180

THE MAGNITUDE OF MORTALITY FROIM ACUTE RESPIRATORYINFECTIONS IN CHILDREN UNDER 5 YEARS IN DEVELOPING COUNTRIES

Michel Garenne,8 Caroline Ronsmansb & Harry Campbell"

The most widespread and fatal of all acute diseases,pneumonia, is now Captain of the Men of Death.

Sir William Osier, 1901.

The mast recent WHO estimates (for 1990) lindicatethat ou1 of nearly 12.9 million children undelr 5 whodie each year in developing countries, about 4.3million die of ARI. Of these, it is estimated that 0.8million (18.6% of all ARI deaths) occur in the firstmonth of life. Other estimates have indicated thatabout two-thirds of ARI deaths occur in the first yearof life (4). The WHO estimates further state that theARI complicatio,ns of measles accounted for 0.48million deaths (11% of ARI deaths and 55~fo of allmeasles deaths) and that the ARI complications ofpertussiis accounted for 0.26 million deaths (6% ofARI deaths and 72% of all pertussis deathsl. Thus,ARI was estimated to be the single largest cause ofdeath in young children, being associated with 330;0of all chlildhood deaths in developing countries."

Introduction

In developed countries, during the last hundredyears, the evolution of mortality due to acute respir-atory infections (ARI) has been dramatic (1). At highlevels of mortality, such as XIXth century Europe,ARI was the category of diseases making the largestcontribution to shortening of life expectancy. Dis-eases due to ARI represented a loss of 7.5 years oflife, more than all other infectious diseases (4.8years) and diarrhoeal diseases (2.9 years). Amonginfants and children, ARI was the first cause of deathoutside the neonatal period. When life expectancywas below 45 years, 25% of all deaths in the agegroup 0-4 years were due to ARI, compared to only4% when life expectancy was higher than 70 years.

These estimates are based on various sources. Themain sources of information have been analysesbased on national cause-of-death statistics Inotifiedto WHO and extrapolations from these data to thosecountries which do not record cause-of-death databut have similar levels of child mortality. The aim ofthis article is to review and discuss the availabledata on mortality from ARI among children under 5in developing countries. For this purpose, 25 studieswith data on ARI deaths were reviewed. They werecompared with historical data from developed coun-tries before 1965.

Recognition of pneumonia and other ARI as animportant public health problem in developing coun-tries is recent. The magnitude of mortality from ARIin childhood in developing countries was docu-mented and published for the first time in the early19605 (2).

More recently, the World Health Organization (WHO)and other international agencies have made ARI oneof their priorities for intervention. Increased concernabout the important contribution of ARI deaths tooverall mortality was raised at the World HealthAssembly in 1976.d In 1983, a Technical AdvisoryGroup on ARI was established by WHO in Geneva(3). The global programme for the control of acuterespiratory infections was officially initiated in 1984as a distinct programme under Disease Preventionand Control in WHO's Seventh General Programmeof Work, covering the period 1984-1989. The centralobjective of the programme is to reduce mortalityfrom ARI, in particular pneumonia. This objective isendorsed in the Declaration of the World Summit forChildren, New York, 30 September 1991, which es-tablished the goal of reducing by one-third thedeaths due to ARI in children under 5 years of ageduring the period 1990-2000.

Data and method

To evalluate the relationship between proportion ofARI deaths and level of mortality in historical popu-lations, the study by Preston et al. (5) was used. Theauthors ana lysed the causes of death by age and sexin 180 data sets from 43 national populations before1965 (a complete review of the data will be pub-lished in a separate paper: Garenne et al. folrthcom-ing)! Causes of death from ARI were coded accord-ing to the International Classification of D'iseases(ICD), Sixth and Seventh Revisions.

To estimate the magnitude of mortality from ARI indevelopling countries, a search of the MEDLINE database from January 1980 to December 1991 wasperformed. The search focused on communitystudies of mortality from all causes and frorr, ARI inchildren under 5 years in developing countries. Re-sults from 2 unpublished studies were provided bythe authors (6, f,l. The data base revealed 21 com-munity-lbased lo!ngitudinal studies in 13 countries(6-26, f): only the studies with detailed calJses ofdeath for children aged < 5 years were kept for thefinal analysis. In 12 studies, the ascertainment of ARIdeaths was part of an overall assessment of cause-specific mortality (5 in Bangladesh, 1 in Kenya, 1 inMorocco, 1 in Nigeria, 2 in Senegal, 1 in The Gambiaand 1 in Guinea-BissauJ. In the 9 remaining studies,the longitudinal surveillance was aimed specificallyat ascertaining deaths due to ARI. The latter studieswere undertaken to assess the impact oln ARI-

.Associate Professor of Demography, Harvard University, Centerfor Population and Development Studies, Cambridge, MA, UnitedStates of America.b DrPH candidate, Harvard University, Center for Population and

Development Studies, Cambridge, MA, United States of America.c Consultant in Public Health (Child Health), Fife County Health

Board, Scotland, United Kingdom.d World Health Organization. Official records, 233: 63-109 (1976).

.World Health Organization. Implementation of the Global Strategyfor Health for All by the Year 2000, second evaluation: and eighthreport on the world health situation. Geneva, WHO, 1992. (Doc-ument A45/3).f Garenne, M. et al. ARI mortality in a rural area of Senegal. Draft

paper,1992.

Rapp. trimest statist sanit monc'., 45 (1992)

181

are coded und~r the acute respiratory infections(code 466). Influenza and pneumonia are groupedunder the same subtitle (code 480-487). In addition,the fourth digit of the ICD makes it possible tospecify pneumonia occurring after certain diseases,such as measles (code 484.0), whooping cough(code 484.3) and varicella (484.8). Earlier revisions ofthe ICD differ slightly from the Ninth Revision (28,31). For instance, in the Eighth Revision pneumonia(480-486) was classified separately from influenza(470-474) and pneumonia after measles, varicella orpertussis was not included in the diseases of the

respiratory system.

mortality of a community-based ARI interventionproject (2 in India, 2 in Nepal, 1 in Pakistan and 1 inthe United Republic of Tanzania); to assess theimpact of a pneumococcal vaccine (2 in Papua NewGuinea); to identify the etiological agents respon-sible for acute lower respiratory infections (ALRI)and to determine risk factors for ARI morbidity (1 inthe Philippines).

In addition, 2 studies based on national death regis-tration systems were included. Puffer & Serranoinvestigated causes of death in children from 13regions in 8 countries in Latin America during theperiod 1968-1971 (27). They used the Eighth Revisionof the ICD for classification of ARI deaths (28). VonSchirnding reviewed national data on mortality fromARI in South Africa for the period 1968-J985 (29).Data on ARI in children from "coloured race" wereincluded in our analysis. Data on black children wereexcluded because of underreporting of deaths in thisgroup. Causes of death were classified according tothe Ninth Revision of the ICD (30). Data from thesestudies were compared to the historical data.

