An evaluation of clinical indicators for severe paediatric illness.

An evaluation of clinical indicators for severe paediatric illness L.A. Paxton,1 S.C. Redd,2 R.W. Steketee,3 J.O. Otieno,4 & B. Nahlen5 To help reduce paediatric morbidity and mortality in the developing world, WHO has developed a diagnostic and treatment algorithm that targets the principal causes of death in children, which include acute respiratory infection, malaria, measles, diarrhoeal disease, and malnutrition. With this algorithm, known as the Sick Child Charts, severely ill children are rapidly identified, through the presence ofany one of 13 signs indicative of severe illness, and referred for more intensive health care. These signs are the inability to drink, abnormal mental status (abnormally sleepy), convulsions, wasting, oedema, chest wall retraction, stridor, abnormal skin turgor, repeated vomiting, stiff neck, tender swelling behind the ear, pallor of the conjunctiva, and comeal ulceration. The usefulness of these signs, both in current clinical practice and within the optimized context of the Sick Child Chart algorithm in a rural district of western Kenya, was evaluated. We found that27% of children seen in outpatient clinics had one or more of these signs and that pallor and chest wall retraction were the signs most likely to be associated with hospital admission (odds ratio (OR) = 8.6 and 5.3, respectively). Presentation with any of these signs led to a 3.2 times increased likelihood of admission, although 54% of hospitalized children had no such signs and21% of children sent home from the outpatient clinic had at least one sign. Among inpatients, 58% of all children and 89% of children who died had been admitted with a sign. Abnormal mental status was the sign most highly associated with death (OR = 59.6), followed by poor skin turgor (OR = 5.6), pallor (OR = 4.3), repeated vomiting (OR = 3.6), chest wall retraction (OR = 2.7), and oedema (OR = 2.4). Overall, the mortality risk associated with having at least one sign was 6.5 times higher than that for children without any sign. While these signs are useful in identifying a subset of children at high risk of death, their validation in other settings is needed. The training and supervision of health workers to identify severely ill children should continue to be given high priority because of the benefits, such as reduction of childhood mortality. Introduction The reduction of child mortality rates in developing countries, which are about 5-15 times higher than those in developed countries (1-3), is one of the greatest public health challenges faced by develop- ing countries. In many countries attempts to achieve this goal are hampered by difficulties in access to health services that provide correct standard case management, or ignorance among families of the signs of severe disease in children or of the most I Chief Technical Advisor, Rakai Project, Entebbe, Uganda. 2 Chief, Measles Elimination Activity, National Immunization Pro- gramme, Centers for Disease Control and Prevention (CDC), 1600 Clifton Rd, Mailstop E61, Atlanta, GA 30333, USA. Correspond- ence should be sent to Dr S.C. Redd at this address. 3 Medical Epidemiologist, Pediatric and Family Studies Section, Division of HIV/AIDS, CDC, Atlanta, GA, USA. 4 Chief Paediatrician, Kisumn District Hospital, Kisumu, Kenya; formerly at Siaya District Hospital, Siaya, Kenya. 5 Director, CDC Kenya Field Station, Kisian, Kenya. Reprint No. 5745 appropriate place to go for treatment (3). At the international level, therefore, several organizations have sought to improve the quality of paediatric health care in developing countries by encouraging the use of standardized approaches to diagnosis and treatment. For some diseases, these recommen- dations have been simplified and promoted as algorithms.a,b These algorithms, which were designed to be used by clinical workers at first-level health facilities, encourage a standard approach to clinical diagnosis based on the duration and severity of the patient's symptoms and on the detection of simple signs by physical examination. a World Health Organization. Technical bases for the WHO rec- ommendations on the management of pneumonia at first-level health facilities. Unpublished document WHO/ARI/91.20, 1991 (available upon request from Division of Child Health and Develop- ment, World Health