Using clinical signs to diagnose anaemia in African children.

Using clinical signs to diagnose anaemia in African children S.P. Luby,1 P.N. Kazembe,2 S.C. Redd,' C. Ziba,3 O.C. Nwanyanwu,3 A.W. Hightower,1 C. Franco,3 L. Chitsulo,3 J.J. Wirima,4 & M.A. Olivar3'5 Anaemia is a serious and common problem among young children in sub-Saharan Africa. As a first step towards developing guidelines for its recognition and treatment, we conducted a study to evaluate the ability of health workers to use clinical findings to identify children with anaemia. Health care workers examined a total of 1104 children under 5 years of age at two hospital-based outpatient clinics in rural Malawi. Blood samples were taken to determine haemoglobin concentrations. Pallor of the conjunctiva, tongue, palm or nail beds was 66% sensitive and 68% specific in distinguishing children with moderate anaemia (haemoglobin concentration, 5-8gidl) and 93% sensitive and 57% specific in distinguishing those with severe anaemia (haemoglobin concentration, <5 gldl). Even without laboratory support, which is often unavailable in rural Africa, clinical findings can identify the majority of children with anaemia. Introduction Paediatric anaemia is common throughout sub-Saha- ran Africa (1). Anaemic children have reduced exer- cise capacity (2, 3), slower growth (4, 5), impaired neurological (6, 7) and cognitive (8) development, delayed wound healing (9), and increased risk of dying (10). The main causes of anaemia in African children - dietary iron deficiency, malaria, helminth infection and malnutrition - are treatable. The treatment of symptomatic anaemic children with a haemoglobin concentration of <5 g/dl by blood transfusion prevents death (10). The treatment of children with severe anaemia before they become symptomatic and of those with moderate anaemia before it becomes severe would probably prevent death, the risks associated with blood transfusion (e.g., human immunodeficiency virus (HIV)) and hepa- titis B virus infection, and the sequelae of anaemia. I Epidemiology Branch, Division of Parasitic Diseases, National Center for Infectious Diseases, Centers for Disease Control and Prevention, Atlanta, GA, USA. Requests for reprints should be sent to Dr S.P. Luby, Community Health Sciences, Aga Khan University, PO Box 3500, Stadium Road, Karachi 74800, Paki- stan. 2 Kamuzu Central Hospital, Lilongwe, Malawi. 3Community Health Sciences Unit, Ministry of Health, Lilongwe, Malawi. 4The Medical College, University of Malawi, Blantyre, Malawi. 5 Deceased. Reprint No. 5622 Diagnosing anaemia in African children is diffi- cult because few rural areas can support laboratory facilities. Physical examination has been evaluated as a means of diagnosis; however, most of the stud- ies have explicitly excluded Blacks (11-13), severely restricted (14) or excluded young children (11-13, 15, 16), or evaluated patients with a less severe spec- trum of anaemia (11-16) than is typical among Afri- can children. Thus, the relevance of these studies to the diagnosis of childhood anaemia in Africa is unclear. We have conducted a study to determine wheth- er African children with severe and moderate anaemia could be identified using clinical signs and report our findings in this article. Methods The study was conducted in the outpatient depart- ments of two hospitals in Malawi, Mangochi District Hospital and Nkhoma Hospital, both of which serve predominately rural areas. A systematic sample was recruited by approaching the parent or guardian of every fourth sick child brought to the clinic for under-5-year-olds in Mangochi between 17 April and 28 May 1993, and every second and third sick child brought to the paediatric clinic in Nkhoma between 28 April and 5 June 1993. Children aged over 5 years were excluded. For patients whose parent or guardian consent- ed, a fingerprick blood sample was collected and the haemoglobin concentration was measured by spectro- Bulletin of the World Health Organization, 1995, 73 (4): 477-482 477 S.P. Luby et al. photometry of a modified sodium azide/methaemo- globin reaction.a The accuracy of this technique has been demonstrated in a variety of settings (17-19). Each child was given a standard physical exam- ination, including