In the community studies reviewed, the lack of con-sistency in the inclusion of the diseases causingdeath from ARI was striking. In 10 studies, deathsdue to ARI were not further differentiated into upperor lower respiratory infections. ALRI, where speci-fied, mostly referred to deaths from pneumonia. ARIdeaths following measles were addressed separatelyin 7 studies. Pertussis was considered as an ARIdeath in 2 studies (9, 11), while 4 studies listedpertussis as a separate cause (6, 12, 23, f). Deathsdue to laryngitis and influenza were listed in 2studies only (6, f). Varicella was mentioned in thesole Senegal study! One study classified ARI underthe heading "symptoms, signs and ill-defined dis-eases" (13).

Definition of ARI deaths

Classification of ARI deaths

The International Classification of Diseases andCauses of Death (ICD) classifies diseases accordingto the biological etiology of the causes of death or,where etiology is not apparent, the anatomical local-ization. The ARI classified under "diseases of therespiratory system" in the Ninth Revison are pre-sented in Table 1. Acute bronchitis and bronchiolitis

Role of pneumonia

Among the diseases listed in Table 1, it is generallyagreed that in developing countries, most ARI

TABLE 1. ARIIN THE INTERNATIONAL CLASSIFICATION OF DISEASES. NINTH REVISION (ICD-9)

TABLEAU 1. LES IRA DANS LA CLASSIFICATION INTERNATIONALE DES MALADIES. NEUVIEME REviSION(CIM-9)

Diseases of the respiratory system -Maladies de I'appareil respiratoire

ACUTE RESPIRATORY INFECTIONS (codes 460-466) -AFFECTIONS AIGUEs DES VOlES RESPIRATOIRES (codes 460-466):

(460) -common cold -rhume banal

(461) -acute sinusitis -sinusite aigue

(462) -acute pharyngitis -pharyngite aigue

(463) -acute tonsillitis -amygdalite aigue{464) -acute laryngitis and tracheitis -laryngite et tracheite aigues(465) -acute upper respiratory infections of multiple or unspecified sites -

infection aigue des voies respiratoires superieures, a localisations multiples ou non

precisees(466) -acute bronchitis and bronchiolitis -bronchite et bronchiolite aigues

PNEUMONIA AND INFLUENZA (codes 480-487) -PNEUMONIE ET GRIPPE (codes 480-487):

(480) -,- viral pneumonia -pneumonie a virus

(481) -pneumoccocal pneumonia -pneumonie a pneumocoque(482-483) -pneumonia due to other bacteria and other organisms -pneumonies dues a d'autres

bacteries ou d'autres organismes(484) -pneumonia in infectious diseases classified elsewhere -pneumonie au cours d'autres

maladies infectieuses classees ailleurs

484.0 .measles (055.1) -rougeole (055.1)

484.3 .whooping cough (033) -coqueluche (033)484.8 .varicella (052) -varicelle (052)

484.* .other infectious diseases -autres maladies infectieuses

(485) -bronchopneumonia, organism unspecified -br.onchopneumonie, micro-organismenon precise

(486) -pneumonia, organism unspecified -pneumonie, micro-organisme non precise

(487) -influenza -grippe

Notes. Codes 470-478 include other URI diseases and chronic conditions (deviated septum and polyps and chronic upper respiratory). Codes 490-496 include chronicobstructive pulmonary disease and allied conditions (chronic bronchitis, emphysema, asthma). Codes 010-018 cover tuberculosi~, includinQ respiratory tuberculo:sis. -Les codes 470 a 478 couvrent d'~utres maladies des voies respiratoires superieures {deviation de la cloison, polypes, affections chromques). Les codes 490 a496 couvrent des maladies pulmonaires obstructives chroniques et affections connexes (bronchite chronique,emphyseme, asthma). Les codes 010 a 018 couvrent la

tuberculose, y compris celle de I'appareil raspiratoire.

Wid hlth statist. quart., 45 (1992)

182

deaths among infants and children may be ascribedto pneumonia, bronchiolitis and acute obstructivelaryngitis (32). However, the similarity in clinicalsymptoms of pneumonia and bronchiolitis haveoften hampered the distinction of these syndromesin developing countries and t1:1e magnitude of mor-tality due to viral bronchiolitis is not well document-ed in populations. Indirect evidence from hospitalstudies suggests that pneumonia is the leadingcause of death from ARI in developing countries.Pneumonia is the primary cause of hospitalizationfor ARI, before bronchiolitis and laryngitis (33-37).Hospital-based data on case-fatality rates (CFR) byclinical syndrome vary widely. Rahman reportedsimilar CFRs for pneumonia and bronchiolitis (CFR =8%) among children under 5 years old in Bangladesh(33). In the United Republic of Tanzania, Mtangoobserved the highest CFRs in children withlaryngotracheitis (CFR = 28%) and bronchiolitis (CFR= 6%), while children with pneumonia had a CFR of3% (34). Weissenbacher observed higher CFRs forpneumonia than for other infections (CFR 5.8 vs. 2.1)(35). In addition, the risk of death is higher inchildren when a bacterial pathogen is identified(33, 35).

Specific problems of verbal autopsies

The ICD classification scheme normally requires aphysician or a laboratory diagnosis. Its application todeveloping countries raises specific difficulties sincemost deaths of children occur outside hospitals.Investigators have therefore developed methods forinterviewing relatives of the deceased person andhave attemped to translate this information into amedical diagnosis. These procedures, called "verbalautopsies", have been reviewed recently (42). Incommunity studies, verbal autopsies have been usedsystematically for more than three decades forassessing causes of death (42, 43).

The quality of verbal autopsies depends upon manyconditions: the design of the interview (structured,semi-structured or open interview), the time elapsedsince death, the person answering the questions, thequality of the interviewer and the qualification of thepersons who review and code the interviews.Q Sen-sitivity and specificity of the criteria used in verbalautopsies depend not only on their own char-acteristics but also on the capacity of the family tonotice and report the symptoms. Clinical case def-initions of ALRI have been validated against con-firmed diagnoses of pneumonia, whether they werefatal or not, and may not be accurate for identifyingdeath due to ARI. Few studies have attempted tovalidate clinical cr1teria against death from ALRI.Kalter (44) validated clinical signs reported by themother after the death of the child in 100 childrenunder 2 years who died from ALRI as diagnosed bya physician. Reports of cough and dyspnea beforedeath had a sensitivity of 86% and a specificity of47%. Including duration of symptoms improved thespecificity but sensitivity decreased to 41%. Navarro(45) found that in 71 children under 5 with autopsy-proven pneumonia, 50 (70%) had clinical signs ofsevere or complicated pneumonia at admission.Shann (46) evaluated clinical signs among children1-59 months of age, admitted with cough and chestin-drawing and compared those who died with thosewho survived. Among the clinical signs evaluated,highest specificity was achieved through identifyingthe severity of the chest in-drawing.

Results from national registration data confirm thesefindings. Von Schirnding (29) found that among3774 "coloured" infants aged 0-11 months who diedfrom ARI, 96.3% had a diagnosis of pneumoniarecorded on the death certificate. Puffer & Serrano(27) also reported a majority of deaths due to pneu-monia in Latin America: among neonatal deathsfrom ARI, 95.10;0 (1092/1148) had pneumonia; thisproportion was 77% (2 591/3359) among infants(1-11 months) and 69.8% (489n01) among childrenaged 12-59 months. In Ecuador, pneumonia accoun-ted for 590;0 and bronchiolitis for 280;0 of the 1 304ARI deaths reported in infants in 1987 (37).

