A household survey of dysentery in Burundi: implications for the current pandemic in sub-Saharan Africa.

A household survey of dysentery in Burundi: implications for the current pandemic in sub-Saharan Africa M.E. Birmingham,1 L.A. Lee,2 M. Ntakibirora,3 F. Bizimana,4 & M.S. Deming5 To characterize the epidemiology of dysentery (defined as bloody diarrhoea) in Burundi, we reviewed national surveillance data and conducted a household cluster survey including two case-control studies: one at the household, the other at the individual level. We estimated that community incidences for dysentery (per 1000 residents) in Kibuye Sector were 15.3 and 27.3, and that dysentery accounted for 6% and 12% of all deaths, in 1991 and 1992, respectively. Factors associated (P - 0.05) with contracting dysentery were being female, using a cloth rag after defecation, a history of recent weight loss, and not washing hands before preparing food. The attributable risk, at the household level, of not washing hands before preparing food was 30%. Secondary household transmission accounted for at most 1 1% of dysentery cases. This study suggests that Shigella dysenteriae type 1 may be one of the leading causes of preventable mortality in Burundi and other African countries where effective antimicrobial agents are no longer afford- able. Since hands were the most important mode of transmission of S. dysenteriae in this study, community- based interventions aimed at increasing hand washing with soap and water, particularly after defecation and before food preparation, may be effective for controlling dysentery epidemics caused by S. dysenteriae type 1 in Africa. Introduction Epidemics of dysentery caused by infection with Shigella dysenteriae type I (1) are occurring in an increasing number of countries in eastern, central, and southern Africa. One of the earliest documented epidemics caused by S. dtysenteriae type 1 in Africa began in what is now Burundi in 1943 and spread to what is now Rwanda, and surrounding countries, be- fore ending in 1952 (2). Few cases were reported from the region until 1979. when a second epidemic began in Zaire, spread in 1981 to Rwanda and Burundi, in 1982 to the United Republic of Tanza- nia, and in 1990 to countries in southern Africa where there had not previously been a problem (3- 7). Dysentery remains an important public health problem in these countries. International Health Program Office, Centers for Disease Control and Prevention, Atlanta, GA, USA. Currently: Global Programme for Vaccines and Immunization, World Health Organization, 1211 Geneva 27, Switzerland. Correspondence should be sent to Dr Birmingham at the latter address. 2 International Health Program Office, Centers for Disease Control and Prevention, Atlanta, GA, USA. 3Ministry of Public Health, Bujumbura, Burundi. 4Ministry of Public Health, Bujumbura, Burundi. I International Health Program Office, Centers for Disease Control and Prevention, Atlanta, GA, USA. Reprint No. 5753 Although dysentery caused by S. dysenteriae type 1 infection is a growing problem in sub-Saharan Africa, the epidemiology of dysentery during the most recent pandemic has not been well character- ized. Most publications have described earlier Afri- can epidemics or the experience in Asia and Latin America. During epidemics in Zaire and Rwanda in the early 1980s, community incidences of dysentery ranged from 5% to 6.4% (8, 9). In Latin America and Asia, the incidences of dysentery caused by S. dysenteriae type 1 infection have generally been highest in young children, and show few sex-specific differences (10-14). In epidemics in Asia, Latin America, Zaire, and Rwanda, case-fatality rates ranged from I % to 7% and were highest among young children and the elderly (15). A study con- ducted in 1992 showed that 7.2% of 795 patients with dysentery seen at nine government health centres in Burundi died; case-fatality rates were the same for children and adults (16). Treatment with an effective antimicrobial can reduce the duration and severity of dysentery; how- ever, increasing levels of antimicrobial resistance in S. dysenteriae type 1 strains are a global problem. In many African countries there are no longer afford- able treatment options. During the 1979 pandemic that started in central Africa, initial isolates of S. dysenteriae type I from patients with dysentery were resistant to ampicillin but sensitive to trimethoprim- Bulletin of the World Health Organization, 1997, 75 (1): 45-53 ©c World Health Organization 1997 45 M.E. Birmingham et al. Fig. 1. Number of monthly health facility visits for dysentery, Burundi, 1980-1993, from national surveillance data. 30 25 S o 20 x X 15 ; 10 6 z 5 0 1980 1982 1984 1986 1988 1990 1992 Year WHO 96876 sulfamethoxazole or nalidixic acid (17). By 1982, plasmid-mediated resistance to trimethoprim- sulfamethoxazole and nalidixic acid had been docu- mented, and by 1990, S. dysenteriae type 1 isolates from this region were highly resistant to all three antimicrobial