7
International Journal of Epidemiology tNTERNATIQNAL REFERENCE CiNTrfE Vol. 23, No. 4 © International Epidemiological Association 1994 FOR COMMUNITY WATFR 'ilf*' 3 ! Y A f ' H Printed in Great Britain SANITATION (¡RC) Family Latrines and Paediatric Shigellosis in Rural Bangladesh: Benefit or Risk? FARUQUE AHMED,*'* JOHN D CLEMENS,* 1 MALLA R RAO*' f AND A K BANIK* Ahmed F (International Centre for Diarrhoeal Disease Research, Bangladesh, Dhaka, Bangladesh), Clemens J D, Rao M R and Banik AK. Family latrines and paediatric shigellosis in rural Bangladesh: Benefit or risk? International Journal of Epidemiology 1994; 23: 856-862. Background. The potential benefits of installing excreta disposal facilities on the burden of paediatric diarrhoea in less-developed settings remain controversial. We conducted a longitudinal study to evaluate whether family latrines are associated with interruption of the transmission of shigellosis to younger children in rural Bangladesh. Methods. We prospectively studied 1529 children under 5 years of age exposed to index cases of Shigella dysentery. In all 219 children with culture-proven shigellosis detected during 1 month of follow-up were compared with 1310 control children who did not develop shigellosis or Shigella-negative dysentery. Results. Overall, the presence of a family latrine appeared to be associated with a higher, not a lower, risk of paediatric shigellosis (adjusted odds ratio (OR a ) = 1 -37, 95% confidence interval (CI): 0.99-1.89). While use of a pit or sanitary latrine revealed no evidence of a protective association (ORa = 0.96, 95% CI: 0.43-2.15), use of a hanging latrine in which faeces were discharged directly onto the ground or into a body of water was associated with a notable increase Of risk (OR a = 1.42, 95% CI: 1.02-1.98, P < 0.05). Conclusions. While cautioning that installation of sanitary latrines may not be sufficient to reduce the burden of paediatric shigellosis in less-developed settings, these data suggest that eliminating unsanitary latrines constitutes a potentially important intervention in its own right in these settings. st c'" A. W j Improvement of facilities for disposal of excreta was one of the major aims of the International Water Supply and Sanitation Decade (1981-1990).' Doubts have been raised, however, concerning the validity of findings of studies that have assessed the impact of excreta disposal facilities on paediatric diarrhoea." Moreover, because the impact of such facilities on diarrhoea may vary ac- cording to the pathogen involved, 1 aetiology-specific studies are desirable. Because morbidity and mortality due to shigellosis is particularly high in less-developed countries, assessments of the impact of latrines on the transmission of Shigella are of great importance. 3 ' 4 In this paper, we describe the results of a large- scale, community-based study of Shigella-cxposed neighbourhoods in rural Bangladesh, addressing the following questions: (1) Was the presence of a latrine * Internationa] Centre for Diarrhoea! Disease Research, Bangladesh, Dhaka, Bangladesh. * Division of Epidemiology, Statistics, and Prevention Research, Na- tional Institute of Child Health and Human Development, Bethesda, MD, USA. Reprint requests to: John D Clemens, Division of Epidemiology, Statis- tics, and Prevention Research, National Institute of Child Health and Human Development, 6100 Executive Plaza Building, Bethesda, MD 20892, USA. associated with a lower risk of the occurrence of shigel- losis among children? (2) Did this association differ ac- cording to whether the latrine was sanitary versus unsanitary? (3) Was the association modified by breastfeeding status? METHODS Overview The association between having a latrine and the occur- rence of shigellosis was determined in 1529 Bangladeshi children exposed to Shigella in their residential neighbourhoods. Each child was prospectively followed by home visits for 1 month, and the presence and type of a family latrine was related to the occurrence of Shigella diarrhoea. Assembly of Index Cases for Initiating Neighbourhood Contact Studies As described earlier 5 ' 6 index cases of Shigella dysentery were assembled from patients presenting to the three diarrhoeal treatment centres serving the Matlab study population of the International Centre for Diarrhoeal Disease Research, Bangladesh. Such cases were poten- tially eligible to initiate follow-up in their residential neighbourhoods if the onset of diarrhoea was less than 1 da\ rv r r 856

Britain Family Latrines and Paediatric Shigellosis in ... · Occurrence of Shigellosis Table 2 shows the relationship between having a family latrine and the occurrence of shigellosis

  • Upload
    others

  • View
    6

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Britain Family Latrines and Paediatric Shigellosis in ... · Occurrence of Shigellosis Table 2 shows the relationship between having a family latrine and the occurrence of shigellosis