Bronchiolitis may be misclassified as an upper res-piratory tract infection (URTI), as has been suggestedby Bulla & Hitze (38). The authors suggest that thehigh proportion of deaths due to URTI reported in 9African countries (64% of all AR/), may have been inpart due to misclassification of lower respiratorytract infections (LRTI), e.g. bronchiolitis classified asURTI. In addition, bronchiolitis is often complicatedwith pneumonia in developing countries.

Standard criteria for post-mortem diagnosis of ARIhave not yet been developed. In the communitystudies reviewed, 7 authors provided no criteria forclassifying ARI deaths. Pandey (20) defined an algo-rithm for classification of cases, but not for deaths.The ICD classification was utilized in 5 studies.Criteria for inclusion in the specific ARI categoriesand lists of ARI codes, however, were not provided.In Matlab (Bangladesh), 3 physicians independentlyassigned the ICD code after reading the post-morteminterview and an additional interview was under-taken if no consensus could be reached (9, 10).

Diagnosis of ARI

The validity of causes of death depends first on thevalidity of the diagnosis. In the case of ARI, theclinical distinction between the various syndromesremains a difficult undertaking. Inter-observer vari-ation in auscultation of the chest is frequent andideally, definitive clinical diagnosis should be basedon X-ray findings, culture of lung aspirates andmeasurement of blood oxygen levels (39-41). Thedistinction between pneumonia and bronchiolitis isparticularly difficult. Clinical signs for both syn-dromes include signs of respiratory distress such astachypnea, nasal flaring and intercostal retractions.The presence of diffuse wheezing characteristic ofbronchiolitis may be difficult to recognize for non-trained observers. Wheeze can also be found inchildren with pneumonia.

ARI in multiple causes of death

Often, ARI do not occur alone, but in associationwith other infections or conditions;such as malnutri-tion, diarrhoea and chronic conditions. The codingof multiple causes has been discussed extensively inother documents (47) and in particular in the ICD.Investigators usually include in causes of death dis-eases where ARI is an underlying (principal, primary)or precipitating (immediate, coprimary) cause. In thestudies reviewed, some authors only consideredsingle causes of death (19, 22) while others assumedthat deaths from ALRI were always the primary orcoprimary cause of death (11). Some authors include

9 Garenne, M. & Fontaine, 0. Assessing probable causes of deathsusing a standardized questionnaire -a study in rural Senegal.Proceedings of the IUSSP seminar on morbidity and mortality,Sienna, 7-10 July 1986.

Rapp. trimest. statist. sanit. mond., 45 (1992)

183

logarithmic scale was utilized to better fit the markeddecline of the proportion of ARI deaths at low levelsof mortality. Results were highly significant and aresummarized in the following equations:

or distinguish ARI as a contributing (associated)cause (7, 21, 25, f) although the criteria used fordetemining when ALRI contributes to -rather thandirectly causes -the death are not provided.

In summary, the accuracy of available data on ARIdeaths can be seriously questioned. The apparentvalidity ot the data on ARI as underlying cause ofdeath is probably due to the fact that pneumoniaand bronchiolitis are the most common causes ofdeath from ARI, that their clinical diagnosis has arelatively high sensitivity and specificity , and thatmothers can easily recognize and accurately recallthe symptoms. A more complete discussion on thevalidity of classification of ARI deaths will be pub-lished separately (Ronsmans et al., forthcoming).

Age 0-11 months:%ARI = -4.446 + 4.823*LN (ASDR-0-11)

Age 12-59 months:%ARI = 14.928 + 3.387*LN (ASDR-12-59)

A multivariate analysis was designed to investigatethe statistical effect of four variables: level of mor-tality, regional patterns (West, North, East andSouth) (48), time and level of economic development(Garenne et al., forthcoming). The level of mortalitywas significant for both infant and child mortality.There were differences according to regional pattern,with higher proportions of ARI deaths in the Eastregional pattern. The proportion of ARI deaths wassignificantly lower after 1950 among children aged1-4 years. This could be interpreted as the effect ofantibiotics on ARI mortality. The proportion of ARIdeaths was significantly lower in the more de-veloped countries at ages 1-4, but not in infancy.This again suggests a probable role of case manage-ment, which is likely to be better in more developedcountries.

Results

Most of the data available refer to underlying causesof death. These are first analysed, both for de-veloped countries and for developing countries. ARIdeaths after measles and pertussis are analysed

separately.

ARI mortality in European populations prior to 1965

In his analysis, Preston (1) found that the proportionof ARI deaths declined with the level of mortality,that the proportion of ARI deaths was slightly higheramong children 1-4 than among infants, and thatthere was no difference by sex outside of infancy(Table 2).

Comparison with developing countries

The data from the community studies and from thevital registrations were compared to the Europeanexperience by combining the two age groups: <1and 1-4 years. Values of (q) and (m), the quotient ofmortality (probability of dying between age 0 and 5years per 1 000 live births), and the age-specificdeath rates (deaths at ages 0-4 years per 1 000person-years at risk), are provided to allow easiercomparisons of mortality levels. The expected pro-portion of deaths from ARI was computed using theregression equation from the European data. Tocalculate the expected proportion of ARI deaths inthe age group 0-59 months, the expected proportionof ARI deaths for the age groups 0-11 months and12-59 months were weighted by the proportion ofchildren dying in each age group.h

The analysis of the proportion of ARI deaths waspursued separately for children <1 and 1-4 in theEuropean populations prior to 1965. The relationshipof the percentage of ARI deaths with the level ofmortality was investigated using a log-linear regres-sion, where the dependent variable was the log-arithm of the age-specific death rate (ASDR). The

" % (EO-59) = [(,Qo*%(EO-11)) + (1-1Qo)*1Q4*(E12-59))1/5QO where nQx= probability of dying between ages x and x+n, and % (Ex-x+n)expected proportion of ARI deaths in age group x to x+n based on

historical population.

TABLE 2. ARI MORTALITY IN EUROPEAN POPULATIONS PRIOR TO 1965: AVERAGE OF EMPIRICAL LIFETABLES, ACCORDING TO LEVEL OF MORTALITY

TABLEAU 2. MORTALITE IRA DANS DES POPULATIONS EUROpEENNES AVANT 1965: MOYENNE DESTABLES DE MORTALITE EMPIRIQUEs, sELON LE NIVEAU DE MORTALITE

~Age-specific death rates -Taux de deces par age

Total/1 000Sex ratio/100

Taux de masculinite1100

PercentageARI{100

PourcentageIRA/100

Levelofmortality (eO)'

Niveau demortalite (eO)'

Average eo

(years)eo moyenne(en annees)

ARI/1 000IRA/1000

117.6116.8121.8117.2118.1

20.418.117.314.610.7

45.41

26.58

12.31

5.71

2.53

223.10146.9571.0939.4823.68

38.649.560.767.571.2

Age 0-11 months -

0 a 11 mois

<4545-54

55-6465-6970-74

Age 12-59 months -

12 a 59 mois

<45

45-5455-6465-69

70-74

97.7100.5102.3

98.2100.7

9.62

4.73

1.46

0.49

0.15

26.622.622.020.613.2

36.1320.95

6.652.391.16

38.649.560.767.571.2

.eo is the life expectancy at birth in years -eo represente I'esperance de vie a la nai55ance, en annee5.