measurement of rectal tempera- ture, by one of three non-physician health workers, previously trained in physical diagnosis. To evaluate conjunctival pallor, the examiner gently everted the lower eyelid and directly inspected the ocular and palpebral conjunctiva. To evaluate tongue pallor, the examiner directed a pen-light and observed the superior surface of the tongue. To evaluate palmar pallor, the examiner opened one of the child's hands by partially extending the child's fingers and inspect- ed the palm. To evaluate nail-bed pallor, the posteri- or surface of the child's hand was gently rotated towards the examiner and the nail beds observed directly without applying any pressure to the nails. The examiners were trained prior to the study for at least one day during which they evaluated chil- dren for pallor and were then told the children's respective measured haemoglobin concentrations. To assess severity of pallor, the examiners were asked to use their own judgement and grade pallor as definite, probable, or absent. During the study, the examiners were blinded to the measured haemoglobin level, though they were allowed to review the haemoglobin results of the children after their clinical evaluation was complete and recorded. This was permitted to allow the exam- iner to use the laboratory information in deciding how to treat the child. Each child was examined by only one examiner. Statistical analysis of the data was carried out using the following: X2 test, for comparing the observed differences in proportions between study groups; Student's t-test, for comparing the means of two categories; one-way analysis of variance (Tukey-Kramer method), for pairwise comparisons of three or more means; multiple linear regression, to evaluate the relationship of measured haemoglobin concentration to various factors; and the X2 test for trend, to evaluate changes in examiner sensitivity and specificity during the course of the study. Statis- tical significance was defined as P <0.05. The calcu- lations were performed using SAS and EpiInfo soft- ware.b Sensitivity was calculated as true posi- tives/(true positives + false negatives), specificity a HemoCueg, Mission Viejo, CA. Inclusion of trade names is for identification only and does not imply endorsement by the U.S. Department of Health and Human Services or the U.S. Public Health Service. b Base SASG, release 6.03, 1988. Cary, NC, USA, SAS Institute Inc; and Epilnfo, version 5.01b, 1990. Atlanta, GA, USA, Centers for Disease Control and Prevention. as true negatives/(true negatives + false positives), positive predictive value as true positives/(true posi- tives + false positives), and negative predictive value as true negatives/(true negatives + false negatives). For children aged 6 months to 6 years anaemia was defined according to the WHO definition, i.e., a haemoglobin concentration of <11 g/dl (20). Severe anaemia was defined as a haemoglobin concentration of <5 g/dl since below this level hospitalized chil- dren are often transfused (10, 21). Moderate anaemia was defined as a haemoglobin concentration of 5-8 g/dl, a level that might be considered a reason- able point at which to intervene to prevent children from developing severe anaemia. The study protocol was approved by the Malawi National Health Sciences Research Committee. Results Of 1225 eligible children, the parent's or guardian's consent to participate was obtained for 1139 (93%). Of these, 1104 (97%) underwent a physical examination of the conjunctiva, tongue, palm and nail bed and a blood test to determine haemoglobin concentration. Data were incomplete for the remaining 35 partici- pants; these children were somewhat older (mean age: 23.5 months versus 17.5 months), but did not differ significantly from those with complete data in terms of their sex, chief complaint, prevalence of pallor or mean haemoglobin concentration. The analy- sis was restricted to the 1104 patients for whom complete data were available. The median age of the enrolled children was 13 months (range, <1 month to 60 months); 580 (53%) were boys, and 590 (53%) were seen at Mangochi District Hospital. The common chief complaints were fever (60%), cough (12%), and diarrhoea (5%). The mean haemoglobin