The economic impact of visceral leishmaniasis on households in Bangladesh
This article is corrected by:
- Errata: Errata Volume 11, Issue 9, 1482, Article first published online: 24 August 2006
Corresponding Author Caryn Bern, Division of Parasitic Diseases (MS F-22), Centers for Disease Control and Prevention, 4770 Buford Highway NE, Atlanta, GA 30341, USA. Tel.: +1 770 488 7654; fax: +1 770 488 7761; E-mail: firstname.lastname@example.org
Objectives To explore current patterns of diagnosis and treatment, quantify household economic impact and identify household strategies to cover the costs of visceral leishmaniasis (VL) care in rural Bangladesh.
Method Structured interviews with 113 VL patients from 87 households documenting all provider visits and expenditures for health care for VL, and the ways in which the expenditures were covered.
Results Patients paid a median of 7 visits to six different providers before beginning VL treatment. All visited the subdistrict government hospital at least once. While health care, including antileishmanial drug therapy, is officially available free of charge at government facilities, 79% of patients reported making informal payments for provider access, diagnostics and drug administration; only 14% of patients received their full drug course from this source. For the 58% of patients who purchased the full treatment course, drug cost constituted 34% of direct expenditure. Median direct expenditure for one VL patient was US$87 and median income lost was $40; median total expenditure was 1.2 times annual per capita income of our study population. Households employed multiple coping strategies to cover expenditures, most commonly sale or rental of assets (62%) and taking out loans (64%).
Conclusions Visceral leishmaniasis treatment causes a major economic burden in affected families. Control strategies for VL should facilitate timely, affordable diagnosis and treatment of patients to decrease the infection reservoir and to alleviate the economic burden of VL on households.
Objectifs Investiguer les profils actuels de diagnostic et de traitement, quantifier l'impacte économique sur les familles et identifier les stratégies des familles pour couvrir les dépenses liées aux soins de la leishmaniose viscérale (LV) en milieu rural au Bangladesh.
Méthode Interviews structurées de 113 patients LV provenant de 87 familles et documentation de toutes les visites chez des praticiens et les dépenses pour les soins de santé liée à la LV et la manière utilisée pour couvrir ces dépenses.
Résultats Le nombre médian des visites était de 7 chez 6 différents praticiens avant le commencement du traitement de la LV. Tous les patients ont visité au moins une fois l'hôpital sous district gouvernemental. Bien que les soins de santé, incluant le traitement médicamenteux de la LV soient officiellement disponibles sans charge dans les services gouvernementaux, 79% des patients rapportaient avoir effectué des paiements informels pour l'accès aux soins, le diagnostic et l'administration des médicaments. Seul 14% des patients ont reçu leur traitement complet de source officielle. Pour les 58% des patients qui ont payé pour leur traitement complet, le coût des médicaments représentait 34% des dépenses directes. La dépense directe médiane pour un patient LV était de 87 dollars US et la perte de revenu médian était de 40 dollars US. La dépense totale médiane correspondait à 1,2 fois le revenu annuel par capita dans la population étudiée. Les familles utilisaient multiples stratégies pour faire face et couvrir ces dépenses, la plupart du temps par la vente ou la location de bien (62% des cas) et par des emprunts (64% des cas).
Conclusions Le traitement de la LV pose un poids économique majeur sur les familles affectées. Les stratégies de contrôle de la LV devraient faciliter le diagnostic et le traitement abordables et à temps des patients, pour réduire le réservoir de l'infection et soulager le poids économique posé par la LV sur les familles.
Objetivos Explorar patrones actuales de diagnóstico y tratamiento, cuantificar el impacto económico sobre el hogar e identificar las estrategias para cubrir los costos del cuidado de la leishmaniasis visceral (LV) en zonas rurales de Bangladesh.
Método Entrevistas estructuradas a 113 pacientes de LV provenientes de 87 hogares, en las que se documentaron todas las visitas a proveedores y los gastos incurridos en cuidados sanitarios relacionados con la LV, así como las formas en las que se cubrieron dichos gastos.