agents (3, 5, 18). Therapeutic complications caused by increasing antimicrobial resistance have highlighted the need for effective preventive strategies, but little is known about the primary modes of transmission of S. dysenteriae type 1 in Africa. To date, there have been no controlled field trials in Africa to evaluate the effectiveness of specific preventive interventions. Strategies to limit secondary household transmission may be effective if such transmission is high. In Burma (now Myanmar), secondary cases accounted for 19% of individuals with dysentery in sampled households of index cases (11). In Lusaka, Zambia, having a family member ill with dysentery and using a cup to obtain drinking-water were risk factors for dysentery during an outbreak of S. dysenteriae type 1 in 1992 (6). In a case-control study of patients with dysentery seen at clinics in Burundi, such patients were more likely than those with non-gastro intesti- nal illnesses to have had recent contact with some- one else with dysentery and to have had another episode of diarrhoea during the past year (5). Burundi, a small, densely populated country in the interior of east-central Africa bordered by Rwanda, the United Republic of Tanzania, and Zaire, is one of the foci of the current African dysen- tery pandemic, with seasonal peaks documented since 1980 (Fig. 1). A study conducted in Gitega Province in 1990 demonstrated that most isolates from patients with dysentery were S. dysenteriae type 1 (5). The majority of the population in Burundi live in rural areas, and an estimated 75% live within 6km of a health facility. National disease surveillance at the National Epidemiology and Statistics Service of the Ministry of Public Health compiles monthly re- ports of the number of health facility visits for spe- cific diseases. To determine the magnitude of dysentery as a cause of morbidity and mortality and to identify risk factors for dysentery in Burundi, we reviewed health-facility surveillance data for dysen- tery and conducted a community survey and two case-control studies in Kibuye Sector. Materials and methods In February 1992, a household cluster survey was conducted in Kibuye Sector, Gitega Province, Burundi. A two-stage sampling scheme was used. First, 30 of Kibuye Sector's 134 administrative units (collines) were chosen with a probability of selection proportional to the number of residents enumerated during the 1990 census. A total of 68 households were then chosen in each selected colline by selecting one household at random from a list of households in the colline and following a next- nearest-household path to 67 additional households. All members of each selected household were in- cluded in the survey sample, which was analysed as a representative sample of households and residents in Kibuye Sector. Surveyors visited selected house- holds to interview an adult family member. If no adult family member was available after two visits, the household nearest to the last household surveyed was substituted. Information on dysentery and potential risk fac- tors for dysentery was obtained by recall for all WHO Bulletin OMS. Vol 75 1997 - 46 Household survey of dysentery in Burundi household members for the 7-month period 1 July 1991 to 31 January 1992. These dates were chosen for ease of recall. The former is Independence Day and the first day of summer school vacation and the latter was in the week preceding the start of the survey. Interviews were conducted in Kirundi. To avoid using the Kirundi word for dysentery, which has shameful connotations, surveyors identified affected individuals by asking whether any household mem- bers had experienced bloody diarrhoea during the recall period. A positive household was defined as one in which one or more family members had experienced bloody diarrhoea during the recall period. Two case-control studies were conducted to identify personal and household risk factors for dys- entery. For each affected individual (and household) identified, a control person and a control household were selected for comparison. The control person was the person closest in age to the affected indi- vidual who did not have dysentery during the recall period and who lived in the same household. The control household was the nearest household in the adjacent family compound in which no one had dys- entery during the recall period. Exposure to poten- tial risk factors for dysentery for case and control patients and households was determined using a standardized data collection instrument. The community incidence of dysentery (defined as bloody diarrhoea) obtained from the survey was compared with the rate of health centre visits in Kibuye Sector for bloody diarrhoea during the same period (which included visits for "bacillary dysen- tery" and "amoebiasis"). The ratio of the two rates was used to estimate community-based dysentery incidence before the recall period. Population figures used for calculating incidences were based on national