Source: Reference (I), Table 5. , -Reference (1), tableau 5. , .

WId hlth statist. quart., 45 (1992)

-184

The comparison is based on slightly different def-initions of underlying causes of death. However, theconsistency of the data from developing countrieswith the experience of developed countries wasstriking (Table 3, Fig. 1). The mean proportion of ARIdeaths was 18.8% and the mean of the predictedvalues was 17.5%. In half of the cases, the propor-tion of deaths could be predicted by the level ofmortality with a maximum relative difference of 25%.Major discrepancies between observed and expectedvalues could be explained either by the definitionsused, by the proportion of unknown causes of deathor by atypical regional patterns of mortality, with theexception of South Africa's coloured population, forwhich the observed values were much higher than

expected. In studies aimed at evaluating the effect ofcommunity-based interventions on ARI mortality,data from the control areas consistently reportedthe highest proportional mortality from ARI(Fig. 1). At the other extreme, the very low pro-portional mortality from ARI reported by Chen et al.in Bangladesh (8) probably reflects the lack of stand-ardization for coding ARI deaths.

Age pattern of mortality

The proportion of deaths occurring in each agegroup gives a picture of the age pattern of mortality.Three age groups were selected: neonatal (0-27

Study -Etude

161176557382

6474

2101

159.5118.575.2

101.9

36.626.216.022.2

65.040.094.0

26.218,819.511.4

31.218.822.4

20.218.716.917.7

1.301.011.150.64

258.0172.3341.5

23.121.224.7

1.35

0.89

0.91

130378

1593

136.8100.0256.3

30.721.764.5

33.120.815.8

20.218.323.8

1.641.130.66

Community studies -

BangladeshIndia -Inde

control -temoinintervention

KenyaMorocco- Maroc

Nepal -Nepalsurveillanceinterventionb

Nepal -NepalPakistan

control" -temoin"intervention

Senegal- SenegalU.-R. of Tanzaniad -

R.-U. de Tanzaniedcontrol -temoin

intervention325873

182.2149.4

40.132.3

35.734.9

4.03.41.4

18.618.618.5

10.510.0

5.6

1.761.863.29

21.0

19.031.1

16.416.832.6

17.617.119.6

18.815.516.315.6

0.930.981.66

0.660.691.021.28

26.913.315.813.8

16.2

12.410.716.620.0

12.5 16.7

Rapp. trimest. statist. sanit. mond., 45 (1992)

Registration systems -Systemes d'enregistrementSouth Africa -

Afrique du Sud1968-1973 13810 19.51974-1979 11079 16.51980-1985 4647 7.1

Argentina -ArgentineChaco 1 701 96.7San Juan 2 558 88.2

Bolivia- Bolivie 4276 138.2Brazil -Bresil

Recife 3635 121.3Ribeirao 1126 63.1SaoPaulo 4312 74.5

Chile- Chili 2714 65.6Colombia -Colombie

Cali 1 627 75.9Cartagena -

Carthagene 1 255 69.0Medellin 1 348 68.4

El Salvador 2210 126.3Jamaica -Jama.ique 1 903 46.9Mexico- Mexique 3953 75.9

Mean- Moyenne 2649 115.7

185

FIG.1

OBSERVED AND PREDICTED PERCENTAGE OF ARI DEATHS BY LEVEL OF MORTALITY"(8 community studies and 14 registration systems)

POURCENTAGE DE DEcEs IRA OBSERVEs ET PREvus PAR NIVEAU DE MORTALITE"(8 etudes communautaires et 14 systemes d'enregistrement)

40

.Tan (i~ Tan (c)

I BPak (c)01

.Nepl (C)

.Ind (c)

-.:~"c:~

.".,~"C

.~".""C.,

.P..,~B""Cc:o.Eo0.eQ.

Ia:-.:

S"~

"C

!F...."

"C

...

Oc:o.Eo0.O~

~

.Nepi~

.

...

-

~

.. -

r. ./

/ ....I

I.I

.Ban

010 20 30 40 50 60 70 80 W

Age-specific death rate in children 0-5 years -Taux de deces par age chez les enfants de 0-5 ans

.(c) refers to the control area, (i) to the intervention area -(c) indique une zone temoin, (i) une zone d'intervention.

Study codes -Codes etude

1000

Nep; Nepal -NepalPak: PakistanTan; U.R. of Tanzania

Ban: BangladeshBol: Bolivia -BolivieInd: India -Inde R.U. de Tanz.

days), post-neonatal (28 days-first birthday) andearly childhood (1-4 years). ARI deaths in childrenunder 5 are usually concentrated in the age group1-11 months (Table 4, Fig. 2). For all studies com-bined, 20.8% of ARI deaths occurred before age 1month, 57.8% at 1-11 months and 21.5% at 12-59months. There was a marked gradient of deaths atages 12-59 months, ranging from low values in Eastregional patterns to high values in extreme Southregional patterns such as Senegal. The share ofneonatal mortality was more mixed, probably reflect-ing inconsistencies in definitions more than realdifferences. In particular, in the Indian study (18),neonatal mortality from ARI seems to have beenlargely overestimated.

plications of measles cases was conducted, pneu-monia usually occurring in the second and thirdweek after the onset of the symptoms accounted for30% of measles deaths, and acute laryngitis, usuallyoccurring in the third or fourth week, for about 2%.Other measles deaths were due mainly to diarrhoea,sometimes with an accompanying pneumonia.

Measles also represents an important proportion ofall ARI mortality. This proportion ranges from 1.5%in Chile to 92.5% in Guinea-Bissau, and the mean forall studies is 18.6%. This proportion depends verymuch on the incidence of measles and the measlesimmunization coverage over the period considered.In Chile, measles immunization coverage was highand few measles deaths were registered. In Guinea-Bissau, Smedman (16) reported a massive outbreakof measles in the year of the study. The two studiesdescribing an unusually high contribution of measlesto ARI (16, 25) assumed that all measles deaths wereassociated with ARI, which is not agreed by otherauthors.