concentration of enrolled patients was 8.8 g/dl (range, 2.1-17.1 g/dl; standard deviation (SD) = 2.4 g/dl): 82% were anaemic according to the WHO definition; 35% had moderate anaemia; and 5% had severe anaemia. For each anatomical site patients with definite pallor had significantly lower haemoglobin levels than those judged to have probable pallor, while the latter had significantly lower levels than patients without pallor (Table 1). Probable pallor at each anatomical site and for each definition of anaemia was more sensitive but less specific than definite pallor in diagnosing anae- mia (Table 2). The lower the haemoglobin threshold used to define anaemia, the more sensitive both defi- nite and probable pallor were in identifying anaemic children (Table 2). Because of the small number of children with severe anaemia, excluding them from 478 WHO Bulletin OMS. Vol 73 1995 Diagnosis of anaemia in African children Table 1: Mean haemoglobin levels, according to physi- cal signs of pallor, by anatomical site, among 1104 paediatric outpatients, Mangochi and Nkhoma, Malawi, 1993a Mean haemoglobin level ± SD (g/dl): Pallor Conjunctiva Tongue Palm Nail bed Definite 4.9 ± 1.5 4.3 ± 1.4 4.8 ± 1.6 5.0 ± 1.7 Probable 7.8 ± 2.0 7.4 ± 2.2 7.9 ± 2.1 8.0 ± 2.1 None 9.2 ± 2.2 9.2 ± 2.2 9.6 ± 2.1 9.5 ± 2.1 a For each anatomical site, the mean haemoglobin concentra- tion was significantly different, by level of pallor (P <0.05); one- way analysis of variance with Tukey-Kramer adjustment for mul- tiple comparisons. the analysis had only a slight effect on the sensitivity of probable pallor at any site to identify children with moderate anaemia. For probable pallor at any site there was an inverse relationship between sensitivity and specific- ity according to the definition of anaemia that was used (Table 2). In general, nail-bed and palmar pal- lor were more sensitive and less specific than pallor of the conjunctiva and tongue. Definite pallor at any of the four sites had a specificity >97% for each of the anaemia definitions. We developed a composite definition for pallor by combining the physical examination findings. Children were classified as having definite pallor if they had such pallor in at least one anatomical site; probable pallor if they had no signs of definite pallor and had probable pallor in at least one site; and no pallor if they had neither definite nor probable pallor at any site. Composite definitions for definite and probable pallor were slightly more sensitive and slightly less specific than definite or probable pallor in individual sites (Table 3). Probable pallor at any anatomical site was 70% sensitive, 68% specific, and had a 54% positive predictive value in diagnosing a haemoglobin concentration of <8 g/dl. To evaluate test performance throughout the population, we constructed multiple linear regression models predicting haemoglobin levels. The sensitiv- ity and specificity of pallor did not vary with age or according to rectal temperature. Thus, the diagnostic performance of pallor was independent of age and temperature. Since each examiner evaluated different chil- dren, the distribution of haemoglobin concentrations varied according to the examiner. Despite this varia- tion, children with definite pallor had significantly lower haemoglobin concentrations than children with probable pallor, and those with probable pallor had significantly lower concentrations than those without pallor (P <0.05 for each comparison, Table 4). E co~ C E2 o-.Q ) C o 0 .5ci) a~~~~~~~~~~~~~~~~~~~~~i 0 .0 a CO-C'- N-(D c) oc I o:0 ) ~~~~~~~~~~~0 ~ ~ ~ ~ ~ ~ - 0 ocoNrco 0 0. .2 ~~~~~0 0 ~~~~~~~E0 0.~~~~0 (1) C .5 a) 0 C co EC .) E a. coo LO' c0t 0 >00 CDo 0 0 a >2 0 )000L a.0/ CD 0~~~~~~~~ o 0) CD. a C CO C0 a% N. 00 a 0cn c~~a -0)~~~~~~~~~~~~a m ~ ~ ~~~0 . Cco NC)C') 0lM c C0 co! o 0o00 0~ ECL ~~~~00-0~~~~~ *S~~~~~~~n coLO a a - a 0 a. 