (2) to provide continuing training of all PERCH staff throughout study enrollment; and (3) to facilitate regular evaluation of all PERCH clinical staff, and comparison of staff performance within and across sites.

rials, thereby supporting the training of any new staff who had missed their initial site training course; (2) to provide continuing training of all PERCH staff throughout study enrollment; and (3) to facilitate regular evaluation of all PERCH clinical staff, and comparison of staff performance within and across sites. The website contained all lectures from the initial training course, which could be streamed or downloaded as lectures with recorded voice-over, or as PowerPoint presentations. When internet speeds were slow, staff accessed training materials from DVDs, which had been distributed to all sites at the start of the study. At several sites, limited access to personal computers meant that project leaders downloaded the MCQs and organized the evaluations as classroom sessions. The website baseline training was supplemented by on-site training in practical skills and ward-based clinical teaching, both coordinated by the local PERCH study leader. On completion of the online course, trainees were required to take the same MCQ examination as those who had participated in face-to-face training. Trainees achieving a score of 80% or more were able to download a certificate from the website. The website also contained 2 additional MCQ examinations and a video quiz (see Evaluation).

f the online course, trainees were required to take the same MCQ examination as those who had participated in face-to-face training. Trainees achieving a score of 80% or more were able to download a certificate from the website. The website also contained 2 additional MCQ examinations and a video quiz (see Evaluation). Evaluation MCQ examinations were conducted after initial baseline training, immediately before and after each refresher course, and online after 10 months and 20 months of enrollment. An online video quiz was used to assess interobserver variation in interpretation of clinical signs at 20 months. Checklist evaluation of practical skills was performed at the end of the first year. MCQs were designed to test knowledge and understanding of the screening, consent and enrollment process, and the recognition and correct interpretation of key clinical signs, particularly those included in the WHO definitions of severe and very severe pneumonia. Each of the 10–20 MCQs contained a typical PERCH case scenario, plus, in most cases, a photograph or short video of a clinical sign. Answers to each question were provided at the end of the quiz, once all of the questions had been answered, with explanatory notes highlighting key learning points. Staff scoring <80% in the MCQ administered after baseline training were required to repeat selected lectures and the quiz, while staff scoring <80% after refresher training received additional training from their site-specific trainer.

uestions had been answered, with explanatory notes highlighting key learning points. Staff scoring <80% in the MCQ administered after baseline training were required to repeat selected lectures and the quiz, while staff scoring <80% after refresher training received additional training from their site-specific trainer. The video quiz assessed the ability of clinical staff to identify 6 clinical signs (lower chest wall indrawing [LCWI], head nodding, deep breathing, central cyanosis, nasal flaring, alert child). Clinical staff were shown 35 video clips (10 videos of LCWI, the defining clinical feature of WHO severe pneumonia, and 5 videos of each of the other clinical signs). Each video lasted approximately 10 seconds, and clinicians had to decide whether a specific clinical sign was present or not. Local clinical standardization trainers observed PERCH nurses and field workers carrying out anthropometry, IS, and NP/OP swab collection. Scored checklists (Supplementary Tables 1–3) were used to award points for key predefined procedural steps, the resulting percentage score providing a measure of procedural competence.

Local clinical standardization trainers observed PERCH nurses and field workers carrying out anthropometry, IS, and NP/OP swab collection. Scored checklists (Supplementary Tables 1–3) were used to award points for key predefined procedural steps, the resulting percentage score providing a measure of procedural competence. Statistical Analysis Median percentage scores and interquartile range (IQR) were calculated for MCQ tests and checklists. Median MCQ scores before and after refresher training were compared using the Wilcoxon signed-rank test. The distribution of results across participants was compared within and across study sites. Results were stratified by professional cadre and by whether staff assessed both cases and controls, or controls only. Differences between groups were examined with the Kruskal-Wallis test. For each of the 6 clinical signs in the video quiz (35 videos in total), individual responses were used to assess the percentage agreement between the clinical staff and the principal trainer, who was the designated “gold standard.” Calculation of Fleiss’ κ and the Gwet AC1 statistic, which is less affected by low prevalence than the κ statistic, were used to measure the degree of interobserver variability [14–16]. Kaplan-Meier curves were constructed to illustrate the proportion of PERCH clinical staff remaining in the study, from the time of baseline clinical standardization training. Curves were censored when staff members left the study, or on completion of PERCH enrollment.

For each of the 6 clinical signs in the video quiz (35 videos in total), individual responses were used to assess the percentage agreement between the clinical staff and the principal trainer, who was the designated “gold standard.” Calculation of Fleiss’ κ and the Gwet AC1 statistic, which is less affected by low prevalence than the κ statistic, were used to measure the degree of interobserver variability [14–16]. Kaplan-Meier curves were constructed to illustrate the proportion of PERCH clinical staff remaining in the study, from the time of baseline clinical standardization training. Curves were censored when staff members left the study, or on completion of PERCH enrollment. RESULTS Training Courses Between March 2011 and August 2013, a total of 32 training courses were conducted at 8 study sites in 7 countries. Of 331 staff attending 1 or more courses, 45 (14%) were interested local clinical staff, not directly involved in the study. Feedback from course participants was positive, with 90% of all course components being graded as “very good” (4/5) or “excellent” (5/5).

g courses were conducted at 8 study sites in 7 countries. Of 331 staff attending 1 or more courses, 45 (14%) were interested local clinical staff, not directly involved in the study. Feedback from course participants was positive, with 90% of all course components being graded as “very good” (4/5) or “excellent” (5/5). Initial (baseline) clinical standardization training took place over a 6-month period between March and September 2011. At each site, training occurred immediately prior to a period of pilot enrollment, and a median of 5 months (range, 4–9 months) before the start of the study. Seventy staff members joined PERCH after the initial training course at their site, and received baseline training from their site project leader and/or the training website. In South Africa, baseline training was repeated 6 months after the start of enrollment, due to extensive staff turnover during the pilot period. The first round of refresher training took place a median of 7 (range, 5–11) months and the second round a median of 18 months (range, 14–21) after the start of the study. A national nurses strike in Kenya during 2012 delayed refresher training by 3 months. At all other sites, training and enrollment continued uninterrupted, despite extensive flooding in Thailand during 2011, civil war in Mali during 2012–2013, and political instability in Bangladesh during 2013.

dBaseline training refers to the training that all staff underwent at the time of joining the study; it does not relate to a specific time-point, as new staff members were recruited throughout the study. eP < .001 with Wilcoxon signed-rank test. fP = .17 with Kruskal-Wallis test (no significant difference in distribution of scores between online MCQ1 and MCQ2). Table 4. Multiple-Choice Question Scores by Cadre and Role Evaluation Time-Point MCQ % Score, Median (IQR) MCQ % Score, Median (IQR) Doctor Clinical Officer Nurse Nurses Sees Cases and Controls Nurses Sees Controls Only No. No. No. P Valuea No. No. P Valuea Postbaseline training 64 100 (90–100) 16 100 (90–100) 64 90 (80–100) .07 49 90 (90–100) 15 90 (70–100) .10 Prerefresher training 1 42 85 (75–90) 13 75 (60–85) 40 75 (62.5–85) .02 31 80 (65–95) 9 65 (55–65) .02 Postrefresher training 1 44 95 (90–100) 13 90 (80–95) 42 90 (75–100) .05 32 90 (80–100) 10 77.5 (65–90) .08 Online MCQ 1 46 100 (90–100) 17 90 (80–90) 40 90 (80–100) .03 30 90 (80–100) 10 90 (80–100) .91 Prerefresher training 2 39 90 (80–100) 15 85 (75–85) 42 70 (60–85) <.001 31 75 (65–85) 11 65 (55–75) .13 Postrefresher training 2 36 100 (95–100) 13 85 (85–95) 44 82.5 (72.5–95) <.001 31 90 (75–95) 13 80 (70–90) .46 Online MCQ 2 42 100 (95–100) 14 65 (60–80) 43 85 (65–95) <.001 33 85 (60–95) 10 87.5 (80–95) .59 Abbreviations: IQR, interquartile range; MCQ, multiple-choice question; PERCH, Pneumonia Etiology Research for Child Health. a P value obtained by Kruskal-Wallis test.

the start of the study. A national nurses strike in Kenya during 2012 delayed refresher training by 3 months. At all other sites, training and enrollment continued uninterrupted, despite extensive flooding in Thailand during 2011, civil war in Mali during 2012–2013, and political instability in Bangladesh during 2013. Evaluation MCQ and video quiz results are presented for the 158 doctors, clinical officers, and nurses who enrolled PERCH cases and/or controls. High rates of staff turnover meant that only 57 of 158 (36.1%) of those who received baseline training completed all 7 MCQs plus the video quiz, but at each evaluation time-point MCQ scores were available for a median of 94% (IQR, 87%–100%) of the eligible staff at all sites (Table 3). Median scores were ≥80% at each point of testing, and improved with refresher training by a median of 10 percentage points. There was significant heterogeneity (P < .001) in the range of baseline training scores between sites, with South Africa and Mali having the greatest range of scores and Thailand the least variability (Figure 1A). Refresher training scores are shown in Figure 1B and 1C. The proportion of staff attaining a score of ≥80% rose from 54.7% and 60.4% before refresher training 1 and 2, respectively, to 84.9% and 82.8% after training (Table 3). The difference between pre- and postcourse scores (excluding those attaining 100% in the precourse MCQ) did not vary significantly (P > .8) between sites. Median precourse MCQ scores were significantly lower among nurses and clinical officers compared to doctors (Table 4); nurses who assessed controls only scored lower than those who assessed both cases and controls, though (with the exception of prerefresher training 1) this failed to reach statistical significance.

es. Median precourse MCQ scores were significantly lower among nurses and clinical officers compared to doctors (Table 4); nurses who assessed controls only scored lower than those who assessed both cases and controls, though (with the exception of prerefresher training 1) this failed to reach statistical significance. Table 3. Multiple-Choice Question Scores for Clinical Staff Assessing Pneumonia Etiology Research for Child Health (PERCH) Cases and/or Controls, by Evaluation Time-Point (All Study Sites) Evaluation (MCQ) Time-Point No. of Clinical Staffa No. of Staff With MCQ Scoreb MCQ Score Improvement With Refresher Trainingc Median % Score (IQR) Percentage Scoring ≥80 Median Difference (Post- Pre) (IQR) Percentage With Improved Scores Postbaseline trainingd 158 144 100 (90–100) 87.5 Prerefresher training 1 110 95 80 (65–90) 54.7 10 (10–20)e 93.1 Postrefresher training 1 110 99 90 (85–100) 84.9 Online MCQ 1 110 103 90 (80–100)f 90.3 Prerefresher training 2 105 96 80 (70–90) 60.4 10 (5–15)e 88.5 Postrefresher training 2 105 93 90 (80–100) 82.8 Online MCQ 2 110 99 90 (75–100)f 74.8 Abbreviations: IQR, interquartile range; MCQ, multiple-choice question; PERCH, Pneumonia Etiology Research for Child Health. aThe reduction in staff numbers after baseline training reflects staff loss, which was greatest during the pilot period and early months of recruitment.

Evaluation (MCQ) Time-Point No. of Clinical Staffa No. of Staff With MCQ Scoreb MCQ Score Improvement With Refresher Trainingc Median % Score (IQR) Percentage Scoring ≥80 Median Difference (Post- Pre) (IQR) Percentage With Improved Scores Postbaseline trainingd 158 144 100 (90–100) 87.5 Prerefresher training 1 110 95 80 (65–90) 54.7 10 (10–20)e 93.1 Postrefresher training 1 110 99 90 (85–100) 84.9 Online MCQ 1 110 103 90 (80–100)f 90.3 Prerefresher training 2 105 96 80 (70–90) 60.4 10 (5–15)e 88.5 Postrefresher training 2 105 93 90 (80–100) 82.8 Online MCQ 2 110 99 90 (75–100)f 74.8 Abbreviations: IQR, interquartile range; MCQ, multiple-choice question; PERCH, Pneumonia Etiology Research for Child Health. aThe reduction in staff numbers after baseline training reflects staff loss, which was greatest during the pilot period and early months of recruitment. bMissing values: (i) Baseline training (n = 14): All 14 staff received baseline training; 9 joined PERCH during the last 6 months of recruitment and trained online, but failed to take the final MCQ; 2 were site trainers, 1 of whom had translated all of the MCQ questions, answers, and explanations into Thai; 3 MCQ scores were mislaid. (ii) Refresher training (median, 11 [range, 7–15]): staff absent from refresher training 1 or 2 or the online MCQs were on sick, compassionate, annual, or maternity leave, or were carrying out essential ward duties. cExcludes staff scoring 100% on the prerefresher training.

bMissing values: (i) Baseline training (n = 14): All 14 staff received baseline training; 9 joined PERCH during the last 6 months of recruitment and trained online, but failed to take the final MCQ; 2 were site trainers, 1 of whom had translated all of the MCQ questions, answers, and explanations into Thai; 3 MCQ scores were mislaid. (ii) Refresher training (median, 11 [range, 7–15]): staff absent from refresher training 1 or 2 or the online MCQs were on sick, compassionate, annual, or maternity leave, or were carrying out essential ward duties. cExcludes staff scoring 100% on the prerefresher training. dBaseline training refers to the training that all staff underwent at the time of joining the study; it does not relate to a specific time-point, as new staff members were recruited throughout the study. eP < .001 with Wilcoxon signed-rank test. fP = .17 with Kruskal-Wallis test (no significant difference in distribution of scores between online MCQ1 and MCQ2). Table 4. Multiple-Choice Question Scores by Cadre and Role

cle indicates outliers (values lying outside 1.5 multiplied by the IQR). Abbreviations: BAN, Bangladesh; GAM, The Gambia; KEN, Kenya; MAL, Mali; MCQ, multiple-choice question; pre, pre-course MCQ; post, post-course MCQ; RF1, refresher training 1; RF2, refresher training 2; SAF, South Africa; THA, Thailand; ZAM, Zambia. Checklist evaluations of practical skills were carried out on 105 of 166 (63%) staff performing NP/OP swabs, 64 of 112 (57%) staff collecting IS samples, and 107 of 166 (64%) staff conducting anthropometry. Analyzing all sites combined, median checklist scores were 92% (IQR, 90%–96) for NP/OP swabs, 96% (IQR, 90%–98) for IS, and 95% (IQR, 88%–100) for anthropometry, with a median score of >82% for each of the 3 skills when analyzing by site. The video quiz took place during the final 4 months of enrollment at each site. Ninety-six of 110 current staff members participated, of whom 42 (44%) were nurses, 40 (42%) doctors, and 14 (14%) clinical officers. Percentage agreement between participants and the clinical trainer was high (≥89%) for all clinical signs (Table 5). Interobserver concordance was moderate for central cyanosis (AC1 statistic, 0.54); substantial for LCWI, deep breathing, nasal flaring, and the alert child (AC1, 0.62–0.82); and excellent for head nodding (AC1, 0.88). Table 5. Agreement With Principal Trainer and Interobserver Agreement for Select Clinical Signs

The video quiz took place during the final 4 months of enrollment at each site. Ninety-six of 110 current staff members participated, of whom 42 (44%) were nurses, 40 (42%) doctors, and 14 (14%) clinical officers. Percentage agreement between participants and the clinical trainer was high (≥89%) for all clinical signs (Table 5). Interobserver concordance was moderate for central cyanosis (AC1 statistic, 0.54); substantial for LCWI, deep breathing, nasal flaring, and the alert child (AC1, 0.62–0.82); and excellent for head nodding (AC1, 0.88). Table 5. Agreement With Principal Trainer and Interobserver Agreement for Select Clinical Signs Agreement With Trainera Interobserver Agreement Clinical Sign No. of Videos Percentage Agreement With Trainer, Median (IQR) AC1b κb LCWI 10 89.1 (85.4–95.8) 0.62 0.62 Head nodding 5 99.0 (95.8–99.0) 0.88 0.87 Deep breathing 5 92.7 (92.7–99.0) 0.82 0.80 Central cyanosis 5 90.2 (75.8–94.6) 0.54 0.54 Nasal flaring 5 95.8 (93.8–99.0) 0.79 0.68 Alert child 5 94.8 (83.3–97.9) 0.62 0.62 Abbreviations: IQR, interquartile range; LCWI, lower chest wall indrawing. aOne hundred ten staff members who assessed Pneumonia Etiology Research for Child Health (PERCH) cases and/controls were available for the video quiz. Ninety-six (87%) staff participated in the video quiz (14 missing values).