Organization, 1211 Geneva 27, Switzerland). bWorld Health Organization. A manual for the treatment of diar- rhoea. Unpublished document WHO/CDD/SER/80.2 Rev. 2, 1990 (available upon request from Division of Child Health and Develop- ment, World Health Organization, 1211 Geneva 27, Switzerland). Bulletin of the World Health Organization, 1996, 74 (6): 613-618 613 L.A. Paxton et al. Algorithms for management of single disease entities such as diarrhoea or acute respiratory infec- tions have been successfully instituted in many coun- tries (4).c.d However, children often suffer from multiple disease processes simultaneously, and the use of disease-specific algorithms may inadvertently make health care workers focus attention on only one disease (5, 6). To overcome this problem WHO has recently developed a composite algorithm, known as the Sick Child Charts, to integrate pre- vious algorithms aimed at the main causes of pre- ventable paediatric morbidity and mortality in the developing world (7). These causes include acute respiratory infection, malaria, measles, diarrhoeal disease, malnutrition, and otitis media, which are estimated to cause 70% of deaths among children in the developing world. This algorithm is organized so that severely ill children can be rapidly identified and managed. Chil- dren with any one of 13 clinical signs of severe illness are to be given initial therapy and referred for more intensive health care. These signs were taken from disease-specific algorithms and were combined by WHO with the aid of expert clinical opinion.e The signs are presumed to identify children at high risk of death or serious disability and include the inability to drink, abnormal mental status (defined as being ab- normally sleepy or difficult to rouse), convulsions (from the carer's history), evidence of malnutrition (wasting, oedema), respiratory distress (chest wall retraction, stridor), severe dehydration (abnormal skin turgor, repeated vomiting), meningitis (stiff neck), mastoiditis (tender swelling behind the ear), severe anaemia (pallor of the conjunctivae), and corneal ulceration or clouding. The usefulness of this system to identify chil- dren with severe illness has not been assessed. As part of an evaluation of a preliminary version of the Sick Child Assess and Classify Chart, we examined the clinical utility of these 13 signs. We recorded the prevalence of the signs among children seen in out- c World Health Organization. Programme for the Control of Acute Respiratory Infections: fifth programme report 1990-1991. Unpublished document WHO/ARI/94.33, 1994 (available upon re- quest from Division of Child Health and Development, World Health Organization, 1211 Geneva 27, Switzerland). dWorld Health Organization. Programme for the Control of Diarrhoeal Diseases: ninth programme report 1992-1993. Unpub- lished document WHO/CDD/94.46, 1994 (available upon request from Division of Child Health and Development, World Health Organization, 1211 Geneva 27, Switzerland). eWorld Health Organization. An integrated approach to man- agement ofchildhood illness: Development and research priorities. Unpublished document WHO/CDR/94.7, 1994 (available upon re- quest from Division of Child Health and Development, World Health Organization, 1211 Geneva 27, Switzerland). patient clinics, assessed their current role in deter- mining hospital admission, and determined the risk of in-hospital mortality associated with each sign. Methods Study site. The outpatient evaluation was conducted from August to December 1993 in four clinics in Siaya District, western Kenya, and in the paediatric outpatient department of Siaya District Hospital, which is the referral hospital for the district. The inpatient study was conducted from mid-June to mid-November 1993 in the inpatient ward of Siaya District Hospital. This inpatient service admits chil- dren from both its own outpatient department and other facilities in the district. Study population and survey methods. The investi- gators selected prospective health workers with a minimum of secondary school education to be trained to identify the 13 signs of severity. The train- ing, over a period of several weeks, involved class- room instruction with draft training materials from WHO and examination of hospitalized children un- der the tutelage of a CDC (U.S. Centers for Disease Control and Prevention)-employed