Contribution of measles to ARI mortality

As reported by various authors, a high proportion ofmeasles deaths seem to be associated with ARI(Table 5). Proportions range from 100% in Guinea-Bissau and the Philippines to 25% in Bangladesh.Few authors, however, define in detail "measlesassociated with ARI". It is possible that the upperrespiratory symptoms accompanying measles havebeen misclassified as pneumonia after measles. InSenegal, where an in-depth analysis. of com-

C;ontribution of pertussis to ARI mortality

According to the ICD, deaths from pertussis areclassified under ARI if they are caused by pneu-

WId hlth statist. Quart.. 45 (1992)

35

30

25

20

15

10

5

186 -

FIG.2

AGE DISTRIBUTION OF DEATHS FROM ARI(4 longitudinal studies, 13 registration systems)

REpARTITION PAR AGE DES DEcEs DUS AUX IRA(4 etudes longitudinales, 13 systemes d'enregistrement)

100

80

60

40

20

"'"'"~"C"

"C"c:.s::10c: ~"~":g"C:C""Ec:...-'E~"«"

E.!!oN~ "

~.c:,,"

.c:«~a:,,-

"C" ."~"0..,c:"C0"

.f "

0"Co.C:00

o:t 0~

0

ARG; BRAZ3 MEXBRAZ2 COL3 ELSA JAM COL: SEN

KEN

TABLE 4. PROPORTION OF ARI DEATHS BY AGE AND STUDY, CHILDREN AGED UNDER 5 YEARSa

TABLEAU 4. PROPORTION DES DEcEs IRA PAR AGE ET PAR ETUDE (ENFANTS DE MOINS DE 5 ANS)a

, Pays

ARG1ARG2BOL

54 (19.4)

88 (20.4:

308 (22.5:

183 (65.6)

306 (71.0)727 (53.1)

62 (14.0) 253 (57.0)29 (24.2) 69 (57.4)

184 (25,9) 421 (59.3)

89 (16.4) 406 (74.9) ~. 54 (26.6) 107 (52.6) 42 (20.7)

25 (20.3) 57 (46.3) 41 (33.4)21 (13.5) 102 (65.9) 32 (20.6)

69 (18.8) 214 (58.5) 83 (22.7)36 (21.7) 82 (49.4) 48 (28.9)

139 (21.7) 383 (59.8) 119 (18.6)

1254 (20.8) 3491 (57.8) 1296 (21.5)

..es chiffres indiquent le nombre (pourcentage) des deces IRA d8ns ch8que groupe d'Age.

BAAl'BAAZ2BAAZ3CHIL

4237

334

12922

105A7

COL1COL2COL3ELSAJAMMEX

India- IndeKenyaMorocco -MarocSenegal- Senegal

Argentina -ArgentineChacoSan Juan

Bolivia- BolivieBrazil- Bresil

RecifeRibeiraosao Paulo

Chile -ChiliColombia- Colombie

CaliCartagena -CarthageneMedellin

El SalvadorJamaica -Jama.iqueMexico- Mexique

All studies -Toutes etudes

ARI deaths in each age grouprhe figures give Jmber (percentagel

Rapp. trim.

(15.1 )( 8.6)(24.4)

(29.0)

(18.4)(14.8)( 8.71

187

TABLE 5. CONTRIBUTION OF MEASLES TO ARI MORTALITY , SELECTED COUNTRIES, VARIOUS YEARS a

TABLEAU 5. CONTRIBUTION DE LA ROUGEOLE AUX DEcEs IRA, DANS UN CERTAIN NOMBRE DE PAYS

ET POUR DIVERSES ANNEESa

\uthor/year/area -Auteur/annee/zone

(7.9)

(20.6)

(3.2)

(25.41

(63.61

(100.0)

(77.91

(75.01

(77.31

(100.0)

(100.01

(30.01

3906831

7388

1313118

250421

122

221

1-596-596-35

6-351-111-110-830-590-590-59

(13.71(10.21(13.0)

(92.51(61.11

(9.2)(25.01

131222621178

(4.8)(7.5)

(18.0)

(18.3)(10.0)

(6.8)(1.5)

(10.6)(11.9)(12.3)

(7.7)(9.0)

(12.5)

{49.1){78.4)

{80.1)

{84.0){91.5)

{85.2){8Z.6)

{87.9){61.5){80.4)

{65.2){45.4)

{75.6)

58810722512

1867429

20261249

57102564

0-590-590-59

40647

16223

83109

92181

11332

0-590-590-590-59

687563603

1534540

2011

0-590-590-590-590-590-59

Spika, 1989Riley,1986Fauveau, 1990 intervention

comparison -

comparaisonBhatia, 1989 intervention

comparison -comparaisonSmedman,1986Tupasi, 1990Garenne, 1992Mtango, 1986Puffer, 1973 Argentina -Argentine

ChacoSan Juan

Bolivia- BolivieBrazil -Bresil

RecifeRibeiraosao Paulo

Chile- ChiliColombia -Colombie

CaliCartagena -CarthageneMedellin

El SalvadorJamaica -JamalqueMexico- Mexique

.Deaths from measles or AlAI as an underlying or associated cause. Pertussis deaths are not included in the AlAI deaths -Deces ayant pour cause immediate 01associee la rougeole ou une IAAI. les deces dus a la coqueluche ne sont pas comptes parmi les deces IAAI.

, AlAI; acute lower respiratory infection -IAAI; infection aigue des voies respiratoires inferieures.

The lack of standardization for ascertaining causes ofdeath is another major limitation for a proper evalu-ation of the magnitude of ARI in mortality amongpreschool children. This involves the methods ofinvestigating causes of deaths, e.g. verbal autopsies,the lack of sensitivity and specificity of clinical diag-noses on which most cause-of-death data in de-veloping countries are based, and the methods ofrecording and coding multiple causes of death. Inparticular, the role of ARI may be underestimated inconsidering only underlying causes of death. It is notsurprising that the studies in which ARI was a majorfocus of research reported the highest proportionalmortality from ARI. Whether these studies representthe true contribution of ARI to mortality, or whetherthe increased attention led to overestimation, is dif-ficult to ascertain. Death is generally preceded bysigns of respiratory distress. Unless specific criteriafor a minimum duration of these symptoms beforedeath are defined as a prerequisite for assigning ARIas the cause, inclusion of non-specific signs of res-piratory distress may lead to overestimation ofmortality from ARI. Few authors, however, defined aminimum duration of respiratory symptoms beforedeath (6, 18, f).

monia. Only one study provided an indication of themagnitude of pneumonia among pertussis deaths. InSenegal, only 12% of all pertussis deaths were es-timated to be due to pneumonia as an immediatecause. For the other studies, we computed the ratioof all pertussis deaths to all ARI and pertussisdeaths. Percentages of pertussis deaths ranged from0.5% in Medellfn (Colombia) to 28.3% in Senegal(Table 6). The studies reporting the highest pro-portions were studies where the epidemiology ofpertossis was a major subject of research (12, f).However, the ratio of pertussis deaths to ARI deathsin community studies depends very much upon theepidemiologyof pertussis during the study period.

Interventions

One way to indirectly validate the ascertainment ofcauses of death is to observe the changes in ARI-specific mortality rates after cause-specific inter-ventions. The community-based treatment trialsshowed consistent declines in overall and ARI-specific mortality rates (Fig. 3).

Associations of ARI with measles and pertussis de-serve particular attention, since effective measuresfor the control of these infections are available.Pneumonia is one of the most important com-plications of measles and has been responsible for alarge proportion of measles deaths in developingcountries (49, 50). Pneumonia typically occurs 2-3weeks after the acute attack of measles and may bedue to the direct effect of the measles virus or to thepulmonary superinfection following the depressiveeffect of the measles virus on the immune system(51, 52). Despite the absence of clear definitions,and the often lacking information on levels of im-

Discussion

The assessment of the magnitude of ARI mortality ishampered by several issues. Firstly, there is nostandard definition of ARI. While there is a generalagreement that most ARI deaths are due to pneu-monia, bronchiolitis and laryngotracheitis, othercauses such as influenza may have been overlooked.Influenza was a significant cause of death In XIXthcentury Europe. Since most of the definitions ofdeaths due to ARI refer to deaths from pneumonia,the reported data should be interpreted as represent-ing primarily mortality from pneumonia.