census data from 1990 and a projected 2.62% annual population growth rate (19). Crude mortality rates used for 1991 and 1992 were estimates made by the United Nations Popu- lation Division (20, 21). The case-fatality rate for patients seen in health facilities was obtained from a study conducted in 1992 in an adjacent province of patients with dysentery seen at health facilities (16). Point estimates and their standard errors from the community survey were calculated using SUDAAN statistical software (22). Univariate analyses of matched case-control data were per- formed using the Mantel-Haenszel weighted odds ratio (23). Multivariate analyses of matched case- control data used a stepwise multiple logistic regres- sion procedure that controlled for age and sex (24). An estimate subjected to a two-sided test yielding a P value <0.05 was considered to be statistically sig- nificant. Attributable risks were calculated according to the method of Bruzzi et al. (25). Results From January 1980 to October 1993, dysentery in Burundi was characterized by seasonal peaks, with the highest number of health facility visits reported between October and January (Fig. 1). During 1992, the nationwide number of health facility visits for dysentery increased dramatically compared with previous years, and showed a larger-than-normal seasonal peak during October and November. Data collection by means of the routine surveillance sys- tem was interrupted in October 1993 because of civil conflict. Information was collected for 9319 residents of 2054 households in Kibuye Sector (estimated 1991 population: 227575). For the 7-month recall period, 127 persons were identified with dysentery, yielding a 7-month incidence of 13.6 cases per 1000 persons (95% confidence interval: 9.5-17.7). Age-specific incidences had a bimodal distribution. The highest rates were seen in children <5 years of age and adults ¢'30 years of age; rates increased with age among adults (Fig. 2). Incidences for the 7-month period were significantly higher for females than for males (14.6 vs 9.9 cases per 1000 persons, P = 0.03). The 127 affected individuals resided in 113 house- holds; 99 of which had one case and 14 of which had two cases. If it is assumed that exposure to a pri- mary case was the only source of infection for the second affected individual in the household, intra- household transmission accounted for at most 11% of the 127 cases of dysentery. This is likely to be an Fig. 2. Dysentery incidences, by age group, with 95% confidence limits, Kibuye Sector, Burundi, 1 July 1991-31 January 1992. 1- U)CD o 80 0. 60 X 40 .D 20 <1 1-4 5-14 15-29 30-59 Age group (years) .60 WHO 96877 WHO Bulletin OMS. Vol 75 1997 +0,X~=~ 47 M.E. Birmingham et al. overestimate, as some of these individuals may have been infected outside the household or have been co-primary cases. As there were 374 potential cases for the households of index cases, the secondary at- tack rate was r3.7% (14/374). In addition to bloody diarrhoea, dysentery in the 127 affected individuals was characterized by abdominal cramps (92%), tenesmus (92%), and fever (54%). The median duration of illness was 3 days; 97 individuals (76%) sought care at a health facility and 23 (18%) were hospitalized. Of the 97 individuals who sought care at a health facility, 63 (65%) visited the facility more than once for the same dysentery episode. The reason most frequently cited for not seeking care at a health facility was lack of money (70%). There was little difference with regard to age, sex, or clinical presentation between affected individuals who did and who did not seek care, with the exception that those who did not seek care at a health facility were more likely than those who did to report having large quantities of blood in their stools (67% vs 47%), although this difference was not significant. A total of 102 cases were matched to a control in the same household (the remainder had no available household-matched control). The proportions of cases and controls with exposure to potential risk factors are shown in Table 1. Cases were more likely than controls to have a history of recent weight loss before dysentery onset (odds ratio (OR) = 7.0, P = 0.003) and to use a cloth rag for anal cleansing fol- lowing defecation (OR = 13.0, P = 0.002). No spe- cific activities during the week before the onset of illness were associated with cases, including attend- ance at a gathering or funeral, travel outside the home, or contact with someone with dysentery. Cases and controls did not differ with regard to hand-washing practices. In a multivariate regression controlling for age and sex, a history of recent weight loss (OR = 1.81, P = 0.02) and use of a cloth rag following defecation (OR = 12.2, P = 0.003) were significantly associated with having dysentery. The estimated attributable risks for weight loss and using a cloth rag were 2% and 5%, respectively. Characteristics of case and control households in adjacent family compounds are shown in Table 