Resultados Los pacientes pagaron una mediana de 7 visitas a 6 proveedores diferentes antes de comenzar tratamiento para la LV. Todos visitaron al menos una vez el hospital gubernamental sub-distrital. Aunque la atención sanitaria, incluido el tratamiento para LV, está disponible de forma gratuita en los centros sanitarios gubernamentales, el 79% de los pacientes reportaron haber realizado pagos informales para acceder a proveedores, para diagnóstico y administración de medicamentos. Solo el 14% de los pacientes recibieron un tratamiento completo por esta vía. Para el 58% de los pacientes que compraron un tratamiento completo, el costo de la medicación constituyó un 34% de los costes directos. La mediana de los costes directos para un paciente con LV fue de US$87, y la media de ingresos perdidos fue de $40. La media del gasto total fue 1.2 veces los ingresos anuales per cápita de la población de estudio. Los hogares utilizaban múltiples estrategias para cubrir los gastos extras, siendo lo más comunes la venta o alquiler de activos (62%) así como el pedir préstamos (64%).
Conclusiones El tratamiento de la LV resulta en una carga económica importante para las familias afectadas. Las estrategias de control deberían facilitar un diagnóstico y tratamiento asequible y a tiempo, para así disminuír el reservorio de infección y aliviar la carga económica que recae sobre las familias.
Since the 1970s, South Asia has experienced a major resurgence of anthroponotic visceral leishmaniasis (VL). Clinical VL illness, also known as kala-azar (‘black fever’ in Hindi), is characterized by prolonged fever, weight loss, hepatosplenomegaly, darkening of the skin and anaemia (Pearson et al. 1999). In South Asia, the disease is transmitted by the female sand fly Phlebotomus argentipes and humans are the only reservoir. Case-fatality rates approach 100% if untreated, and often exceed 10% among treated cases (Berman 1997).
India, Bangladesh and Nepal together account for an estimated 250 000 to 300 000 cases per year, comprising 60% of the global burden of VL (Desjeux 1992; Murray & Lopez 1996). Recently, the Government of India has expressed its commitment to the elimination of VL as a public health problem (Government of India 2002); Bangladesh and Nepal are part of the South Asian regional VL control initiative. The proposed strategy relies primarily on early VL case detection and treatment, plus vector control through residual insecticide application in houses. However, delays in diagnosis and treatment are common in South Asia, and result in part from the burden of payments for provider visits, diagnostic tests and antileishmanial drug therapy (Ahluwalia et al. 2003). Recently, the economic impact of specific diseases such as tuberculosis, malaria and dengue has begun to be quantified at the household level (Ettling et al. 1994; Kamolratanakul et al. 1999; Van Damme et al. 2004), and out-of-pocket healthcare expenditure is increasingly recognized as an important contributor to impoverishment (Kawabata et al. 2002; Meessen et al. 2003; Van Damme et al. 2003). A small study in Nepal suggested that direct and indirect costs of VL treatment totalled twice the per capita income in the 18 households studied (Adhikari & Maskay 2003). However, empirical data for the economic impact of VL are sparse. For this reason, we collected data to better understand where VL patients obtain diagnosis and treatment in rural Bangladesh, and to investigate the household economic impact and coping strategies to pay for VL care.
The site of the study was Mymensingh District, the most highly endemic district in Bangladesh, reporting 4000 VL cases per year. The study protocol was approved by the ethical review committees of CDC, ICDDR, B and Emory University; each participant gave informed consent before the interview. The economic impact study included households with VL patients in two villages, ‘village 1’, the site of the epidemiological study described in earlier publications (Ahluwalia et al. 2003; Bern et al. 2005) and ‘village 2’, located about 7 km away. Both villages were chosen based on high reported VL incidence in surveillance data. The Thana (subdistrict) Health Complex (THC) is the main official provider of health care, including antileishmanial treatment; the THC is located approximately 3 km from the centre of village 1 and 8 km from the centre of village 2. Village 2 was included in the current study to provide more recent data on health care seeking and expenditures; the epidemiological study had provided the first-line antileishmanial drug sodium antimony gluconate (SAG) free of charge for patients in village 1 from January 2002 to June 2004, consequently altering behaviour and expenditures. For village 1, only VL patients treated before January 2002 were included in the study. By contrast, all but five of the patients from village 2 were treated after January 2002.
We performed all analyses for each village separately and for the two villages combined. As there were no significant differences in the results by village (with the exception of the time period in which patients were treated, as noted above), we present the aggregate data here. During the epidemiological study in village 1, we collected data on total household expenditure as a proxy for income. The median household size was five members, and the estimated annual household income was $540. The median annual per capita income was $108 ($1.49 per day per household, or $0.30 per day per person). Both villages were predominantly agricultural and the field team judged them to be similar in socio-economic distribution. We therefore used the median annual per capita income of village 1 to provide perspective on the magnitude of the economic impact of the disease in the aggregate data.