Agreement With Trainera Interobserver Agreement Clinical Sign No. of Videos Percentage Agreement With Trainer, Median (IQR) AC1b κb LCWI 10 89.1 (85.4–95.8) 0.62 0.62 Head nodding 5 99.0 (95.8–99.0) 0.88 0.87 Deep breathing 5 92.7 (92.7–99.0) 0.82 0.80 Central cyanosis 5 90.2 (75.8–94.6) 0.54 0.54 Nasal flaring 5 95.8 (93.8–99.0) 0.79 0.68 Alert child 5 94.8 (83.3–97.9) 0.62 0.62 Abbreviations: IQR, interquartile range; LCWI, lower chest wall indrawing. aOne hundred ten staff members who assessed Pneumonia Etiology Research for Child Health (PERCH) cases and/controls were available for the video quiz. Ninety-six (87%) staff participated in the video quiz (14 missing values). bFor both the AC1 and κ statistic, a value of 0 indicates no agreement beyond chance, while a value of 1 denotes perfect agreement. Values of ≤0.40 are generally indicative of poor agreement, 0.41–0.60 moderate agreement, 0.61–0.80 substantial agreement, and >0.80 excellent agreement. Staff Retention

aOne hundred ten staff members who assessed Pneumonia Etiology Research for Child Health (PERCH) cases and/controls were available for the video quiz. Ninety-six (87%) staff participated in the video quiz (14 missing values). bFor both the AC1 and κ statistic, a value of 0 indicates no agreement beyond chance, while a value of 1 denotes perfect agreement. Values of ≤0.40 are generally indicative of poor agreement, 0.41–0.60 moderate agreement, 0.61–0.80 substantial agreement, and >0.80 excellent agreement. Staff Retention Figure 2 provides an intersite comparison of staff retention for 137 staff members who attended the initial training course at their site, prior to the start of study enrollment. Retention varied by site (log-rank test for equality of survivor function across sites: P < .001) and cadre, with retention of clinical officers (86% [18/21] of whom were at the Kenya site) being significantly higher than retention of nurses and doctors (log-rank test, P = .016). Retention over 24 months of enrollment was high (>80%) in Kenya and Mali; moderate (40%–70%) in Bangladesh, Thailand, Zambia, and The Gambia; and low (<30%) in South Africa.

l officers (86% [18/21] of whom were at the Kenya site) being significantly higher than retention of nurses and doctors (log-rank test, P = .016). Retention over 24 months of enrollment was high (>80%) in Kenya and Mali; moderate (40%–70%) in Bangladesh, Thailand, Zambia, and The Gambia; and low (<30%) in South Africa. Figure 2. Staff retention during the course of the Pneumonia Etiology Research for Child Health (PERCH) study, by site. Kaplan-Meier graph displaying the proportion of staff attending the initial baseline training (N = 137) who remained with the study; analysis includes all staff members regardless of whether or not they enrolled study participants. Drops represent staff members leaving the study over time. The table beneath the graph indicates the number of staff members who remained in the study over time. Abbreviations: BAN, Bangladesh; GAM, The Gambia; KEN, Kenya; MAL, Mali; SAF, South Africa; THA, Thailand; ZAM, Zambia.

r not they enrolled study participants. Drops represent staff members leaving the study over time. The table beneath the graph indicates the number of staff members who remained in the study over time. Abbreviations: BAN, Bangladesh; GAM, The Gambia; KEN, Kenya; MAL, Mali; SAF, South Africa; THA, Thailand; ZAM, Zambia. DISCUSSION Although staff training is an important component of all clinical trials, most studies fail to document its content or evaluate and report on its effectiveness [17]. By means of MCQs, a video quiz, and checklist evaluation of practical skills, we assessed key knowledge and clinical skills of PERCH staff throughout the duration of the study. Despite considerable challenges posed by staff turnover, language differences, intersite variation in the number and cadre of staff performing clinical assessments, and political and geographic factors beyond our control, a satisfactory level of clinical standardization was achieved within and across all study sites. Because of clinical standardization, we consider that the variable proportion of very severe pneumonia cases at different PERCH sites, from 10% in Bangladesh, where screening took place in an outpatient clinic, to approximately 50% among hospitalized children in Mali and Kenya, is a true reflection of intersite differences in case severity.

andardization, we consider that the variable proportion of very severe pneumonia cases at different PERCH sites, from 10% in Bangladesh, where screening took place in an outpatient clinic, to approximately 50% among hospitalized children in Mali and Kenya, is a true reflection of intersite differences in case severity. MCQs were administered at the end of baseline training and at regular intervals throughout the study. To answer questions correctly, staff needed thorough knowledge of the PERCH case definition and inclusion and exclusion criteria, and the ability to recognize and interpret key clinical signs from the accompanying video clips. The lower MCQ scores attained in Mali following baseline training may have been because the 3-day course concluded 1 day early due to extenuating circumstances, and was delivered in French by a nonnative speaker. In Thailand and Bangladesh, courses were delivered in both English and Thai or Bangla, and MCQ scores at these sites were comparable to the scores from countries where English is more widely spoken. At all sites and time points, doctors attained significantly higher MCQ scores than clinical officers and nurses, who generally spend a shorter period of time in professional clinical training. The nurses who only assessed healthy controls scored worse than nurses assessing both cases and controls, probably because they were exposed to fewer children with clinical signs.

icantly higher MCQ scores than clinical officers and nurses, who generally spend a shorter period of time in professional clinical training. The nurses who only assessed healthy controls scored worse than nurses assessing both cases and controls, probably because they were exposed to fewer children with clinical signs. Clinical video has been shown previously to be an effective way of testing agreement between clinicians on the presence or absence of clinical signs [6], despite the obvious difference from the “real-life” clinical situation, in which a clinician’s judgement is affected by information other than an isolated clinical sign. The same study showed that health workers of different cadres and varying levels of clinical experience could correctly identify clinical signs from video recordings for which there was high proportionate agreement between experts [6]. Clinical signs are not, however, always clear-cut in real-life. To this end, the PERCH video quiz included a random selection (approximately 20%) of “gray” cases—namely, those in which a clinical sign (eg, LCWI) was present but subtle, making it genuinely difficult to decide on its presence or absence. Despite this, percentage agreement between staff and the trainer was ≥89% for all 6 clinical signs in the quiz, while interobserver agreement (agreement between participants) varied from “moderate” for central cyanosis, a clinical sign which is easily missed in African children and which is difficult to photograph or film successfully, to “substantial” or “excellent” for the other clinical signs. Good-quality clinical video clips are a valuable and scarce resource, and we hope that the video clips available on the PERCH clinical standardization training website (www.perchtraining.org) will be useful for other clinical researchers.

film successfully, to “substantial” or “excellent” for the other clinical signs. Good-quality clinical video clips are a valuable and scarce resource, and we hope that the video clips available on the PERCH clinical standardization training website (www.perchtraining.org) will be useful for other clinical researchers. Although the PERCH clinical standardization program successfully attained its objectives, a number of useful lessons have been learned. It would have been informative to evaluate staff knowledge and skills prior to the initial training course, as this would have provided a useful baseline comparator for the subsequent MCQ scores. The improvement in MCQ scores with refresher training suggests that it would have been valuable to have had more regular refresher training courses at each site, coordinated by local site trainers. Limited availability of personal computers and slow internet speeds reduced the utility of the training website at several of the study sites. These shortcomings are not shared by mobile phone technology, which could provide a useful alternative platform for training and evaluation. It took time to obtain a sufficient number of good-quality video clips of relevant clinical signs, and consequently the video quiz took place during the final 4 months of enrollment, by which time many of the original PERCH staff had left the study. It would have been preferable to organize the quiz at the start of enrollment, and repeat it during the second year. Although the checklist evaluations of practical skills were useful training and evaluation tools, they were time-consuming and were consequently performed on approximately 60% of the relevant study staff. High rates of staff turnover emphasized the importance of establishing a robust system for training new staff outside of the regular training schedule. Turnover was lowest among clinical officers, which may reflect their longer-term clinical attachments.

ently performed on approximately 60% of the relevant study staff. High rates of staff turnover emphasized the importance of establishing a robust system for training new staff outside of the regular training schedule. Turnover was lowest among clinical officers, which may reflect their longer-term clinical attachments. There is increasing recognition that public health policy should be based on data that are globally representative. Enhanced connectivity, the widespread availability of powerful computing, statistical and data management tools, and the advent of funders willing to pay for large networked studies have increased the feasibility of conducting large, multicountry research studies. Ensuring that the clinical and laboratory data obtained during the course of such studies are robust, standardized, and comparable is of paramount importance. The results of the PERCH clinical standardization program give us confidence that any etiological or clinical differences observed across the study sites are true differences, and not attributable to differences in application of the clinical case definition or differences in techniques used for clinical measurements or specimen collection. We hope that the methods, results, and lessons learned from the PERCH clinical standardization program will usefully inform other researchers embarking on large-scale clinical or epidemiological studies of pneumonia or other major causes of childhood morbidity and mortality.

linical measurements or specimen collection. We hope that the methods, results, and lessons learned from the PERCH clinical standardization program will usefully inform other researchers embarking on large-scale clinical or epidemiological studies of pneumonia or other major causes of childhood morbidity and mortality. Supplementary Data Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author. Supplementary Material Supplementary_Material Click here for additional data file. Notes Author contributions. J. C. designed the clinical standardization materials and was the principal trainer. O. S. L., K. L. O., D. R. F., D. R. M., M. D. K., L. L. H., H. C. B., W. A. B., S. R. C. H., K. K. L., S. A. M., J. A. G. S., D. M. T., S. C. M., and R. A. K. conceived and designed the study and supervised study conduct. C. P., A. N. D., M. D. K., J. A. G. S., L. L. H., and D. R. F. participated in the analysis and interpretation of results and drafting of the manuscript. J. O. A., C. B., A. N. D., A. J. D., B. E. E., D. G., M. M. H., R. A. K., S. K., N. K., G. M., D. P. M., A. M., M. M., K. N., D. E. P., B. P., P. S., and M. S. were involved in the study conduct, data collection, and/or data management. All authors review and approved the manuscript. J. C. had final responsibility for the decision to submit for publication.

E. E., D. G., M. M. H., R. A. K., S. K., N. K., G. M., D. P. M., A. M., M. M., K. N., D. E. P., B. P., P. S., and M. S. were involved in the study conduct, data collection, and/or data management. All authors review and approved the manuscript. J. C. had final responsibility for the decision to submit for publication. Acknowledgments. We thank the following individuals and organizations for providing clinical video clips and training materials: WHO, Department of Child and Adolescent Health and Department of Nutrition; Professor Mike English; Professor Kathryn Maitland; Professor Karen Kotloff; GlaxoSmithKline; Rale Repository; COPAN Diagnostics. We are grateful to Dr David Peel and Dr Irwin Shorr for their valuable advice on pulse oximetry and anthropometry, respectively. We thank Incuna Ltd for their invaluable support in the development of the PERCH clinical standardization website. We offer sincere thanks to the patients and families who participated in this study. We are grateful to the following members of the original study teams for running the studies and collecting data: Disclaimer. The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention, Department of Health and Human Services, or the US government. This paper is published with the permission of the Director of the Kenya Medical Research Institute.

onclusions in this report are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention, Department of Health and Human Services, or the US government. This paper is published with the permission of the Director of the Kenya Medical Research Institute. Financial support. PERCH was supported by the Bill & Melinda Gates Foundation (grant number 48968 to the International Vaccine Access Center, Department of International Health, Johns Hopkins Bloomberg School of Public Health). J. A. G. S. was supported by a clinical fellowship from The Wellcome Trust of Great Britain (award number 098532). Supplement sponsorship. This article appears as part of the supplement “Pneumonia Etiology Research for Child Health (PERCH): Foundational Basis for the Primary Etiology Results,” sponsored by a grant from the Bill & Melinda Gates Foundation to the PERCH study of Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland.

onsorship. This article appears as part of the supplement “Pneumonia Etiology Research for Child Health (PERCH): Foundational Basis for the Primary Etiology Results,” sponsored by a grant from the Bill & Melinda Gates Foundation to the PERCH study of Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland. Potential conflicts of interest. M. D. K. has received funding for consultancies from Merck, Pfizer, and Novartis, and grant funding from Merck. L. L. H. has received grant funding from Pfizer and GlaxoSmithKline. K. L. K. has received grant funding from Merck Sharp & Dohme. S. A. M. has received honoraria for advisory board membership from the Bill & Melinda Gates Foundation, Pfizer, Medimmune, and Novartis; has received institutional grants from GSK, Novartis, Pfizer, Minervax, and the Bill & Melinda Gates Foundation; and has served on speaker’s bureaus for Sanofi Pasteur and GSK. K. L. O. has received grant funding from GSK and Pfizer and participates on technical advisory boards for Merck, Sanofi Pasteur, PATH, Affinivax, and ClearPath. All other authors report no potential conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

For diagnosing viral pneumonia, upper respiratory tract (URT) specimens have become the most common specimen type due to their logistical ease of collection [1, 2]. However, detection of viruses in URT specimens has low specificity as this finding might simply reflect an URT infection without lower respiratory tract involvement or coincidental asymptomatic or past infection [2–4]. A possible solution to the lack of specificity of simply detecting the presence or absence of a virus in the URT of pneumonia patients is to determine whether the density of a virus in the URT can better distinguish its causative role in pneumonia. There are reports that a higher pathogen load in the URT is associated with pneumonia, for both Streptococcus pneumoniae and some respiratory viruses [5–8]. In addition, for some viruses, higher viral load in the URT has been associated with worse outcomes [7, 9–11]. In this analysis, we describe viral load in nasopharyngeal/oropharyngeal (NP/OP) specimens from cases and community controls from a large multicountry childhood pneumonia study (Pneumonia Etiology Research for Child Health [PERCH]), as well as demographic and clinical characteristics associated with higher viral load and disease severity. An overarching aim was to explore whether the incorporation of viral load data into the main PERCH etiology analysis might improve the assignment of the etiology of pneumonia cases.

rch for Child Health [PERCH]), as well as demographic and clinical characteristics associated with higher viral load and disease severity. An overarching aim was to explore whether the incorporation of viral load data into the main PERCH etiology analysis might improve the assignment of the etiology of pneumonia cases. METHODS The PERCH study design and enrollment strategy has been previously described [12]. In brief, PERCH is a case-control study of the etiology of World Health Organization (WHO)–defined severe and very severe pneumonia among hospitalized children aged 1–59 months and age frequency-matched community controls. Enrollment took place during August 2011–January 2014 for 24 months at each of 9 study sites located in 7 countries—Dhaka and Matlab, Bangladesh; Basse, The Gambia; Kilifi, Kenya; Bamako, Mali; Soweto, South Africa; Nakhon Phanom and Sa Kaeo, Thailand; and Lusaka, Zambia [13]. Case and Control Definitions For this analysis, we included only cases with evidence of pneumonia on chest radiograph, defined as consolidation and/or any other infiltrate assessed according to the WHO radiological pneumonia criteria [14]. A control participant was considered to have a respiratory tract illness (RTI) if cough or runny nose was reported. RTI was also considered present if a child had (1) at least 1 of ear discharge, wheezing, or difficulty breathing and (2) either a measured temperature of ≥38.0°C within the previous 48 hours or a history of sore throat. Controls who did not meet the definition of RTI are referred to as non-RTI controls.

was reported. RTI was also considered present if a child had (1) at least 1 of ear discharge, wheezing, or difficulty breathing and (2) either a measured temperature of ≥38.0°C within the previous 48 hours or a history of sore throat. Controls who did not meet the definition of RTI are referred to as non-RTI controls. Specimen Collection and Laboratory Testing Nasopharyngeal and oropharyngeal swabs were collected from PERCH cases and controls at enrollment. Nasopharyngeal specimens were collected by inserting flocked swabs (Copan ETC) into the posterior nasopharynx and rotating 180° for 2–3 seconds. Oropharyngeal specimens were then collected by rubbing Rayon swabs (Fisher Scientific) over both tonsillar pillars and the posterior oropharynx for several seconds. Following collection, swabs were placed together in the same 3-mL vial of universal transport media (Copan) and processed within 24 hours of collection. Specimens were left at room temperature for no more than 2 hours or at 4°C for no more than 24 hours, before freezing at –70°C.

sterior oropharynx for several seconds. Following collection, swabs were placed together in the same 3-mL vial of universal transport media (Copan) and processed within 24 hours of collection. Specimens were left at room temperature for no more than 2 hours or at 4°C for no more than 24 hours, before freezing at –70°C. All specimens were tested in-country using a standardized methodology, and details are described elsewhere [15]. Specimens were evaluated using the Fast-track Diagnostics Respiratory Pathogens 33 test (FTD Resp 33, Fast-track Diagnostics, Sliema, Malta), a 33-target, 8-multiplex real-time polymerase chain reaction (PCR) platform for the detection of viruses, bacteria, and fungi. The 18 viruses or virus classes included influenza A, B, and C viruses; parainfluenza virus (PIV) types 1, 2, 3, and 4; coronaviruses NL63, 229E, OC43, and HKU1; human metapneumovirus (HMPV) A and B (A and B not differentiated); rhinovirus; respiratory syncytial virus (RSV) A and B (A and B not differentiated); adenovirus; enterovirus and parechovirus (not differentiated); human bocavirus (HBOV); and cytomegalovirus. Cytomegalovirus is not included in this analysis but is discussed in a separate publication of the pathogen load of pathogens commonly detected in both cases and controls [16]. Some sites (Bangladesh, The Gambia, Mali, South Africa) collected lung aspirates from children with consolidation on chest radiograph who met clinical and radiologic criteria for the procedure [17]. Lung aspirate specimens were tested for viral targets using the same method described for NP/OP specimens.