International Journal of Epidemiology t N T E R N A T I Q N A L R E F E R E N C E C i N T r f E Vol. 23, No. 4© International Epidemiological Association 1994 F O R C O M M U N I T Y W A T F R ' i l f * ' 3 ! Y A f ' H P r i n t e d in G r e a t Britain

SANITATION (¡RC)

Family Latrines and PaediatricShigellosis in Rural Bangladesh:Benefit or Risk?FARUQUE AHMED,*'* JOHN D CLEMENS,* 1 MALLA R RAO*' f AND A K BANIK*

Ahmed F (International Centre for Diarrhoeal Disease Research, Bangladesh, Dhaka, Bangladesh), Clemens J D,Rao M R and Banik AK. Family latrines and paediatric shigellosis in rural Bangladesh: Benefit or risk? InternationalJournal of Epidemiology 1994; 23: 856-862.Background. The potential benefits of installing excreta disposal facilities on the burden of paediatric diarrhoea inless-developed settings remain controversial. We conducted a longitudinal study to evaluate whether family latrinesare associated with interruption of the transmission of shigellosis to younger children in rural Bangladesh.Methods. We prospectively studied 1529 children under 5 years of age exposed to index cases of Shigella dysentery.In all 219 children with culture-proven shigellosis detected during 1 month of follow-up were compared with 1310control children who did not develop shigellosis or Shigella-negative dysentery.Results. Overall, the presence of a family latrine appeared to be associated with a higher, not a lower, risk of paediatricshigellosis (adjusted odds ratio (ORa) = 1 -37, 95% confidence interval (CI): 0.99-1.89). While use of a pit or sanitarylatrine revealed no evidence of a protective association (ORa = 0.96, 95% CI: 0.43-2.15), use of a hanging latrine inwhich faeces were discharged directly onto the ground or into a body of water was associated with a notable increaseOf risk (ORa = 1.42, 95% CI: 1.02-1.98, P < 0.05).Conclusions. While cautioning that installation of sanitary latrines may not be sufficient to reduce the burden ofpaediatric shigellosis in less-developed settings, these data suggest that eliminating unsanitary latrines constitutes apotentially important intervention in its own right in these settings.

stc'"

A.

W

j

Improvement of facilities for disposal of excreta was oneof the major aims of the International Water Supply andSanitation Decade (1981-1990).' Doubts have beenraised, however, concerning the validity of findings ofstudies that have assessed the impact of excreta disposalfacilities on paediatric diarrhoea." Moreover, becausethe impact of such facilities on diarrhoea may vary ac-cording to the pathogen involved,1 aetiology-specificstudies are desirable. Because morbidity and mortalitydue to shigellosis is particularly high in less-developedcountries, assessments of the impact of latrines on thetransmission of Shigella are of great importance.3'4

In this paper, we describe the results of a large-scale, community-based study of Shigella-cxposedneighbourhoods in rural Bangladesh, addressing thefollowing questions: (1) Was the presence of a latrine

* Internationa] Centre for Diarrhoea! Disease Research, Bangladesh,Dhaka, Bangladesh.* Division of Epidemiology, Statistics, and Prevention Research, Na-tional Institute of Child Health and Human Development, Bethesda,MD, USA.Reprint requests to: John D Clemens, Division of Epidemiology, Statis-tics, and Prevention Research, National Institute of Child Health andHuman Development, 6100 Executive Plaza Building, Bethesda, MD20892, USA.

associated with a lower risk of the occurrence of shigel-losis among children? (2) Did this association differ ac-cording to whether the latrine was sanitary versusunsanitary? (3) Was the association modified bybreastfeeding status?

METHODSOverviewThe association between having a latrine and the occur-rence of shigellosis was determined in 1529 Bangladeshichildren exposed to Shigella in their residentialneighbourhoods. Each child was prospectively followedby home visits for 1 month, and the presence and type ofa family latrine was related to the occurrence of Shigelladiarrhoea.

Assembly of Index Cases for Initiating NeighbourhoodContact StudiesAs described earlier5'6 index cases of Shigella dysenterywere assembled from patients presenting to the threediarrhoeal treatment centres serving the Matlab studypopulation of the International Centre for DiarrhoealDisease Research, Bangladesh. Such cases were poten-tially eligible to initiate follow-up in their residentialneighbourhoods if the onset of diarrhoea was less than 1

da\

rvrr

856

Page 2: Britain Family Latrines and Paediatric Shigellosis in ... · Occurrence of Shigellosis Table 2 shows the relationship between having a family latrine and the occurrence of shigellosis

! • ! ? ,^••••'•7 <"••:• ; ' • • - • • ' . • : i - v ^ V ' " - : i r í ' , í r t . ' •'••* ' ' • „ ' * • ' » * • ' • . * * ! • • - v i - f i r i - í . 1 ; ^ ' : ' ! - , * , • £ ( » « . - '

-.í

,<l

•I1

FAMILY LATRINES AND PAËDIATRIC SHIGELLOSIS

week before initiation of the neighbourhood contactstudy. As detailed previously,5 240 index cases werechosen among the 507 cases that were eligible during theperiod 1 November 1987 to 30 November 1989.

