Clinical signs of pneumonia in children attending a hospital outpatient department in Lesotho.

Clinical signs of pneumonia in children attending a hospital outpatient department in Lesotho S.C. Redd,' R. Vreuls,2 M. Metsing,3 P.H. Mohobane,4 E. Patrick,5 & M. Moteetee6 To determine the value of clinical findings for the diagnosis of pneumonia, we evaluated 950 children who presented with respiratory illness to the outpatient department of the Queen Elizabeth II Hospital, Maseru, Lesotho. Those children at high risk for pneumonia and a systematically selected 20% sample of children at low risk were examined in turn by a nurse, a general practitioner, and a paediatrician; a chest radiograph was recorded for each child. Pneumonia was defined as radiographic findings compat- ible with the disease as interpreted by a paediatric radiologist. A respiratory rate .50 breaths/minute was a sensitive sign for pneumonia among infants (sensitivity range for the three examiners: 59-79%/o), but identified a progressively smaller proportion of children with pneumonia in older age groups. Adjust- ing the respiratory rate for age using a threshold of .40 breaths/minute for children aged .12 months improved the sensitivity, but identified <30% of children with pneumonia ageq .24 months. No drop in sensitivity with age was found when respiratory rate thresholds were evaluated for children with more severe radiographic evidence of pneumonia. Introduction Acute respiratory infections (ARI) are a major cause of mortality among children in developing countries, accounting for over 4 million deaths per year (1) - nearly one-third of all childhood deaths (2). In addi- tion to promoting routine childhood immunization against measles and pertussis, the WHO Control of Acute Respiratory Infections programme stresses improved case management to prevent deaths from ARI, principally by treating children with pneumonia with antimicrobial drugs. For the case management strategy to be effective, episodes of pneumonia must be promptly identified and treated, often by persons with limited training and access to advanced medical technology. I Medical Epidemiologist, Malaria Branch, Centers for Disease Control, mailstop F12, Atlanta, GA 30333, USA. Requests for reprints should be sent to Dr Redd at this address. 2 Paediatrician, Medical Spectrum Twente, Enshede, Nether- lands. Formerly: Chief of Paediatrics, Ministry of Health, Maseru, Lesotho. 3 Paediatrician, Ministry of Health, Maseru, Lesotho. 4 Nurse Clinician, Ministry of Health, Maseru, Lesotho. 5 Radiologist, Henrietta Egleston Hospital, Emory University School of Medicine, Atlanta, GA, USA. 6 Physician, UNICEF, Mogadishu, Somalia. Formerly: Head, Family Health Division, Ministry of Health, Maseru, Lesotho. Reprint No. 5459 To determine the value of clinical findings to identify children with pneumonia, particularly the findings recommended by WHO, we evaluated 950 children with respiratory illness who had been brought to a hospital outpatient department. Chest radiographs were used to document pneumonia, and children who, on preliminary screening, appeared to be at low risk for pneumonia were also included. Three health workers examined most of the children, enabling us to compare the findings of different observers. Methods The study was conducted in Maseru, Lesotho, at the Queen Elizabeth II Hospital, the central referral hos- pital for Lesotho, which also serves as a source of primary health care for many residents of Maseru. Approximately 40 under-5-year-olds, not limited to those with respiratory illness, were seen at the out- patient department each work day during the study. Children aged 3 months to 5 years with a cough, blocked or runny nose, ear pain, or breathing diffi- culties, who were brought to the outpatient depart- ment over a 3-month-period were eligible for enrol- ment in the study. After the verbal, informed consent of the child's carer had been obtained, a trained nurse assessed each child. In order to identify as many cases of pneumonia as possible, but minimize the radiation exposure to children at low risk for Bulletin of the World Health Organization, 1994, 72 (1): 11 3-118 113 S.C. Redd et al. pneumonia, we classified children into high-risk and low-risk groups based on this initial assessment. Children with a history of rapid breathing, a history of difficulty in drinking, elevated respiratory rate (>40 breaths/minute for children