paediatrician. Trained health workers examined every child aged 2-60 months seen at these sites during the study periods, and administered a questionnaire to the child's guardian. Information collected included the child's age and sex, symptoms during the cur- rent illness (including presence and duration of cough, fever, ear pain, diarrhoea, and rash), and the presence on physical examination of each of the 13 signs. Children admitted to the inpatient ward underwent a more extensive examination that in- cluded determination of haemoglobin concentration and inspection of Giemsa-stained thick blood smears to diagnose malaria. Each child was independently examined by the staff physician or a paramedical clinical officer who had not been trained in the recognition of the signs of severity. The staff physician or a medical officer provided clinical care and made all the diagnoses and treatment decisions. The outcome for each out- patient (discharged, admitted, transferred, or ab- sconded) and each hospitalized patient (discharged alive, died, transferred, absconded) was recorded. Each sign of severe illness was evaluated as a predictor of hospitalization (outpatient portion of the study) and as a predictor of in-hospital death (inpatient portion), using both univariate and multivariate logistic regression analysis. Epi Info version 5 (8) and SAS® software (9) were used to analyse the data. 614 WHO Bulletin OMS. Vol 74 1996 Clinical indicators of severe paediatric illness The protocol for this evaluation was approved by the Human Subjects Review Board of CDC and by the Kenya Medical Research Institute. Results Outpatient evaluation Questionnaires were completed for 3013 children (49.4% female; median age, 13 months). The out- come for 214 children (7% of the total) was not known because the child absconded either before examination by the staff physician or medical officer or before the findings were recorded. The records of these children were excluded from further analysis after it had been verified that they did not differ from the remaining population in age, sex, or frequency of the 13 signs of severe illness. The most frequently cited reasons for seeking care were fever (87%), cough (73%), and diarrhoea (38%). Repeated vomiting was the most frequently reported sign of severity, which was present in 13% of children evaluated (Table 1). Overall, 28% of chil- dren in the outpatient department had at least one sign of severe illness. Of the 2799 children for whom complete information was available, 779 (28%) were referred for hospitalization. Children with any of the 13 signs had a 3.2 times higher odds of hospitaliza- tion than children without such signs (Table 1). However, among the 779 admitted, 424 (54%) had none of the 13 signs of severity. Of the 2020 children who were sent home from the outpatient depart- ment, 415 (21%) had at least one of the 13 signs of severe illness. In univariate analyses, the signs among outpa- tients most highly associated with hospitalization included pallor (odds ratio (OR) = 8.6), abnormal mental status (OR = 7.9), chest wall retraction (OR = 5.3), oedema (OR = 3.8), and wasting (OR = 3.5)) (Table 1). In a multivariate logistic regression model, pallor remained the danger sign most highly associated with hospital admission (OR = 5.7), followed by chest wall retraction (OR = 4.4), oedema (OR = 3.1), convulsions (OR = 2.1), abnormal skin turgor (OR = 1.7), and repeated vomiting (OR = 1.6). Inability to drink, abnormal mental status, and wasting were not associated with hospitalization in the multivariate model. Inpatient evaluation Forms were completed for 1262 children (48.4% fe- male; median age, 11 months) who were admitted to the inpatient ward of Siaya District Hospital during the survey. Almost 10% of these children (123) ab- sconded from the hospital after physical examination on admission. The prevalence of signs of severe ill- ness was similar among the children who absconded and those who did not, with the following exceptions: absconders were more likely to have malnutrition (OR = 3.5, P < 0.005) or severe dehydration (OR = 1.9, P < 0.05), and were less likely to have had con- vulsions (OR = 0.36, P < 0.05) than non-absconders. Table 1: Likelihood of referral for hospitalization from the outpatient clinics in Siaya District, by 13 physical signs of severe illness No. among No. among