Assembly of Subjects for the StudyWorkers visited the neighbourhood of each index casewithin 48 hours of presentation to the treatment centre.Informed consent for participation was obtained from aresponsible adult for each participating child. All fami-lies situated around the courtyard of the residence of theindex case and around two adjacent courtyards consti-tuted the neighbourhood for study. Tn all 1934 childrenaged 0-59 months from 240 neighbourhoods were as-sembled, out of which 1756 (91%), from 1218 families,were present at baseline.

Acquisition of Baseline DataOn day 1, a reliable adult member was questioned aboutselected sociodemographic characteristics of the family,including ownership and use of a latrine. The design ofthe latrine was observed by field workers, with queriesbeing made regarding the underground structure. Moth-ers (or caretakers, if mothers were absent) were askedabout diarrhoea in their children occurring in the previ-ous week, as well as about antibiotic use and the feedingpractices of each child. Heights of children were mea-sured to the nearest 1 mm. Height-for-age was assessedin relation to National Center for Health Statistics refer-ence data.7 A rectal swab was also collected from eachparticipating child and immediately placed in bufferedglycerol saline transport medium.8 Plating was done onMacConkey and Salmonella-Shigella media within a fewhours of swab collection, and evaluated for Shigdlausing standard techniques.9

Surveillance for Post-Baseline DiarrhoeaAlter baseline, study children were visited on alternatedays for 10 days to obtain interval diarrhoeal histories.After day 10, three additional visits were made, com-mencing from day 17, during which 7-day diarrhoealrecall histories were obtained. To improve the 7-dayrecall, mothers were encouraged to record diarrhoea oncalendars designed for use by illiterate adults. Rectalswabs were collected if diarrhoea, as defined below, wasreported in the interval since the previous visit. Fieldworkers were instructed to empirically treat all dysentericillnesses with nalidixic acid.10

DefinitionsSeveral definitions were formulated before analysis ofthe data. A diarrhoeal day was defined as a 24-hourperiod with s 3 non-bloody loose or liquid motions, or

with s 1 bloody loose or liquid motions. The onset of anepisode of diarrhoea was the first day with diarrhoeabefore which there were at least 3 consecutive diarrhoea-free days. The end of the episode occurred with the onsetof & 3 consecutive diarrhoea-free days, the day of termi-nation of the episode being the last diarrhoeal daypreceding the diarrhoea-free days. ' ' Persistent diarrhoeawas defined as an episode that lasted & 14 days whileepisodes of shorter duration were termed acute.11 Shig-ellosis denoted isolation of Shigella from a faecalspecimen anytime from 3 days before the onset of adiarrhoeal episode to 3 days after the end of the episode.Dysentery denoted diarrhoeal episodes described asbloody. If a child had episodes of both shigellosis andculture-negative dysentery, the child was classified ashaving shigellosis.

A family was considered to have a latrine if it ownedand used a latrine. In the rare event that a family hadmore than one latrine, the one primarily used by thefamily was classified. A hanging latrine referred to aplatform resting on pillars, with walls made of brick, tin,bamboo, or jute stick, and from which faeces fell directlyonto the land surface, embankment, or water. A pitlatrine consisted of a concrete slab with a squat-type panplaced over an underground pit. A latrine with a septictank was denoted as a sanitary latrine. A child wasclassified as breastfed if breast milk constituted any por-tion of the child's diet, as ascertained on the day beforethe onset of illness for children having diarrhoea begin-ning at or before baseline, and at baseline for otherchildren.

Statistical AnalysisBefore analysis of the data, we decided to exclude 227children detected during follow-up as having culture-negative dysentery, since we were interested in the impactof latrines upon transmission of Shigella and since wecould not exclude Shigella as an aetiological pathogen inthese culture-negative episodes, due to the known insen-sitivity of culture techniques for Shigella}1 After theseexclusions, a study population of 1529 children re-mained. For the examination of associations betweenexposure variables and acute versus persistent Shigelladiarrhoea, only episodes of shigellosis having onsets be-fore day 19 of follow-up were included to ensure that allepisodes had an equal opportunity of being detected asbecoming persistent, regardless of the day of onset (fol-low-up extended only to day 31).