aged .12 months; >50 breaths/minute for children aged 3-12 months), wheezing, nasal flaring, or chest indrawing were defined as being at risk for pneumonia. Children without any of these findings were classified as being at low risk for pneumonia. All high-risk chil- dren and a systematically selected 20% sample of the low-risk children underwent further standard clinical examinations. Children were examined, in order, by a general practitioner and a paediatrician, and lastly a chest radiograph was recorded. The examinations were conducted within approximately a two-hour interval. The complete examinations by the general prac- titioner and paediatrician included measurement of respiratory and pulse rates, rectal temperature, assessment for chest indrawing and nasal flaring, auscultation of the chest, otoscopy, and measurement of height and weight. The respiratory rate was meas- ured for one minute using electronic sounding timers on calm, awake children. The proportion of children who were crying and could not be consoled at the time of examination ranged between 1% and 4% for the three examiners; the results of these examina- tions were included in the analyses. Chest indrawing was defined as either intercostal, subcostal, or lower chest wall indrawing, in accord with WHO recom- mendations from 1988, which differ from the current definition that includes only lower chest wall indrawing.a The radiographs were reviewed in the USA after the end of patient enrolment. Pneumonia was defined as the presence of a pulmonary parenchymal density compatible with pneumonia on chest radiography as interpreted by the paediatric radiologist (E.L.P) in the USA. Technically poor radiographs were classi- fied as unreadable, while those with equivocal find- ings were classified as uninterpretable. Those chest radiographs with pulmonary findings compatible with pneumonia and a systematically selected 20% sample of those without pneumonia were re-read by the same paediatric radiologist. Radiographs for which the two interpretations did not agree were reviewed a third time and assigned to an indetermi- nate category if the readings could not be definitive- ly reconciled. Radiographs with abnormalities that were consistent with pneumonia were also rated as a Acute respiratory infections in children: case management in small hospitals in developing countries. Unpublished document WHO/ARI/90.5, 1990. being severe (lobar or more than one lobe of involve- ment) or not severe. The frequencies of the clinical findings were compared in children with and without radiographic evidence of pneumonia, excluding unreadable and uninterpretable radiographs. The results were weight- ed to calculate the sensitivity and specificity of single and combined clinical findings for identify- ing children with radiographic pneumonia as if all low-risk patients had been examined. The definitions of sensitivity and specificity used were standard (3). Continuous variables were assessed using Wilcoxon's rank sums test (4). The mean difference in respira- tory rate measurements between examiners was cal- culated for the children assessed by each pair of examiners. The study protocol was approved by the Lesotho National Research Review Committee and the Cen- ters for Disease Control Institutional Review Board. Results A total of 950 children with respiratory infection were potentially eligible for the study (277 at high risk and 673 at low risk for pneumonia). All the high-risk children and 128 of the 134 eligible low- risk children (96%) were enrolled. A total of 382 (94% of those enrolled) were examined by the gen- eral practitioner and 251 (62% of those enrolled) by the paediatrician. There were no differences in the age, sex distribution, or proportion of children diagnosed as having pneumonia among those exam- ined or not examined by the paediatrician. Chest radiographs were available for 393 children (97% of those enrolled). The median age of the enrolled children was 11.8 months (range, 3-59 months); high-risk chil- dren were significantly younger (median age, 9.9 months) than low-risk children (median age, 19.5 months, P <0.0001, Wilcoxon's rank sums test). Of the 391 enrolled children whose sex was recorded, 228 (58%) were male. Presenting complaints, indi- cating that enrolled children had respiratory illness, included cough (403 (99%)), runny nose (323 (80%)), blocked nose (217 (54%)), problems in breathing (70 (17%)), and