2. Members of case households were more likely than control households not to wash their hands routinely before preparing food (79% vs 66%, P = 0.05). Case households were more likely not to have soap, but the difference was not statistically significant (29% vs 19%, P = 0.09). The types of container used to transport and store water, the presence and use of a latrine and other household hygienic practices were not associated with household members having dys- entery. The median amount of water transported daily to the household for drinking, cooking, and bathing was 5 litres per household member, and showed no significant difference between case and control households. In a multivariate regression analysis that controlled for age and sex, not washing hands before preparing food was a significant risk factor for households (OR = 1.8, P = 0.05), with an attributable risk of 30%. Table 1: Exposure to potential dysentery risk factors among 102 household-matched case-control pairs, Kibuye Sector, Burundi, 1 July 1991-31 January 1992 Potential risk factors reported: Weight loss during the 3 months prior to illness During the week prior to illness having: Gone to a party Gone to a market Attended a funeral Cared for a sick person Travelled outside the home Had contact with someone with dysentery Generally not having washed hands: Before eating Before preparing food After defecating Having used for anal cleansing after defecating: Leaves Cloth rag Paper a Figures in parentheses are percentages. Exposure: Mantel-Haenszel - weighted Cases Controls odds ratio 19(19) 56(55) 5(5) 4(4) 21(21) 13(13) 26(25) 62(61) 4(4) 5(5) 18(18) 9(9) 21(21) 14(14) 78(76) 77(75) 85(83) 81(79) 79(77) 78(76) 19(19) 7(7) 10(10) 15(15) 7.0 0.5 0.7 1.3 0.8 1.3 3.0 2.2 1.1 1.8 1.1 13.0 0.4 95% Confidence interval 1.6-63.5 0.1-1.3 0.3-1.5 0.2-9.1 0.1-4.4 0.5-3.1 0.8-17.2 0.8-7.0 0.4-2.7 0.5-6.8 P-value 0.003 0.11 0.32 0.50 0.71 0.55 0.08 0.11 0.83 0.29 WHO Bulletin OMS. Vol 75 1997 0.4-3.1 0.82 2.0-552.5 0.002 0.1-1.6 0.17 18(18)- 7(7) 48 Household survey of dysentery in Burundi Table 2: Exposure to potential dysentery risk factors among 113 matched case-control household pairs, Kibuye Sector, Burundi, 1 July 1991-31 January 1992 Potential risk factors reported: Not having soap in the house at time of interview Household members not having washed hands: Before eating Before preparing food After defecating Having used to carry household water: 40-I plastic jug Narrow-mouthed clay pot Wooden trough Having used to store household water: 40-I plastic jug Narrow-mouthed clay pot Bucket Having transported -_51 of water per household member to the household daily Household members having defecated: In a latrine In a shallow hole In bushes a Figures in parentheses are percentages. Exposure: Mantel-Haenszel Case Control weighted households households odds ratio 33(29)a 22(19) 4(4) 89(79) 84(74) 65(58) 30(27) 23(20) 63(56) 34(30) 9(8) 56(50) 96(85) 13(12) 4(4) 6(5) 75(66) 95(84) 67(59) 26(23) 23(20) 66(58) 33(29) 1 0(9) 57(50) 98(87) 1 0(9) 5(4) 1.7 0.7 1.8 0.5 0.9 1.3 1.0 0.9 1.1 0.9 1.0 0.9 1.3 0.8 During the recall period, 3087 health centre vis- its for dysentery in Kibuye Sector were reported to the national surveillance system, i.e. 17.7 visits per 1000 residents. On the basis of the community inci- dence of 13.6 cases per 1000 persons obtained from the survey, during tI,e recall period there were 0.77 dysentery cases for every health centre visit for dys- entery. This ratio was used to estimate annual com- munity incidences (Table 3). Numbers of health centre visits for dysentery reported from Kibuye Table 3: Estimated dysentery incidence, mortality rate, and proportional mortality from dysentery, Kibuye Sector, Burundi, in 1991 and 1992 Variable 1991 1992 Mid-year populationa 227575 233537 No. of health facility visits for "bacillary dysentery" and "amoebiasis"b 4 513 8 289 No. of reported dysentery cases in the community No. of health facility visitsc 0.77 0.77 No. of reported dysentery cases in the communityd 3475e 6383e Community dysentery incidence (per 1000)' 1 5.3e 27.3e No. of dysentery-associated deathsg 250e 460e Dysentery mortality rate (per 1000)h 1.1e 2.0e Crude mortality rate (per 1000)' 17 17 % of deaths due to dysentery' 6e 12e a Projected from 1990 census data and a 2.62% annual growth rate. b No. of health facility visits as reported to national surveillance system. c Ratio = 13.6 reported cases per 1000 residents: 17.7 visits per 1000 residents. d 1991: 4513 x 0.77 = 3475; 1992: 8289 x 0.77 = 6383. e Estimated. 1991: 3475 - 227575 = 15.3; 1992: 6383 - 233537 = 27.3. 9 Case-fatality rate = 7.2% (16). 