Inter-group comparison of means were statisticallycompared with the Student's t-test, comparisons of pro-portions were evaluated with the %2 or Fisher's exact test,and medians were assessed with Wilcoxon's rank sumtest. To evaluate the association between presence of

Page 3: Britain Family Latrines and Paediatric Shigellosis in ... · Occurrence of Shigellosis Table 2 shows the relationship between having a family latrine and the occurrence of shigellosis

INTERNATIONAL JOURNAL OF EPIDEMIOLOGY

latrines and the occurrence of paediatric shigellosis, wecompared cases (n = 219), neighbourhood contacts aged<60 months with culture-confirmed shigellosis; with con-trols (n = 1310), those in the same age group in whomneither shigellosis nor Shigella-ncg&tive dysentery wasdetected. The magnitude of the association was ex-pressed as odds ratio to incorporate simultaneousanalysis of outcomes detected at baseline or thereafter.To obtain odds ratios adjusted for the simultaneousconfounding effect of relevant baseline variables, se-lected on the basis of either statistical significance orbiological importance, multiple logistic regression mod-els were fitted, taking case-control status as thedependent variable (PROC LOGISTIC, Version 6.07,Statistical Analysis Software, Cary, NC).'3 Possible in-teractions between variables were also evaluated. The95% confidence intervals (CI) for odds ratios were com-puted with Woolf s method in simple analyses, and withuse of the standard error of coefficients in logistic mod-els. Adjustment for the clustering of responses amongfamilies was done by a software program developed byResearch Triangle Institute (Shah B V, Folsorn R E,Harrell F E, Dillard C N. RT/LOG/T: Procedure forLogistic Regression on Survey Data, January 1987. Re-search Triangle Institute, PO Box 12194, ResearchTriangle Park, NC 27709, USA). All test statistics wereinterpreted in a two-tailed fashion to estimate P valuesand confidence intervals.

RESULTSBaseline ComparabilityIn all 219 cases of microbiologically-confirmed shigello-sis were found among 204 families. Bivariatecomparisons of cases and controls for several relevantbaseline characteristics revealed that older age (medianage of 33 versus 27 months, P < 0.001), stunting (meanheight-for-age z scores of-2.9 versus -2.5, P < 0.001),lack of breastfeeding (34% of cases versus 11% of con-trols among under-3 year olds, P < 0.001), Muslimreligion (98% of cases versus 95% of controls, P < 0.05),family ownership of land (89% of cases versus 82% ofcontrols, P < 0.05), and lower age of the index case(median age 4 versus 5 years, P < 0.01) were signifi-cantly associated with the occurrence of Shigella cases(Table 1).

Association between Type of Latrine and theOccurrence of ShigellosisTable 2 shows the relationship between having a familylatrine and the occurrence of shigellosis. Overall, use ofa family latrine appeared to be associated with a higher,not a lower, risk of paediatric shigellosis (odds ratioadjusted for covariates [ORJ =1.37,95% CI: 0.99-1.89).

TABLE I Comparison ojcases andcontrols for selectedbaseline feaivtxBangladesh, ¡987-1989*

Feature

Median age of child (months)(range)

MaleMuslimHeight-for-age z score(MeaniSD)

Breastfed1-'Median years of schooling of

mother or caretaker (range)Median family size (range)

Median per capita yearlyincome of family (range)d

Family owns landHouse with tin or

brick walls1

Functioning tubewell incourtyard

Residence in index familyShigella species of index casejiexneridysenteriae 1 (shiga)Other

Cases(n = 219)

33(5-59)

108 (49)215(98)-2.9 ±1.1

82(66)0

(Chi 2)7

(2-20)3007(833-18 743)194(89)72(33)

62(28)

15(7)

106(48)95 (43)18(8)

Median age of index case (years) 4(range)

Male index caseIndex case is mother'Season of selection*

Pre-monsoonMonsoonPost-monsoon

(0-68)141 (64)

0(0)

54(25)56(26)

109(50)

Controls *-%•(n = 1310) .',

2 7 * "(0-59)

670(51)1246(95)' . •

-2.5 i U * "J±-

730(89)"* •0

(0-12)6

(2-24)3000(383-29 000)1069(82)*384 (29)

407(31)

70(5)

675 (52)493 (38)142(11)

5 "(0-79)

887(68)10(1)

295 (23)431 (33)584(45)

#

fi

• "i.