earache (8 (2%)). Only 16 (4%) of the children had received treatment with an antimicrobi- al drug before entering the study. The radiographs of 78 of the children (20%) were initially interpreted as being compatible with pneumonia; upon review, 64 (82%) were categorized as having evidence of pneumonia. The radiographs of 52 children initially held not to show evidence of pneumonia were re-examined, and one child (2%) was interpreted as having pneumonia. The 15 radio- graphs that had different first and second interpreta- 114 WHO Bulletin OMS. Vol 721994 Clinical signs of pneumonia in children tions were examined a third time; five were interpret- ed as indicating atelectasis rather than pneumonia; five were judged to be uninterpretable; four as not showing pneumonia; and one as showing pneumonia. Thus, 65 children (17%) had radiographic findings consistent with pneumonia-53 in the high-risk group and 12 in the low-risk group. The radiographs of 299 of the children (76%) did not show evidence of pneu- monia; 17 (4%) were uninterpretable; and 12 (3%) could not be read because their quality was poor. To assess how well the 20% sample of children at low risk for pneumonia who were enrolled in the study was representative of all the children at low risk for pneumonia, we compared the characteristics of the 128 enrolled; children with those of the 236 children not enrolled and for whom questionnaires were available. There were negligible differences in the respiratory rates (mean, 34.2 breaths/minute for enrolled, compared with 33.5 breaths/minute for non- enrolled children), age (mean, 22.5 months for enrol- led; 21.7 months for non-enrolled), and sex distribu- tion (59% male for enrolled; 55% for non-enrolled), suggesting that the low-risk children not enrolled in the study had the same clinical manifestations of ARI and risk for pneumonia as the low-risk children who were enrolled. We therefore, weighted each observa- tion from a low-risk child by a factor of five to esti- mate the distribution of pneumonia and clinical find- ings among all the study children attending the outpatient department with a respiratory illness. The respiratory rate was the best sign for dis- tinguishing children who had pneumonia from those who did not. Children with pneumonia had a mean + SD respiratory rate of 53.0 ± 13.9 breaths/minute, as measured by the nurse, compared with 44.3 ± 11.1 for children without pneumonia (unweighted data, P = 0.0001, Wilcoxon's rank sums test). Similar mean rates were reported by the general practitioner and the paediatrician. We evaluated different respiratory rate cut-offs to identify children with pneumonia, as proposed by the WHO Control of Acute Respiratory Infections programme (Table 1). For children under 12 months of age, a respiratory rate .50 breaths/minute was 59-79% sensitive for identifying those with pneumo- nia, according to the three examiners. This cut-off identified a progressively smaller proportion of chil- dren with pneumonia in the older age groups (Table 1). A respiratory rate threshold of .40 breaths/min- ute identified a higher proportion of children with pneumonia than a threshold of .50, but for children under 12 months of age the specificity for the lower respiratory rate threshold was <50% for all three examiners. The paediatrician exhibited the largest gain in sensitivity for the 12-23-month age group (Table 1). Neither of these respiratory rate thresholds classified correctly more than 30% of children over 24 months old with pneumonia. The specificity increased with the age of the child for all three examiners. Of other clinical findings, neither crepita- tions nor nasal flaring was more than 50% sensitive at identifying children with pneumonia (Table 2). A history of rapid breathing, according to the mother, was more sensitive, but for young children was less than 50% specific (Table 2). Combined definitions of respiratory rate cut-offs with other clinical findings improved the sensitivity minimally (by 2-5%, data not shown). Table 1: Prevalence of elevated respiratory rates, as measured by a nurse, general practitioner, and paediatrician, and radiographic evidence of pneumonia among the study children. Age group (months) Sensitivityc 3-11 12-23 .24 Specificityd 3-11 12-23 224 Rate measured by nursea flb .50 >40 22/2 59 84 19/4 41 49 11/6 24 27 132/29 44/32 16/45 72 90 97 44 64 87 Rate measured by Gpa flb >50 >40 21/2 65 18/4 40 11/6 15 