'W~~~ 00.0U~~~~~~~~~~~~ Z~ 0 >CL a 0 a 'o~ 00 0 ~ 0~G -C .CC o~OC C)C)O0 0 9 .oO~~~~~~~MCI -L co WI C-C~0 C 0 aL OW CL~0 0C -o0c ~~ aC0)~~o m co 0 aE C~~~~3~~ it) (Ho<0 v vi/)v~~~~~~~~~U WHO Bulletin OMS. Vol 73 1995 cn0) U') v 0) C 3 0 0) 0 E 0) a .C .3 0 ,It 479 S.P. Luby et al. Table 3: Sensitivity, specificity, positive predictive value, and negative predictive value of definite or probable pallor at any site In predicting anaemia among 1104 paedlatric out- patients, Mangochl and Nkhoma, Malawi, 1993 Anaemia definition Sensitivity (%) Specificity (%) PpVa NPVb (haemoglobin conc.) Def.c Prob.d Def. Prob. Def. Prob. Def. Prob. <5 g/dl 59 93 97 57 49 10 98 99 <8 g/dl 17 70 100 68 98 54 69 81 5-8 g/dl 10 66 100 68 97 48 71 82 <11 g/dl 7 51 100 83 100 93 19 27 8 PPV = positive predictive value. b NPV = negative predictive value. c, d See footnotes b and c, Table 2. We evaluated each examiner's sensitivity and specificity in diagnosing moderate anaemia for each week of the study using the X2 test for trend. Neither sensitivity nor specificity significantly improved for any of the health workers during the course of the study. Discussion Among the study children, 82% had haemoglobin concentrations <11 g/dl and were therefore anaemic according to the WHO definition; in principle, these children would benefit from diagnosis and treatment of their condition. It is currently impractical to con- sider treating all cases of anaemia among African children, so that focusing upon those most in need of therapy would appear to be a reasonable strategy. In the study, health workers with a minimum of additional training successfully identified a subgroup of African children at high risk for complications of anaemia. Screening for probable pallor of the con- junctiva, tongue, palm or nail bed proved to be a good test for anaemia among these children. This approach identified 66% of children with moderate anaemia (haemoglobin concentration, 5-8 g/dl) and, as has been reported previously (13, 15), was more sensitive (93%) in distinguishing patients with severe anaemia. In contrast to prior studies of White adults, our study suggests that, in settings where paediatric anaemia is highly prevalent, signs detected on physi- cal examination are useful in identifying those chil- dren who are most anaemic. Our results differ from previous studies primarily because we used a more extreme definition of anaemia (haemoglobin concen- tration <8 g/dl, rather than <1 1 g/dl); as a result, pal- lor proved a more sensitive sign for identifying anae- mia than had previously been reported (14, 15). Furthermore, the high prevalence of anaemia in the study population resulted in a higher positive predic- tive value for pallor than that reported by previous workers (14, 15). Of note is that the prevalence of anaemia in the study population is comparable to that in other locations in Africa (1), and hence the test performance should be similar in those settings. In our study, evaluating all four anatomical sites conjunctiva, tongue, palm and nail beds- was the most sensitive technique for recognizing anaemia. Previous investigators have evaluated palmar-crease pallor (11-13), but we rejected this approach because of the normally increased pigmentation in the palmar creases of Blacks. Although some researchers have Table 4: Mean haemoglobin concentration, by physical signs of pallor and examiner, among 1104 paediatric outpatients, Mangochi and Nkhoma, Malawi, 1993 Mean haemoglobin conc. ± SD (g/dl):a Examiner 1 Examiner 2 Examiner 3 All children seen by this 8.7 ± 2.6 9.2 ± 2.1 8.8 ± 2.0 examiner Children with at least one 4.7 ± 1.6 6.3 ± 1.3 5.1 ± 1.3 sign of definite pallor Children with at least one sign of 7.5 ± 2.1 8.9 ± 2.0 8.1 ± 1.6 probable but no sign of definite pallor Children with no signs of 9.6 ± 2.3 9.9 ± 1.9 9.7 ± 1.6 pallor a For each examiner, mean haemoglobin concentration was significantly different, by level of pallor (P<0.05); one-way analysis of variance using Tukey-Kramer adjustment for multiple comparisons. WHO Bulletin OMS. Vol 73 1995480 Diagnosis of anaemia in African children concluded that nail-bed evaluation is not useful (11-13), in our study nail-bed pallor performed almost as well as combined evaluation of all four sites. However, the examiners assessed nail beds for pallor at the same time as the other anatomical sites, and it is therefore unclear whether assessment of nail-bed pallor alone would be as diagnostically use- ful. If further study demonstrates that evaluation of nail-bed pallor independently is nearly as sensitive and specific as the composite