WId hlth statist. quart., 45 (19921

188

FIG.3

INTERVENTION EFFECTS ON MORTALITY FROM ALL CAUSES AND ARIIN CHILDREN UNDER 5 YEARSa(4 community-based intervention studies)

EFFET DES INTERVENTIONS SUR LE NIVEAU DE MORTAllTE ET LE POURCENTAGE DES DEcEs IRA CHEZ LESENFANTS DE MOINS DE 5 ANSa (4 etudes d'intervention a base communautaire)

40

Tan (c)

---~~

~ 35"

....."

"0..

..,"

..,"0.."

~ 30o

'E00,0

11:

Ia:~

Tan (i)

.-.

Pak (c)

;- Nep (c

..j

I'//

Ind (c)

~25

//".,

.!0~"

"0- / /

psf)/

/ //"

8. 20e

I~ ~

1520 7030 40 50 60

Age-specific death rate in children 0-5 yearsl1 000 years -Taux de deces par age chez des enfants de 0 a 5 ansl1 000 annees

(c) refers to the control area, and ii) to the intervention area -(c) indique une zone temoin, et (i) une zone d'intervention.

Study codes -Codes etude

Ind: India -Inde Pak: Pakistan

Nep: Nepal -Nepal Tan: U.R. of Tanzania -R.U. de Tanzanie

TABLE 6. CONTRIBUTION OF PERTUSSIS TO ARI MORTALITY, SELECTED COUNTRIES, VARIOUS YEARS"

TABLEAU 6. CONTRIBUTION DE LA COQUELUCHE AUX DEcEs IRA, DANS UN CERTAIN NOMBRE DE PAYSET POUR DIVERSES ANNEES"

PertCIssis/ALRlbdeaths

oeCeS par

coqueluche/IARI'

Age group{months)

.roupe d'ige{mois!

% Pertussis among ALRIpertussis death.

% de coqueluche parmiles deces

IARlb + coqualuche

Author/year/area- Auteur/annee/zone

1-351-351-59

1-590-59

0-59

4/969/1597/543/41

67/31010/431

(4.2)

(5.7)

(13.0)

(7.3)

(21.6)

(2.3)

Fauveau, 1990 interventioncomparison -comparaison

Omondi-Odhiambo, 1984Darkaoui, 1989Garenne, 1992Mtango, 1986Puffer, 1973 Argentina -Argentine

ChacoSan Juan

Bolivia- BolivieBrazil -Bresil

RecifeRibeiraosao Paulo

Chile -ChiliColombia -Colombie

CaliCartagena -CarthageneMedellin

El SalvadorJamaica -JamaOjqueMexico- Mexique

0-590-590-59

0-590-590-590-59

24/61221/109358/2570

33/190G8/437

36/20627/1256

(3.91

(1.91

(2.31

0-590-590-590-590-590-59

12/69921/5843/407

35/156£21/561

27/2038

.Death from pertussis or ALRI as an underlying or associated cause. Measles deaths are not included in the ALRI deathsassociee la coqueluche ou une IARI. Les deces dus a la rougeole ne sont pas comptes parmi les deces IARI.b ALRI: acute lower respiratory infection -IARI: Affection aigue des voies respiratoires inferieures.

(1.71(1.81(1.71

(0.61

(1.71(3.61(0.51(2.21(3.71

(1.31

.Deces avant pour cause immediate

Rapp. trimest. statist. sanit. mond., 45 (199:

-189 -

munization coverage or the presence of majormeasles epidemics in the studies reviewed, themean estimate of 16.8% of ARI deaths that can beascribed to measles comes remarkably close to thecommonly-used WHO estimate of 15%. With thecurrent progress in world immunization againstmeasles, this proportion has decreased to 10-12%.

carried out, its contribution to ARI will remain un.known.

Ascertainment of ARI deaths due to pertussis posesmore problems. Before immunization was com-monly practised in industrialized countries, majorepidemics occurred every 4-6 years (53). The epi-demic pattern of pertussis makes difficult the es-tablishment of its importance in short-term sur-veillance. In verbal autopsies, a history of the typicalwhoop will only be elicited if specific questions areasked. In addition, very young infants show a lowerfrequency of paroxysms and typical whoop, com-plicating the diagnosis in this age group. The highproportion of ARI deaths attributed to pertussis inthe studies in Senegal and Kenya, where deathsfrom pertussis were addressed with particular atten-tion, may merely reflect this fact, though the contri-bution of major epidemics during the study periodcannot be excluded. Pertussis may also cause de-layed mortality through its effect on the nutritionalstatus of the child (54). The epidemiology of pertus-sis has received less attention in developing count-ries than measles or ARI, and unless precise casedefinitions are developed and long-term surveillance

Data based on underlying causes of death suggestthat, in developing countries, approximately 1 out of6 deaths of children aged 0-4 years are due topneumonia. This estimate matches what is knownfrom developed countries at similar levels of mor-tality in the past, To this major underlying cause ofdeath, one shou1d add other ARI deaths, ARI deathsafter measles, pertussis or other infectious diseasesas well as in association with acute malnutrition.Without more accurate data, it seems to be difficultto give a final estimate, but the WHO figure of 1 outof 3 deaths due to -or associated with -ARI maybe close to the real range of ARI proportional mor-tality in children of developing countries.

ARI mortality has been declining steadily with im-proving living conditions in developed countries andhas been declining very rapidly since 1950 whenantibiotics became availabJe. Perhaps the best wayto estimate the current burden of ARI diseases indeveloping countries is to compare ARI mortality tothe lowest values recorded in developed countries.This would provide a number of deaths that couldbe averted if the best medical technology wereprovided to every child. Such a comparison and thehigh values of ARI mortality found in many develop-ing countries indicate that more efforts should bemade to better control ARI.

SUMMARY

This article reviews the available evidence of mor-.tality from acute respiratory infections (ARI) amongchildren aged under 5 years in contemporary devel-oping countries and compares the findings withEuropean populations before 1965. In Europeanpopulations before 1965, the level of mortality wasfound to be a determinant of the proportion ofdeaths due to ARI. There were marked differencesaccording to regional patterns of mortality. Deathsfrom ARI played a smaller role after 1950, when theuse of antibiotics became generalized.

In developing countries, the role of ARI mortalityseems to be similar to the European experience. Theage pattern is very marked. In absolute values, ARImortality is highest in the neonatal period and de-creases with age. In relative values, ARI mortality ishighest in the postneonatal period.

ARI, mainly pneumonia, accounts for about 18% ofunderlying causes of death in developing countries.Pneumonia and other ARI are frequent complicationsof measles and pertussis; ARI is also commonlyfound after other infections and in association withsevere malnutrition. Virtually no data are available indeveloping countries to provide final estimates ofthe role of ARI in mortality of children aged under5 years. However, the WHO figure of 1 out of 3deaths due to -or associated with -ARI may beclose to the real range of the ARI-proportional mor-tality in children of developing countries.

