Deaths From Rotavirus Disease in Bangladeshi Children

Estimates From Hospital-Based Surveillance

Abstract

Background: 

To assess the potential health benefits of introducing new rotavirus (RV) vaccines, we estimated mortality from RV gastroenteritis in Bangladeshi children <5 years of age.

Methods: 

We examined data from ongoing diarrhea surveillance in a systematic 2% sample (4% until 1995) of patients visiting the International Centre for Diarrheal Disease Research, Bangladesh, Dhaka Hospital during 1993–2004 and all patients visiting the rural Matlab Hospital during 2000–2004. To estimate deaths from RV, we multiplied the proportion of diarrhea visits attributable to RV with 2004 estimates of diarrhea deaths in Bangladeshi children.

Results: 

At Dhaka Hospital, RV was detected in 33% of 18,300 children with diarrhea. The proportion of diarrhea attributable to RV nearly doubled during 2002–2004 compared with 1993–1995 (42% versus 22%, P < 0.001). At Matlab Hospital, RV was detected in 35% of 4597 children with diarrhea. At both sites, most RV cases were among children age 3–24 months and the number of cases peaked during the cool and dry months from December through February. Of the 325,600 deaths among children <5 years that occur each year, we estimated 5600 to 9400 (2–3%) were attributable to RV. Thus, between 1 in 390 and 1 in 660 children born in Bangladesh each year die of RV infection by age 5.

Conclusions: 

These data clearly demonstrate the tremendous health burden of RV gastroenteritis. The increasing proportion of severe diarrhea cases underscores the need for specific interventions against RV, such as vaccines, to further reduce diarrhea mortality and morbidity.

Rotavirus (RV) is the leading cause of severe gastroenteritis in infants and young children worldwide and is estimated to account for 600,000 deaths in children <5 years of age.¹ Two new vaccines have recently been demonstrated in large studies to be safe and effective in preventing severe RV gastroenteritis in children.^2,3 Because those studies were conducted exclusively among children in high and middle-income countries, additional evaluations are ongoing or being planned to assess whether the vaccines would perform equally well in low-income settings, where more than 80% of all RV deaths are estimated to occur.⁴

Estimates of the health burden of RV disease, particularly deaths associated with RV infection would allow policy makers and donors to better assess the potential value of these new vaccines. We used data from ongoing pathogen-specific diarrhea surveillance at 2 International Centre for Diarrheal Disease Research, Bangladesh (ICDDR,B) hospitals and national estimates of total and disease-specific mortality for Bangladeshi children <5 years of age to estimate the current mortality from RV disease. Our findings have yielded useful information for policy makers in Bangladesh and provided insights into the changing etiologic trends of severe childhood diarrhea in developing countries during the past 2 decades.

PATIENTS AND METHODS

Study Population.

ICDDR,B Dhaka Hospital provides care for around 100,000 patients with diarrhea each year, 55% of whom are children <5 years of age, coming mostly from urban slum areas.⁵ Dhaka city is densely populated with a total population of 11 million residents. The rural Matlab Hospital of ICDDR,B maintains a cohort of 220,000 persons at the 149 villages in the hospital catchment area under active surveillance. Matlab Hospital provides care to 15,000 patients each year from this surveillance population each year, 60% of whom are children <5 years of age.

Specimen Collection and Testing.

Fecal bulk stool samples were collected from a 2% systematic sample, that is, every 50th outpatient (every 25th patient, 4% until 1995) who visited Dhaka Hospital during 1993–2004 with diarrhea as either an inpatient (60%) or an outpatient (40%) and all patients who visited Matlab Hospital during 2000–2004 were tested in the Laboratory Sciences Division of ICDDR,B for a variety of enteric agents. RV antigen was detected using an enzyme-linked immunosorbent assay (ELISA) modeled after the DAKOPATTS commercial kit.⁶ The same ELISA for RV detection was used throughout the study period during 1990–2004. The same specimens were also tested routinely for Salmonella (enteric media), Shigella (MacConkey agar), Vibrio cholerae O1 and O139 [thiosulfate citrate bile salts (TCBS) agar], and parasites such as Giardia lamblia, Entaemoeba histolytica, and Cryptosporidium using standard assays.^5–7

Analysis.

Because most severe RV disease and deaths occur among children <5 years of age, we limited our analysis to this age group. Patients whose stool specimens tested positive by ELISA for RV were compared with those infected with other pathogens with respect to age and seasonal patterns. In developing countries, even severe and fatal case-patients in the community rarely go to a hospital, because of limited access to health care.⁸ In the absence of appropriate therapy such as intravenous rehydration, they may die of RV diarrhea at the same or higher rate as among severe diarrhea cases admitted to a hospital.^9,10 Thus, to derive national estimates of RV deaths in children, we applied the detection rate of RV among children with diarrhea at the 2 ICDDR,B hospitals to national statistics on overall and diarrhea-specific mortality for children <5 years. To estimate the fraction of diarrhea deaths attributable to RV, we applied a weighted average of RV detection rates from the urban Dhaka data (20% weight) and the rural Matlab data (80% weight), because around four-fifths of the Bangladesh population lives in rural areas.^5,11

Ethical Consideration.