1991: 3475 x 7.2% 250; 1992: 6383 x 7.2% 460. h 1991: 250 + 227575 = 1.1; 1992: 460 - 233537 2.0. Figures from United Nations Population Division (20, 21). '1991:1.1 17 = 6%; 1992: 2.0 17 = 12%. WHO Bulletin OMS. Vol 75 1997 95% Confidence interval 0.9-3.4 0.1-2.8 1.0-3.4 0.3-1.1 0.5-1.8 0.6-3.1 0.5-2.0 0.5-1.7 0.5-2.3 0.3-2.6 0.6-1.6 0.4-1.9 0.5-3.3 0.2-3.7 P-value 0.09 0.53 0.05 0.07 0.74 0.44 1.00 0.21 0.86 0.81 0.90 0.72 0.53 0.74 49 M.E. Birmingham et al. Fig. 3. Number of monthly health centre visits for dys- entery, Kibuye Sector, Burundi, 1991 and 1992, from national surveillance data. 2 000 1 500 Un U11, _D -1 0000 6 z 500 J FMAM J J ASO ND J FMAM J J ASO ND 1991 1992 WHO 96878 Sector to the national surveillance system increased from 4513 (1991) to 8289 (1992) (Fig. 3), reflecting the concurrent nationwide increase. Using the ratio of community dysentery incidence to health centre visits, we estimated that community incidence rates for Kibuye Sector in 1991 and 1992 were 15.3 and 27.3 cases per 1000 residents, respectively. To calculate the dysentery mortality rate and the proportion of deaths due to dysentery in Kibuye Sector during 1991 and 1992, data from other sources were used (Table 3). Multiplying the number of cases of dysentery in Kibuye Sector by the case- fatality rate for patients with dysentery seen at health facilities in an adjacent province in 1992 (7.2%), we estimated that 250 deaths were associ- ated with dysentery during 1991 and 460 deaths dur- ing 1992, yielding a dysentery mortality rate of 1.1 deaths per 1000 persons per year during 1991, and 2.0 deaths per 1000 persons per year during 1992. With the assumption of a crude mortality rate of 17 deaths per 1000 persons for both 1991 and 1992 in Burundi (19), dysentery accounted for approxi- mately 6% of deaths in 1991 and 12% of deaths in 1992. Discussion A possible limitation of this study is the 7-month period of recall, which may have resulted in under- reporting of dysentery episodes. However, as the survey was conducted shortly after the 1991 seasonal dysentery peak in Burundi, the median recall time was less than 7 months. The reported symptoms and health-care behaviour of dysentery patients did not vary significantly with month of onset, and the sea- sonal pattern of dysentery visits reported to the national surveillance system closely paralleled the pattern of dysentery onset in the community survey. Furthermore, the dysentery incidence obtained in the survey was only 6% greater than that derived from visits reported to the national surveillance sys- tem, after adjusting for the proportion of patients who sought care and the proportion who made more than one visit. In addition, the severe symptoms as- sociated with dysentery make it less subject to recall bias than other common illnesses. The incidences of dysentery found in this study are consistent with those reported from other S.- dysenteriae-type-1-endemic countries in Africa dur- ing epidemics in the early 1980s (8, 9). Incidences as high as 33% have been reported in some countries, generally for localized areas during an epidemic peak (12, 13). The highest incidences in Kibuye Sec- tor occurred among adults and women. Other com- munity studies of shigellosis caused by S. dysenteriae type 1 have generally shown the highest incidences among young children. In Zaire during the 1981- 1982 seasonal peak, dysentery incidences based on hospitalizations were higher for women than men, and higher for adults than children (9). In Teknaf, Bangladesh, incidences for shigellosis due to S. dysenteriae type 1 and S. flexneri were higher among women than men but showed little age-specific dif- ferences (26). Women and girls may be at a higher risk than men of contracting shigellosis because of the potential for exposure to infected faeces when caring for young children (4). In contrast, during S. dysenteriae type I epidemics in Guatemala during 1969-1970 and in Rangoon, Burma, in 1984-85, com- munity-based studies have found no increased risk of morbidity for adults and no differences in incidence rates by sex (10, 11). Secondary household transmission accounted for a small proportion of the dysentery cases in Kibuye Sector. This finding is similar to that of other studies of S. dysenteriae type 1, but the secondary transmission rate found in this study is much lower than that reported for other Shigella serogroups in developed countries. During an epidemic of S. dysenteriae type I in Rangoon, Burma, in 1984-85, secondary cases accounted for 19% of all affected individuals in 109 randomly selected households of index cases (11). During the 1992 S. dysenteriae type 1 outbreak in Zambia, only 6 of 42 cases (14%) were found in a household in which another household member had had dysentery during the month before the index case became ill (6). In contrast, secondary household transmission accounted for 39% of all dysentery cases in a study of S. sonnei infections in the USA (27). Differences in rates of intra- household transmission of S. dysenteriae type 1 com- WHO Bulletin OMS. Vol 75 1997 7-month recall period - 1 start of survey . , * r11 11 ' ..