124/29 42/32 16/41 60 76 96 81 42 27 Rate measured by paediatriciana nb .50 .40 14/1 79 13/4 21 6/3 14 24 48 83 90/21 59 30/22 85 10/24 97 100 73 24 25 52 88 a Rate in breaths per minute. b No. of children at high risk/No. of children at low risk. c Sensitivity to identify children with radiographic evidence of pneumonia (calculated by counting each observation for a high-risk child and weighting each observation for a low-risk child by 5). d Specificity to identify children with no radiographic evidence of pneumonia (calculated by counting each observation for a high-risk child and weighting each observation for a low-risk child by 5). WHO Bulletin OMS. Vol 72 1994 115 S.C. Redd et al. Table 2: Clinical findings, as reported by a nurse, general practitioner, and paediatrician, for identification of the study children with radiographic evidence of pneumonia Nurse GP Paediatrician Age group (months) Fast breathinga Nasal flaring Nasal flaring Crepitations Nasal flaring Crepitations Sensitivityb 3-11 69 19 42 19 32 32 12-23 49 24 26 13 27 27 .24 24 8 17 32 14 38 Specificityc 3-11 51 93 93 93 95 96 12-23 71 97 95 89 94 87 .24 92 99 90 92 93 87 a History reported by mother. b Sensitivity to identify children with radiographic evidence of pneumonia (calculated by counting each observation for a high-risk child and weighting each observation for a low-risk child by 5). c Specificity to identify children with no radiographic evidence of pneumonia (calculated by counting each observation for a high-risk child and weighting each observation for a low-risk child by 5). We also evaluated the two above-mentioned re- spiratory rate thresholds in children with more severe radiographic evidence of pneumonia, defined as evi- dence of lobar pneumonia or involvement of more than one lobe. Although there were few cases in any specific age stratum, elevated respiratory rate thresh- olds identified a high proportion of cases of radio- graphically severe pneumonia, especially among children aged .12 months (Table 3). There was no drop in sensitivity for older children; however, a re- spiratory rate threshold of .50 breaths/minute iden- tified less than half the children under 12 months of age with radiographically severe pneumonia who were examined by the nurse and general practitioner. In evaluating children with radiographic evidence of pneumonia, but who did not have radiographically severe pneumonia, both respiratory rate thresholds were less sensitive for older children (Table 2). Comparison of the respiratory rate measure- ments made by the three examiners indicated that the the nurse's mean measurement was 2.3 breaths/min- ute less than the general practitioner's, and 1.2 breaths/minute less than the paediatrician's; the gen- eral practitioner's mean measurement was 0.9 breaths/minute higher than the paediatrician's. Discussion The results of this study support the recommenda- tions of the WHO Control of Acute Respiratory Infections programme for identifying children with pneumonia;b clinical signs can be used to classify b Supervisory skills: management of the young child with acute respiratory infection. Unpublished WHO document, 1991. children who have or do not have pneumonia without recourse to expensive laboratory tests that are fre- quently not available in developing countries. Our data confirm the reports of other workers (5-11) that rapid breathing in young children with a cough is the best sign for identifying those with pneumonia. Our findings extend these previous observations by inclu- ding multiple examiners in a single study and by analysing in more detail the effect of age on the value of rapid breathing as a marker for children with pneumonia. The weighting system we used avoided verification bias (12) and incomplete assess- ment of the population under study. The low preva- lence of pneumonia in our study population (13% (125/950)) is much less than that reported in pre- vious studies and could indicate that we investigated a less severely ill population than other workers. For infants, the clinical definition of pneumonia (respiratory rate .50 breaths/minute) proposed by WHO effectively identified children with radio- graphic evidence (Table 1). The reported sensitivities (6-8, 10) for pneumonia in infants lie in the range 71-89%, with the lowest sensitivity (10) being from a study that compared clinical signs with radiograph- ic findings, as did our study. Other clinical findings