Visceral leishmaniasis is a chronic illness that usually progresses in severity over a period of months and is ultimately fatal without treatment. We defined a VL illness episode as lasting from the time symptoms started until clinical cure. For the purpose of this analysis, the episode of VL illness refers only to an episode of kala-azar and its initial treatment course. We excluded patients with kala-azar treatment failure or post-kala-azar-dermal leishmaniasis; these conditions are less frequent and more costly to treat. VL cases from village 1 were identified during the epidemiological study through previously described methods (Bern et al. 2005). Cases in village 2 were ascertained through the THC laboratory registry; additional participants identified themselves as treated VL patients and volunteered to be interviewed during visits to village 2. Village 2 patients were included if they described a positive diagnostic test for VL at the local THC, and received 20 injections of SAG, corresponding to the first-line treatment regimen recommended by official government guidelines.
The economic impact questionnaires were administered from 28 January to 25 March, 2004. All interviews were conducted in Bengali by two of the authors (R.C. and D.A.S.). We collected data for the VL illness episode regarding all healthcare provider visits, all expenditures related to care, the source of SAG, income lost by the household and coping mechanisms to pay for expenditures. Lost income was estimated by multiplying days of work lost due to the illness by the wage-earner's reported average daily income. Daily income for farmers was estimated to be the same as that reported for an average agricultural worker during the period of the study, or $1.75 per day. Most households had only one wage-earner, thus, this figure represents the daily household income, not the daily per capita income. Agricultural workers were not the lowest paid members of this community; daily labourers in other areas of work earned even less.
We interviewed the VL patient or a member of the household who served as a caretaker and could provide the relevant information. We also conducted surveys of private pharmacies and laboratories in the Mymensingh district capital and in the subdistrict town nearest our study sites on several occasions to document market prices and availability of diagnostic tests and antileishmanial drug. Data were analysed using SAS (v. 9.1; SAS Institute Inc., Cary, NC, USA).
We interviewed 113 VL patients from 87 households, 69 from village 1, and 44 from village 2; 59 were male and 54 female. The median age was 15 years (range 1–80); 53 (47%) patients were younger than 15, 54 (48%) 15–44 years and six (5%) 45 years or older. Patients were treated for VL during the years 2000 (22 patients), 2001 (49), 2002 (28) and 2003 (14). We grouped providers into four categories: village doctors (people with little or no training in medical care who work as village providers), traditional healers (providers of herbal medicine or of traditional Islamic healing methods, which involve spiritual healing, prayers, amulets and taking holy water), government facilities and private providers (physicians, laboratories, pharmacists and drug vendors). Many of the physicians on the THC staff also conducted a private practice in the afternoons and evenings.
The number and order of provider visits were available for 112 VL patients. Patients visited a median of six different healthcare providers (range 1–11) and paid a median of seven visits (range 1–21) before beginning antileishmanial treatment. Seventy patients (63%) sought care first from a village doctor, 17 (15%) from a traditional healer, 17 (15%) from a private provider and eight (7%) from the THC. After the initial visit, there were many variations in the order and type of providers consulted. However, all patients visited the THC at least once. Final diagnosis and treatment occurred at the THC for 65/112 (58%) of patients, while 26 (23%) had the diagnosis made at the THC, but antileishmanial drug injections were administered by a village doctor (22), private doctor (3) or an unspecified provider (1). The other 21 patients used various combinations of government and private facilities for diagnosis and treatment. The number of providers, total number of visits, type of provider consulted first and the location of the final treatment did not vary by income level.
The median direct expenditure for care for a VL patient was US$87 (Table 1), or 81% of the median annual per capita income. The costs fell into four categories: provider expenses (consultation and diagnostic fees, medications other than SAG), expenditure to buy SAG, informal payments and miscellaneous costs (special food for the patient, transport costs, etc). Households spent significantly more for the care of male patients than for female patients (Table 2). A longer duration of illness and higher income were associated with higher expenditures.