-1 I. ,1 I11 I. .J.I. 50 Household survey of dysentery in Burundi pared to those for S. sonnei may be due to the higher rates of secondary transmission found when the in- dex case is a young child (27). In the USA, 40% of index cases with S. sonnei were children <5 years of age, whereas in Kibuye Sector, children <5 years of age accounted for only 17% of cases. Other possi- ble explanations for the relatively low proportion of secondary cases in Burundi include hygiene educa- tion given to dysentery patients at health centres and high levels of immunity due to subclinical infection. Our findings suggest that efforts to prevent second- ary household transmission will have little impact in Burundi. Unpredictable, explosive epidemics exceeding normal seasonal increases have been observed in several countries where infections caused by S. dysenteriae type 1 are endemic (8-10, 15, 26). The cause of these aperiodic increases is unknown but may be fluctuations in the immunity of the host population. Epidemiological evidence and challenge studies in primates indicate that there is incomplete immunity of unknown duration following infection with Shigella spp. (28). Other possible causes include changes in living conditions that facilitate the spread of infection or the introduction of a strain (i.e. serotype variant) to which the host population has little acquired immunity. Pulse-field gel electrophoresis of S. dysenteriae type 1 indicates that strains isolated from patients in Burundi in 1993 may have been introduced from Zambia or Mozambique (29). However, there is now considerable evidence that protective immunity to Shigella spp. observed after vaccination or natural infection is due primarily to response to the LPS antigen, which is serotype rather than strain specific (30). This study suggests that dysentery may have ac- counted for as much as 6% of deaths in Burundi during 1991 and 12% during 1992. Even higher mor- tality associated with dysentery has been reported in refugee camps with severe crowding and poor sani- tary conditions. In a 1994 survey of newly arrived Rwandan refugees in a refugee camp in Zaire, dys- entery caused by S. dysenteriae type 1 accounted for 40% of all deaths (18). In the Matlab region of Bang- ladesh, dysentery accounted for 36% of deaths in children aged 1-4 years during the 1983-1984 S. dysenteriae type 1 epidemic (15). The high propor- tional mortality found here is striking, in that it suggests dysentery is one of the leading causes of preventable mortality in an S.-dysenteriae-type-1- endemic African country, outside of a refugee or emergency context. Our estimates of mortality rates and proportional mortality were based on the as- sumption that those who did and those who did not seek health care experienced the same mortality. Even if no deaths occurred among those not seeking care, our estimate of proportional mortality would be reduced by only 24%. A recent history of weight loss was strongly as- sociated with contracting dysentery in this study. A study of dysentery patients in Peru found lower lev- els of IgA to Shigella invasion plasmid antigen (Ipa) among malnourished children than well nourished children, which suggests that higher mucosal immu- nity to this antigen in well nourished individuals lim- its the severity and spread of Shigella disease (31). Acquired immunodeficiency syndrome (AIDS) in Africa has been characterized by decreased cellular immunity and profound weight loss and may also increase the risk for infection with S. dysenteriae type 1. These explanations could provide an immunologi- cal basis for our findings but need to be confirmed through additional studies. Not washing hands before preparing food was the factor with the highest attributable household risk in this study. Hand washing with soap and water had a significant impact on reducing secondary household transmission of Shigella in Bangladesh (32); however, the impact of the intervention on reducing transmission of S. dysenteriae type I was much lower than for other Shigella serotypes. In a randomized hand-washing intervention in Burma, a group of children <5 years of age and their mothers were provided with soap and asked to wash their hands after defecating and before preparing or eat- ing meals; incidences of dysentery in the hand- washing group were 40% lower than in the control group, but this difference was not significant at the chosen level (33). Although rates of secondary household transmission in Burundi appear low, the promotion of hand washing with soap and water may reduce the rate of primary infection by preventing the contamination of household food and water prior to consumption. S. dysenteriae type 1 infections are increasing in incidence and spreading geographically in sub-Saha- ran Africa. S. dysenteriae is an important cause of morbidity and may be one of the leading causes of preventable mortality in Burundi, as well as in other S.