í%M'ñS.Ü%ST

11*M•:ê

. ¥

" Values represent number (per cent).•P < 0.05; **P < 0.01; ••* < 0.001, for comparison of cases versuscontrols.b For 215 cases and 1267 controls with known information.' Under-3 year olds only (124 cases and 819 controls).d Bangladesh tafca (1 US dollar = 32 taka). Three controls withunknown income were excluded.e Excluded one control with unknown wall type.

No caretaker was an index case.* Pre-monsoon = March~May; monsoon ~ June-September;post-monsoon = October-February.

Compared to having no latrine, having a hanging latrinewas significantly associated with the occurrence of shig-ellosis (ORa = 1.42,95% CI: 1.02-1.98, P < 0.05), whereashaving a pit or sanitary latrine was not (ORa - 0.96,95%CI: 0.43-2.15). Additional adjustment for the clusteringof responses among families yielded adjusted oddsratios of 1.42 (95% CI: 1.02-1.99, P < 0.05) and 0.96(95% CI: 0.44-2.10) for hanging or pit/sanitary latrines,respectively.

t

Page 4: Britain Family Latrines and Paediatric Shigellosis in ... · Occurrence of Shigellosis Table 2 shows the relationship between having a family latrine and the occurrence of shigellosis

FAMILY LATRINES AND PAEDIATRIC SHIGELLOSIS 859

/ ' • ' . ' • • "

Z--_(.•'•

? : * • • •

È:!¿f_"..

T*BLE2 Overallassociation bi

No latrineLUnne

Hanging latrine

Pit or sanitarylatrine1

Cases

No. (%)

131 (60)88(40)

79(36)

9(4)

Hween latrine type and Í

Controls

No, (%)

905(69)405(31)

348 (27)

57(4)

episodes of shigellosis: Bangladesh, 1987-1989

Crude oddsratio

(95% confidence interval)

1.00b

1.50**(1.12-2.02)1.57"

(1.16-2.13)1.09

(0.53-2.26)

Adjusted oddsratio^

(95% confidence interval)

1.00*1.37

(0.99-1.89)1.42*

(1.02-1.98)0.96

(0.43-2,15)

'Adjusted for age (coded directly, in months), height-tbr-age (z scores), breastfeeding (0 = no, 1 = yes), mother's education (in years), religion (0 =Hindu. 1 = Muslim), land ownership (0 = no, 1 = yes), tubewell in courtyard (0 = no, 1 = yes), index age (in years), and season (coded as dummyvariables) in multiple logistic models fitted to the 215 cases and 1267 controls with complete information. Children aged s 36 months classified asnon-breastfed.''Referent group.'Families of nine controls had sanitary latrines, while none of the case families had sanitary latrines.•/• Í 0.05: **P< 0,01, for comparison of cases versus controls.

Modification of the Association by Breastfeeding StatusBecause breastfeeding is documented to be a potentprotective factor against Shigella diarrhoea inBangladesh,5 and because excreta disposal im-provements have been observed to have a substantiallygreater impact on infant mortality among non-breastfedthan among breastfed infants in other settings,14 weexamined the possibility that breastfeeding was an im-portant modulator of the association between latrinesand the risk of Shigella diarrhoea. Among non-breastfedchildren, the association between having a hanging la-trine and paediatric shigellosis, compared to having nolatrine, was substantial (ORa = 1.82, 95% CI: 1.18-2.80,P < 0.01), whereas no association was observed for hav-ing a pit or sanitary latrine (Table 3). The elevated riskassociated with hanging latrines was primarily attribut-able to an increase in the occurrence of persistent Shigellaepisodes (ORa = 3.10,95% CI: 1.55-6.20, P<0,01)ratherthan to an increase in the occurrence of acute episodes ofshigellosis (ORa = 1.44, 95% CI: 0.82-2.52). In contrast,no such associations were seen among breastfed children(ORa for shigellosis = 0,97, 95% CI: 0.55-1.69; ORa forpersistent shigellosis^ 1.18, 95% CI: 0.47-2.97; ORa foracute shigellosis = 0.94, 95% CI: 0.42-2.12).

Other Modifications of the AssociationTo further define the association between hanging la-trines and Shigella in non-breastfed children, severaladditional subgroup analyses were performed. The asso-ciation seemed to increase with increasing age of thechild, with a peak in the fourth year of life (ORa = 3.62,95% CI:1.73-7.57, P < 0.001), though the association

declined in the fifth year (OR, = 1.02,95% Œ0.49-2.13).The association appeared to be somewhat greater if themothers had no schooling (ORa = 2.17, 95% CI: 1.17-4.02, P < 0.05) than if the mothers had some schooling(ORa = 1.58, 95% CL0.86-2.90). The association alsoseemed to be more pronounced during the monsoonseason (ORa = 2.77, 95% Œ 1.09-7.00, P < 0.05) thandurmgthe other seasons (ORa = 1.50, 95% CI:0.55-4.08for the pre-monsoon season; (ORa = 1.68, 95% CI:0.93-3.03 for the post-monsoon season).