All specimens were tested in-country using a standardized methodology, and details are described elsewhere [15]. Specimens were evaluated using the Fast-track Diagnostics Respiratory Pathogens 33 test (FTD Resp 33, Fast-track Diagnostics, Sliema, Malta), a 33-target, 8-multiplex real-time polymerase chain reaction (PCR) platform for the detection of viruses, bacteria, and fungi. The 18 viruses or virus classes included influenza A, B, and C viruses; parainfluenza virus (PIV) types 1, 2, 3, and 4; coronaviruses NL63, 229E, OC43, and HKU1; human metapneumovirus (HMPV) A and B (A and B not differentiated); rhinovirus; respiratory syncytial virus (RSV) A and B (A and B not differentiated); adenovirus; enterovirus and parechovirus (not differentiated); human bocavirus (HBOV); and cytomegalovirus. Cytomegalovirus is not included in this analysis but is discussed in a separate publication of the pathogen load of pathogens commonly detected in both cases and controls [16]. Some sites (Bangladesh, The Gambia, Mali, South Africa) collected lung aspirates from children with consolidation on chest radiograph who met clinical and radiologic criteria for the procedure [17]. Lung aspirate specimens were tested for viral targets using the same method described for NP/OP specimens. Statistical Analysis Human immunodeficiency virus (HIV)–positive cases were excluded from analyses unless stated otherwise. PCR quantification was log10-transformed. Demographic characteristics of cases and controls were compared using the χ2 test. All controls, both RTI and non-RTI, were included in the main analysis. All analyses of viral load were restricted to children positive for each virus. Among children positive for each virus, t tests adjusted for site and age were performed to compare mean cycle threshold (Ct) values between cases and controls. For each virus, a trend analysis, using simple linear regression, was performed to test if viral density increased with age for cases and for controls. Among cases only, mean Ct values were also compared by days since onset, severity, vital status, and HIV status. Multivariable logistic regression, adjusting for age and site, was performed to compare odds of being a case for each 3.4-unit drop in Ct value, which was approximately equivalent to a 1 log10 increase in viral copies/mL; Ct values instead of viral density were used for regression because viral density was only accurate within the linear range of the assay (104–108 copies/mL). Kernel density distribution plots were created to show distributions of viral density by case/control status. Receiver operating characteristic (ROC) curves and the corresponding area under the curve (AUC) were generated to investigate the performance of viral load in determining case status among children positive by NP/OP PCR for each virus, and the Youden index was calculated to determine the optimized diagnostic cutoffs to differentiate cases and controls [18]. To guard against bias in the estimates of sensitivity due to having a small number of some viruses detected among cases, the Youden index was calculated using leave-one-out cross-validation where sample size was sufficient [19]. Redefining positivity using the optimal cutpoints, we calculated odds ratios associated with case status for children above vs below the optimal cutpoint including negatives.

of some viruses detected among cases, the Youden index was calculated using leave-one-out cross-validation where sample size was sufficient [19]. Redefining positivity using the optimal cutpoints, we calculated odds ratios associated with case status for children above vs below the optimal cutpoint including negatives. The proportion of radiographically confirmed cases attributable to each virus [population attributable fraction: population prevalence × (1 – 1 / OR)] was calculated using 2 methods: (1) any positive vs negative and (2) positive above vs below the optimal cutpoint, the former method being optimal for laboratory sensitivity and the latter for a balance of epidemiological sensitivity and specificity. All analyses were performed using SAS software version 9.4 (SAS Institute, Cary North Carolina) and R Statistical Software 3.2.1 (R Foundation for Statistical Computing, Vienna, Austria). All P values are 2 sided. Ethical Considerations The PERCH study protocol was approved by the institutional review board or ethical review committee at each of the study site institutions and at the Johns Hopkins Bloomberg School of Public Health. Parents or guardians of all participants provided written informed consent. RESULTS Of 1935 radiographically confirmed cases and 5325 controls in PERCH, 1733 cases (1227 severe, 506 very severe) and 4986 controls (1185 RTI and 3801 non-RTI) were not known to be HIV infected and had viral density results available for analysis (Table 1).

Ethical Considerations The PERCH study protocol was approved by the institutional review board or ethical review committee at each of the study site institutions and at the Johns Hopkins Bloomberg School of Public Health. Parents or guardians of all participants provided written informed consent. RESULTS Of 1935 radiographically confirmed cases and 5325 controls in PERCH, 1733 cases (1227 severe, 506 very severe) and 4986 controls (1185 RTI and 3801 non-RTI) were not known to be HIV infected and had viral density results available for analysis (Table 1). Table 1. Characteristics of Chest Radiograph–Positive Children With Severe and Very Severe Pneumonia and Controls—Pneumonia Etiology Research for Child Health (PERCH) Study, August 2011–January 2014 Characteristic CXR+Casesa (n = 1733) All Controls (n = 4986) χ2P Valueb Site Kenya 282 (16.3) 855 (17.2) <.001 The Gambia 273 (15.8) 624 (12.5) Mali 239 (13.8) 724 (14.5) Zambia 189 (10.9) 535 (10.7) South Africa 433 (25.0) 823 (16.5) Thailand 98 (5.7) 657 (13.2) Bangladesh 219 (12.6) 768 (15.4) Age 1–5 mo 680 (39.2) 1555 (31.2) <.001 6–11 mo 415 (24.0) 1187 (23.8) 12–23 mo 424 (24.5) 1235 (24.8) 24–59 mo 214 (12.4) 1009 (20.2) Female sex 756 (43.6) 2477 (49.7) <.001 Prior antibiotic usec 597 (42.4) 84 (1.7) <.001 Respiratory tract illnessd NA 1185 (23.8) NA No. of viruses detected 0 viruses 180 (10.4) 1048 (21.0) 1 virus 628 (36.2) 1928 (38.7) <.001 2 viruses 616 (35.6) 1420 (28.5) ≥3 viruses 309 (17.8) 590 (11.8) Data are presented as No. (%) unless otherwise indicated. Abbreviations: CXR+, chest radiograph positive; NA, not applicable.

Respiratory tract illnessd NA 1185 (23.8) NA No. of viruses detected 0 viruses 180 (10.4) 1048 (21.0) 1 virus 628 (36.2) 1928 (38.7) <.001 2 viruses 616 (35.6) 1420 (28.5) ≥3 viruses 309 (17.8) 590 (11.8) Data are presented as No. (%) unless otherwise indicated. Abbreviations: CXR+, chest radiograph positive; NA, not applicable. aCXR+ defined as having radiographic evidence of pneumonia. bComparing distribution of characteristics between CXR+ cases and controls. Bolded values are significant (P < .05). cPrior antibiotic use: administered antibiotics at the study facility prior to the collection of specimens (cases only), antibiotics at a referral facility (cases only), or positive serum bioassay (cases and controls). dSee Methods for respiratory tract illness definition. Overall, 89.6% of cases and 79.0% of controls had at least 1 virus detected, and 53.4% and 40.3%, respectively, had ≥2 viruses detected (Table 1). Among the 17 viruses tested, RSV was the most commonly detected among cases (27%) but was uncommon among controls (3%) (Table 2). Rhinovirus was the next most commonly detected virus in cases but was present at a similar frequency among controls (21% for both). Table 2. Mean Nasopharyngeal/Oropharyngeal Polymerase Chain Reaction Cycle Threshold Values for Chest Radiograph–Positive Cases and Controls and Odds Ratios for Viral Load Being Predictive of Case Status—Pneumonia Etiology Research for Child Health (PERCH) Study, August 2011–January 2014

Overall, 89.6% of cases and 79.0% of controls had at least 1 virus detected, and 53.4% and 40.3%, respectively, had ≥2 viruses detected (Table 1). Among the 17 viruses tested, RSV was the most commonly detected among cases (27%) but was uncommon among controls (3%) (Table 2). Rhinovirus was the next most commonly detected virus in cases but was present at a similar frequency among controls (21% for both). Table 2. Mean Nasopharyngeal/Oropharyngeal Polymerase Chain Reaction Cycle Threshold Values for Chest Radiograph–Positive Cases and Controls and Odds Ratios for Viral Load Being Predictive of Case Status—Pneumonia Etiology Research for Child Health (PERCH) Study, August 2011–January 2014 Virus CXR+ Casesa (n = 1733) All Controls (n = 4986) P Valued OR per 1 Log10 Increase, Copies/mL (95% CI)e No.b (%)b Ct Value Meanc (95% CI) No.b (%)b Ct Value Meanc (95% CI) Adenovirus 164 (9.5) 27.7 (26.8–28.5) 594 (11.9) 29.5 (29.2–29.8) <.001

Table 2. Mean Nasopharyngeal/Oropharyngeal Polymerase Chain Reaction Cycle Threshold Values for Chest Radiograph–Positive Cases and Controls and Odds Ratios for Viral Load Being Predictive of Case Status—Pneumonia Etiology Research for Child Health (PERCH) Study, August 2011–January 2014 Virus CXR+ Casesa (n = 1733) All Controls (n = 4986) P Valued OR per 1 Log10 Increase, Copies/mL (95% CI)e No.b (%)b Ct Value Meanc (95% CI) No.b (%)b Ct Value Meanc (95% CI) Adenovirus 164 (9.5) 27.7 (26.8–28.5) 594 (11.9) 29.5 (29.2–29.8) <.001 1.27 (1.10–1.46) Coronavirus 229 18 (1.1) 31.1 (28.6–33.5) 54 (1.1) 30.2 (28.3–32.0) .58 0.89 (.62–1.26) Coronavirus 43 38 (2.2) 26.4 (24.4–28.5) 192 (3.9) 28.0 (27.1–28.8) .30 1.13 (.91–1.39) Coronavirus 63 36 (2.1) 27.0 (25.3–28.7) 158 (3.2) 28.5 (27.7–29.3) .26 1.18 (.90–1.55) Coronavirus HKU 37 (2.2) 29.2 (27.0–31.4) 111 (2.2) 27.7 (26.5–28.9) .40 0.91 (.74–1.13) Influenza A 62 (3.6) 28.5 (27.7–29.4) 57 (1.2) 29.8 (28.4–31.2) .31 1.21 (.85–1.72) Influenza B 18 (1.1) 27.6 (25.7–29.5) 29 (0.6) 28.5 (26.7–30.3) .82 1.07 (.63–1.83) Influenza C 10 (0.6) 28.1 (24.8–31.4) 29 (0.6) 27.3 (25.3–29.3) .14 0.44 (.17–1.15) HBOV 231 (13.4) 30.5 (29.6–31.3) 660 (13.3) 31.7 (31.3–32.1) .007 1.13 (1.03–1.24) HMPV A/B 185 (10.8) 28.1 (27.6–28.7) 206 (4.1) 28.9 (28.2–29.5) .02 1.23 (1.03–1.46) Parainfluenza 1 89 (5.2) 26.1 (24.9–27.2) 49 (1.0) 29.4 (27.6–31.2) .008 1.37 (1.08–1.74) Parainfluenza 2 23 (1.3) 34.0 (31.7–36.3) 53 (1.1) 35.1 (33.9–36.3) .70 1.10 (.71–1.69) Parainfluenza 3 104 (6.1) 25.0 (24.0–25.9) 142 (2.9) 29.0 (28.0–30.0) <.001

1.13 (1.03–1.24) HMPV A/B 185 (10.8) 28.1 (27.6–28.7) 206 (4.1) 28.9 (28.2–29.5) .02 1.23 (1.03–1.46) Parainfluenza 1 89 (5.2) 26.1 (24.9–27.2) 49 (1.0) 29.4 (27.6–31.2) .008 1.37 (1.08–1.74) Parainfluenza 2 23 (1.3) 34.0 (31.7–36.3) 53 (1.1) 35.1 (33.9–36.3) .70 1.10 (.71–1.69) Parainfluenza 3 104 (6.1) 25.0 (24.0–25.9) 142 (2.9) 29.0 (28.0–30.0) <.001 1.47 (1.22–1.77) Parainfluenza 4 44 (2.6) 31.7 (30.3–33.1) 86 (1.7) 32.2 (31.3–33.1) .88 0.98 (.73–1.31) PV/EV 131 (7.6) 30.1 (29.5–30.8) 423 (8.5) 30.4 (30.0–30.7) .45 1.08 (.89–1.31) Rhinovirus 365 (21.2) 31.7 (31.3–32.0) 1056 (21.2) 32.4 (32.3–32.6) .003 1.21 (1.08–1.35) RSV 461 (26.8) 22.2 (21.8–22.5) 140 (2.8) 27.0 (26.1–28.0) <.001 2.02 (1.71–2.37) Abbreviations: CI, confidence interval; Ct, cycle threshold; CXR+, chest radiograph positive; HBOV, human bocavirus; HMPV, human metapneumovirus; OR, odds ratio; PV/EV, parechovirus/enterovirus; RSV, respiratory syncytial virus. aCXR+ defined as having radiographic evidence of pneumonia. bNo. (%) positive in the nasopharynx/oropharynx among those with available results for the given virus. cAmong those with a positive density. dComparing mean cycle threshold value of CXR+ cases vs all controls using linear regression adjusting for age and site. Bolded values are significant (P < .05). eOdds ratio is for approximately each 3.4-unit drop in Ct value (equivalent to approximately 1 log10 increase in copies/mL) adjusting for age and site using logistic regression. Bolded values are significant (P < .05).

dComparing mean cycle threshold value of CXR+ cases vs all controls using linear regression adjusting for age and site. Bolded values are significant (P < .05). eOdds ratio is for approximately each 3.4-unit drop in Ct value (equivalent to approximately 1 log10 increase in copies/mL) adjusting for age and site using logistic regression. Bolded values are significant (P < .05). Analysis of Viral Load Among Cases and Controls RSV had the highest mean viral load among cases (7.3 log copies/mL; Figure 1); no viruses other than RSV had a mean viral load >6 log copies/mL. Among controls, no viruses had a mean viral load >6 log copies/mL. Eight viruses among cases (RSV, influenza C, PIV1, PIV3, PIV4, coronavirus 43, coronavirus 63, and HMPV) had mean viral loads >5 log copies/mL vs 5 viruses among controls (RSV, influenza C, PIV4, coronavirus 43, and coronavirus 63). There were 7 viruses that had significantly higher mean viral density among cases after adjusting for site and age—adenovirus, HBOV, HMPV, PIV1, PIV3, rhinovirus, and RSV. After adjusting for age and site, there was a significant increase in the odds of being a case (vs a control) for each 3.4-unit drop in Ct value (approximately 1 log increase in copies/mL) for the same 7 viruses, ranging from a 13% increased odds for HBOV to a 102% increased odds for RSV (Table 2).

rhinovirus, and RSV. After adjusting for age and site, there was a significant increase in the odds of being a case (vs a control) for each 3.4-unit drop in Ct value (approximately 1 log increase in copies/mL) for the same 7 viruses, ranging from a 13% increased odds for HBOV to a 102% increased odds for RSV (Table 2). Figure 1. Nasopharyngeal/oropharyngeal viral load (log10 copies/mL) for chest radiograph–positive (CXR+) cases and all controls among those in which the virus was detected—Pneumonia Etiology Research for Child Health (PERCH) study, August 2011–January 2014. CXR+ defined as having radiographic evidence of pneumonia. Box-and-whiskers plot features include the following: central line in box is median, bottom line of box is first quartile (25%), top line of box is third quartile (75%), diamond is mean, and top and bottom of whiskers represent 95% confidence intervals. Area above the upper dotted line and below the lower dotted line indicate areas outside the linear range of the assay for calculation of viral load from cycle threshold (Ct) values where there is a greater degree of uncertainty in viral density calculations. Numbers on x-axis indicate number of positive results for that virus. *P value comparing mean Ct value between controls and CXR+ cases <.05 after adjusting for age and site. Abbreviations: Adeno, adenovirus; Boca, human bocavirus; CXR, chest radiograph; Flu, influenza virus; HCoV, human coronavirus; HMPV, human metapneumovirus; Para, parainfluenza virus; PV/EV, parechovirus/enterovirus; Rhino, rhinovirus; RSV, respiratory syncytial virus.

d CXR+ cases <.05 after adjusting for age and site. Abbreviations: Adeno, adenovirus; Boca, human bocavirus; CXR, chest radiograph; Flu, influenza virus; HCoV, human coronavirus; HMPV, human metapneumovirus; Para, parainfluenza virus; PV/EV, parechovirus/enterovirus; Rhino, rhinovirus; RSV, respiratory syncytial virus. Viral load was similar between RTI and non-RTI controls for most viruses with the exception of RSV, where the mean viral load was significantly higher for RTI controls (Supplementary Figure 1). Despite the differences in viral load between cases and controls noted above, there was substantial overlap in the viral density distribution between cases and controls in which virus was detected, as shown in the box plots and kernel density distribution plots (Figures 1 and 2). Kernel density distribution plots were examined for a bimodal distribution with a smaller subset of cases having viruses at a higher viral load that might be indicative of those with pneumonia due to that virus. Adenovirus, coronavirus 229, and PIV1–3 had a suggestion of a bimodal distribution among cases. The NP/OP viral load for the 2 cases with viruses detected in PCR of lung aspirates (ie, HMPV and adenovirus) fell within the distribution of viral loads for other cases, as well as controls, positive for that virus (Figure 2).