-dysenteriae-type-1-endemic African countries. Antimicrobials that are effective against resistant strains of S. dysenteriae type 1 are no longer afford- able in many endemic countries. Although several candidate vaccines show promise, an efficacious and affordable vaccine suitable for wide-scale use in de- veloping countries is unlikely to be available in the near future. In this study, secondary household transmission contributed little to the spread of the illness. Transmission of S. dysenteriae by hands was the most important mode. These findings suggest that community-based interventions aimed at inter- rupting faecal-oral transmission by increased hand WHO Bulletin OMS. Vol 75 1997 51 M.E. Birmingham et al. washing with soap and water, particularly after def- ecation and before food preparation, may be effec- tive measures for controlling dysentery epidemics caused by S. dysenteriae type 1 in Africa. Controlled field trials to assess the effectiveness of specific pre- ventive interventions are an urgent public health priority. Acknowledgments We thank Dr T. Wyant and Ms N. Mendoz for assistance in data analysis; Dr B. Hersh and Dr S. Nkurikiye for providing surveillance data; the residents of Kibuye Sector for participating in this survey; and the 12 interviewers who collected the data. Resume Enquete dans les menages sur la dysenterie au Burundi et pandemie actuelle en Afrique subsaharienne Le Burundi est l'un des foyers de d6part de la pand6mie actuelle de dysenterie a Shigella dysen- teriae type 1 en Afrique subsaharienne. Afin de caract6riser l'6pid6mie de dysenterie au Burundi, nous avons examin6 les donn6es nationales de surveillance et effectu6 une enquete par sondage dans les menages avec deux 6tudes cas-t6moins, l'une au niveau des m6nages et l'autre au niveau de l'individu. Sur les 9319 habitants enquet6s dans le secteur de Kibuye, 127 ont declar6 avoir souffert de dysenterie (d6finie comme diarrh6e sanglante) entre le 1er juillet 1991 et le 31 janvier 1992. Les taux d'incidence par age presentaient une distribu- tion bimodale, les taux les plus eleves s'observant chez les enfants de 1 a 4 ans et chez les adultes a partir de 30 ans. Chez ces derniers, l'incidence augmentait avec l'age. Pendant la p6riode de rappel s'6tendant sur 7 mois, l'incidence 6tait significativement plus forte chez les sujets de sexe f6minin que chez les sujets de sexe masculin (14,6 cas pour 1000 contre 9,9 pour 1000, p = 0,03). Sur les 97 malades trait6s dans un etablissement de soins, 63 (65%) sont venus a plusieurs reprises pour le meme 6pisode de dysenterie. La raison la plus fr6quemment cit6e par les malades n'ayant pas consulte 6tait le manque d'argent (70%). 11 y avait peu de diff6rences entre les malades ayant consulte et les autres en ce qui concerne l'age, le sexe ou le tableau clinique, sauf que les malades n'ayant pas consult6 rapportaient plus fr6quem- ment la pr6sence de quantites importantes de sang dans les selles (67% contre 47%). Dans la communaut6 6tudi6e, il y a eu 0,77 cas de dysenterie par visite pour ce motif a un centre de sante enregistr6e dans les donn6es nationales de surveillance. D'apres ce rapport, et en prenant le taux de letalit6 de 7,2% obtenu dans une 6tude r6alis6e dans une province contigue, nous avons estim6 l'incidence de la dysenterie dans la com- munaut6 du secteur de Kibuye a 15,3 cas pour 1000 habitants en 1991 et 27,3 cas pour 1000 en 1992, et que la dysenterie peut avoir 6t6 a l'origine de 6% des d6ces enregistr6s au Burundi en 1991 et 12% en 1992. Les facteurs de risque associ6s a la dysenterie (p S 0,05) 6taient I'appartenance au sexe f6minin, l'utilisation d'un chiffon en tissu pour s'essuyer apres la d6f6cation, des ant6c6dents de perte de poids r6cente, et le non-lavage des mains avant de pr6parer les repas. Au niveau des m6- nages, le risque attribuable 6tait de 30% pour le non-lavage des mains avant de pr6parer les repas. La transmission secondaire au sein du foyer etait responsable d'au maximum 1 1% des cas. Cette 6tude laisse a penser que S. dysenteriae type 1 peut etre l'une des principales causes de mortalit6 6vitable au Burundi ainsi que dans d'autres pays africains ou les antimicrobiens efficaces contre cet agent pathogene ne sont plus accessibles aux malades pour des raisons de coOt. Comme, dans cette 6tude, il apparait que S. dysenteriae est principalement transmis par les mains des personnes contamin6es, des inter- ventions a base communautaire ax6es sur la pro- motion du lavage des mains a l'eau et au savon, en particulier apres la d6f6cation et avant de preparer les repas, peuvent etre des mesures efficaces de lutte contre les epid6mies de dysenterie dues a S. dysenteriae type 1 en Afrique. La r6alisation d'essais pratiques contr6les de telles interventions est une priorit6 de sant6 publique. References 1 Shiga K. The trend of prevention, therapy and epide- miology of dysentery since the discovery of its causa- tive organism. New England journal of medicine, 1936, 215: 1205-1211. 2. Jadin J, Resseler J. La dysenterie bacillaire au Ruanda-Urundi et au Kivu. Annales de la Societ6 belge de M6decine tropicale, 1957, 37: 347-369. 3. Frost JA, Rowe B, Vandepitte J. Acquisition of trimethoprim resistance in an epidemic strain of Shigella dysenteriae type 1 from Zaire. Lancet, 1982, 1: 963. 