outpatients hospitalized Odds ratio Sign (n = 2799) children (n = 779) (95% Cl) Pallor 112 (4)8 84 (11) 8.6; 5.5-13.6b Abnormal mental status 12 (0.4) 9 (1) 7.9; 2.0-36.6 Chest wall retraction 166 (6) 107 (14) 5.3; 3.8-7.5 Oedema 44 (2) 26 (3) 3.8; 2.0-7.4 Wasting 55 (2) 31 (4) 3.5; 2.0-6.1 Inability to drink 55 (2) 30 (4) 3.2; 1.8-5.7 Abnormal skin turgor 163 (6) 82 (11) 2.8; 2.0-3.9 Convulsions 128 (5) 61 (8) 2.5; 1.7-3.6 Repeated vomiting 358 (13) 140 (18) 1.8; 1.4-2.3 Stiff neck 0 0 - Stridor 0 0 - Corneal ulceration 0 0 - Tender swelling 0 0 - behind the ear Any of the 13 signs 770 (28) 355 (46) 3.2; 2.7-3.9 a Figures in parentheses are percentages. b Figures in italics are the 95% confidence intervals. WHO Bulletin OMS. Vol 74 1996 615 L.A. Paxton et al. Table 2: Risk of death during hospitalization among inpatients in Siaya District Hospital, by 13 physical signs of severe illness No. among hospital No. among children inpatients who died Odds ratio Sign (n = 1 139) (n = 75) (95% Cl) Abnormal mental status 9 (1)a 6 (8) 30.8; 6.7-159b Inability to drink 25 (2) 15 (20) 23.9; 9.9-58.7 Repeated vomiting 22 (2) 9 (12) 11.0; 4.2-12.1 Wasting 23 (2) 6 (8) 5.4; 1.8-15.0 Abnormal skin turgor 119 (10) 25 (33) 5.2; 3.0-9.0 Pallor 297 (26) 42 (56) 4.0; 2.4-6.7 Chest wall retraction 252 (22) 34 (45) 2.9; 1.9-28.8 Oedema 65 (6) 9 (12) 2.5; 1.1-5.4 Convulsions 136 (12) 7 (9) 0.8; 0.3-1.7 Stiff neck 2 (0.02) 1 (1) -c Stridor 1 (0.001) 0 _C Corneal ulceration 6 (0.05) 1 (1) -c Tender swelling 2 (0.02) 1 (1) -c behind the ear Any of the 13 signs 666 (58) 67 (89) 6.5; 3.0-14.8 a Figures in parentheses are percentages. b Figures in italics are the 95% confidence intervals. c Odds ratio invalid. After excluding the 123 absconders, the records of 1139 children were analysed. Among hospitalized children, fever was the commonest complaint (94%), followed by cough (81%) and diarrhoea (43%). Pallor was the most frequently identified sign of severe illness, followed by chest wall retraction and convulsions (Table 2). Of the 1139 hospitalized children, 666 (58%) pre- sented with at least one sign and 75 died (7%), 67 (89%) of whom had at least one of the 13 signs at the time of admission. Common symptoms and signs identified in children who died included pallor (56%), chest wall retraction (45%) and abnormal skin turgor (33%). In the univariate analysis, abnor- mal mental status (OR = 30.8) and inability to drink (OR = 23.9) carried the highest risk of death, al- though only 27 children (2.4%) were admitted with these signs. There was no increased risk associated with a history of convulsions. In the final multiple logistic regression model, with adjustments for con- founding among the signs of severity, abnormal men- tal status remained the sign most highly associated with death (OR = 59.6), followed by poor skin turgor (OR = 5.6), pallor (OR = 4.3), repeated vom- iting (OR = 3.6), chest wall retraction (OR = 2.7), and oedema (OR = 2.4). Too few children presented with stridor, stiff neck, mastoiditis, or corneal cloud- ing to draw any conclusions about their associated mortality risks. Overall, the risk of dying while hospi- talized was 6.5 times greater for children with at least one sign of severe illness, compared with children without any such sign. Eight children, 11% of those who died, had none of the signs of severity. We reviewed again their admission records for symptoms, signs, and laboratory examinations not incorporated in the Sick Child Charts and were unable to find any sign or combination of signs that reliably distinguished them from children who did not die. Their only common symptom was fever, which was present in over 90% of hospitalized children. Discussion During their clinical training, health workers learn about signs and symptoms that suggest serious, po- tentially life-threatening disease which, especially in children, may require immediate and intensive care. WHO's Sick Child Charts have incorporated these clinical signs into a functional algorithm for use at first-level health facilities in developing countries. Our study shows that the system used in these charts to identify severely ill children did identify those at high risk of death in hospital. We also observed that, within the context of current clinical practice in a rural district of Kenya, children presenting with one or more of these signs were more