findings from all the anatomical sites, this simpler approach could be taught to rural health workers. Many investigators have documented substantial inter-observer variation in the use of pallor as a diag- nostic test for anaemia (11, 13, 15, 22). The present study was not designed to measure such variation. In the clinical setting, validity is a more important cri- terion, and we found that children who had pallor, as determined by any of the three examiners, had lower haemoglobin concentrations than children without pallor. Although anaemia causes substantial morbidity and mortality in Africa, guidelines for managing the condition are rudimentary. Our study demonstrates that physical examination is sufficient to diagnose moderate and severe anaemia. In this setting, prob- able pallor at any site was a sensitive and specific measure for identifying children with moderate anae- mia, and definite pallor had high specificity in dis- tinguishing those with severe anaemia. Resume Utilisation de signes cliniques pour diagnostiquer I'an6mie chez les enfants africains L'anemie pediatrique est courante dans toute l'Afrique subsaharienne. 11 serait probablement possible de prevenir la mortalit6 due a l'anemie et les sequelles qu'elle entraine chez les survivants en traitant les enfants atteints d'anemie grave avant qu'elle ne devienne symptomatique et en traitant ceux qui presentent une anemie moderde avant qu'elle ne devienne grave. Le diagnostic de l'anemie chez les enfants africains est difficile car peu de regions rurales peuvent disposer d'instalia- tions de laboratoire. L'examen physique a d6ja e evalue comme moyen de diagnostic de l'ane- mie; toutefois, la plupart des etudes excluaient les Noirs, limitaient tres strictement ou excluaient les jeunes enfants, et portaient sur des patients pre- sentant des tableaux d'anemie moins graves que ceux que l'on rencontre couramment chez les jeunes enfants africains. Nous avons realise une 6tude visant a d6terminer si des enfants africains atteints d'anemie moderee a grave peuvent etre identifies au moyen de criteres cliniques. Trois agents de sante non m6decins d6ja for- mes au diagnostic clinique ont e recrutes et ont req;u une formation d'au moins une journ6e sur la reconnaissance de la paleur, avec acces a poste- riori aux taux d'hemoglobine mesures chez les enfants examin6s. L'6tude a porte sur 1104 enfants de moins de 5 ans recrutes dans deux services de consultations externes d'hopitaux situes en zone rurale au Malawi. Les agents de sante ainsi formes ont recherch6 les signes de paleur en examinant la conjonctive, la langue, les paumes et les lits ungueaux des patients et ont classe les resultats en paleur nette, probable ou absente. Les taux d'h6moglobine ont ete mesures par spectrophotom6trie sur prelevement de sang capillaire au bout du doigt. Le taux d'hemoglobine moyen des sujets recrut6s dans l'etude 6tait de 8,8 g/dl. Pour cha- cun des quatre sites anatomiques examin6s, les patients ayant une paleur nette avaient des taux d'hemoglobine significativement plus bas que ceux chez qui la paleur avait ete classee comme probable, lesquels avaient a leur tour des taux plus bas que ceux chez qui la paleur 6tait not6e comme absente. Par exemple, pour la paleur des lits ungu6aux, le taux d'hemoglobine moyen etait de 5,0 g/dl chez les enfants ayant une paleur nette, de 8,0 g/dl chez ceux ayant une paleur pro- bable et de 9,5 g/dl lorsque la paleur etait absen- te. En g6neral, la paleur des lits ungueaux et des paumes etait un signe plus sensible et moins sp6- cifique que la pAleur de la conjonctive ou de la langue. La pAleur nette de l'un quelconque de ces quatre sites avait une specificit6 .97% pour cha- cune des d6finitions de I'an6mie. La pAleur pro- bable de la conjonctive, de la langue, des paumes ou des lits ungu6aux avait une sensibilit6 de 66% et une specificit6 de 68%, une valeur predictive positive de 48% et une valeur predictive negative de 82% pour l'identification des enfants atteints d'anemie mod6r6e (h6moglobine 5-8 g/dl), et une sensibilite de 93%, une sp6cificit6 de 57%, une valeur pr6dictive positive de 10% et une valeur predictive n6gative de 99% pour l'identification des enfants atteints d'an6mie grave (hemoglobine <5/dl). La sensibilit6 et la specificite de la paleur etaient independantes de I'Age et de la temp6ratu- re rectale de 1'enfant. 