Results are discussed in light of the definitions ofARI used in various studies, the difficulties in ascer-taining and coding multiple causes of death and thequality of data from some sources.

REsUME

Ampleur de la mortalite due aux affections aigues des voies respiratoireschez les enfants de moins de 5 ans dans les pays en developpement

de mortalite presentaient de tres nettes differencesselon les regions. Le role des deces par IRA adiminue a partir de 1950, avec la generalisation deI'usage des antibiotiques.

Cet article passe en revue les donnees disponiblesconcernant la mortalite actuelle par affection aiguedes voies respiratoires (ou infection respiratoire aigue-IRA) chez les enfants de moins de 5 ans dans les paysen developpement, et etablit une comparaison avecla situation en Europe avant 1965. On avait constatequ'avant cette date, daris les populations euro-peennes, le niveau de mortalite etait un determinantde la proportion de deces dus aux IRA. Les tableaux

Dans les pays en developpement, le role de lamortalite par IRA paraTt similaire a ce qu'il etait jadisen Europe. La repartition par age est tres nette. Envaleur absolue, c'est au cours de la periode neona-

Wid hlth statist. quart.,45 (1992)

190

tale que la mortalite par IRA est la plus forte; el lediminue ensuite avec I'age. En valeur relative, cettemortalite est plus forte durant la periode post-neonatale.

sur le role des IRA dans la martalite des enfants demains de 5 ans. Tautefais, le chiffre qu'indiqueI'OMS -1 deces sur 3 dO au assacie aux IRA -daitetre assez prache de la realite, s'agissant de lamartalite prapartiannelle par IRA chez les enfantsdes pays en develappement.Les IRA, et en particulier les pneumonies, sont la

cause initiale d'environ 18% des deces oans les paysen developpement. Elles sont une complication fre-quente de la rougeole et de la coqueluche; on lesrencontre frequemment aussi apres d'autres infec-tions, ou associees a la malnutrition grave. II n'y adans les pays en developpement que tres peu dedonnees qui puissent fournir des estimations finales

L'analyse des resultats tient compte des definitionsdes IRA utilisees dans differentes etudes, des diffi-cultes a determiner les causes multiples de deces eta les coder, et de la qualite des donnees provenantde certaines sources.

REFERENCESREFERENCES

1. PRESTON, S. H. Mortality patterns in national popu-lations. New York, Academic Press, 1976.

2. GORDON, J. E. ET AL. Death rates and causes ofdeath in eleven Punjab villages: an epidemiolo-gical study. Indian journal of medical research,49: 568-594 (1961).

3. A programme for controlling acute respiratoryinfections in children: memorandum from aWHO meeting. Bulletin of the World Health Or-ganization, 62: 47-58 (1984).

4. LEOWSKI, J. Mortality from acute respiratory infec-tions in children under 5 years of age: globalestimates. World health statistics quarterly, 39(2): 138-144 (1986).

LEOWSKI, J. Mortalite due aux infections aiguescdes voies respiratoires chez les enfants de moinsde 5 ans: estimations a I'echelle mondiale. Rap-port trimestriel de statistiques sanitaires mon-diales, 39 (2): 138-144 (1986).

5. PRESTON, S. H. ET AL. Life tables for nationalpopulations. New York, Seminar Press, 1972.

6. DARKAOUI, N. & GARENNE, M. National survey oncauses and circumstances of infant and childdeaths. Rabat, Ministere de la sante publique duMaroc, 1989.

7 o BHATIA, S. Patterns and causes of neonatal andpostneonatal mortality in rural Bangladesh.Studies in family planning, 20: 136-146 (1989).

8. CHEN, L. Co ET AL. Epidemiology and causes ofdeath among children in a rural area ofBangladesh. International journal of epidemio-logy, 9: 25-33 (1980).

9. FAUVEAU, V. ET AL. The effect <if maternal andchild health. and family planning services onmortality: is prevention enough? British medicaljournal, 301: 103-107 (1990):

10. FAUVEAU, V. ET ALo Excess female deaths amongrural Bangladeshi children: an examination ofcause-specific mortality and morbidity. Inter-national journal of epidemiolog~ 20: 729-735(1991).

11. SPlKA, J. S. ET AL. Acute lower respiratory infec-tions: a major cause of death in children inBangladesh. Annals of tropical paediatrics, 9:33-39 (1989).

12. OMONDI-ODHIAMBO, ET AL. Age-specific infant andchildhood mortality and causes of death. In: VanGinneken, J. K. & Muller, A. S. (eds), Maternaland child health in rural Kenya: an epidemiologi-cal study. London, Croom Helm, 1984.

13. BRADLEY, A. K. & GILLES, Ho M. Malumfashi en-demic diseases research project, XXI. Pointers tocauses of death in the Malumfashi area, north-

ern Nigeria. Annals of tropical medicine andparasitology; 78: 265-271 (1984).

14. CANTRELLE, P. ET AL. The profile of mortality and itsdeterminants in Senegal, 1960-1980. In: Deter-minants of mortality change and differentials indeveloping countries. New York, United Nations,1986. (UN populationstudies~E. 85.XIII.4).

15. GREENWOOD, B. M. ET AL. Deaths in infancy andearly childhood in a wel'-vaccinated, rural WestAfrican population. Annals of tropical paedia-trics, 7: 91-99 (1987).

16. SMEDMAN, L. ET AL. Survival 0-6 years of age in aperiurban community in Guinea-Bissau: a lon-gitudinal assessment. Annals of tropical paedia-trics, 6: 67-72 (1986).

17. MCCORD, C. & KIELMANN, A. A. A successful pro-gramme for medical auxiliaries treating child-hood diarrhoea and pneumonia. Tropical doctor,8: 220-225 (1978).

18. BANG, A. T. ET AL. Reduction in pneumonia mor-tality and total childhood mortality by means ofcommunity-based intervention trial in Gadchiroli,India. Lancet, ii: 201-206 (1990).

19. DATTA, N. ET AL. Application of case managementto the control of acute respiratory infections inlow-birthweight infants: a feasibility study.Bulletin of the World Heath Organization, 65 (1 ):77-82 (1987).

DATTA, N. ET AL. La prise en charge des casappliquee a la lutte contre les infections res-piratoires aigues chez les enfants de poids in-suffisant a la naissance: etude de faisabilite[resume]. Bulletin de f'Organisation mondiale dela Sante, 65 (1): 81 (1987).

20. PANDEY, N. R. ET AL. Impact of a pilot acuterespiratory infection (ARI) control programme ina rural community of the hill region of Nepal.Annals of tropical paediatrics, 9: 212-220 (1989).

21. PANDEY, N. R. ET AL. Reduction in total under-fivemortality fn western Nepal through community-based antimicrobial treatment of pneumonia.Lancet, ii: 993-997 (1991).

22. KHAN, A. J. ET AL. Acute respiratory infections inchildren: a case management intervention inAbbottabad district, Pakistan. Bulletin of theWorld Health Organization, 68: 577-585 (1990).