4. Mhalu FS, Moshi WK, Mbaga I. A bacillary dysentery epidemic in Dar es Salaam, Tanzania. Journal of diarrhoeal diseases research, 1984, 2: 217-222. 5. Ries AA et al. Epidemic Shigella dysenteriae type 1 in Burundi: panresistance and implications for preven- 52 WHO Bulletin OMS. Vol 75 1997 Household survey of dysentery in Burundi tion. Journal of infectious diseases, 1994,169:1035- 1041. 6. Tuttle J et al. Antimicrobial-resistant epidemic Shigella dysenteriae type 1 in Zambia: modes of transmission. Journal of infectious diseases, 1995, 171: 371-375. 7. Guidelines for the control of epidemics due to Shigella dysenteriae 1. Geneva, World Health Organization, 1995 (unpublished document WHO/CDR/95.4; avail- able upon request from the Division of Child Health and Development, World Health Organization, 1211 Geneva 27, Switzerland). 8. Huppertz HI. An epidemic of bacillary dysentery in western Rwanda 1981-1982. Central African journal of medicine, 1986, 32: 79-82. 9. Malengreau M et al. Outbreak of Shigella dysentery in eastern Zaire, 1980-1982. Annales de la Societ6 belge de M6decine tropicale, 1983, 63: 59-67. 10. Gangarosa EJ et al. Epidemic Shiga bacillus dysen- tery in Central America. II. Epidemiologic studies in 1969. Journal of infectious diseases, 1970, 22: 181- 190. 11. Han AM, Aye T, Hlaing T. An outbreak of dysentery due to Shigella dysenteriae type 1 in Rangoon, Burma. Journal of diarrhoeal diseases research, 1987, 5: 30-35. 12. Mathan VI et al. Epidemic dysentery caused by the Shiga bacillus in a southern Indian village. Journal of diarrhoeal diseases research, 1984, 2: 27-32. 13. Rahaman MM et al. An outbreak of dysentery caused by Shigella dysenteriae type 1 on a coral island in the Bay of Bengal. Journal of infectious diseases, 1975, 132: 15-19. 14. Sengupta PG et al. Multidrug resistant epidemic shigellosis in a village in west Bengal, 1984. Indian journal of public health, 1990, 34: 15-19. 15. Bennish ML, Wojtyniak BJ. Mortality due to shigellosis: community and hospital data. Reviews of infectious diseases, 1991, 13(suppl. 4): S245- S251. 16. Murray JCS et al. Mortality from dysentery, Burundi. Paper presented at: Forty-third Annual Epidemic Intel- ligence Service Conference, 18-22 April 1994, Atlanta, GA, USA. 17. Rogerie F, Vimont-Vicary P. Etude bacteriologique des shigelloses dans la r6gion du lac Kivu (Afrique centrale). Evolution au cours des quinze dernieres annees (1968-1983). Bulletin de la Soci6t6 de Pathologie exotique et de ses filiales, 1986, 79: 435- 446. 18. Goma Epidemiology Group. Public health impact of Rwandan refugee crisis: what happened in Goma, Zaire, in July, 1994? Lancet, 1995, 345: 339-344. 19. World population prospects 1994. New York, United Nations, 1995 (Population Studies No. 145). 20. State of the world's children 1993. Oxford, Oxford University Press, 1993. 21. State of the world's children 1994. Oxford, Oxford University Press, 1994. 22. Shah B et al. Survey data analysis (SUDAAN), ver- sion 6.0. Research Triangle Park, NC, USA, Research Triangle Institute, 1992. 23. Robins J, Greenland S, Breslow NE. A general es- timator for the variance of the Mantel-Haenszel odds ratio. American joumal of epidemiology, 1986, 124: 719-723. 24. SAS system, release 6.04. Cary, NC, USA, SAS Insti- tute, 1990. 25. Bruzzi P et al. Estimating the population attributable risk for multiple risk factors using case-control data. American journal of epidemiology, 1985, 122: 904- 914. 26. Hossain MA, Albert MJ, Hasan KZ. Epidemiology of shigellosis in Teknaf, a coastal area of Bangladesh: a 10-year survey. Epidemiology and infection, 1990, 105: 41-49. 27. Wilson R et al. Family illness associated with Shigella infection: the inter-relationship of age of the index patient and the age of household members in acquisition of illness. Journal of infectious diseases, 1981, 143: 130-132. 28. Hale TL, Keren DF. Pathogenesis and immunology in shigellosis: applications for vaccine development. Current topics in microbiology and immunology, 1992, 180:117-137. 29. Strockbine NA et al. Analysis of Shigella dysenteriae 1 strains by pulsed-field gel. In: The 94th General Meeting of the American Society of Microbiology, 23- 27 May 1994. Las Vegas, NV, American Society for Microbiology, 1994. 30. Phalipon A, Sansonetti P. Live attenuated Shigella flexneri mutants as vaccine candidates against shigellosis and vectors for antigen delivery. Biologicals, 1995, 23: 125-134. 31. Oberhelm RA et al. Prospective study of systemic and mucosal immune response in dysenteric patients to specific Shigella invasion plasmid antigens and lipopolysaccharides. Infection and immunity, 1991, 59: 2341-2350. 32. Khan MU. Interruption of shigellosis by hand wash- ing. Transactions of the Royal Society of Tropical Medicine and Hygiene, 1982, 76: 164-168. 33. Han AM, Hlaing T. Prevention of diarrhoea and dys- entery by hand washing. Transactions of the Royal Society of Tropical Medicine and Hygiene, 1989, 83: 128-131. WHO Bulletin OMS. Vol 75 1997 53