novirus, coronavirus 229, and PIV1–3 had a suggestion of a bimodal distribution among cases. The NP/OP viral load for the 2 cases with viruses detected in PCR of lung aspirates (ie, HMPV and adenovirus) fell within the distribution of viral loads for other cases, as well as controls, positive for that virus (Figure 2). Figure 2. Kernel density distribution plots comparing nasopharyngeal/oropharyngeal (NP/OP) viral load among chest radiograph–positive (CXR+) cases and all controls for each viral polymerase chain reaction (PCR) target—Pneumonia Etiology Research for Child Health (PERCH) study, August 2011–January 2014. Tick marks across the top of each plot indicate viral load of each individual (first row of black ticks for cases and second row of gray ticks for controls). Dashed curves indicate areas outside the linear range of the assay for calculation of viral load from cycle threshold values. Dotted dashed vertical lines indicate optimal cutpoint distinguishing cases and controls calculated using Youden index. Black arrows in adenovirus and human metapneumovirus plots indicate NP/OP viral load of cases whose lung aspirate specimen was available and PCR positive for that virus. Abbreviations: Adeno, adenovirus; Boca, human bocavirus; CXR, chest radiograph; Flu, influenza virus; HCoV, human coronavirus; HMPV, human metapneumovirus; Para, parainfluenza virus; PV/EV, parechovirus/enterovirus; Rhino, rhinovirus; RSV, respiratory syncytial virus.

imen was available and PCR positive for that virus. Abbreviations: Adeno, adenovirus; Boca, human bocavirus; CXR, chest radiograph; Flu, influenza virus; HCoV, human coronavirus; HMPV, human metapneumovirus; Para, parainfluenza virus; PV/EV, parechovirus/enterovirus; Rhino, rhinovirus; RSV, respiratory syncytial virus. When constructing ROC curves, no virus had an AUC >0.8 (Table 3). RSV had the highest AUC at 0.76, with only 3 other viruses having an AUC between 0.6 and 0.7 (influenza A, PIV1, and PIV3). Despite the low values for the AUC, when redefining positive for a virus as those with viral loads above the ROC optimal cutpoint value as determined by the Youden index, the odds ratio for predicting case status increased substantially for some viruses, approximately doubling for adenovirus, coronavirus 63, PIV2, and RSV (Table 4). Although the odds ratios increased, the population attributable fraction for most viruses did not change substantially, or even decreased (eg, influenza A, RSV) due to the lower frequency of cases with densities above the optimal cutpoint (Figure 3). This is because while the odds ratios are higher at the higher density cutoff, the prevalence of cases above the higher threshold was lower, and thus the proportion of PERCH cases assigned to the virus would not change appreciably by using the higher cutoff.

f cases with densities above the optimal cutpoint (Figure 3). This is because while the odds ratios are higher at the higher density cutoff, the prevalence of cases above the higher threshold was lower, and thus the proportion of PERCH cases assigned to the virus would not change appreciably by using the higher cutoff. Table 3. Receiver Operating Characteristic Areas Under the Curve, Optimal Nasopharyngeal/Oropharyngeal Polymerase Chain Reaction Density Cutpoints for Determining Case Status, and Associated Positive Rate in Cases and Negative Rate in Controls by Virus Among Chest Radiograph–Positive Cases and Controls With Positive Densities—Pneumonia Etiology Research for Child Health (PERCH) Study, August 2011–January 2014 Virusa AUC Optimal Cutpointb, (Log10 Copies/ mL) Proportion of CXR+c Cases Above Cutpoint Proportion of Controls Below Cutpoint Adenovirus 0.60 4.88 0.44 0.78 Coronavirus 43 0.57 6.94 0.36 0.74 Coronavirus 63 0.58 7.24 0.22 0.89 Influenza A 0.61 5.12 0.50 0.68 Influenza B 0.55 3.79 0.89 0.28 HBOV 0.54 5.81 0.20 0.89 HMPV A/B 0.54 3.9 0.91 0.20 Parainfluenza 1 0.65 4.62 0.75 0.55 Parainfluenza 2 0.54 5.64 0.26 0.91 Parainfluenza 3 0.69 4.75 0.81 0.54 PV/EV 0.53 4.38 0.40 0.71 Rhinovirus 0.56 3.64 0.55 0.56 RSV 0.76 6.30 0.84 0.59 Abbreviations: AUC, area under the curve; CXR+, chest radiograph positive; HBOV, human bocavirus; HMPV, human metapneumovirus; PV/EV, parechovirus/enterovirus; RSV, respiratory syncytial virus.

0.26 0.91 Parainfluenza 3 0.69 4.75 0.81 0.54 PV/EV 0.53 4.38 0.40 0.71 Rhinovirus 0.56 3.64 0.55 0.56 RSV 0.76 6.30 0.84 0.59 Abbreviations: AUC, area under the curve; CXR+, chest radiograph positive; HBOV, human bocavirus; HMPV, human metapneumovirus; PV/EV, parechovirus/enterovirus; RSV, respiratory syncytial virus. aViruses with adjusted odds ratios <1 (see Table 2) were excluded from table (coronavirus 229, coronavirus HKU, influenza C, and parainfluenza 4). bCalculated using Youden index and, where possible, leave-one-out cross-validation. Leave-one-out cross-validation was not performed for influenza B, parainfluenza 2, or parechovirus/enterovirus. cCXR+ defined as having radiographic evidence of pneumonia. Table 4. Nasopharyngeal/Oropharyngeal Prevalence of Viruses in Chest Radiograph–Positive Cases and Controls, Defining Positive as Any Detection of Virus and Detection of Virus Above an Optimal Viral Load Cutpoint as Determined by Receiver Operating Characteristic Curves; Odds of Determining Case Status Using Both Definitions of Positive— Pneumonia Etiology Research for Child Health (PERCH) Study, August 2011–January 2014

ols, Defining Positive as Any Detection of Virus and Detection of Virus Above an Optimal Viral Load Cutpoint as Determined by Receiver Operating Characteristic Curves; Odds of Determining Case Status Using Both Definitions of Positive— Pneumonia Etiology Research for Child Health (PERCH) Study, August 2011–January 2014 Virusa Negative Weak Positivec (Below Optimal Cutpoint) Strong Positived (Above Optimal Cutpoint) AORe (95% CI) Any Positive vs Negativef AORe (95% CI) Strong Positive vs Weak Positive/ Negativeg CXR+ Casesb Controls CXR+ Casesb Controls CXR+ Casesb Controls Adenovirus 1556 (90.5) 4384 (88.0) 92 (5.3) 470 (9.4) 72 (4.2) 126 (2.5) 0.88 (.73–1.07) 1.74 (1.28–2.36) Coronavirus 43 1681 (97.7) 4785 (96.1) 25 (1.5) 149 (3.0) 14 (0.8) 43 (0.9) 0.52 (.36–.74) 0.81 (.44–1.49) Coronavirus 63 1684 (97.9) 4819 (96.8) 28 (1.6) 147 (3.0) 8 (0.5) 11 (0.2) 0.63 (.43–.91) 1.78 (.71–4.47) Influenza A 1658 (96.4) 4920 (98.9) 31 (1.8) 41 (0.8) 31 (1.8) 16 (0.3) 3.11 (2.15–4.49) 5.32 (2.87–9.85) Influenza B 1702 (99.0) 4948 (99.4) 2 (0.1) 8 (0.2) 16 (0.9) 21 (0.4) 1.82 (1.00–3.32) 2.24 (1.15–4.36) HBOV 1488 (86.6) 4316 (86.7) 184 (10.7) 593 (11.9) 47 (2.7) 68 (1.4) 1.11 (.94–1.31) 2.02 (1.38–2.96) HMPV A/B 1534 (89.2) 4771 (95.9) 16 (0.9) 47 (0.9) 169 (9.8) 159 (3.2) 2.59 (2.09–3.21) 3.02 (2.40–3.82) Parainfluenza 1 1630 (94.8) 4928 (99.0) 23 (1.3) 27 (0.5) 66 (3.8) 22 (0.4) 5.19 (3.60–7.49) 8.09 (4.92–13.32) Parainfluenza 2 1697 (98.7) 4927 (98.9) 18 (1.0) 48 (1.0) 5 (0.3) 5 (0.1) 1.2 (.73–1.97) 2.55 (.72–9.04) Parainfluenza 3 1616 (94.0) 4838 (97.1) 21 (1.2) 78 (1.6) 83 (4.8) 64 (1.3) 2.13 (1.63–2.77) 3.52 (2.51–4.92) PV/EV 1589 (92.4) 4555 (91.5) 79 (4.6) 303 (6.1) 52 (3.0) 122 (2.4) 0.91 (.74–1.12) 1.22 (.87–1.71) Rhinovirus 1260 (73.3) 3675 (73.8) 211 (12.3) 743 (14.9) 249 (14.5) 559 (11.2) 0.94 (.82–1.07) 1.26 (1.07–1.48) RSV 1259 (73.2) 4840 (97.2) 73 (4.2) 83 (1.7) 388 (22.6) 57 (1.1) 12.55 (10.24–15.38) 24.72 (18.52–33.01) Data are presented as No. (%) unless otherwise indicated.

0) 122 (2.4) 0.91 (.74–1.12) 1.22 (.87–1.71) Rhinovirus 1260 (73.3) 3675 (73.8) 211 (12.3) 743 (14.9) 249 (14.5) 559 (11.2) 0.94 (.82–1.07) 1.26 (1.07–1.48) RSV 1259 (73.2) 4840 (97.2) 73 (4.2) 83 (1.7) 388 (22.6) 57 (1.1) 12.55 (10.24–15.38) 24.72 (18.52–33.01) Data are presented as No. (%) unless otherwise indicated. Abbreviations: AOR, adjusted odds ratio; CI, confidence interval; CXR+, chest radiograph positive; HBOV, human bocavirus; HMPV, human metapneumovirus; PV/EV, parechovirus/enterovirus; RSV, respiratory syncytial virus. aViruses with adjusted odds ratios <1 for association of density (log copies/mL) with case status excluded from table (coronavirus 229, coronavirus HKU, influenza C, and parainfluenza virus 4 as noted in Table 2). bCXR+ defined as having radiographic evidence of pneumonia. cWeakly positive: positive density below optimal cutpoint determined by Youden index. See Table 3. dStrongly positive: density above optimal cutpoint determined by Youden index. See Table 3. eOdds ratios adjusted for site and age. Bolded values are significant (P < .05). fAny positive includes those below and above optimal cutoff. gStrong positives are compared with combined negatives and weak positives.

cWeakly positive: positive density below optimal cutpoint determined by Youden index. See Table 3. dStrongly positive: density above optimal cutpoint determined by Youden index. See Table 3. eOdds ratios adjusted for site and age. Bolded values are significant (P < .05). fAny positive includes those below and above optimal cutoff. gStrong positives are compared with combined negatives and weak positives. Figure 3. Adjusted population attributable fraction (PAF) for chest radiograph–positive cases using 2 methods: any positive vs negative (AF1) and positive above optimal cutpoint vs positive below optimal cutpoint and negative (AF2)— Pneumonia Etiology Research for Child Health (PERCH) study, August 2011–January 2014. PAF = population prevalence × (1 – 1 / odds ratio). Odds ratio (OR) is adjusted for other viruses, site, and age. Confidence intervals calculated using bootstrapping method. PAF not shown where adjusted OR was <1 resulting in negative PAF. Abbreviations: Adeno, adenovirus; AF, attributable fraction; Boca, human bocavirus; Flu, influenza virus; HCoV, human coronavirus; HMPV, human metapneumovirus; Para, parainfluenza virus; PV/EV, parechovirus/enterovirus.

ulated using bootstrapping method. PAF not shown where adjusted OR was <1 resulting in negative PAF. Abbreviations: Adeno, adenovirus; AF, attributable fraction; Boca, human bocavirus; Flu, influenza virus; HCoV, human coronavirus; HMPV, human metapneumovirus; Para, parainfluenza virus; PV/EV, parechovirus/enterovirus. Predictors of Viral Load We explored several potential predictors of viral density among cases. When NP/OP specimens were collected earlier in the course of illness, mean viral load was higher for RSV, PIV1, and PIV3 and was lower for adenovirus; no significant difference was observed for the other viruses (Supplementary Table 1). The viral load among cases did not vary by age for most viruses, including RSV, but a significant trend toward decreasing viral load with increasing age was observed for a few viruses, including adenovirus and RSV, which was also observed among controls (Supplementary Figure 2). A slight, but significant trend toward increasing viral load with increasing age was seen for rhinovirus among cases, but a significant trend in the opposite direction was observed for controls. In general, viral load did not vary by study site. One notable exception was higher PIV1 viral load in The Gambia site (the only site with a PIV1 outbreak, data not shown). HIV-infected cases had a higher mean viral load for coronavirus 43 and a lower viral load for HMPV, PIV3, and RSV (Supplementary Table 2). There were no significant differences in the viral load between cases who were normally nourished vs malnourished (except PIV1 viral load was higher in normally nourished, data not shown).

ected cases had a higher mean viral load for coronavirus 43 and a lower viral load for HMPV, PIV3, and RSV (Supplementary Table 2). There were no significant differences in the viral load between cases who were normally nourished vs malnourished (except PIV1 viral load was higher in normally nourished, data not shown). We assessed whether viral load was associated with pneumonia severity. Rhinovirus was the only virus with higher mean viral load for very severe pneumonia compared to severe pneumonia (Supplementary Figure 3). Influenza A was the only virus with higher mean viral load in fatal compared with surviving cases (Supplementary Figure 4). Furthermore, we compared mean viral densities between cases with evidence of an other infiltrate (without alveolar consolidation) on chest radiograph to cases with evidence of alveolar consolidation (without evidence of an other infiltrate). No significant differences were found for any virus after adjusting for site and age.

more, we compared mean viral densities between cases with evidence of an other infiltrate (without alveolar consolidation) on chest radiograph to cases with evidence of alveolar consolidation (without evidence of an other infiltrate). No significant differences were found for any virus after adjusting for site and age. DISCUSSION In the PERCH study, the evidence for the utility of NP/OP viral load in distinguishing radiographically confirmed cases of severe or very severe pneumonia from controls was mixed. On the one hand, we found a higher mean viral load in NP/OP samples from severe and very severe pneumonia cases than from community controls without pneumonia for several respiratory viruses. Moreover, for many viruses, using a higher viral load threshold to define positivity that maximized the combination of sensitivity and specificity increased the odds ratio for case status over a simple binary (presence/absence) definition of positivity based on viral detection, which has high sensitivity but low specificity. On the other hand, there was substantial overlap in the distribution of NP/OP viral load densities among cases and controls, even for RSV, the virus most strongly associated with case status. No cutoffs clearly distinguished cases from controls.

based on viral detection, which has high sensitivity but low specificity. On the other hand, there was substantial overlap in the distribution of NP/OP viral load densities among cases and controls, even for RSV, the virus most strongly associated with case status. No cutoffs clearly distinguished cases from controls. Previous studies have also shown that the median or mean RSV concentration of NP/OP specimens among children is higher in cases of severe illness than among a healthy or mildly ill control population [7–10, 20, 21]. However, few of these studies compared the distribution of viral loads between severe cases and controls. Those that did compare these 2 groups showed an overlapping distribution similar to our study [8, 22]. Some studies of RSV viral load failed to show an association with severe lower respiratory tract infection [23, 24], but these studies included older children and adolescents in whom the pathogenic significance of detecting RSV in the NP/OP is less clear. One study of RSV viral load in infants showed no association with severe bronchiolitis [25] while another suggested that viral load only influences clinical severity for first RSV infections in young infants [10].

children and adolescents in whom the pathogenic significance of detecting RSV in the NP/OP is less clear. One study of RSV viral load in infants showed no association with severe bronchiolitis [25] while another suggested that viral load only influences clinical severity for first RSV infections in young infants [10]. In some studies, influenza viral load was associated with severe disease [4, 26–29], while in others there was no association [8, 23, 27, 30–32]. A few studies have shown a higher viral load in severe cases for HMPV [24, 33, 34]. Higher viral loads of HBOV in nasopharyngeal aspirates were associated with greater severity of illness among Chinese children [35]. Rhinovirus viral load has been associated with more severe illness [22], but not in some studies [8, 21]. Again, the majority of these studies looked only at the central tendency of the viral load and did not demonstrate a clear dichotomy in the distribution of viral loads based on case status or severity category. We did not find a higher viral load associated with greater severity among pneumonia cases for most viruses. Cases who died had a similar viral load as those who survived, and those with very severe pneumonia had similar viral loads to those children with severe pneumonia. This is in contrast with some other studies of viruses in which higher viral load was observed among RSV-infected children requiring mechanical ventilation [9], and severe acute respiratory syndrome coronavirus– and Middle East respiratory syndrome coronavirus–infected adults who died [36, 37].

n with severe pneumonia. This is in contrast with some other studies of viruses in which higher viral load was observed among RSV-infected children requiring mechanical ventilation [9], and severe acute respiratory syndrome coronavirus– and Middle East respiratory syndrome coronavirus–infected adults who died [36, 37]. We undertook this analysis, in part, to determine if viral loads of NP/OP specimens could be incorporated into the PERCH analysis to identify etiologies of severe/very severe pneumonia. Using a higher density threshold also did not have an appreciable effect on the population attributable fraction for most viruses, suggesting that using higher thresholds to assign viral etiology to cases would likely have little impact on the analysis of the etiologic distribution among the population of PERCH cases [38]. In the final PERCH analyses, we will be able to run sensitivity analyses to assess the impact of incorporating viral density thresholds on the assessment of etiology.