23. MTANGO, F. D. E. & NEUVIANS, D. Acute respiratoryinfections in children under five years ~ controlproject in Bagamoyo district, Tanzania. Trans-actions of the Royal Society of Tropical Medicineand Hygiene, 80: 851-858 (1986).

24. RILEY, I. D. ET AL. Immunization with a polyvalentpneumococcal vaccine ~ effect on respiratory~

Rapp. trimest. statist. sanit. mond., 45 (19921

191 -

ORGANISATION MONDIALE DE LA SANTE. Annuaire destatistiques sanitaires mondiales. Geneve, OMS,1989.

38. BULLA, A. & HITZE, K. L. Acute respiratory infec-tions: a review. Bulletin of the World HealthOrganization, 56 (3): 481-498 (1978).

BULLA, A. & HITZE, K. L. Panorama des infectionsaigues de I'appareil respiratoire [resume].Bulletin de /'Organisation mondiale de la Sante,56 (3): 497 (1978).

39. MULROW, C. D. ET AL. Observer variability in thepulmonary examination. Journal of general in-ternal medicine, 1: 365-367 (1986).

40. STEINHOFF, M. C. Nosology and standards for ARIstudies. In: Gadomski, A. G. (ed.), ALRI and childsurvival in developing countries. Proceedings ofa workshop held in August 1989 in Washington,D.C. Baltimore, JHU Design Publications, 1990.

41. BERMAN, S. Overview of pneumonia in early in-fancy. In: Gadomski, A. G. (ed.), ALRI and childsurvival in developing countries. Proceedings ofa workshop held in August 1989 in Washington,D.C.Baltimore, JHU Design Publications, 1990.

42. GRAY, R. H. Verbal autopsy: using interviews todetermine causes of death in children.Baltimore, The Johns Hopkins University Press,1991. {Occasional paper No.14).

43. KIELMANN, A. A. ET AL. Child and mater:nal healthservices in rural India -the Narangwal experi-ment. Baltimore, The Johns Hopkins UniversityPress, 1983.

44. KALTER, H. D. ET AL. Validation of postmorteminterviews to ascertain selected causes of deathin children. International journal of epidemio-logy, 19: 380-386 (1990).

45. NAVARRO, E. E. ET AL. Clinicopathologic studies ofchildren who die of acute lower respiratory tractinfections: mechanism of death. Reviews of in-fectious diseases, 12 (Supplement 8): 51065-51073 (1990).

46. SHANN, F. ET AL. Clinical signs that predict deathin children with severe pneumonia. Paediatricinfectious disease journal, 8: 852-855 (1989).

47. MANTON, K. G. & STALLARD, E. Recent trends inmortality analysis. New York, Academic Press,1984.

48. COALE, A. & DEMENY, P. Regional model life tablesand stable populations. Princeton, Princeton Uni-versity Press, 1966.

49. MORLEY, D. C. ET AL. Measles in West Africa. WestAfrican medical journal, 16: 24-31 (1967).

50. VOORHOEVE, A. M. ET AL. Agents affecting thehealth of mother and child in a rural area ofKenya ~ III. The epidemiology of measles. Trop-ical and geographical medicine, 29: 428-440

(1977).51. MARKOVITZ, L. A. & NIEBURG, P. The burden of

acute respiratory infection due to measles indeveloping countries and the potential impact ofmeasles vaccine. Reviews of infectious diseases,13 (Supplement 13): 5555-5561 (1991).

52. DOVER, A. S. ET AL. Pneumonia associated withmeasles. JAMA, 234: 612-614 (1975).

53. WRIGHT, P. F. Pertussis in developing countries:definition of the problem and prospects for con-trol. Reviews of infectious diseases, 13(Supplement 16): 5528-5534 (1991).

54. MORLEY, D. C. ET AL. Whooping cough in Nigerianchildren. Tropical geographical medicine, 18:

169-182 (1966).

mortality in children living in the New Guineahighlands. Archives of disease in childhood, 56:

354-357 (1981).

25. RILEY, I. D. ET AL. Pneumococcal vaccine preventsdeath from acute lower-respiratory-tract .infec-tions in Papua New Guinean children. Lancet, ii:

877-881 (1986).

26. TUPASI, T. K. ET AL. Patterns of acute respiratorytract infection in children: a longitudinal study ina depressed community in Metro Manila. Re-views of infectious diseases, 12 (Supplement 8):5940-5949 (1990).

27. PUFFER, R. R. & SERRANO, C. V. Patterns of mor-tality in childhood. Washington, D.C., PanAmerican Health Organization, 1973. (Scientificpublication No.262).

28. WORLD HEALTH ORGANIZATION. Manual of the inter-national statistical classification of diseases, in-juries and causes of death. Eighth Revision.Geneva, WHO, 1967.

ORGANISATION MONDIALE DE LA SANTE. Manuel de laclassification statistique internationale des mala-dies, traumatismes et causes de deces. Huitiemerevision. Geneve, OMS, 1967.

29. VON SCHIRNDING, Y. E. R. ET AL. Acute respiratoryinfections as an important cause of childhooddeaths in South Africa. South African medicaljournal, 80: 79-82 (1991).

30. WORLD HEALTH ORGANIZATION. Manual of the inter-national statistical classification of diseases, in-juries and causes of death. Ninth Revision.Geneva, WHO, 1977.ORGANISATION MONDIALE DE LA SANTE. Manuel de laclassification statistique internationale des mala-dies, traumatismes et causes de deces. Neu-vieme revision. Geneve, OMS, 1977.

31. WORLD HEALTH ORGANIZATION. Manual of the inter-national statistical classification of diseases, in-juries and causes of death. Seventh Revision.Geneva, WHO, 1957.ORGANISATION MONDIALE DE LA SANTE. Manuel de laclassification statistique internationale des mala-dies, traumatismes et causes de deces. Septiemerevision. Geneve, OMS, 1957.

32. PIO, A. ET AL. The magnitude of the problem ofacute respiratory infections. In: Douglas, R. M. &Kirby-Eaton, E. (eds), Acute respiratory infectionsin childhood. Proceedings of an internationalworkshop. Sydney, University of Adelaide, 1985.

33. RAHMAN, M. ET AL. Acute lower respiratory tractinfections in hospitalized patients with diarrhoeain Dhaka, Bangladesh. Review of infectious dis-eases, 12 (Supplement 8): 5899-5906 (1990).

34. MTANGO, F. D. E. ET AL. Magnitude, presentation,management and outcome of acute respiratoryinfections in children under the age of five inhospitals and rural health centres in Tanzania.Tropical medicine and parasitology; 40: 97-102

(1989).35. WEISSENBACHER, M. ET AL. Etiologic and clinical

evaluation of acute lower respiratory tract infec-tions in young Argentinian children: an over-view. Reviews of infectious diseases, 12(Supplement 8): 5889-5898 (1990).

36. SUNAKORN, P. ET AL. Epidemiology of acute res-piratory infections in young children from Thai-land. Paediatric infectious disease journal, 9:873-877 (1990).

37. WORLD HEALTH ORGANIZATION. World health statis-tics annual. Geneva, WHO, 1989.

Wid hlth statist. quart", 45 (1992)