likely to be referred for inpatient care than children who did not have them. However, current practice did not result in the referral and hospitaliza- tion of all children with one or more signs of severe illness, nor did all the referred children have these signs. WHO Bulletin OMS. Vol 74 1996616 Clinical indicators of severe paediatric illness The risk of death during hospitalization was more than six times higher among children admitted with at least one sign of severe illness than those without such signs. Among non-absconding children, the four signs most highly associated with death were inability to drink, abnormal mental status, repeated vomiting, and wasting. A history of convulsions was not associated with increased risk of death in this study; if further investigation shows that conditions causing convulsions can be managed without hospi- talization, convulsions might be removed from the list of signs of severe illness. As currently formulated, the signs of severity in the Sick Child Charts appear to be comprehensive. They identified 89% of children who subsequently died during hospitalization, and we were unable to detect any other clinical indicators that would have reliably identified the few children who died without any of the 13 signs of severity. The Sick Child Charts algorithm therefore ap- pears to be consistent with and probably improves current practice. From this evaluation we know that these signs were more commonly observed in chil- dren who were selected by the staff physician or clinical officer for hospitalization compared with children not admitted, which suggests that these signs are already being used. However, over half (424/779, 54%) of all children admitted from the outpatient clinic had none of the signs of severity. and 415 children with these signs were discharged from the clinic against 355 who were admitted (Table 1). Children admitted without signs of severe illness, but with the staff physician's or clinical officer's as- sessment that hospitalization was required, must be more carefully evaluated; if such children are not to be admitted, we need to be assured that they are not at risk of death or serious morbidity. Similarly, be- fore recommending hospitalization for the large number of children who had signs of severe illness but were sent home from the outpatient department, we need to be sure that they required hospitalization to avert serious morbidity or death. Because of higher hospitalization costs and limited health care resources, the implications of in- creased rates of admission of children with signs of severity must be examined. A simple evaluation of our data from western Kenya suggests that if all children with a sign of severe illness and no children without such signs were referred for inpatient care, 28% of all children attending outpatient depart- ments would be admitted, the same proportion as under current practice. However, in view of the strik- ing difference in mortality rates between children admitted with and without signs of severity (67/666 versus 8/473, Table 2), we expect that the adoption of this system would more than double the number of children at high risk of death who are admitted, at the same time halving the number of children at high risk who are discharged from the outpatient department. Our evaluation of the system proposed in the Sick Child Charts to identify high-risk children in need of more intensive health care is encouraging. We were, however, unable to identify additional signs or symptoms from the inpatient component of our study that would have improved the recognition of additional children at risk. We are also encour- aged by the fact that the proposed signs are currently being used, at least partially, in clinical practice to refer children for further care. Additional validation of the signs to identify children at risk of severe morbidity, death, and disability in other settings is needed. The current study highlights the importance of continuing to include the clinical judgment of health workers as part of this assessment of severity. In the current Sick Child Assess and Classify Chart, health workers are encouraged to refer "any sick children whose illness you cannot manage." The training and supervision of health workers in identi- fying severely ill children should continue to be given high priority because of its potential