DISCUSSIONOur results indicate that having a family latrine wasassociated with an elevated, not a reduced, risk of theoccurrence of shigellosis among rural Bangladeshi chil-dren; that this elevation of risk was attributable both toan apparent lack of protection associated with pit orsanitary latrines and to an elevation of risk associatedwith hanging latrines; and that children weaned frombreastfeeding were particularly vulnerable to the envi-ronmental risk of hanging latrines. The results of aprevious study, although not reaching statistical signifi-cance, also suggest that open latrines increase the risk ofShigella infections among family contacts of Shigellaindex cases.15 Before considering the implications ofthese Findings, several limitations of our study requirediscussion.

Potential LimitationsOur study reflects the experiences of subjects in closeresidential proximity to index cases with culture-con-firmed Shigella dysentery. It is uncertain whether

Page 5: Britain Family Latrines and Paediatric Shigellosis in ... · Occurrence of Shigellosis Table 2 shows the relationship between having a family latrine and the occurrence of shigellosis

860 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY

TABLE 3 Association between latrine type and episodes of shigellosis, by child's diet: Bangladesh, 1987-1989

CasesNo. in stratum(% with cited

feature)

ControlsNo. in stratum(% with cited

feature)

Crudeodds ratio for

stratum(95% confidence interval)

Adjustedodd ratio "for

stratum(95% confidence interval)

Non breastfedNo latrineAny latrine

1.001.73*

(1.13-2.63)1.82"

(1.18-2.80)1.16

(0.44^3.06)Pit or sanitarylatrine

BreastfedNo latrineAny latrine

1.00b

1.06(0.65-1.74)

.09(0.65-1.82)0.91

(0.27-3.07)

1.000,96

(0.55-1.65)0,97

(0.55-1.69)0.82

(0.17-3.86)Pit or sanitarylatrine

aSee Table 2.Referent group.

*P<0.05, "P<0.0l, ***/><0.001.

observations about these intensely exposed children canbe generalized to settings with less intense exposure. The¡ncreased occurrence of shigellosis associated with hang-ing latrines could have been due to a bias if such latrinestended to be used by families that were otherwise moresusceptible to shigellosis. However, this seems unlikelysince the association was restricted only to non-breastfedchildren, and since the overall association persisted de-spite control for several potential confounding variables.Nevertheless, it is still possible that other unmeasuredfactors could have partially accounted for the findings.Bias could also have arisen if shigellosis had been moreintensely detected in children of families having hanginglatrines than those of families having no latrines. Thispossibility seems unlikely for several reasons. Firstly,field workers collected diarrhoeal histories using struc-tured data forms. Secondly, workers were unaware of thestudy hypothesis. Thirdly, the proportion of potentialfollow-up days during which diarrhoeal histories wereobtained were similar among families in the differentlatrine categories. Finally, among non-breastfed chil-dren, the association between having a hanging latrineand the occurrence of severe episodes of shigellosis withvisible blood, which were unlikely to have been missed,remained substantial (ORa = 1.75, 95% CLO.93-3.32).

The purpose of the study was to evaluate whetherfamily latrines interrupted Shigella transmission among

Shigella-sxposed neighbourhoods and therefore it waiconceptually appropriate to include all shigellosis casesdetected among the contacts. However, the inclusion offamilies with multiple shigellosis cases could have crea tedinterpretational problems due to the potential non-inde-pendence of these intra-family episodes. We evaluatedthis potential problem in two ways. Firstly, we repeatedthe analysis including only initial cases occurring in anyparticular family. This analysis was similar to the origi-nal analysis. For example, among non-breastfedchildren, ORa for hanging latrines versus no latrine =1.75 (95% CI: 1.12-2.72, P< 0.05). Secondly, we repeatedthe analysis using a statistical technique that adjustedestimates of variance for odds ratios to account forintra-family clustering of shigellosis cases. As describedearlier, this analysis differed little from the original anal-ysis.

The exclusion of the 227 neighbourhood children withShigella-nega.tive dysentery, which represented shigello-sis cases in which Shigella could not be isolated due tothe vagaries of antibiotic use and specimen process-.ing,8'16 as well as other cases of dysentery, could bequestioned. However, bias due to their exclusion seemsunlikely for the following reasons. Firstly, the differentialuse of antibiotics was unlikely, since among children withshigellosis or culture-negative dysentery at baseline, theingestion of antibiotics during the 7 days prior to baseline

"s : - ; Í .