Table 1. Median expenditure by category for 113 visceral leishmaniasis patients, Bangladesh, 2000–2003
|Provider charges||23 (1–206)|
|Sodium antimony gluconate treatment*||28 (1–96)|
|Informal payments||2 (0–37)|
|Miscellaneous costs||23 (4–386)|
|Total expenditure||87 (21–690)|
Table 2. Total expenditure in US dollars for visceral leishmaniasis care by duration of illness, income level and gender, Bangladesh, 2000–2003
|Duration illness ≤4 months||56||78 (25–419)|
|Duration illness >4 months||46||89 (21–690)|
|Monthly household income ≤$70||65||84 (21–386)|
|Monthly household income >$70||47||91 (25–690)|
|Female patient||54||79 (21, 331)|
|Male patient||59||102 (25, 690)*|
Of 113 VL illness episodes, 90 (80%) were reported to result in a loss of income. The median income lost was $40 (range $1–616). The illness had permanent repercussions for some households. For example, as a result of a one-month absence to get care for his son, the father of one patient lost local market space to sell his produce, thereby suffering a permanent loss of income. The median total cost for one VL patient, including direct and indirect costs, was $126 or 1.2 times the median annual per capita income.
All health care at the THC, including VL diagnostic tests and SAG, was officially stated to be free of charge. However, 89 of 113 (79%) patients reported making informal payments for access to a provider, SAG, administration of SAG, and/or diagnostic tests at the THC. Each payment was usually small (for example, $0.18 per injection or diagnostic test), but nearly all patients had multiple visits, tests and injections. The median total for informal payments was $1.75 ($0.18–37).
The direct agglutination test (DAT), a sensitive and specific serologic test for VL, was widely used during the 1990s, but DAT antigen was no longer available after 1999. Diagnosis at the THC returned to reliance on the aldehyde (or formol-gel) test, a qualitative measure of high-globulin content in use since the 1920s. Aldehyde test sensitivity was estimated at 40% in a study in Nepal (Boelaert et al. 2004). Our study patients reported having up to three aldehyde tests before a positive result was found and treatment could begin. A new rapid immunochromatographic strip test using the recombinant K39 antigen was available in some private laboratories beginning in 2002. However, in a survey of five private laboratories in 2002, the charge per rK39 strip test was between $5 and $9. The rK39 strip test is both sensitive and specific for VL confirmation in South Asia (Sundar et al. 1998; Bern et al. 2000a; Boelaert et al. 2004), and would be a practical alternative to DAT, if an affordable price could be ensured.
Antileishmanial drug treatment, availability and cost
During the study period (2000–2003), there were ongoing shortages of SAG in the THCs, and drug scarcity and consequent high prices in the private sector. There was only one licensed producer of SAG in Bangladesh, which stopped production in 2003. Periodically during the study period, emergency procurements of SAG were provided through the auspices of the World Health Organization. The official market price for Bangladeshi manufactured SAG was $1.90 per 10-cc vial, and a 20-day course for a 34-kg patient (the median weight for VL patients in our epidemiological study) required 14 vials. Thus, the anticipated cost for a course of SAG should be $27.
In our study population, only 14% of patients received their full course of SAG at the THC (Table 3). When SAG supplied by the Ministry of Health was insufficient, THC physicians were forced to ration the drug, giving higher priority to children and to patients with the most severe disease. The THC physicians might give the first few days of SAG and then send patients’ families to buy the remainder of the drug in private pharmacies. However, more than half of patients bought the full course from a private pharmacy. The cost of SAG accounted for 34% of the total expenditure for these patients (Table 3). Among our study patients, the highest expenditure for a course of SAG was $88, more than three times the official price.
Table 3. Sodium antimony gluconate (SAG) cost by source, Bangladesh, 2000–2003
|Full course from Thana Health Complex (THC)||16 (14)||0||0|
|Part THC, part private pharmacy||31 (28)||12 (2–63)||17|
|Full course from private pharmacy||64 (58)||31 (3.5–88)||34|
To better understand drug availability and prices in the private sector, Bangladeshi study team members performed periodic surveys of pharmacies and drug vendors in the district capital. These surveys documented intermittent scarcity and prices that greatly exceeded the official price (Table 4). Antileishmanial drugs other than SAG were extremely difficult to obtain in endemic districts of Bangladesh. The THC never had pentamidine (the official second-line drug for SAG treatment failures) during the 3 years we worked nearby. Amphotericin B was available only in a few specialized pharmacies in Dhaka at a price of >$150 for an adult treatment course.
Table 4. Availability and price of sodium antimony gluconate (SAG) from private pharmacies and drug vendors in Mymensingh district capital, Bangladesh, 2001–2004
Households used four basic coping mechanisms: spending available cash, sale or rental of assets, taking loans and the receipt of financial help or consumables from family members or friends (Table 5). For the 104 VL patients whose illness was not concurrent with another case in the household, the costs for 63 (61%) were defrayed using more than one of these strategies. Households had cash available at the time of illness for 32 (31%) VL patients. In five of these, the cash was sufficient to cover costs of illness, but expenditure for these households was well below the median for the study population. The most common coping strategies were the sale or rental of assets (62%) and taking loans (64%); 23% of households took loans from more than one source.