to assign viral etiology to cases would likely have little impact on the analysis of the etiologic distribution among the population of PERCH cases [38]. In the final PERCH analyses, we will be able to run sensitivity analyses to assess the impact of incorporating viral density thresholds on the assessment of etiology. In contrast to the accompanying analyses of bacterial pneumonia, our conclusions about the interpretation of viral load and whether to include it in the main PERCH etiology analysis were limited by the lack of a gold standard to diagnose viral pneumonia [5, 16]. There were few PERCH cases who underwent lung aspirate procedures and even fewer who had a lung aspirate in which a virus was detected in their lungs. Among a population of pneumonia cases in which a virus was detected in the NP/OP, there was likely a mixture of those in whom the virus had a causal role in pneumonia and those in whom it did not. The inability to identify which children had pneumonia due to which virus hampered the study’s ability to determine if higher viral loads in the NP/OP were associated with pneumonia.

detected in the NP/OP, there was likely a mixture of those in whom the virus had a causal role in pneumonia and those in whom it did not. The inability to identify which children had pneumonia due to which virus hampered the study’s ability to determine if higher viral loads in the NP/OP were associated with pneumonia. Besides the lack of a gold standard, other limitations might have affected our results. First, specimens were taken at one point in time on admission to the hospital. We observed that viral load varied with the time since illness onset, with higher viral load earlier in the course of symptomatic illness for some viruses. Taking sequential samples in which we could compare the peak viral load between cases of different clinical severity would have been optimal. Second, our design precluded us from assessing the role of viral load in the lung. Upper respiratory tract viral load might reflect the amount of replication in the local epithelial cells rather than the viral burden in the lung parenchyma. Evaluation of viral load of specimens from the lung, either through lung aspirates or bronchoalveolar lavage, would provide more direct evidence of the role of viral load in pneumonia severity.

oad might reflect the amount of replication in the local epithelial cells rather than the viral burden in the lung parenchyma. Evaluation of viral load of specimens from the lung, either through lung aspirates or bronchoalveolar lavage, would provide more direct evidence of the role of viral load in pneumonia severity. The widespread use of sensitive PCR assays for testing NP/OP specimens has led to a higher reported prevalence of pneumonias attributed to respiratory viruses in both adults and children [2]. Due to the high prevalence of viral infections of the URT itself, however, it is difficult to conclude that detection of a virus in the URT of a pneumonia patient is equivalent to having pneumonia due to that virus. In the PERCH study, the viral loads in the NP/OP of pneumonia patients are unlikely to further clarify the role of that virus in causing pneumonia. However, the PERCH study design was not optimal to answer this question definitively. Further research, such as longitudinal studies and animal models, is needed to better elucidate the interpretation of viral load in the diagnosis and clinical management of viral pneumonia. Supplementary Data Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the author to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the author, so questions or comments should be addressed to the author. Supplementary Material Supplemental Tables Figures Click here for additional data file. Notes

Supplementary Data Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the author to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the author, so questions or comments should be addressed to the author. Supplementary Material Supplemental Tables Figures Click here for additional data file. Notes Author contributions. D. R. F. led the analysis, interpreted results, and drafted initial manuscript. W. F. and Q. S. performed the analysis. W. F., Q. S., D. E. P., M. M. H., and D. R. M. provided significant guidance on the development of the manuscript. O. S. L., K. L. O., D. R. F., L. L. H., R. A. K., D. R. M., M. D. K., H. C. B., W. A. B., S. R. C. H., K. L. K., S. A. M., J. A. G. S., D. M. T. conceived and designed the study and supervised study conduct. W. F., D. E. P., Q. S., M. M. H., P. V. A., M. A., J. O. A., V. L. B., A. N. D., A. J. D., B. E. E., D. G., M. L., S. C. M., J. M., J. M., C. P., P. S., S. O. S., M. D. T., T. W., and K. Z. were involved in study conduct. S. L. Z. provided statistical expertise and led the integrated etiology analysis. All authors reviewed and approved the manuscript. D. R. F. had full access to all the data in the study and had final responsibility for the decision to submit for publication.

., M. D. T., T. W., and K. Z. were involved in study conduct. S. L. Z. provided statistical expertise and led the integrated etiology analysis. All authors reviewed and approved the manuscript. D. R. F. had full access to all the data in the study and had final responsibility for the decision to submit for publication. Acknowledgments. We acknowledge the significant contributions of the PERCH Study Group and all PERCH investigators. We offer our gratitude to the members of the Pneumonia Methods Working Group, PERCH Expert Group, and PERCH Chest Radiograph Reading Panel for their time and lending expertise to assist the PERCH Study Group. See Supplementary Materials for a full list of names. We offer sincere thanks to the patients and families who participated in this study. Disclaimer. The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention, Department of Health and Human Services, or the US government. This article is published with the permission of the Director of the Kenya Medical Research Institute. Financial support. PERCH was supported by the Bill & Melinda Gates Foundation (grant number 48968 to the International Vaccine Access Center, Department of International Health, Johns Hopkins Bloomberg School of Public Health). J. A. G. S. was supported by a clinical fellowship from the Wellcome Trust of Great Britain (award number 098532).

onsorship. This article appears as part of the supplement “Pneumonia Etiology Research for Child Health (PERCH): Foundational Basis for the Primary Etiology Results,” sponsored by a grant from the Bill & Melinda Gates Foundation to the PERCH study of Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland. Potential conflicts of interest. M. D. K. has received funding for consultancies from Merck, Pfizer, and Novartis, and grant funding from Merck. L. L. H. has received grant funding from Pfizer and GlaxoSmithKline. K. L. K. has received grant funding from Merck Sharp & Dohme. S. A. M. has received honoraria for advisory board membership from the Bill & Melinda Gates Foundation, Pfizer, Medimmune, and Novartis; has received institutional grants from GSK, Novartis, Pfizer, Minervax, and the Bill & Melinda Gates Foundation; and has served on speaker’s bureaus for Sanofi Pasteur and GSK. K. L. O. has received grant funding from Pfizer and GlaxoSmithKline and has served on technical advisory boards for Merck, Sanofi Pasteur, PATH, Affinivax, and ClearPath. All other authors report no potential conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

… ≥3.5 5 (38.5) 6 (21.4) 4 (28.6) 0 (0.0) … Died in hospital (row %) 0 3 (17.6) .34 174 (5.1) .03 70 (4.7) .09 24 (28.6) .85 … <2.2 0 (0.0) 2 (2.8) 2 (5.9) 2 (100) … 2.2–3.5 2 (18.2) .45 10 (8.5) .05 3 (5.1) .26 2 (40.0) .16 … ≥3.5 5 (35.7) 4 (13.3) 3 (20.0) 0 (0.0) … Table excludes human immunodeficiency virus (HIV)–i nfected children; children with unknown HIV status are included. Abbreviations: CRP, C-reactive protein; CXR, chest radiograph; MCPP, microbiologically confirmed pneumococcal pneumonia; NP/OP, nasopharyngeal/oropharyngeal; PCR, polymerase chain reaction; PCV, pneumococcal conjugate vaccine; WBC, white blood cell; …, not applicable for controls. aMCPP defined as pneumococcus isolated from culture of blood, lung aspirate, pleural fluid, PCR of lung aspirate or pleural fluid, or detection of Streptococcus pneumoniae antigen on BinaxNOW testing of pleural fluid. bCXR positive (CXR+) defined as radiographic evidence of pneumonia (consolidation and/or other infiltrates). cCase with any nonpneumococcal bacterial pathogen detected by blood culture, by lung aspirate culture or PCR, or by pleural fluid culture or PCR. dPercentages in the “Total” row represent column percentages. In all subsequent rows, the number and percentage represent children in the corresponding case/control group and whole-blood pneumococcal load category who had the characteristic.

cCase with any nonpneumococcal bacterial pathogen detected by blood culture, by lung aspirate culture or PCR, or by pleural fluid culture or PCR. dPercentages in the “Total” row represent column percentages. In all subsequent rows, the number and percentage represent children in the corresponding case/control group and whole-blood pneumococcal load category who had the characteristic. eTest for trend from Cochran-Armitage in binomial proportions: The first P value listed is across all 4 whole-blood PCR quantity categories and the second P value is across the last 3 PCR quantity categories for which pneumococcus was detected in the blood. fFour sites had introduced PCV prior to start of enrollment: Kenya, The Gambia, Mali, and South Africa. Results restricted to PCV-using sites only are shown in Supplementary Table 4A, where there was no longer an association observed among controls. gPrior antibiotic use defined as serum bioassay positive (cases and controls), antibiotic administration at the referral facility, or antibiotic administration prior to whole-blood specimen collection at the study facility (cases only).

fFour sites had introduced PCV prior to start of enrollment: Kenya, The Gambia, Mali, and South Africa. Results restricted to PCV-using sites only are shown in Supplementary Table 4A, where there was no longer an association observed among controls. gPrior antibiotic use defined as serum bioassay positive (cases and controls), antibiotic administration at the referral facility, or antibiotic administration prior to whole-blood specimen collection at the study facility (cases only). There was some variation by site of the performance of high blood pneumococcal load to detect MCPP cases, although numbers were too small to detect any statistical significance (Figure 2B). Pneumococcal blood load varied significantly among the sites for both nonconfirmed cases and controls, with Mali having the highest load (median, 0.59 and 0.40 × 103 copies/mL for nonconfirmed cases and controls, respectively) and Bangladesh the lowest (median, 0.09 and 0.1 × 103copies/mL, respectively; Table 1). However, data for the 2 Asian countries are limited because of the extremely small numbers (n = 3–7) of pneumococcal PCR–positive nonconfirmed cases and controls. The proportion of nonconfirmed cases with high load ranged in African sites from 2.3% in Kenya to 7.6% in Mali (P < .001; Figure 2B).

The conventional method to identify the cause of pediatric bacterial pneumonia is by blood culture, but this has low sensitivity (<10%) to detect pneumococcal pneumonia, a leading bacterial cause of pneumonia [1, 2]. Detection of pneumococcus in blood by polymerase chain reaction (PCR) may be more sensitive for detecting bloodstream infection, especially in cases treated with antibiotics prior to specimen collection [2]. If so, this test could identify blood culture–negative pneumococcal pneumonia cases. However, when blood pneumococcal PCR was evaluated in the Pneumonia Etiology Research for Child Health (PERCH) study, we found that whole-blood lytA positivity had poor diagnostic accuracy, in terms of both sensitivity and specificity [3]. Others have observed that higher pneumococcal PCR load in the blood is associated with greater severity of disease [4–7], so it is possible that the load of pneumococcus in blood is higher in children with pneumonia than in well children. We evaluated the utility of quantitative whole-blood pneumococcal PCR as a diagnostic test for pneumococcal pneumonia in the PERCH study.

n the blood is associated with greater severity of disease [4–7], so it is possible that the load of pneumococcus in blood is higher in children with pneumonia than in well children. We evaluated the utility of quantitative whole-blood pneumococcal PCR as a diagnostic test for pneumococcal pneumonia in the PERCH study. METHODS Study Design As described and presented elsewhere [8, 9], PERCH is a case-control study evaluating the etiology of severe and very severe pneumonia conducted in 9 sites in 5 African and 2 Asian countries: Basse, The Gambia; Bamako, Mali; Kilifi, Kenya; Soweto, South Africa; Lusaka, Zambia; Nakhon Phanom and Sa Kaeo, Thailand; and Dhaka and Matlab, Bangladesh. Enrollment occurred for 24 consecutive months at each site during August 2011–January 2014. Cases were children aged 1–59 months hospitalized with 2005 World Health Organization (WHO)–defined severe or very severe pneumonia. Controls were selected randomly from the community and were frequency-matched to cases for age (1 to <6 months, 6 to <12 months, 12 to <24 months, and 24–59 months) and month of enrollment. Controls were also matched on human immunodeficiency virus (HIV) status at the 2 sites with high HIV prevalence (South Africa and Zambia).

were selected randomly from the community and were frequency-matched to cases for age (1 to <6 months, 6 to <12 months, 12 to <24 months, and 24–59 months) and month of enrollment. Controls were also matched on human immunodeficiency virus (HIV) status at the 2 sites with high HIV prevalence (South Africa and Zambia). Cases and controls were evaluated at enrollment for clinical signs and symptoms as well as risk factors for pneumonia. Severe pneumonia was defined as having cough or difficulty breathing and lower chest wall in-drawing that, in the subset of children presenting with wheeze, did not resolve after administration of bronchodilators; very severe pneumonia was defined as having cough or difficulty breathing and at least 1 of the following: central cyanosis, difficulty breastfeeding/drinking, vomiting everything, multiple or prolonged convulsions, lethargy, unconsciousness, or head nodding. Controls were enrolled regardless of presence of respiratory symptoms, as long as they did not have case-defining severe or very severe pneumonia. Chest radiographs (CXRs) performed at the time of admission were defined as CXR-positive (CXR+) if they were classified as either alveolar consolidation or other infiltrates using WHO methods [10, 11]. Respiratory tract illness (RTI) in controls was defined as having cough or runny nose. RTI was also defined as having (1) at least 1 of ear discharge, wheezing, or difficulty breathing and (2) either a measured fever (temperature ≥38.0°C) within the previous 48 hours or a history of sore throat. Prior antibiotic exposure was defined as either a positive serum bioassay (cases and controls) or documentation of antibiotics administered at the referral or study hospital prior to specimen collection (cases only) [2].

er a measured fever (temperature ≥38.0°C) within the previous 48 hours or a history of sore throat. Prior antibiotic exposure was defined as either a positive serum bioassay (cases and controls) or documentation of antibiotics administered at the referral or study hospital prior to specimen collection (cases only) [2]. Pneumococcal conjugate vaccine (PCV) was introduced prior to PERCH in The Gambia, Kenya, Mali, and South Africa. PCV was introduced in July 2013 in Zambia, 18 months after enrollment started. In Bangladesh and Thailand, PCV was available but only in the private market, with almost no usage reported in the study areas.

er a measured fever (temperature ≥38.0°C) within the previous 48 hours or a history of sore throat. Prior antibiotic exposure was defined as either a positive serum bioassay (cases and controls) or documentation of antibiotics administered at the referral or study hospital prior to specimen collection (cases only) [2]. Pneumococcal conjugate vaccine (PCV) was introduced prior to PERCH in The Gambia, Kenya, Mali, and South Africa. PCV was introduced in July 2013 in Zambia, 18 months after enrollment started. In Bangladesh and Thailand, PCV was available but only in the private market, with almost no usage reported in the study areas. All specimen collection and laboratory methods were standardized across all sites and have been described elsewhere [12]. Pleural fluid was collected from cases when clinically indicated at all sites, and lung aspirates were collected at select sites (The Gambia, South Africa, Mali, and Bangladesh) when relevant and feasible. At all sites, nasopharyngeal and oropharyngeal (NP/OP) swabs were obtained and combined, and whole blood was collected, from both cases and controls at the time of enrollment. Blood, lung aspirates, and pleural fluid from cases were cultured for detection of bacterial organisms. NP/OP, lung aspirate, and pleural fluid specimens were tested for 33 pathogens including pneumococcus using the Fast-track Diagnostics Respiratory Pathogens 33 test (Fast-track Diagnostics [FTD], Sliema, Malta). Quantification data were generated through creation of standard curves using 10-fold serial dilutions of plasmid standards, with calculation of pathogen load (copies/mL) from the sample cycle threshold values. High PCR load in NP/OP specimens was defined as >6.9 log10 copies/mL, which was the optimal colonization load threshold for discriminating microbiologically confirmed pneumococcal pneumonia (MCPP) cases from all controls [13].

ndards, with calculation of pathogen load (copies/mL) from the sample cycle threshold values. High PCR load in NP/OP specimens was defined as >6.9 log10 copies/mL, which was the optimal colonization load threshold for discriminating microbiologically confirmed pneumococcal pneumonia (MCPP) cases from all controls [13]. Whole-blood specimens from cases and controls were evaluated for the presence of the Streptococcus pneumoniae autolysin gene lytA using a quantitative real-time PCR assay based on the US Centers for Disease Control and Prevention method, as described previously [3, 14]. Quantification standards consisting of lytA plasmids (FTD) diluted 1:10 from 107 copies/mL to 102 copies/mL were run in triplicate on every plate. Data points with detected pneumococcal concentration below the lower limit of linearity and lower limit of detection of the assay were retained in the data set, with the understanding that accuracy and precision are affected in that range.

1:10 from 107 copies/mL to 102 copies/mL were run in triplicate on every plate. Data points with detected pneumococcal concentration below the lower limit of linearity and lower limit of detection of the assay were retained in the data set, with the understanding that accuracy and precision are affected in that range. Cases were defined as MCPP if they had pneumococcus detected by culture of blood, lung aspirate, or pleural fluid, PCR of lung aspirate or pleural fluid, or detection of pneumococcal antigen (BinaxNOW, Alere, Orlando, Florida) in pleural fluid. Cases were defined as “confirmed for a nonpneumococcal pathogen” if they were not MCPP cases but were culture positive (blood, lung aspirate, or pleural fluid) or PCR positive in lung aspirate or pleural fluid for nonpneumococcal pathogenic bacteria. Nonconfirmed cases had no bacterial pathogens detected by culture of blood, lung aspirate, or pleural fluid, PCR of lung aspirate or pleural fluid, or detection of pneumococcal antigen on pleural fluid.

spirate, or pleural fluid) or PCR positive in lung aspirate or pleural fluid for nonpneumococcal pathogenic bacteria. Nonconfirmed cases had no bacterial pathogens detected by culture of blood, lung aspirate, or pleural fluid, PCR of lung aspirate or pleural fluid, or detection of pneumococcal antigen on pleural fluid. Analysis Median blood pneumococcal load was compared between groups using the Kruskal-Wallis test. The optimal threshold for discriminating all MCPP cases from all controls (ie, including the blood pneumococcal PCR-negative children) was identified with receiver operating characteristic (ROC) analyses using the Youden index to maximize sensitivity and specificity [15]. To guard against bias in the estimates of sensitivity due to having a small number of MCPP cases, the Youden index was calculated using leave-one-out cross-validation [16]. The proportion with high load, defined as those above the threshold, was determined for cases stratified by bacteremia status, MCPP, and CXR findings, and for controls by RTI status. Analyses were performed overall and stratified by site, HIV status, and prior antibiotic exposure. Comparisons of proportions were made using the χ2 test or Fisher exact test. Ethical Considerations The PERCH study protocol was approved by the institutional review board or ethical review committee at each of the study site institutions and at the Johns Hopkins Bloomberg School of Public Health. Parents or guardians of all participants provided written informed consent.