' (,•;•• T

••?;•:•*•

..¿ai:

• î . ' l

§1

Page 6: Britain Family Latrines and Paediatric Shigellosis in ... · Occurrence of Shigellosis Table 2 shows the relationship between having a family latrine and the occurrence of shigellosis

í ? . s - '

M,-H

FAMILY LATRINES AND PAF.DIATRIC SHIGELLOSIS 861

did not differ significantly by type of family latrine.Secondly, the procedure for specimen collection andprocessing was similar for all children and laboratorypersonnel were unaware of the type of latrine. Althoughrepeating the analysis with the 227 culture-negative dys-entery cases included in the case definition reduced themagnitude of the association (among non-breastfedgroup, ORa for hanging latrines versus no latrine = 1.23,95% CI: 0.86-1.76), we believe that this was most prob-ably due to the heterogeneous aet iology ofculture-negative dysentery among the under-5 year olds.

Biological ImplicationsThe apparent lack of protection associated with pit orsanitary family latrines could suggest that small numbersof such facilities in a highly contaminated environmentare not sufficient to reduce paediatric shigellosis or thatunhygienic behavioural practices may have counteractedthe potential benefits of such latrines.1'17 The associationbetween hanging latrines and paediatric shigellosis prob-ably occurred in part because hanging latrines directlydeposited faeces into, or near, canals and other bodies ofwater, contaminating drinking and cooking water fromthese water sources, and exposing individuals duringswimming, bathing, or other water-related activities. ' 'Although discharge of uncovered contaminated faeceson land by hanging latrines could have promoted trans-mission via mechanical carriage by flies,20 or by directexposure of children playing in these areas, it seems to bean unlikely explanation for our observation because thecomparison group comprised families without latrines.

The apparent increase in the magnitude of the associ-ation between hanging latrines and shigellosis with ageof the child could reflect greater environmental exposureto faecal contamination with increasing mobility. On theother hand, the absence of a demonstrable association inthe fifth year of life could be because of the protectiveeffect of acquired immunity resulting from prior infec-tions.' The stronger association observed in the rainyseason could have been due to monsoon-induced disper-sion of faecal material from the grossly contaminatedhanging latrines, leading to greater contamination ofwater sources. The fact that hanging latrines were asso-ciated with greater increases in the occurrence ofpersistent shigellosis than of acute shigellosis is likelyrelated to the small inoculum required to transmit Shi-gella, and the possibility that increasing the inoculumSKe via environmental contamination has a greater effecton diarrhoeal severity than diarrhoeal incidence.22

Practical ImplicationsThe failure to observe a protective effect of pit or sanitary•atrines, does not preclude a role in interrupting the

transmission of shigellosis, but suggests that these mea-sures must be supplemented by other interventions forthe control of shigellosis. Our data indicate that elimina-tion of unsanitary latrines constitutes a clear publichealth priority. The potential benefits of eradication ofunsafe latrines are especially cogent in view of the de-monstrable association between unsafe latrines andpersistent Shigella diarrhoea, a major cause of malnutri-tion and death among children in less-developed

3 °3

countries. "

ACKNOWLEDGEMENTSThis research was supported by the International Centrefor Diar rhoea l Disease Research, Bangladesh(ÍCDDR,B), and funded by the Diarrhoeal DiseasesControl Program of the World Health Organization andthe US Agency for International Development (USAIDgrant # DPE-5928-A-00-6002-00). ICDDR, B is sup-ported by countries and agencies which share its concernfor the health problems of developing countries. Currentdonors include the aid agencies of the Governments ofAustralia, Bangladesh, Belgium, Canada, Denmark,France, Japan, The Netherlands, Norway, Saudi Arabia,Sweden, Switzerland, the UK, and the US; internationalorganizations including the United Nations Develop-ment Programme, the United Nations Children's Fund,and the World Health Organization; and private foun-dations including the Ford Foundation and theSasakawa Foundation.

The authors gratefully acknowledge the technical as-sistance and support of ICDDR,B staff members. DrIsabelle de Zoysa and Dr David Sack gave helpful advicefor the design and analysis of the study. The Demo-graphic Surveillance System project of ICDDR,Bprovided demographic information.

REFERENCES1 Esrey S A, Feachem R G, Hughes J M. Interventions for the control of

diarrhoeal diseases among young children: improving watersupplies and excreta disposal facilities. Bull World Health Organ1985:63:757-72.

2 Blum D, Feachem R G. Measuring the impact of water supply andsanitation investments on diarrhoeal diseases1, problems of meth-odology. Int J Epidemiol 1983; 12: 357-65.