to reduce sub- stantially childhood mortality. Acknowledgements This work was primarily funded by the United States Agency for International Development. A grant from the World Health Organization was given to CDC (Centers for Disease Control and Prevention) for a joint collaborative (CDC/Kenya Medical Research Institute (KEMRI)) evalua- tion of the Sick Child Charts. Resume Evaluation des indicateurs cliniques de maladies pediatriques graves Pour aider a reduire la morbidit6 et la mortalite chez l'enfant dans les pays en developpement, I'Orga- nisation mondiale de la Sant6 a etabli un algorithme de diagnostic et de traitement ax6 sur les principales causes de deces chez l'enfant infections respira- toires aigues, paludisme, rougeole, maladies diar- rh6iques et malnutrition. A l'aide de cet algorithme, (Sick Child Charts), les enfants gravement malades sont rapidement identifies par la pr6sence de l'un quelconque de 13 signes indicateurs, et sont orientes sur un 6tablissement sp6cialis6 pour y recevoir des soins plus intensifs. Les signes indicateurs sont: incapacit6 de boire, 6tat mental WHO Bulletin OMS. Vol 74 1996 617 L.A. Paxton et al. anormal (somnolence anormale ou inconscience), convulsions, emaciation, oedeme, tirage respira- toire, stridor, s6cheresse anormale de la peau, vomissements a r6petition, raideur de la nuque, tum6faction douloureuse derriere l'oreille, paleur de la conjonctive, et ulc6ration corneenne. L'utilit6 de ces signes a 6t6 evalu6e dans la pratique clinique actuelle et dans le contexte optimise d'utilisation de l'algorithme dans un district rural de l'ouest du Kenya. Nous avons observe que 27% des enfants vus dans les services de consulta- tions externes presentaient un ou plusieurs de ces signes et que la paleur de la conjonctive et le tirage respiratoire etaient les signes les plus souvent associ6s a une hospitalisation (odds ratio (OR) = 8,6 et 5,3 respectivement). La pr6sence de l'un quelconque de ces signes augmentait de 3,2 fois la probabilit6 d'hospitalisation, bien que 54% des enfants hospitalis6s n'aient pr6sent6 aucun de ces signes et que 21% des enfants renvoyes chez eux apres la consultation aient present6 au moins un signe. Parmi les cas hospitalis6s, 58% de 1'ensemble des enfants et 89% des enfants d6ced6s presentaient un signe a l'hospitalisation. L'etat mental anormal 6tait le signe pr6sentant la plus forte association avec le d6ces (OR = 59,6); venaient ensuite la secheresse anormale de la peau (OR = 5,6), la paleur de la conjonctive (OR = 4,3), les vomissements a r6p6tition (OR = 3,6), le tirage respiratoire (OR = 2,7) et l'oedeme (OR = 2,4). Globalement, le risque de mortalite associ6 a la presence d'au moins un signe etait 6,5 fois plus 6lev6 que chez les enfants ne presentant aucun signe. Bien que ces signes soient utiles pour iden- tifier un sous-groupe d'enfants a haut risque de mortalit, leur validation dans d'autres contextes est n6cessaire. La formation et la supervision des agents de sante en vue de l'identification des en- fants gravement malades doivent rester prioritaires en raison de leurs avantages, notamment en ce qui concerne la r6duction de la mortalit6 juveno- infantile. References 1. Grant JP. State of the world's children 1994. New York, Oxford University Press, 1994: 82. 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. World Bank. World development report 1993. Invest- ing in health. New York, Oxford University Press, 1993: 1-16. 4. El-Rafie M et al. Effect of diarrheal disease control on infant and child mortality in Egypt. Lancet, 1990, 335: 334-338. 5. Redd SC et al. Usefulness of clinical case definitions in guiding treatment decisions for African children with suspected malaria or pneumonia. Lancet, 1992, 340: 1140-1143. 6. O'Dempsey TJ et al. Overlap in the clinical features of pnenumonia and malaria in African children. Transac- tions of the Royal Society of Tropical Medicine and Hygiene, 1993, 87: 662-665. 7. Gove S et al. Usefulness of clinical case-definitions in treatment of childhood malaria or pneumonia. Lancet, 1993, 341: 304-305. 8. Dean AG et al. Epi Info, version 5: a word processing, database, and statistics program for epidemiology on microcomputers. Atlanta, GA, Centers for Disease Control, 1990. 9. SAS Institute Inc. SAS® Companion for the Micro- soft Windows environment, version 6. Cary, NC, 1993. 618 WHO Bulletin OMS. Vol 74 1996