Analysis Median blood pneumococcal load was compared between groups using the Kruskal-Wallis test. The optimal threshold for discriminating all MCPP cases from all controls (ie, including the blood pneumococcal PCR-negative children) was identified with receiver operating characteristic (ROC) analyses using the Youden index to maximize sensitivity and specificity [15]. To guard against bias in the estimates of sensitivity due to having a small number of MCPP cases, the Youden index was calculated using leave-one-out cross-validation [16]. The proportion with high load, defined as those above the threshold, was determined for cases stratified by bacteremia status, MCPP, and CXR findings, and for controls by RTI status. Analyses were performed overall and stratified by site, HIV status, and prior antibiotic exposure. Comparisons of proportions were made using the χ2 test or Fisher exact test. Ethical Considerations The PERCH study protocol was approved by the institutional review board or ethical review committee at each of the study site institutions and at the Johns Hopkins Bloomberg School of Public Health. Parents or guardians of all participants provided written informed consent. RESULTS Whole-Blood Pneumococcal PCR Load Distribution Blood was tested for pneumococcus by PCR in 3995 (94.4%) cases and 4987 (93.7%) community controls. Pneumococcal load was evaluated in all children who were blood pneumococcal PCR positive, including 290 (7.3%) cases (36 MCPP cases, 242 nonconfirmed cases, and 12 cases confirmed for a nonpneumococcal pathogen) and 273 (5.5%) controls; 1 case who was pneumococcal PCR positive was missing load data. MCPP cases were identified only at the African sites (range, n = 3–19 MCPP cases who were blood PCR positive; Table 1).

290 (7.3%) cases (36 MCPP cases, 242 nonconfirmed cases, and 12 cases confirmed for a nonpneumococcal pathogen) and 273 (5.5%) controls; 1 case who was pneumococcal PCR positive was missing load data. MCPP cases were identified only at the African sites (range, n = 3–19 MCPP cases who were blood PCR positive; Table 1). Table 1. Median Pneumococcal Polymerase Chain Reaction (PCR) Load (103 Copies/mL) in Whole Blood Among Children Who Were Whole-Blood PCR Positive, by Case/Control Group and Characteristic Characteristic MCPP Casesa (n = 56) Nonconfirmed Casesb (n = 3832) Nonconfirmed CXR+c Cases (n = 1745) Confirmed Nonpneumococcal Casesd (n = 107) All Controls (n = 4987) PCR+, No. Median (IQR) Loade PCR+, No. Median (IQR) Load PCR+, No. Median (IQR) Load PCR+, No. Median (IQR) Load PCR+, No. Median (IQR) Load Overall 36 4.1f (1.1–77.4) 242 0.29f (0.14–0.93) 127 0.30f (0.14–1.0) 12 1.6f (0.32–5.7) 273 0.19f (0.11–0.48) PERCH sites P = .51 P < .001 P = .31 P = .41 P < .01

Characteristic MCPP Casesa (n = 56) Nonconfirmed Casesb (n = 3832) Nonconfirmed CXR+c Cases (n = 1745) Confirmed Nonpneumococcal Casesd (n = 107) All Controls (n = 4987) PCR+, No. Median (IQR) Loade PCR+, No. Median (IQR) Load PCR+, No. Median (IQR) Load PCR+, No. Median (IQR) Load PCR+, No. Median (IQR) Load Overall 36 4.1f (1.1–77.4) 242 0.29f (0.14–0.93) 127 0.30f (0.14–1.0) 12 1.6f (0.32–5.7) 273 0.19f (0.11–0.48) PERCH sites P = .51 P < .001 P = .31 P = .41 P < .01 Kenya 3 83.9 (4.6–419.8) 25 0.17 (0.14–0.37) 15 0.23 (0.12–0.48) 3 0.21 (0.06–1.5) 48 0.24 (0.12–0.45) The Gambia 6 2.2 (0.63–8.8) 51 0.21 (0.11–0.52) 20 0.24 (0.10–0.46) 4 2.6 (1.1–7.5) 47 0.23 (0.13–0.52) Mali 19 2.4 (0.53–96.8) 56 0.59 (0.21–3.1) 25 0.53 (0.17–2.1) 2 43.2 (0.11–86.2) 38 0.40 (0.08–1.1) Zambia 4 40.0 (4.1–98.9) 37 0.30 (0.14–1.2) 22 0.30 (0.15–1.4) 2 1.6 (0.43–2.8) 31 0.17 (0.11–0.38) South Africa 4 4.8 (1.9–20.6) 66 0.27 (0.16–0.93) 44 0.30 (0.16–0.94) 1 7.8 (7.8–7.8) 98 0.16 (0.09–0.34) Bangladesh 0 … 5 0.09 (0.06–0.30) 1 25.1 (25.1–25.1) 0 … 6 0.10 (0.06–0.10) Thailand 0 … 2 0.05 (0.01–0.09) 0 … 0 … 5 0.27 (0.06–0.86) Age P = .27 P = .57 P = .71 P = .35 P = .71 1–5 mo 7 144.6 (1.2–731.6) 96 0.25 (0.15–0.53) 51 0.30 (0.17–0.53) 3 0.43 (0.06–1.5) 88 0.19 (0.11–0.49) 6–11 mo 10 2.4 (1.1–4.2) 68 0.32 (0.14–1.2) 33 0.30 (0.13–0.93) 3 2.8 (0.21–86.2) 72 0.20 (0.10–0.41) 12–23 mo 9 7.1 (1.8–12.1) 52 0.27 (0.13–1.4) 32 0.30 (0.13–1.6) 3 0.49 (0.11–11.4) 64 0.17 (0.11–0.33) 24–59 mo 10 2.1 (0.53–7.7) 26 0.31 (0.13–3.8) 11 1.3 (0.12–17.0) 3 3.6 (1.7–7.8) 49 0.19 (0.11–0.88) HIV infected P = .68 P = .27 P = .16 P = .12 P = .96 Yesg 10 3.9 (0.53–12.1) 24 0.32 (0.17–21.3) 19 0.47 (0.29–25.6) 3 7.8 (2.8–86.2) 19 0.16 (0.11–0.46) No 23 2.8 (1.2–83.9) 197 0.27 (0.14–0.93) 98 0.28 (0.14–0.93) 8 1.0 (0.27–2.6) 224 0.19 (0.11–0.46) Prior antibioticsh P = .25 P = .21 P = .03

.8) 49 0.19 (0.11–0.88) HIV infected P = .68 P = .27 P = .16 P = .12 P = .96 Yesg 10 3.9 (0.53–12.1) 24 0.32 (0.17–21.3) 19 0.47 (0.29–25.6) 3 7.8 (2.8–86.2) 19 0.16 (0.11–0.46) No 23 2.8 (1.2–83.9) 197 0.27 (0.14–0.93) 98 0.28 (0.14–0.93) 8 1.0 (0.27–2.6) 224 0.19 (0.11–0.46) Prior antibioticsh P = .25 P = .21 P = .03 P = .47 P = .49 Yes 9 3.9 (1.9–126.8) 100 0.30 (0.15–1.1) 64 0.39 (0.18–1.5) 5 0.43 (0.11–3.6) 9 0.23 (0.19–0.61) No 25 2.4 (0.73–12.1) 131 0.28 (0.12–0.86) 58 0.22 (0.12–0.86) 6 1.6 (0.49–7.8) 253 0.19 (0.11–0.47) Pneumococcal NP/OP PCR load >6.9 log10 copies/mL P = .97 P = .03 P = .05 P = .10 P = .19 Yes 26 4.1 (0.73–96.8) 48 0.46 (0.16–5.1) 33 0.50 (0.16–13.1) 3 7.8 (1.5–86.2) 27 0.28 (0.13–0.88) No 9 4.6 (1.5–33.6) 191 0.28 (0.14–0.67) 92 0.30 (0.14–0.69) 8 0.46 (0.16–3.2) 243 0.18 (0.11–0.45) All P values obtained by Kruskal-Wallis test; P values within cells represent comparison within the case/control group for that characteristic. Bold indicates P < .05. Abbreviations: HIV, human immunodeficiency virus; IQR, interquartile range; MCPP, microbiologically confirmed pneumococcal pneumonia; NP/OP, nasopharyngeal/oropharyngeal; PCR+, polymerase chain reaction positive for lytA gene; PERCH, Pneumonia Etiology Research for Child Health. aMCPP defined as pneumococcus isolated from culture of blood, lung aspirate, pleural fluid, PCR of lung aspirate or pleural fluid, or detection of Streptococcus pneumoniae antigen in pleural fluid specimens on BinaxNOW.

respectively; Table 1). However, data for the 2 Asian countries are limited because of the extremely small numbers (n = 3–7) of pneumococcal PCR–positive nonconfirmed cases and controls. The proportion of nonconfirmed cases with high load ranged in African sites from 2.3% in Kenya to 7.6% in Mali (P < .001; Figure 2B). Factors Associated With Whole-Blood Pneumococcal Load Overall and by site, there were no notable associations between pneumococcal load and age, sex, HIV, or PCV use among those who were PCR positive (Tables 1 and 2 and Supplementary Table 3). Although the distribution of blood pneumococcal PCR load did not differ by receipt of antibiotics prior to sample collection among MCPP cases or controls, load was higher among nonconfirmed CXR+ cases if they had prior antibiotics (median, 0.39 vs 0.22 × 103 copies/mL, P = .03; Table 1 and Supplementary Figure 2). Children with high NP/OP loads were more likely to have higher loads in blood (Table 3 and Figure 3). Despite this, among nonconfirmed cases above the NP/OP threshold of >6.9 log10 copies/mL, only 8.1% were above the blood threshold, and the majority (77.4%) of nonconfirmed cases above the blood threshold were below the NP/OP threshold. Table 3. Percentage With Pneumococcal Whole-Blood Polymerase Chain Reaction (PCR) Load ≥2.2 Log10 Copies/mL by Nasopharyngeal/Oropharyngeal PCR Load >6.9 Log10 Copies/mL

Abbreviations: HIV, human immunodeficiency virus; IQR, interquartile range; MCPP, microbiologically confirmed pneumococcal pneumonia; NP/OP, nasopharyngeal/oropharyngeal; PCR+, polymerase chain reaction positive for lytA gene; PERCH, Pneumonia Etiology Research for Child Health. aMCPP defined as pneumococcus isolated from culture of blood, lung aspirate, pleural fluid, PCR of lung aspirate or pleural fluid, or detection of Streptococcus pneumoniae antigen in pleural fluid specimens on BinaxNOW. bNonconfirmed cases defined as cases without isolation of bacteria from culture of blood, lung aspirate or pleural fluid, or PCR of lung aspirate or pleural fluid. cCXR positive (CXR+) defined as radiographic evidence of pneumonia (consolidation and/or other infiltrates). dConfirmed nonpneumococcal bacterial case was defined as a case with any nonpneumococcal bacterial pathogen detected by blood culture, by lung aspirate culture or PCR, or by pleural fluid culture or PCR. eMedian load = median whole-blood lytA load (103 copies/mL) among children with PCR-positive whole-blood specimens. f P value for MCPP vs nonconfirmed, all controls, and nonconfirmed CXR+ cases, <.001 for all; P value for MCPP vs confirmed non-pneumococcal cases, .06; P value for nonconfirmed cases vs confirmed non-pneumococcal cases, .05; P value for confirmed non-pneumococcal vs all controls, .003; P value for nonconfirmed cases vs all controls, <.001. gControls were matched on HIV status at the 2 sites with high HIV prevalence (South Africa and Zambia).

f P value for MCPP vs nonconfirmed, all controls, and nonconfirmed CXR+ cases, <.001 for all; P value for MCPP vs confirmed non-pneumococcal cases, .06; P value for nonconfirmed cases vs confirmed non-pneumococcal cases, .05; P value for confirmed non-pneumococcal vs all controls, .003; P value for nonconfirmed cases vs all controls, <.001. gControls were matched on HIV status at the 2 sites with high HIV prevalence (South Africa and Zambia). hPrior antibiotics defined as serum bioassay positive (cases and controls), antibiotic administration at the referral facility, or antibiotic administration prior to whole-blood specimen collection at the study facility (cases only). Blood pneumococcal PCR load was higher among MCPP cases (median, 4.0 × 103 copies/mL) compared with controls (median, 0.19 × 103copies/mL; P < .001; Table 1), but some MCPP cases had low load (range, 0.16–989.9 × 103 copies/mL) and some controls had high load (range, 0.01–551.9 × 103 copies/mL; Figure 1A and Supplementary Figure 1). Load among MCPP cases was also higher than among nonconfirmed cases (median, 0.29 × 103 copies/mL; P < .001). Interestingly, load among cases confirmed for nonpneumococcal pathogens (median, 1.6 × 103 copies/mL; Table 1 and Supplementary Table 1) was also higher than that among nonconfirmed cases (P = .05) and controls (P = .003). Median load was similar in controls with and without RTI (Supplementary Table 2).

/mL; P < .001). Interestingly, load among cases confirmed for nonpneumococcal pathogens (median, 1.6 × 103 copies/mL; Table 1 and Supplementary Table 1) was also higher than that among nonconfirmed cases (P = .05) and controls (P = .003). Median load was similar in controls with and without RTI (Supplementary Table 2). Figure 1. Comparison of pneumococcal polymerase chain reaction (PCR)a load in whole blood between microbiologically confirmed pneumococcal pneumonia (MCPP)b cases and community controls. A, load distribution among PCR-positive children. B, Receiver operating characteristic analysis among all children, including PCR-negative children. aPneumococcal load (density) by PCR for the lytA gene (log10 copies/mL) in whole-blood specimens. bMCPP defined as pneumococcus detected by culture of blood, lung aspirate, or pleural fluid, PCR of lung aspirate or pleural fluid, or BinaxNOW of pleural fluid. Whole-Blood Pneumococcal PCR Load Threshold ROC analyses determined that ≥2.2 log10 copies/mL was the optimal threshold that maximized sensitivity and specificity to distinguish MCPP cases from controls (Figure 1B); 62.5% of MCPP cases were above the threshold vs 3.1% of controls (Figure 2A). The threshold was similar when restricted to HIV-negative children and when determined using controls without RTI (data not shown).

imal threshold that maximized sensitivity and specificity to distinguish MCPP cases from controls (Figure 1B); 62.5% of MCPP cases were above the threshold vs 3.1% of controls (Figure 2A). The threshold was similar when restricted to HIV-negative children and when determined using controls without RTI (data not shown). Figure 2. Percentage of children with whole-blood pneumococcal polymerase chain reaction (PCR) load >2.2 log10 copies/mL by case/control group overall (A) and by site (B). Denominator numbers for (B) are provided in Supplementary Table 5. Abbreviations: BCx+, blood culture positive; Conf non-Spn, a case with any nonpneumococcal bacterial pathogen detected by blood culture, by lung aspirate culture or PCR, or by pleural fluid culture or PCR; CXR+, findings of alveolar consolidation or other infiltrates on chest radiograph; CXR-AC, findings of alveolar consolidation (with or without other infiltrates) on chest radiograph; MCPP, microbiologically confirmed pneumococcal pneumonia (defined as pneumococcus isolated from culture of blood, lung aspirate, pleural fluid, PCR of lung aspirate or pleural fluid, or detection of Streptococcus pneumoniae antigen in pleural fluid on BinaxNOW); PCR+, polymerase chain reaction positive for lytA gene; RTI, respiratory tract illness.