3 Henry F J. The epidemiologjc importance of dysentery in communities.Rev Infea Dis 1991; 13: (Suppl.4): S238-44.

4 Keusch G T, Bennish M L Shigdlosis: recent progress, persistingproblems and research issues. Pediatr Infect Dis J 1989; 8: 713—19.

5 Ahmed F, Clemens J D, Rao M R, Sack D A, Khan M R, Haque E.Community-based evaluation of the effect of breast feeding onthe risk of microbiologically confirmed or clinically presumptiveshigellosis in Bangladeshi children. Pediatrics 1992; 90:406-11.

6 Ahmed F, Clemens J D, Rao M R, Khan M R, Haque E. Initiation offood supplements and slopping of breast-feeding as determinantsof weanling shigellosis. Bull World Health Organ 1993; 71:571-78.

Page 7: Britain Family Latrines and Paediatric Shigellosis in ... · Occurrence of Shigellosis Table 2 shows the relationship between having a family latrine and the occurrence of shigellosis

862 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY

7 Dibley M J, Staehlitlg N, Nieburg P, Trowbridge F L. Interpretationof Z-score anthropométrie indicators derived from the inter-national growth reference. Am J Clin Nutr 1987; 46: 749-62.

8 Stypulkowska-Misiurewie* H. Problems in bacteriological diagnosis ofshigellosis. In: Rahman M M, Greenough W B, Novae N et al-(eds). Shigellosis: A Continuing Global Problem. Dhaka: Interna-tional Centre for Diarrhoeal Disease Research, Bangladesh,1983. (Special publication no. 20: 87-98).

Edwards P R, Ewing W H . LientijicutionofEnterobacteriaceae. 3 rdedn .New York, NY: Burgess Publishing Company , 1972.

10 Kaplan S L, Feigin R D. Shigellosis (Bacillary Dysentery). In: NelsonW E, Vaughan III V C, Behrman R E (eds). Nelson Textbook ofPediatrics. Philadelphia, PA: W B Saunders Company , 1983pp.670-71.

1 ' Schorling J B, Wanfce C A, Schorling S K, McAulifte J F , Souza M A,Guer ran t R L. A prospective study of persistent d iar rhea a m o n gchildren in an urban Brazilian slum: pat terns of occurrence andetiologic agents. Am J Epidemiol 1990; 132: 144-56.

12 Harris J C, Herbert L D, Hornick R B. Fecal leukocytes in diarrhea!illness. Ann Intern Med 1972; 76: 697-703.

KJeinbaum D , Kuppe r L, Morgenstern H. Epidemiológica!Research.Principles and Quantitative Methods. Belmont, M A : LifetimeLearning Publications, 1983.

14 Butz W P, Habicht J, Davanzo J. Environmental factors in therelationship between breast feeding and infant mortal i ty: the roleof sanitat ion and water in Malaysia. Am J Epidemiol 1984; 119:516-25.

K h a n M, Shahidul lah M . Cont ras t ing epidemiology of Shigella

dysenteriae and Shigellaflexneri, Dacca . Trans R Soc Trop MniHyg 1980; 74: 528-33.

16 Butler T, Mahmoud A A F, Warren K. S. Algori thms in the d iagnomand management of exotic diseases. XXVTÍ, Shi^ellosis. J InftctDis 1977; 136:465-68.

17 Briscoe J. Intervention studies and the definition of dominant tram-mission routes. Am J Epidemiol 1984; 120: 449-55.

18 Rajasekaran P, Dut t P R, Pisharoti K A. Impact of water supply onthe incidence of diarrhoea and shigellosis among children in ruralcommunities in Madurai . Indian J Med Res 1977; 66: 189-99.

19 Makintubce S, Mallonee J, Istre Ci R. Shigellosis outbreak associatedwith swimming. Am J Public Health 1987; 77: 166-68.

Cohen D, Green M, Block C et al. Reduct ion of transmission ofshigellosis by control of houseflies (Musca domestica). Lancet1991; 337: 993-97.

' Cruickshank R. The influence of age and nutri t ion on the incidenceand control of enteric infections. Med Clin North Am 1967; 51:643-52.

22 Black R E. Brown K H, Becker S. Malnutr i t ion is a determining factorin diarrheal durat ion, but not incidence, among young childrenin a longitudinal study in rural Bangladesh. AmJClinNutr 1984;39: 87-94.

23 Persistent diarrhoea in children in developing countries: Memorandumfrom a WHO meeting. Bull World Health Organ 1988; 66: 709-17.

(Revised version received December 1993)

Il. . . * • •