All patients’ parents were given information about this surveillance program and children's names were kept anonymous throughout the data analysis. The research protocol was exempted from the Internal Review Board of Emory University and ICDDR,B.

RESULTS

Dhaka Data.

From January 1993 through December 2004, 1.7 million patients with diarrhea visited Dhaka Hospital. Of these, 33,920 (2%) were automatically enrolled in the surveillance system and submitted a fecal specimen to be screened for enteric pathogens. Approximately 55% (N = 18,544) of all submitted specimens were from children <5 years of age. In this age group, RV was the most common pathogen detected in 6056 (33%) specimens, followed by Vibrio cholerae (12%) and Shigella species (7%) (Table 1). The detection rate of RV nearly doubled from 22% during the first 3 years of the study period (1993–1995) to 42% during the last 3 years (2002–2004) (P < 0.001) because of the reduction in hospitalization on non-RV etiology. The detection rate of other pathogens did not change significantly during this period, but the rate to detect unknown pathogens decreased from 53% to 37% (P < 0.001). Rates of RV detection increased over time from 1993 to 2004 (Fig. 1, χ² for trend P = 0.001), but the increase was not steady and the absolute number of RV visits did not change substantially with 30% decrease of total hospital visits during the same period.

TABLE 1: Etiologic Agents Detected in Fecal Specimens From Children <5 years of Age, Admitted With Diarrhea to ICDDR, B Dhaka Hospital, Overall 12 Years During 1993–2004, for the First 3 Years (1993–1995), and the Last 3 Years (2002–2004), and to Matlab Hospital

FIGURE 1.: Number and detection rate of rotavirus in fecal specimens from children <5 years of age admitted to ICDDR,B Dhaka Hospital during 1993–2004, by year.

Of the 6056 RV cases at Dhaka Hospital, 87% (n = 5717) were <2 years of age, but only 3% (n = 191) were infants <3 months old (Fig. 2). RV was present year-round but had a seasonal peak in the cool dry months of December through February; during this period the RV detection rate was 41% (Fig. 3). The peak detection rate of RV (53%) was in the month of January.

FIGURE 2.: Number and detection rate of rotavirus in fecal specimens from children <5 years of age, admitted with diarrhea to ICDDR,B Dhaka Hospital during 1993–2004, by 3 month age groups.

FIGURE 3.: Number and detection rate of rotavirus in fecal specimens from children <5 years of age admitted with diarrhea to ICDDR,B Dhaka Hospital from 1993–2004, by month of year.

The median duration of hospitalization for RV diarrhea was 22 hours, slightly longer than for other diarrhea causes (19 hours, P < 0.001 by the Mann-Whitney U test). Between 1993 and 2004, 9 children <5 years of age with confirmed RV died in the hospital. Because this figure represented RV deaths in only 2% of patients (4% until 1995) who had their stool tested for RV, we estimated by extrapolation that approximately 400 children died at Dhaka Hospital from RV during the study period.

Matlab Data.

From March 2000 to February 2005, fecal samples from all the 7633 diarrhea patients who were registered in the surveillance area and visited Matlab Hospital were screened for enteric pathogens. Of these, 61% (n = 4597) were from children <5 years of age. RV was again the most common pathogen and was detected in 1617 (35%) of specimens. The distribution of other etiologic agents was similar to that seen in Dhaka Hospital (Table 1).

Among the 1617 children <5 years of age with RV diarrhea, most (90%, n = 1536) were <2 years, but only 1% (20 cases) were infants <3 months old. Similar to Dhaka Hospital, a seasonal peak of RV was seen in Matlab during December through February, when the detection rate was 41%. The peak detection rate of RV (54%) was in February.

Estimation of Deaths From RV Diarrhea in Children <5 Years of Age in Bangladesh.

A total of 3.7 million infants are born in Bangladesh each year and the mortality rate for those under 5 years of age is estimated to be 88 deaths per 1000 live births per year.^11,12 Thus, we calculated a total of 325,600 deaths in children <5 years of age each year. Cause-specific data on childhood deaths from the 2004 Bangladesh Demographic and Health Survey (nationally-representative household surveys with more than 10,000 samples) indicated that 5.1% were attributed primarily to “diarrhea,” so we used this attributable fraction as our conservative lower estimate of proportion of child deaths caused by diarrhea. Bangladesh Demographic and Health Survey also indicated that an additional 6.8% of child deaths were attributed to a combination of “acute respiratory tract infection, diarrhea, and possible serious infection.” For a second maximum attributable fraction of child deaths caused by diarrhea, we assigned half of these deaths to diarrhea and thus arrived at an overall fraction of 8.5%. Multiplying this range (5.1–8.5%) with the overall estimate of 325,600 child deaths yielded diarrhea death figures of 16,600–27,700 in children <5 years of age. Finally multiplying the range of estimated diarrhea deaths by the weighted average of diarrhea visits at Dhaka and Matlab Hospitals that were attributable to RV (36.7%) yielded as estimated RV mortality of 5600–9400.