Children with high NP/OP loads were more likely to have higher loads in blood (Table 3 and Figure 3). Despite this, among nonconfirmed cases above the NP/OP threshold of >6.9 log10 copies/mL, only 8.1% were above the blood threshold, and the majority (77.4%) of nonconfirmed cases above the blood threshold were below the NP/OP threshold. Table 3. Percentage With Pneumococcal Whole-Blood Polymerase Chain Reaction (PCR) Load ≥2.2 Log10 Copies/mL by Nasopharyngeal/Oropharyngeal PCR Load >6.9 Log10 Copies/mL Characteristic MCPP Casesa Nonconfirmed Casesb Nonconfirmed CXR+c Cases Confirmed Nonpneumococcal Cased All Controls Total No.e WB PCR ≥2.2 Log10 Copies/mL, No. (%) Total No. WB PCR ≥2.2 Log10 Copies/mL, No. (%) Total No. WB PCR ≥2.2 Log10 Copies/mL, No. (%) Total No. WB PCR ≥2.2 Log10 Copies/mL, No. (%) Total No. WB PCR ≥2.2 Log10 Copies/mL, No. (%) Overall 56 35 (62.5) 3832 166 (4.3) 1745 91 (5.2) 107 10 (9.3) 4987 154 (3.1) Pneumococcal NP/OP PCR load >6.9 log10 copies/mL All sites P = .07 P < .01 P < .01

med pneumococcal pneumonia (defined as pneumococcus isolated from culture of blood, lung aspirate, pleural fluid, PCR of lung aspirate or pleural fluid, or detection of Streptococcus pneumoniae antigen in pleural fluid on BinaxNOW); PCR+, polymerase chain reaction positive for lytA gene; RTI, respiratory tract illness. When evaluating all cases and controls as opposed to just those who were positive by blood PCR, the proportion of children who were positive but had low blood load (ie, <2.2 log10 copies/mL) was similar regardless of case or control group (range, 1.8%–2.4%; Figure 2A). However, among those positive by blood PCR, the proportion of children with high pneumococcal load differed by case and control group. Only 1 (2.8%) MCPP case with pneumococcus detected in blood by PCR had low load; thus, the impact of the threshold on sensitivity (62.5%) was minimal. However, 56% (154/273) of PCR-positive controls had high blood pneumococcal load (3.1% of all controls); using the threshold improved specificity to detect MCPP cases by 2.4% (to 96.9%) over that of any PCR positivity. The proportion of RTI controls with high load (3.6%) was not statistically greater than that of non-RTI controls (2.9%; P = .21; Supplementary Table 4A). Among 12 blood pneumococcal PCR–positive cases confirmed for a nonpneumococcal etiology, 10 (83.3%) had high load (Figure 2A). Of these 10 cases, pneumococcus was detected in the nasopharynx by culture in 4 and by PCR in all 9 of those with results (Supplementary Table 1). There was a wide diversity among the organisms cultured from blood in this group, some of which have previously been observed in coinfections with pneumococcus [17]. Two cultures revealed Haemophilus influenzae, and the remaining were Candida species, which is not thought to cause pneumonia; Acinetobacter species, which can cause pneumonia but is also a common skin contaminant; Neisseria meningitidis; Escherichia coli; Pseudomonas aeruginosa; Staphylococcus aureus; and Salmonella species plus Streptococcus pyogenes.

nfluenzae, and the remaining were Candida species, which is not thought to cause pneumonia; Acinetobacter species, which can cause pneumonia but is also a common skin contaminant; Neisseria meningitidis; Escherichia coli; Pseudomonas aeruginosa; Staphylococcus aureus; and Salmonella species plus Streptococcus pyogenes. Among nonconfirmed cases who were pneumococcal PCR positive in blood, 68.3% had high load (71.7% of CXR+-nonconfirmed cases), resulting in 4.3% and 5.2% of all nonconfirmed cases and CXR+-nonconfirmed cases, respectively, with high load (Figure 2A). Although the percentage of nonconfirmed cases with high load was greater than that of controls (4.3% vs 3.1%; unadjusted P < .01), the association was not strong (odds ratio adjusted for site [AOR], 1.3; 95% confidence interval [CI], 1.0–1.6). Because of this, and in an effort to increase specificity, we evaluated whether a higher threshold could identify whether some of the 166 individual nonconfirmed cases that were above the 2.2 log10 copies/mL threshold had a higher likelihood of having pneumococcal pneumonia than others. A threshold of >3.5 log10 copies/mL that visually distinguished MCPP cases from controls among PCR-positive children was evaluated (Figure 1A). Sensitivity among HIV-uninfected MCPP cases was greatly diminished (32.6%) and positivity in all other groups, including nonconfirmed cases, was ≤2% (Table 2). Although the ability to distinguish nonconfirmed cases from controls was improved (AOR, 2.7; 95% CI, 1.4–5.3), only n = 30 (0.8%) nonconfirmed cases had load >3.5 log10 copies/mL, reducing the value of a higher threshold to identify hidden pneumococcal pneumonia cases.

, including nonconfirmed cases, was ≤2% (Table 2). Although the ability to distinguish nonconfirmed cases from controls was improved (AOR, 2.7; 95% CI, 1.4–5.3), only n = 30 (0.8%) nonconfirmed cases had load >3.5 log10 copies/mL, reducing the value of a higher threshold to identify hidden pneumococcal pneumonia cases. Table 2. Characteristics Associated With Whole-Blood Pneumococcal Polymerase Chain Reaction Load in Human Immunodeficiency Virus–Uninfected Cases and Controls at All Sites Characteristic Pneumococcal Blood PCR Load (Log10 Copies/mL) MCPPa Cases (n = 43) Nonconfirmed Cases (n = 3621) Nonconfirmed CXR+b Cases (n = 1601) Confirmed Nonpneumococcal Casesc (n = 93) Controls (n = 4779) No. (%d) P Valuee No. (%) P Value No. (%) P Value No. (%) P Value No. (%) P Value Total (column %d) 0 17 (39.5) 3402 (94.0) 1493 (93.3) 84 (90.3) 4525 (94.7) <2.2 1 (2.3) 72 (2.0) 34 (2.1) 2 (2.2) 110 (2.3) 2.2–3.5 11 (25.6) 117 (3.2) 59 (3.7) 5 (5.4) 131 (2.7) ≥3.5 14 (32.6) 30 (0.8) 15 (0.9) 2 (2.2) 13 (0.3) Female sex (row %d) 0 5 (29.4) .04 1420 (41.7) .61 649 (43.5) .47 49 (58.3) .74 2240 (49.5) .16 <2.2 0 (0.0) 25 (34.7) 12 (35.3) 1 (50.0) 65 (59.1) 2.2–3.5 4 (36.4) .08 50 (42.7) .60 30 (50.8) .30 4 (80.0) .63 69 (52.7) .39 ≥3.5 10 (71.4) 11 (36.7) 7 (46.7) 0 (0.0) 7 (53.8) At least 1 dose of PCVf (row %) 0 14 (82.4) .69 1730 (52.7) .002 788 (54.6) .09 37 (45.7) .33 2190 (49.6) <.001

49 (58.3) .74 2240 (49.5) .16 <2.2 0 (0.0) 25 (34.7) 12 (35.3) 1 (50.0) 65 (59.1) 2.2–3.5 4 (36.4) .08 50 (42.7) .60 30 (50.8) .30 4 (80.0) .63 69 (52.7) .39 ≥3.5 10 (71.4) 11 (36.7) 7 (46.7) 0 (0.0) 7 (53.8) At least 1 dose of PCVf (row %) 0 14 (82.4) .69 1730 (52.7) .002 788 (54.6) .09 37 (45.7) .33 2190 (49.6) <.001 <2.2 1 (100) 45 (64.3) 24 (72.7) 2 (100) 83 (77.6) 2.2–3.5 6 (60.0) .71 71 (64.0) >.99 33 (60.0) .53 2 (50.0) >.99 99 (78.6) >.99 ≥3.5 11 (78.6) 19 (65.5) 10 (66.7) 1 (100) 10 (76.9) Preceding antibioticsg (row %) 0 3 (17.6) .34 1283 (39.4) .91 581 (40.8) .04 39 (48.1) .63 78 (1.8) .05

49 (58.3) .74 2240 (49.5) .16 <2.2 0 (0.0) 25 (34.7) 12 (35.3) 1 (50.0) 65 (59.1) 2.2–3.5 4 (36.4) .08 50 (42.7) .60 30 (50.8) .30 4 (80.0) .63 69 (52.7) .39 ≥3.5 10 (71.4) 11 (36.7) 7 (46.7) 0 (0.0) 7 (53.8) At least 1 dose of PCVf (row %) 0 14 (82.4) .69 1730 (52.7) .002 788 (54.6) .09 37 (45.7) .33 2190 (49.6) <.001 <2.2 1 (100) 45 (64.3) 24 (72.7) 2 (100) 83 (77.6) 2.2–3.5 6 (60.0) .71 71 (64.0) >.99 33 (60.0) .53 2 (50.0) >.99 99 (78.6) >.99 ≥3.5 11 (78.6) 19 (65.5) 10 (66.7) 1 (100) 10 (76.9) Preceding antibioticsg (row %) 0 3 (17.6) .34 1283 (39.4) .91 581 (40.8) .04 39 (48.1) .63 78 (1.8) .05 <2.2 0 (0.0) 23 (34.3) 12 (36.4) 2 (100) 2 (1.9) 2.2–3.5 3 (27.3) .71 46 (40.7) .51 28 (50.0) .05 1 (20.0) .63 5 (3.9) .23 ≥3.5 4 (33.3) 12 (41.4) 10 (66.7) 1 (50.0) 1 (7.7) Pneumococcal NP/ OP PCR load >6.9 log10 copies/mL (row %) 0 10 (62.5) .71 379 (11.3) <.001 175 (11.9) <.001 22 (26.8) .46 344 (7.8) .37 <2.2 0 (0.0) 11 (15.5) 8 (24.2) 0 (0.0) 10 (9.2) 2.2–3.5 8 (80.0) > .99 20 (17.4) .09 12 (20.7) .24 1 (25.0) > .99 13 (10.0) > .99 ≥3.5 9 (64.3) 10 (33.3) 7 (46.7) 0 (0.0) 1 (7.7) Very severe pneumonia (row %) 0 8 (47.1) .56 1064 (31.3) .04 423 (28.3) .018 37 (44.0) > .99 … <2.2 1 (100) 23 (31.9) 9 (26.5) 2 (100) … 2.2–3.5 7 (63.6) .73 42 (35.9) 0.13 24 (40.7) .13 3 (60.0) .16 … ≥3.5 8 (57.1) 15 (50.0) 7 (46.7) 0 (0.0) … CXR+ (row %) 0 14 (100) .02 1493 (51.2) .05 1493 (100) 41 (62.1) .59 … <2.2 1 (100) 34 (50.0) 34 (100) 0 (0.0) … 2.2–3.5 7 (87.5) .40 59 (59.0) .14 59 (100) 3 (60.0) .52 … ≥3.5 7 (63.6) 15 (68.2) 15 (100) 2 (100) … CXR with alveolar consolidation (row %) 0 11 (78.6) .13 678 (23.3) <.001 678 (45.4) <.001 28 (42.4) > .99 … <2.2 1 (100) 16 (23.5) 16 (47.1) 0 (0.0) … 2.2–3.5 6 (75.0) .26 38 (38.0) <.001 38 (64.4) .003 2 (40.0) > .99 … ≥3.5 5 (45.5) 14 (63.6) 14 (93.3) 1 (50.0) … Hypoxemia (row %) 0 4 (23.5) .23 1220 (35.9) .20 637 (42.7) .78 43 (51.8) .07 … <2.2 0 (0.0) 19 (26.4) 13 (38.2) 1 (50.0) … 2.2–3.5 5 (45.5) > .99 54 (46.2) .11 28 (47.5) .77 1 (20.0) .61 … ≥3.5 6 (42.9) 11 (36.7) 4 (26.7) 0 (0.0) … CRP ≥40 mg/L (row %) 0 12 (75) .23 726 (24.7) <.001 413 (31.8) <.001 53 (72.6) .79 … <2.2 1 (100) 18 (29.0) 13 (41.9) 0 (0.0) … 2.2–3.5 8 (100) .58 38 (38.8) <.001 22 (44.0) .06 2 (100) > .99 … ≥3.5 9 (90.0) 17 (73.9) 10 (83.3) 1 (50.0) … WBC count >15/mm3 (row %) 0 9 (52.9) .48 1208 (37.2) .19 581 (40.9) .91 34 (42) .32 … <2.2 1 (100) 25 (37.9) 12 (40.0) 0 (0.0) … 2.2–3.5 5 (50.0) .49 39 (35.1) .26 25 (44.6) .75 2 (40.0) > .99 … ≥3.5 5 (38.5) 6 (21.4) 4 (28.6) 0 (0.0) … Died in hospital (row %) 0 3 (17.6) .34 174 (5.1) .03 70 (4.7) .09 24 (28.6) .85 … <2.2 0 (0.0) 2 (2.8) 2 (5.9) 2 (100) … 2.2–3.5 2 (18.2) .45 10 (8.5) .05 3 (5.1) .26 2 (40.0) .16 … ≥3.5 5 (35.7) 4 (13.3) 3 (20.0) 0 (0.0) … Table excludes human immunodeficiency virus (HIV)–i

Characteristic MCPP Casesa Nonconfirmed Casesb Nonconfirmed CXR+c Cases Confirmed Nonpneumococcal Cased All Controls Total No.e WB PCR ≥2.2 Log10 Copies/mL, No. (%) Total No. WB PCR ≥2.2 Log10 Copies/mL, No. (%) Total No. WB PCR ≥2.2 Log10 Copies/mL, No. (%) Total No. WB PCR ≥2.2 Log10 Copies/mL, No. (%) Total No. WB PCR ≥2.2 Log10 Copies/mL, No. (%) Overall 56 35 (62.5) 3832 166 (4.3) 1745 91 (5.2) 107 10 (9.3) 4987 154 (3.1) Pneumococcal NP/OP PCR load >6.9 log10 copies/mL All sites P = .07 P < .01 P < .01 P = .42 P = .12 Yes 36 26 (72.2) 457 37 (8.1) 231 25 (10.8) 27 3 (11.1) 392 17 (4.3) No 18 8 (44.4) 3306 127 (3.8) 1481 65 (4.4) 77 6 (7.8) 4490 135 (3.0) Kenya … P = .18 P = .09 P > .99 P = .45 Yes 0 0 (0) 26 1 (3.8) 10 1 (10.0) 3 1 (33.3) 14 1 (7.1) No 4 3 (75.0) 529 12 (2.3) 229 7 (3.1) 3 1 (33.3) 735 30 (4.1) The Gambia P = .23 P > .99 P > .99 P = .51 P > .99 Yes 10 4 (40.0) 84 4 (4.8) 42 2 (4.8) 4 0 (0.0) 49 2 (4.1) No 5 0 (0.0) 474 26 (5.5) 201 10 (5.0) 10 3 (30.0) 532 27 (5.1) Mali P = .52 P = .47 P = .27 P = .46 P > .99 Yes 21 17 (81.0) 140 13 (9.3) 54 7 (13.0) 12 1 (8.3) 112 4 (3.6) No 3 2 (66.7) 476 34 (7.1) 174 13 (7.5) 14 0 (0.0) 602 21 (3.5) Zambia P = .40 P = .71 P = .23 P > .99 P = .29 Yes 3 3 (100) 41 1 (2.4) 23 0 (0.0) 1 0 (0.0) 36 2 (5.6) No 3 1 (33.3) 406 23 (5.7) 179 16 (8.9) 21 2 (9.5) 508 14 (2.8) South Africa P > .99 P < .01 P < .01 P = .26 P = .13 Yes 2 2 (100) 100 17 (17.0) 70 14 (20.0) 7 1 (14.3) 94 8 (8.5) No 3 2 (66.7) 781 31 (4.0) 428 19 (4.4) 20 0 (0.0) 864 40 (4.6) Thailand … … … …

P = .42 P = .12 Yes 36 26 (72.2) 457 37 (8.1) 231 25 (10.8) 27 3 (11.1) 392 17 (4.3) No 18 8 (44.4) 3306 127 (3.8) 1481 65 (4.4) 77 6 (7.8) 4490 135 (3.0) Kenya … P = .18 P = .09 P > .99 P = .45 Yes 0 0 (0) 26 1 (3.8) 10 1 (10.0) 3 1 (33.3) 14 1 (7.1) No 4 3 (75.0) 529 12 (2.3) 229 7 (3.1) 3 1 (33.3) 735 30 (4.1) The Gambia P = .23 P > .99 P > .99 P = .51 P > .99 Yes 10 4 (40.0) 84 4 (4.8) 42 2 (4.8) 4 0 (0.0) 49 2 (4.1) No 5 0 (0.0) 474 26 (5.5) 201 10 (5.0) 10 3 (30.0) 532 27 (5.1) Mali P = .52 P = .47 P = .27 P = .46 P > .99 Yes 21 17 (81.0) 140 13 (9.3) 54 7 (13.0) 12 1 (8.3) 112 4 (3.6) No 3 2 (66.7) 476 34 (7.1) 174 13 (7.5) 14 0 (0.0) 602 21 (3.5) Zambia P = .40 P = .71 P = .23 P > .99 P = .29 Yes 3 3 (100) 41 1 (2.4) 23 0 (0.0) 1 0 (0.0) 36 2 (5.6) No 3 1 (33.3) 406 23 (5.7) 179 16 (8.9) 21 2 (9.5) 508 14 (2.8) South Africa P > .99 P < .01 P < .01 P = .26 P = .13 Yes 2 2 (100) 100 17 (17.0) 70 14 (20.0) 7 1 (14.3) 94 8 (8.5) No 3 2 (66.7) 781 31 (4.0) 428 19 (4.4) 20 0 (0.0) 864 40 (4.6) Thailand … … … … P > .99 Yes 0 0 (0) 3 0 (0.0) 2 0 (0.0) 0 0 (0) 8 0 (0.0) No 0 0 (0) 213 0 (0.0) 94 0 (0.0) 6 0 (0.0) 606 3 (0.5) Bangladesh … P = .24 P = .15 … … Yes 0 0 (0) 63 1 (1.6) 30 1 (3.3) 0 0 (0) 79 0 (0.0) No 0 0 (0) 427 1 (0.2) 176 0 (0.0) 3 0 (0.0) 643 0 (0.0) All P values obtained from Fisher exact or χ2 test. Abbreviations: CXR, chest radiograph; MCPP, microbiologically confirmed pneumococcal pneumonia; NP/OP, nasopharyngeal/oropharyngeal; PCR, polymerase chain reaction; WB, whole blood.