Table 5. Coping strategies used to finance care for visceral leishmaniasis patients, Bangladesh, 2000–2003
|Cash available||33 (32)†|
|Sale or rental of assets||64 (62)‡|
|Gifts in cash or kind||15 (14)¶|
Among the 64 VL episodes for which households sold or rented assets, the sale of animals (36%) and agricultural products (42%) were the most common. In our study villages, ‘rental’ of land was actually the use of land as collateral for a loan. The villager exchanged temporary ownership of the land for cash. Until the loan was repaid in full, the lender cultivated and harvested the land.
Households borrowed a median of $53 ($9–246) per illness episode, and incurred median interest of $18 ($0–246), or one-third of the loan amount. Some study households had to make interest payments of 10% of the principal every month until they were able to save up the total amount of the principal to be repaid in a lump sum. This practice resulted in interest payments amounting to more than 100% of the principal for five study participants.
Households with concurrent VL cases employed multiple simultaneous coping strategies to cover expenditures for all cases. These included the sale of large quantities of land and the acquisition of several loans with high interest rates. In a household with four VL cases, the family sold a cow, took a loan, rented land as collateral for another loan, sold rice and received cash from a family member. Despite these attempts to cover costs, one of the children of the household was sent to live with her grandparents to shift the expenses of caring for her. At the time of interview, 6 months after the most recent case of VL, this household had not repaid their loans or recovered their land.
Our analysis provides insight into the heavy economic burden posed by VL, all the more devastating because of the impoverished populations affected. Like tuberculosis and HIV, VL is a chronic illness, ultimately lethal if untreated, whose diagnosis and drug treatment are extremely expensive compared with the income of affected families. In another parallel with tuberculosis, the pool of untreated VL patients provides the reservoir for ongoing transmission. Timely, effective drug treatment is therefore an essential public health intervention as well as a life-saving cure for the individual. For tuberculosis and HIV, governments and international donors increasingly underwrite the costs of care, due to the widespread recognition that ensuring universal treatment of these diseases is in the interests of the community, and indeed, the world (Kaul & Faust 2001). The global disease burden for VL is much smaller than for tuberculosis or HIV, but the burden is concentrated in certain regions and communities: in highly endemic areas, the impact of VL may be greater than for either (Boelaert et al. 2000). In village 1, 23% of all recorded deaths from 2000–2003 were attributable to VL (C. Bern, unpublished data ). Because of its transmission characteristics, VL tends to disproportionately affect some communities, and VL is clustered within households (Bern et al. 2005). Thus, households often have more than one VL case within a few years, and the economic impact on the household is even more profound.
In May 2005, a Memorandum of Understanding was signed among the ministries of health of India, Bangladesh, Nepal and the South East Asia Regional Office of the World Health Organization declaring a regional commitment to decrease the incidence of VL below 1 case per 10 000 population by 2015 (World Health Organization Southeast Asia Regional Office 2005). For all of these countries, rapid diagnosis and treatment of VL is a cornerstone of their control programs. In order to decrease the reservoir of VL in South Asia, the time from symptom onset to treatment, now an average of 4 months (Bern et al. 2000b; Thakur 2000; Bern et al. 2005), must be decreased substantially.
Our data demonstrate that VL patients used a combination of the private and public sectors in seeking care, but that ultimately nearly all arrived at the government-run THC for diagnosis and recommendation for antileishmanial treatment. Drug scarcity at the THC was likely a key factor that shifted the later phases of healthcare seeking and drug procurement to the private sector for many patients. However, most patients do not begin with the THC, despite high costs in the private sector. Part of the explanation for this lies in the negative experiences which the poor villagers had at the THC (Ahluwalia et al. 2003), and part in the opportunity costs of seeking care outside the village. To seek care at the THC generally required the man of the household to take time from work and lose income, pay for transportation, give informal payments to facilitate access and wait for many hours before being seen.