DISCUSSION

Our data demonstrate the tremendous burden of RV disease in Bangladeshi children and highlight the potential value of introducing effective RV vaccines. RV accounted for more than one-third of childhood diarrhea visits in both the urban and rural hospitals. Extrapolation of this proportion to national data on diarrhea deaths yielded an estimate of 5600–9400 annual deaths from RV disease. In other words, between 1 in 390 and 1 in 660 children born in Bangladesh each year die of RV diarrhea by the age of 5, or 2–3% of all childhood deaths are caused by RV. Our data on the age distribution of RV cases indicate that <3% of all cases occur in the first 3 months of life, likely because of a possible protective of maternal antibodies.¹³ Thus, if effective RV vaccines were included in the routine schedule of the Expanded Program on Immunization (ie, at 6, 10, and 14 weeks of age), most severe cases of RV should be preventable by vaccination.

Our estimate of 5600 to 9400 annual RV deaths in Bangladeshi children is lower than the 1995 estimate by Unicomb and Glass of 14,850–17,550 annual deaths from RV disease.⁷ The lower estimate from our study is in part attributable to a decline in child mortality under 5 years of age in Bangladesh from 133 per 1000 in 1993–1994 to 88 per 1000 in 2004. In addition, Unicomb and Glass⁷ estimated that 25% of child deaths were attributable to diarrhea, whereas recent national data from Bangladesh indicate that 5.1–8.5% of child deaths are caused by diarrhea. These 2 factors combined yielded an almost 5-fold lower estimate of total diarrhea deaths (16,600–27,700 in our study compared with 135,000 for Unicomb and Glass⁷). The fraction of severe diarrhea cases attributable to RV in our study was almost twice that in the report by Unicomb and Glass⁷ and therefore the estimated RV mortality has only declined by half during the past decade. A recent review of global data reported a similar trend of increasing detection rates of RV in children with severe diarrhea in the context of a decline in overall mortality from diarrhea.¹⁴ This changing etiologic trend underscores the need for specific interventions against RV, such as vaccines, to further reduce diarrhea mortality.

Some limitations should be considered in the interpretation of our data. First, the analysis depends on the continuous surveillance data from the 2 major diarrhea treatment hospitals. The patients treated at ICDDR,B Hospitals, which are well known for their care of children with diarrhea, may represent a selected population of the most severely ill children who are able to travel to receive care. A disease that kills more quickly such as cholera or shigellosis might cause deaths out of proportion to hospital visits, because the cases might be more likely to die before they reach the hospital. Thus, this method could overestimate the mortality from RV.

However, RV is consistently identified more commonly among patients with more severe illness, compared with less severely ill patients,^1,13 and so it is reasonable to assume that they will cause a similar proportion of fatal diarrheal episodes, which means that the rates of RV among children who do seek care would be a reasonable proxy for the rate of RV among severe diarrhea cases who die in settings without access to medical care. A second limitation of this analysis is the uncertainty over attribution of child deaths to diarrhea on the basis of interviews with next of kin or other caregivers of those who died. To account for this uncertainty, we included a conservative lower estimate as well as an upper estimate based on different assumptions and present a range of estimated RV mortality.

Despite these limitations, data from this unique longitudinal diarrhea surveillance system with more than 2 decades of information collected using the same methodology clearly demonstrate that the health burden of RV diarrhea in Bangladeshi children is great. The lack of decline in RV visits despite a substantial overall reduction in diarrhea mortality underscores the need for targeted interventions against RV. If ongoing trials of the new RV vaccines prove that they are also safe and effective in low-income settings with high levels of malnutrition and poor sanitation, and if they are available at an affordable price, then their inclusion in the routine childhood vaccination program in Bangladesh would be expected to save thousands of lives and hundreds of thousands of hospital visits each year. In the future, the hospital surveillance used at ICDDR,B enables monitoring the trend of RV diarrhea after the implementation of an immunization program.

ACKNOWLEDGMENTS

The authors thank Shams El Arifeen, Public Health Sciences Division, ICDDR,B and Prof. Philip S. Brachman, Rollins School of Public Health (RSPH), Emory University, for their helpful suggestions.

ICDDR,B is a major center for research of diarrhea, developing and promoting realistic solutions to the major health, population, and nutrition problems for the poor of Bangladesh and in the other developing countries. They are supported by countries and agencies that share its concern for the health problems of developing countries. Current donors providing core support includes Australian International Development Agency (AusAID), Government of Bangladesh, Canadian International Development Agency (CIDA), The Kingdom of Saudi Arabia, The Government of the Netherlands, Government of Sri Lanka, Swedish International Development Cooperative Agency (SIDA), Swiss Development Cooperation (SDC), and Department for International Development (DFID), UK; and international organizations, including Asian Development Bank (ADB), the United Nations Children's Fund (UNICEF), and the World Health Organization (WHO).