Yes 0 0 (0) 63 1 (1.6) 30 1 (3.3) 0 0 (0) 79 0 (0.0) No 0 0 (0) 427 1 (0.2) 176 0 (0.0) 3 0 (0.0) 643 0 (0.0) All P values obtained from Fisher exact or χ2 test. Abbreviations: CXR, chest radiograph; MCPP, microbiologically confirmed pneumococcal pneumonia; NP/OP, nasopharyngeal/oropharyngeal; PCR, polymerase chain reaction; WB, whole blood. aMCPP defined as isolation of pneumococcus from blood culture, culture or PCR of lung aspirate or pleural fluid, or BinaxNOW antigen detection on pleural fluid. bNonconfirmed cases defined as cases without isolation of bacteria from culture of blood, lung aspirate or pleural fluid, or PCR of lung aspirate or pleural fluid. cCXR positive (CXR+) defined as radiographic evidence of pneumonia (consolidation and/or other infiltrates). dConfirmed nonpneumococcal case defined as a case with any nonpneumococcal bacterial pathogen detected by blood culture, by lung aspirate culture or PCR, or by pleural fluid culture or PCR. eSome children who were pneumococcal WB above the threshold were missing NP/OP pneumococcal results; these children are captured in the “Overall” row but excluded from subsequent rows.

dConfirmed nonpneumococcal case defined as a case with any nonpneumococcal bacterial pathogen detected by blood culture, by lung aspirate culture or PCR, or by pleural fluid culture or PCR. eSome children who were pneumococcal WB above the threshold were missing NP/OP pneumococcal results; these children are captured in the “Overall” row but excluded from subsequent rows. Figure 3. Association between pneumococcal polymerase chain reaction (PCR) load (log10 copies/mL) in nasopharyngeal/oropharyngeal (NP/OP) and whole-blood (WB) specimens among cases without evidence of prior antibiotic exposure (Abx), by microbiologically confirmed pneumococcal pneumonia (MCPP) status. PCR load (density) in NP/OP specimens >6.9 log10 copies/mL (horizontal line) demarks the optimal colonization load (density) threshold for discriminating MCPP cases from all controls [13]. The shaded area to the right denotes specimens with WB pneumococcal PCR load ≥2.2 log10 copies/mL. MCPP was defined as pneumococcus isolated from culture of blood, lung aspirate, pleural fluid, PCR of lung aspirate or pleural fluid, or detection of Streptococcus pneumoniae antigen in pleural fluid on BinaxNOW. Nonconfirmed cases were defined as cases without isolation of bacteria from culture of blood, lung aspirate or pleural fluid, or PCR of lung aspirate or pleural fluid. Prior antibiotic exposure was defined as serum bioassay positive, antibiotic administration at the referral facility, or antibiotic administration prior to WB specimen collection at the study facility.

isolation of bacteria from culture of blood, lung aspirate or pleural fluid, or PCR of lung aspirate or pleural fluid. Prior antibiotic exposure was defined as serum bioassay positive, antibiotic administration at the referral facility, or antibiotic administration prior to WB specimen collection at the study facility. Because associations between pneumococcal PCR load and pneumonia and pneumococcal risk factors may differ by HIV status, we restricted evaluations of factors associated with load distribution to HIV-negative children only in Table 2; however, associations above the threshold in Supplementary Table 4A and 4B are shown for all children. Among nonconfirmed PCR positives, higher load was not associated with having very severe pneumonia, hypoxemia, CXR+, or white blood cell count >15/mm3, but there was a nonsignificant trend for proportion of children who died to increase with increasing load (Table 2). However, clinical measures considered suggestive of bacterial pneumonia were associated with increasing load, including findings of alveolar consolidation on CXR (23.5%, 38.0%, and 63.6% for nonconfirmed cases with positive but <2.2, 2.2–3.5, and ≥3.5 log10 copies/mL, respectively) and C-reactive protein ≥40 mg/L (29.0%, 38.8%, and 73.9% for nonconfirmed cases with positive but <2.2, 2.2–3.5, and ≥3.5 log10 copies/mL, respectively; Table 2).

ngs of alveolar consolidation on CXR (23.5%, 38.0%, and 63.6% for nonconfirmed cases with positive but <2.2, 2.2–3.5, and ≥3.5 log10 copies/mL, respectively) and C-reactive protein ≥40 mg/L (29.0%, 38.8%, and 73.9% for nonconfirmed cases with positive but <2.2, 2.2–3.5, and ≥3.5 log10 copies/mL, respectively; Table 2). DISCUSSION To our knowledge, the PERCH study is the only published evaluation of pneumococcal load as measured by lytA in blood to compare pneumonia patients with community controls, mostly because previous studies that evaluated healthy controls did not observe any controls to be pneumococcal PCR positive [18, 19]. We found that among children who had pneumococcus in the blood as measured by PCR, median load was higher in children with MCPP than in community control children or children with pneumonia confirmed for a nonpneumococcal organism. However, even though all but 1 pneumococcal PCR-positive MCPP case had high load (≥2.2 log10 copies/mL), 37.5% of all MCPP cases did not because sensitivity for any positivity was low [3]. Additionally, many children without evidence of pneumococcal pneumonia, including controls, had high load, demonstrating the poor specificity of blood load for use as a clinical diagnostic to identify specific cases with pneumococcal pneumonia. Of 273 community controls who were blood pneumococcal PCR positive, 56% had high load (3.1% overall). In addition, 10 of 12 (83.3%) PCR-positive cases confirmed for another etiology had high load (9.3% overall), which may mean either poor specificity or that pneumococcus was a coinfecting agent in these cases, just not cultured from blood. Despite this poor sensitivity and specificity, having high pneumococcal blood load may have some value in ascribing etiology in epidemiological studies as the overall proportion in controls (3.1%) was lower than that in nonconfirmed cases (4.3%), but the association was not strong (AOR, 1.3; 95% CI, 1.0–1.6). Pneumococcal load among nonconfirmed cases was lower than among MCPP cases, suggesting that if PCR positives in nonconfirmed cases are pneumococcal cases, they are different from MCPP cases. Caution is needed in interpreting high load in pneumonia cases as having etiologies attributed to pneumococcus because of the relatively large proportion (9.3%) of cases confirmed for a nonpneumococcal pathogen that were also pneumococcal PCR positive and had high load.

coccal cases, they are different from MCPP cases. Caution is needed in interpreting high load in pneumonia cases as having etiologies attributed to pneumococcus because of the relatively large proportion (9.3%) of cases confirmed for a nonpneumococcal pathogen that were also pneumococcal PCR positive and had high load. The gain in specificity in using high pneumococcal load in blood compared to any positivity was at little cost to sensitivity as all but 1 PCR-positive MCPP case had high load. However, overall the sensitivity was poor (62.5%). This may be due in part to such small specimen volumes being tested, increasing the likelihood of cases with low bacterial counts in blood to be missed. Repeat testing may improve the sensitivity of PCR to detect cases with pneumococcus in the blood, but because these cases are more likely to be the cases with low bacterial load, repeat testing may not improve the sensitivity of detecting high load. Two small studies evaluated pneumococcal bacteremia in nonhospitalized controls as measured by culture in organisms per milliliter; they compared blood pneumococcal culture load in pneumonia cases compared to controls with RTI, otitis media, or fever without focus of infection [5, 7]. One study found that pneumococcal culture load was similar between pneumonia cases and controls, with 1 of 4 (25%) pneumonia cases having ≥10 organisms/mL compared with 5 of 25 (20%) controls (investigators observed similar findings for H. influenzae), whereas the other found high load in the single pneumonia case compared to 1 of 19 controls.

mococcal culture load was similar between pneumonia cases and controls, with 1 of 4 (25%) pneumonia cases having ≥10 organisms/mL compared with 5 of 25 (20%) controls (investigators observed similar findings for H. influenzae), whereas the other found high load in the single pneumonia case compared to 1 of 19 controls. There are few studies that describe the factors associated with pneumococcal PCR load in blood. Only 2 studies in children were found [20, 21], 1 of which evaluated a different gene target, pneumolysin (ply), in only 11 blood culture–negative Malawian children with CXR+ pneumonia [20].The distribution of load in the Malawian study (median, 0.34 × 103 copies/mL) was similar to that in comparable PERCH cases (median, 0.3 × 103 copies/mL), whereas the distribution in the other South African study (median in children, approximately 2.8 log10 [0.63 × 103] copies/mL) was twice as high as in PERCH. Both studies observed higher load in HIV-infected children, in contrast to PERCH, which found that load did not differ by HIV status; however, the Malawian results were not statistically significant and included meningitis cases, which drove much of the difference between HIV-positive and HIV-negative cases, and the evaluation of HIV in the South African study included nonsevere pneumonia cases and adults, unlike PERCH. PERCH also had relatively few HIV-infected cases relative to the South African study.

gnificant and included meningitis cases, which drove much of the difference between HIV-positive and HIV-negative cases, and the evaluation of HIV in the South African study included nonsevere pneumonia cases and adults, unlike PERCH. PERCH also had relatively few HIV-infected cases relative to the South African study. Three studies, all in adults, evaluated the association between pneumococcal load in blood by PCR using the lytA target and pneumococcal blood culture–positive pneumonia, and all observed higher pneumococcal blood load by PCR among blood culture–positive patients compared with blood culture–negative patients [4, 22, 23]. The PERCH study results support these findings; among blood pneumococcal PCR-positive pneumonia cases, 29 of 30 (96.7%) that were blood culture positive had high load compared with 166 of 242 (68.6%) that were blood culture negative (P < .01). However, as stated above, these findings are not specific to pneumococcal pneumonia, as 154 of 273 (56.4%) pneumococcal PCR-positive community controls had high load, as did 10 of 12 (83.3%) cases confirmed for a nonpneumococcal pathogen, although the latter could indicate coinfection. A finding of high blood pneumococcal load or viable organisms in the blood may be an indication of more advanced or severe disease, as all of the above cited studies that evaluated disease severity found higher load among the most severe cases as assessed by pneumonia severity index risk class, intensive care unit admission, mental status, or mortality [4, 20–23]. Studies that evaluated pneumococcus by syndrome type (ie, meningitis vs pneumonia) as a measure of severity found higher load among meningitis compared with pneumonia cases [5–7]. The PERCH study also observed higher pneumococcal load in children who died (P = .05), despite the fact that high load was not associated with other indications of severity, such as low oxygen saturation or pneumonia danger signs, perhaps because all children enrolled in PERCH had severe or very severe pneumonia by design. Additionally, children with WHO-defined very severe pneumonia syndrome enrolled in PERCH may have been suffering from illness caused by organisms other than pneumococcus.

uch as low oxygen saturation or pneumonia danger signs, perhaps because all children enrolled in PERCH had severe or very severe pneumonia by design. Additionally, children with WHO-defined very severe pneumonia syndrome enrolled in PERCH may have been suffering from illness caused by organisms other than pneumococcus. We were only able to assess the value of blood pneumococcal load in the diagnosis of pneumococcal pneumonia at the African sites since so few cases and controls had positive blood pneumococcal PCR at the Asian sites. However, the load distributions of the 8 cases and 11 controls at the Asian sites with positive blood pneumococcal PCR detected overlapped, as was observed at the African sites. The PERCH study highlights this regional difference in positivity and the need for further study in Asia to interrogate the association of pneumococcal NP/OP and blood densities in cases and controls.

rols at the Asian sites with positive blood pneumococcal PCR detected overlapped, as was observed at the African sites. The PERCH study highlights this regional difference in positivity and the need for further study in Asia to interrogate the association of pneumococcal NP/OP and blood densities in cases and controls. Our finding that some of the community controls had high pneumococcal loads in blood is intriguing. It is possible that this might indicate early signs of or increased risk of developing disease in some of these children. Because we did not follow controls longitudinally, we could not determine if any went on to become severely ill. However, we did monitor whether cases enrolled in PERCH had previously been enrolled as a control, and none of the controls with high load were the ones we know of that later became a case. We also did not find that children with RTI were more likely than well controls to have high pneumococcal blood load, which would have been expected if high load indicated early disease.

had previously been enrolled as a control, and none of the controls with high load were the ones we know of that later became a case. We also did not find that children with RTI were more likely than well controls to have high pneumococcal blood load, which would have been expected if high load indicated early disease. It was surprising to find that a greater proportion of cases confirmed for a nonpneumococcal pathogen had high pneumococcal PCR load compared with cases not confirmed for any pathogen (9.3% vs 4.3%; P = .03) and that, among pneumococcal PCR-positive cases, the proportion with high load was similar to that in MCPP cases. Perhaps some had true dual infection with >1 pathogen causing their pneumonia. It is also possible that pneumococcus colonizing the nasopharynx entered the bloodstream because of reduced immune protection or some other mechanism related to the pneumonia caused by the other nonpneumococcal pathogen. A major limitation of any blood test in the attribution of pneumococcus as the etiologic cause of a pneumonia episode is the fact that existing methodologies can only identify pneumococcal cases in which pneumococci have entered the bloodstream, and will miss those where bacteremia is absent or transient. Therefore, other measurements such as high load in the nasopharynx may help identify the pneumococcal cases in whom bacteremia is not detected.

that existing methodologies can only identify pneumococcal cases in which pneumococci have entered the bloodstream, and will miss those where bacteremia is absent or transient. Therefore, other measurements such as high load in the nasopharynx may help identify the pneumococcal cases in whom bacteremia is not detected. The associations of pneumococcal load in blood with MCPP case status, blood culture positivity, and severity are intuitive. However, blood pneumococcal load by itself cannot be used for diagnosing pneumococcal pneumonia in individual children 1–59 months of age. This signals that the determinants of pneumococcal load in blood by PCR are complex and incompletely understood. Despite this, pneumococcal load in blood may be informative in attributing the burden of disease caused by pneumococcus in epidemiological studies. Supplementary Data Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author. Supplementary Material cix149_suppl_Supplemental_Tables_Figures Click here for additional data file. Notes

Supplementary Data Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author. Supplementary Material cix149_suppl_Supplemental_Tables_Figures Click here for additional data file. Notes Author contributions. M. D. K. and S. C. M. led the analysis, interpreted results, and drafted the manuscript. J. A. G. S., D. R. M., and S. A. M. assisted with interpretation of results and drafting of the manuscript. N. L. W. and D. E. P. performed analyses and assisted with interpretation of results. M. D. K., J. A.G. S., H. C. B., W. A. B, D. R. F., L. L. H., S. R. C. H., K. L. K., O. S. L., K. L. O., D. M. T., R. A. K., D. R. M., and S. A. M. conceived and designed the study and supervised study conduct. D. A., M. A., J. O. A., V. L. B., J. C., A. N. D., M. D., A. J. D., M. M. H., A. J., R. M., D. P. M., J. M., S. N., C. P., P. S., D. S., S. O. S., and B. T. were involved in study conduct, data collection, and/or data management. S. Z. provided statistical guidance. All authors reviewed and approved the manuscript. M. D. K. and S. C. M. had full access to the data and had final responsibility for the decision to submit for publication.

P., P. S., D. S., S. O. S., and B. T. were involved in study conduct, data collection, and/or data management. S. Z. provided statistical guidance. All authors reviewed and approved the manuscript. M. D. K. and S. C. M. had full access to the data and had final responsibility for the decision to submit for publication. Acknowledgments. We acknowledge members of the following groups who contributed to the study design, conduct, and analysis phases of PERCH (see Supplementary Materials for full list of names): Pneumonia Methods Working Group, PERCH Expert Group, PERCH Contributors, and the PERCH Chest Radiograph Reading Panel. We offer sincere thanks to the patients and families who participated in this study. Disclaimer. The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention, Department of Health and Human Services, or the US government. This article is published with the permission of the Director of the Kenya Medical Research Institute. Financial support. PERCH was supported by the Bill & Melinda Gates Foundation (grant number 48968 to the International Vaccine Access Center, Department of International Health, Johns Hopkins Bloomberg School of Public Health). J. A. G. S. was supported by a clinical fellowship from the Wellcome Trust of Great Britain (award number 098532).

onsorship. This article appears as part of the supplement “Pneumonia Etiology Research for Child Health (PERCH): Foundational Basis for the Primary Etiology Results,” sponsored by a grant from the Bill & Melinda Gates Foundation to the PERCH study of Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland. Potential conflicts of interest. M. D. K. has received funding for consultancies from Merck, Pfizer, and Novartis, and grant funding from Merck. L. L. H. has received grant funding from Pfizer and GlaxoSmithKline. K. L. K. has received grant funding from Merck Sharp & Dohme. S. A. M. has received honoraria for advisory board membership from the Bill & Melinda Gates Foundation, Pfizer, Medimmune, and Novartis; has received institutional grants from GSK, Novartis, Pfizer, Minervax, and the Bill & Melinda Gates Foundation; and has served on speaker’s bureaus for Sanofi Pasteur and GSK. K. L. O. has received grant funding from GSK and Pfizer and participates on technical advisory boards for Merck, Sanofi Pasteur, PATH, Affinivax, and ClearPath. All other authors report no reported conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.