were insufficiently sensitive to be useful for identify- ing children with pneumonia, and combined defini- tions did not improve detection of children with pneumonia. We further documented (6-10) that a respiratory rate of .50 breaths/minute was less sensitive at iden- tifying pneumonia in older children than in infants. By evaluating the findings of the three examiners separately, we were able to determine the effect of reducing the respiratory rate threshold to 40 breaths/minute for children aged .12 months. For all 116 WHO Bulletin OMS. Vol 72 1994 Clinical signs of pneumonia In children Table 3: Association of elevated respiratory rate, as reported by a nurse, general practitioner, and paediatrician, and radiographic evidence of lobar or multi-lobe pneumonia, among the study children Rate measured Rate measured Rate measured by nursea by Gpa by paediatriciana Age group (months) rnb >50 >40 rib .50 .40 rib .50 .40 Sensitivity (severe pneumonia)c 3-11 10/1 38 67 9/1 38 64 5/0 60 100 12-23 8/0 88 100 8/0 75 88 6/0 67 67 .24 7/0 100 100 7/0 57 100 5/0 60 80 Sensitivity (non-severe pneumonia)d 3-11 22/1 74 100 20/1 83 92 15/1 75 85 12-23 19/7 29 44 18/7 30 30 11/6 15 68 >24 5/8 9 11 5/8 6 12 1/4 0 5 a Rate in breaths per minute. b No. of children at high risk/No. of children at low risk. c Sensitivity for children with radiographic evidence of lobar or multi-lobe pneumonia (calculated by counting each observation for a high-risk child and weighting each observation for a low-risk child by 5). d Sensitivity for children with radiographic evidence of non-severe pneumonia (calculated by counting each observation for a high-risk child and weighting each observation for a low-risk child by 5). the examiners the sensitivity increased for the 12-23-month age group (Table 1); the increase was 52% for the paediatrician but only 2% for the gener- al practitioner. The differences in the patient popula- tions that each examiner evaluated, changes in re- spiratory rate that may have occurred between the first and last examination, and the large effect of misclassifying low-risk patients who had pneumonia account for the differences in sensitivity. Previous studies have reported increases in sensitivity of between 9% (7) and 21% (8) caused by lowering the respiratory rate threshold to 40 breaths/minute for the age group .12 months. For children aged .24 months neither respiratory rate threshold (40 or 50 breaths/minute) was adequate for identifying pneu- monia, and for all the examiners the sensitivities were below 30%. Our evaluation of children with radiographic evidence of severe pneumonia is not an assessment of the recommendations made by the WHO Control of Acute Respiratory Infections programme for refer- ral of children with severe pneumonia. Radiographic evidence of the severity of pneumonia may not be a good indicator of the likelihood of a poor outcome (13), and may be even less useful for the youngest children. Our appraisal of severity may be more use- ful in assessing the likelihood that a child really has pneumonia, those with the most severe changes most probably having bacterial pneumonia (14). For chil- dren aged .12 months with radiographic evidence of severe pneumonia, we found that either the 40 or 50 breaths/minute theshold would have correctly classi- fied almost all children as having pneumonia. Although there were few patients in any particular age stratum, it is reassuring that even if the overall sensitivity is low most children with radiographic evidence of severe pneumonia would be identified. Our data suggest that a single respiratory rate threshold of 50 breaths/minute may be adequate for identifying children with severe pneumonia, although the sensitivity can be increased by using an age- adjusted respiratory rate threshold. These findings could explain why intervention studies that have used a single respiratory rate threshold to identify cases of pneumonia for all age groups have demon- strated substantial reduction in mortality rates.c Min- istries of health in countries with substantial child mortality from pneumonia should develop control programmes based on the case management strategy. An assessment algorithm for pneumonia based on measuring the respiratory rate in children with re- spiratory illness is the key to this strategy. c Technical bases for the WHO recommendations on the man- agement of pneumonia in children at first-level health facilities. Unpublished