Although our study relied on recall of costs for an illness that occurred up to 3 years before the interview, nearly all our interviewees could recount their experiences and expenditures in impressive detail. VL in a family member was clearly a major event in their lives. Caution should be used in generalizing the specific results of this study to other settings, as factors such as reliance on the aldehyde test and drug shortages may not apply elsewhere. However, the role of the private sector is even greater in India than in Bangladesh: private expenditure accounts for 82.1% of the total health expenditure in India, compared with 55.8% in Bangladesh (World Health Organization 2001). In addition, Indian villagers face similar challenges in gaining access to appropriate health care (Dugger 2004). The study of 18 households in Nepal suggests that the magnitude of the economic impact of VL is similar to what we found in Bangladesh (Adhikari & Maskay 2003). Although the details vary, the predominant coping mechanisms in our study population (selling assets, loans, reliance on informal support networks) are the same as those reported in studies from Africa and other parts of Asia (Sauerborn et al. 1996; Van Damme et al. 2004). In order to have a more comprehensive understanding, systematic collection of accurate economic data is needed from other VL-endemic settings, and for other key components in VL control, such as the provision of health services and vector control interventions.
The progression of coping strategies in some of our participant households, for example, sale of animals and crops, then sale of durable assets such as land and jewellery, followed by taking high-interest loans, indicates that families may liquidate large portions of their asset base and go into debt to pay for care. VL in a household member caused some study households to lose market access, such as stalls in which to sell their produce, and participants reported that VL contributed to absenteeism from school and food insecurity within households (Ahluwalia et al. 2003). This situation may worsen nutritional status, making previously unaffected household members even more vulnerable to VL (Cerf et al. 1987). The economic burden of caring for VL illness and the subsequent coping mechanisms contribute to a cycle of poverty and further illness, and may undermine the development in endemic communities.
The importance of illness and healthcare costs as major contributors to poverty have been increasingly recognized in recent years (Kawabata et al. 2002; Meessen et al. 2003; Van Damme et al. 2004). Health economists at the World Health Organization have defined ‘catastrophic health expenditures’ as expenditures for healthcare totalling 40% of a household's income after basic subsistence (Kawabata et al. 2002). However, in our study population, with a median household income of $1.49 per day, few households had any income beyond basic subsistence. Add to this a disease for which the minimum cost of the drug alone is $27, and VL inevitably entails catastrophic health expenditure. These facts need to be taken into account when programs are designed to improve access to diagnosis and drug treatment for VL. Because VL is expensive to treat even in the public sector, and because kala-azar patients are the major reservoir of ongoing transmission of disease, an effective means of ensuring rapid treatment of patients has the potential to substantially decrease future public expenditures for this disease, as well as avoiding the productivity losses associated with illness and care-seeking.
Control programs need to develop innovative ways to ensure that VL patients reach competent diagnostic facilities more rapidly and can receive an effective treatment course without making their families even more vulnerable. In Bangladesh, one option might be a partnership between government facilities and a non-governmental organization (NGO), as is currently the case for tuberculosis and leprosy treatment in our study area. VL patients could be diagnosed at the THC with improved diagnostic tests (e.g. DAT or rK39 dipstick) provided through the VL control program, and then referred to the specialized NGO for directly observed treatment and follow-up along the model used for tuberculosis.
Appropriate designs for VL diagnosis and treatment delivery will need to be developed in each country, and whatever the specific solution, control programs must find a way to address the catastrophic cost to families of caring for VL patients. Successful elimination programs require technical solutions, political will and delivery of services to those who need it most. We now have effective tools and South Asian countries are demonstrating that they have the political will to control VL. We must ensure that program implementation provides access to those who need it in a way that sustains the long-term health and welfare of endemic communities.
Anoopa Sharma died on 14 March 2005, at the age of 24. Her co-authors dedicate this article to her life and work, a career cut short by her untimely death. We thank our fieldworkers for their contributions and the residents of the study communities for their collaboration. We are grateful to Dilara Sultana, Milton Quiah, Emily Gurley and Dr David Sack of ICDDR,B; Dr K. R. Talukdar, Dr E. B. Yunus and Dr M. Rahman, the Directorate General of Health Services, Government of Bangladesh; Dr A. Akbar, Former Civil Surgeon, and Dr M. M. Hossain, Civil Surgeon, Mymensingh District; Dr M. G. Datta, Dr A. Hamid and Dr S. M. Alam of the Fulbaria Thana Health Complex; and Brigidier General Dr Md. W. R. Chowdhury, Dr K. Chowdhury, Dr Momen, Dr Utpal and Dr I. Khalil of the Mymensingh Medical College Hospital; and LeAnne Fox, Bob Wirtz, Ray Arthur and Steve Blount of CDC. This research was funded by a grant from the CDC Emerging Infections Initiative. ICDDR,B acknowledges with gratitude the commitment of CDC to the Centre's research efforts.