11 est actuellement impossible en pratique d'envisager de traiter tous les enfants africains anemiques, aussi semble-t-iI raisonnable d'axer les interventions sur ceux qui ont le plus grand WHO Bulletin OMS. Vol 73 1995 481 S.P. Luby et al. besoin d'etre trait6s. Contrairement aux conclu- sions des etudes anterieures sur les adultes blancs, nos resultats semblent indiquer que dans les contextes de forte pr6valence de I'anemie pediatrique, les signes observ6s a l'examen phy- sique sont utiles pour identifier les enfants les plus gravement anemiques. Dans les circons- tances de cette etude, une pAleur probable de l'un quelconque des sites examines 6tait un signe sensible et specifique pour l'identification des enfants atteints d'anemie mod6rde, tandis qu'une paleur nette etait tres specifique pour l'identifica- tion des enfants atteints d'an6mie grave. References 1. Demaeyer E, Adiels-Tegman M. The prevalence of anaemia in the world. World health statistics quar- terly, 1985, 38: 302-316. 2. Bhatia D, Seshadri S. Anemia, undernutrition and physical work capacity of young boys. Indian pediat- rics, 1987, 24: 133-139. 3. Grelsen G. Mild anaemia in African school children: effect on running performance and an intervention trial. Acta paediatrica Scandinavica, 1986, 75: 662-667. 4. Aukett MA et al. Treatment with iron increases weight gain and psychomotor development. Archives of disease in childhood, 1986, 61: 849-857. 5. Obladen M, Sachsenweger M, Stahnke M. Blood sampling in very low-birth-weight infants receiving different levels of intensive care. European journal of pediatrics, 1988, 147: 399-404. 6. Lozoff B et al. Iron-deficiency anemia and iron ther- apy; effects on infant developmental test perfor- mance. Pediatrics, 1987, 79: 981-995. 7. Cantwell RJ. The long-term neurological sequelae of anemia in infancy. Pediatric research, 1974, 8: 342. 8. Palti H, Pevsner B, Adler B. Does anemia in infan- cy affect achievement on developmental and intelli- gence tests? Human biology, 1983, 55: 189-194. 9. Viljanto J. Effect of haematocrit on wound healing in children. Zeitschrift fOr Kinderchirurgie, 1990, 45: 146-147. 10. Lackritz EM et al. Effect of blood transfusion on survival among children in a Kenyan hospital. Lancet, 1992, 340: 524-528. 11. Jacobs HD et al. Observer bias and error in the integumentary clinical diagnosis of chronic anaemia. South African medical journal, 1979, 55: 1031-1034. 12. Strobach RS et al. The value of the physical exam- ination in the diagnosis of anemia. Correlation of the physical findings and the hemoglobin concentration. Archives of internal medicine, 1988, 148: 831-832. 13. Nardone DA et al. Usefulness of physical examina- tion in detecting the presence or absence of anemia. Archives of internal medicine, 1990, 150: 201-204. 14. Wurapa FK, Bulsara MK, Boatin BA. Evaluation of conjunctival pallor in the diagnosis of anaemia. Journal of tropical medicine and hygiene, 1986, 89: 33-36. 15. Gj0rup T et al. A critical evaluation of the clinical diagnosis of anemia. American journal of epidemiol- ogy, 1986, 124: 657-665. 16. Glass R et al. The value of simple conjunctival examination in field screening for anemia. Nutrition reports international, 1980, 21: 405-412. 17. von Schenck H, Falkensson M, Lundberg B. Evaluation of "Hemocue", a new device for deter- mining hemoglobin. Clinical chemistry, 1986, 32: 526-529. 18. Cohen A, Seidl Friedman J. HemoCue® system for hemoglobin measurement, evaluation in anemic and nonanemic children. American journal of clinical pathology, 1988, 90: 302-305. 19. Zwart A et al. Determination of total hemoglobin in whole blood: further tests of the "Hemocue" method. Clinical chemistry, 1987, 33: 2307-2308. 20. Nutritional anaemias. Report of a WHO Scientific Group. Geneva, World Health Organization, 1968 (WHO Technical Report Series, No. 405). 21. Gumodoka B et al. Blood transfusion practices in Mwanza Region, Tanzania. AIDS, 1993, 7: 387-392. 22. Sanchez-Carrillo Cl. Bias due to conjunctiva hue and the clinical assessment of anemia. Journal of clinical epidemiology, 1989, 42: 751-754. 482 WHO Bulletin OMS. Vol 73 1995