document WHO/ARI/91.20, 1991. Acknowledgements We thank Dr S. Gove, Dr H. Campbell, Dr P. Margolis, and Dr 1. de Zoysa (WHO); and Mr D. Gittelman, Mr G. Stroh, Dr R. Waldman, and Dr A. Vernon (CDC) for helpful discussions on the study design and for critically reading the manuscript. This study was supported by the U.S. Agency for International Development through the Africa Child Survi- val Initiative-Combatting Childhood Communicable Dis- eases project. WHO Bulletin OMS. Vol 72 1994 117 S.C. Redd et al. R6sum6 Signes cliniques de pneumopathie chez des enfants vus dans le service de consultations externes d'un hopital au Lesotho La strategie de prise en charge des cas visant a 6viter les deces par pneumopathie chez l'enfant dans les pays en d6veloppement exige une methode simple d'identification des cas. Afin de d6terminer la valeur des observations cliniques conduisant au diagnostic de pneumopathie, nous avons evalu6 950 enfants amen6s pour une mala- die respiratoire a la consultation externe de l'H6pi- tal Queen Elizabeth II, Maseru, Lesotho. Les en- fants a haut risque ont e examines tour a tour par une infirmiere, un m6decin g6neraliste et un p6diatre; pour chaque enfant, une radiographie pulmonaire a ete faite. La pneumopathie etait d6fi- nie par des observations radiographiques compa- tibles avec cette maladie, interpr6t6es par un radiologue sp6cialis6 en p6diatrie. Une fr6quence respiratoire .50 respirations par minute 6tait un signe sensible chez les nourrissons (la sensibilit6 du diagnostic pour les trois examinateurs allait de 59 a 79%), mais la proportion d'enfants atteints de pneumopathie qu'elle permettait d'identifier diminuait progressivement avec l'age. Lorsqu'on corrigeait la frequence respiratoire de l'age en prenant un seuil de 40 respirations par minute pour les enfants de 12 mois et plus, on am6liorait la sensibilit6, mais on identifiait moins de 30% des cas chez les enfants de 24 mois et plus. Aucune diminution de la sensibilit6 avec l'age n'a 6te trou- v6e lorsque les seuils de frequence respiratoire ont et6 6valu6s chez les enfants presentant des signes radiographiques plus severes. Ces obser- vations vont dans le sens des recommandations actuelles de l'OMS pour l'identification des enfants atteints de pneumopathie. Bien que la sensibilit6 chez l'enfant de 12 mois et plus soit plus faible dans notre 6tude que dans les 6tudes anterieures, les enfants presentant des signes radiographiques severes ont Wt6 facilement identifies. References 1. Leowski J. Mortality from acute respiratory infec- tions in children under 5 years of age: global esti- mates. World health statistics quarterly, 1986, 39: 138-144. 2. Gwatkin DR. How many die? A set of demographic estimates of the annual number of infant and child deaths in the world. American journal of public health, 1980, 70: 1286-1289. 3. Griner PF et al. Selection and interpretation of diagnostic tests and procedures. Annals of internal medicine, 1981, 94: 553-600 4. Siegel S. Non-parametric statistics for the behavior- al sciences. New York, McGraw Hill, 1956: 127- 136. 5. Levanthal JM. Clinical predictors of pneumonia as a guide to ordering chest roentgenograms. Clinical pediatrics, 1982, 21: 730-734. 6. Shann F et al. Acute lower respiratory tract infec- tions in children: possible criteria for selection of patients for antibiotic therapy and hospital admis- sion. Bulletin of the World Health Organization, 1984, 62: 749-753. 7. Cherian T et al. Evaluation of simple clinical signs for the diagnosis of acute lower respiratory tract infection. Lancet, 1988, 2:125-128. 8. Campbell H et al. Simple clinical signs for diagno- sis of acute lower respiratory tract infections. Lan- cet, 1988, 2: 742-743. 9. Hortal M et al. Simple clinical signs for diagnosis of acute respiratory infections. Lancet, 1989, 2: 1275. 10. Harari M et al. Clinical signs of pneumonia in chil- dren. Lancet, 1991, 338: 928-930. 11. Mulholland EK et al. Standardized diagnosis of pneumonia in developing countries. Pediatric infec- tious disease journal, 1992, 11: 77-81. 12. Begg CB. Methodologic standards for diagnostic test assessment studies. Journal of general internal medicine, 1988, 3: 518-520. 13. Shann F. WHO definition of tachypnoea in children. Lancet, 1992, 339: 176-177. 14. Wall RA et al. The etiology of lobar pneumonia in the Gambia. Bulletin of the World Health Organiza- tion, 1986, 64: 553-558. 118 WHO Bulletin OMS. Vol 721994