Bulletin of the World Health Organization

Rotavirus mortality in India: estimates based on a nationally representative survey of diarrhoeal deaths

Shaun K Morris a, Shally Awasthi b, Ajay Khera c, Diego G Bassani d, Gagandeep Kang e, Umesh D Parashar f, Rajesh Kumar g, Anita Shet h, Roger I Glass i, Prabhat Jha j & for the Million Death Study Collaborators

a. Division of Infectious Diseases, Hospital for Sick Children, University of Toronto, 555 University Ave, Toronto, ON M5G1X8, Canada.
b. Department of Paediatrics, King George’s Medical University, Lucknow, India.
c. Ministry of Health and Family Welfare, Government of India, New Delhi, India.
d. Department of Paediatrics, Hospital for Sick Children, University of Toronto, Toronto, Canada.
e. Christian Medical College, Vellore, India.
f. National Center for Immunization and Respiratory Diseases, Centers for Disease Control and Prevention, Atlanta, United States of America (USA).
g. School of Public Health, Post Graduate Institute of Medical Education, Chandigarh, India.
h. Department of Paediatrics, St John’s National Academy of Health Sciences, Bangalore, India.
i. Fogarty International Center, Bethesda, USA.
j. Centre for Global Health Research, Li Ka Shing Knowledge Institute, St Michael’s Hospital, University of Toronto, Toronto, Canada.

Correspondence to Shaun K Morris (e-mail: shaun.morris@mail.utoronto.ca).

(Submitted: 03 January 2012 – Revised version received: 04 April 2012 – Accepted: 10 April 2012 – Published online: 16 July 2012.)

Bulletin of the World Health Organization 2012;90:720-727. doi: 10.2471/BLT.12.101873

Introduction

Diarrhoeal diseases are a major cause of hospitalizations and child deaths globally. Together they account for approximately one in six deaths among children younger than five years.1 Of India’s more than 2.3 million annual deaths among children, about 334 000 are attributable to diarrhoeal diseases.13 Rotavirus is the leading cause of severe diarrhoea in children in developed and developing countries.3,4 Two rotavirus vaccines have been shown to be effective against rotavirus and have been licensed in more than 100 countries, including India.57 Introduction of these vaccines is expected to reduce child mortality by decreasing the incidence of severe diarrhoea and the frequency of death from diarrhoeal disease.

There is considerable controversy in India regarding the introduction of rotavirus vaccines, especially because of uncertainty surrounding the estimates of deaths and hospitalizations caused by rotavirus and because of the current cost of the vaccines.8 Studies estimate that 90 000–153 000 children die from rotavirus infection in India each year but these numbers are not based on nationally representative samples.3,9,10 Here, we present new estimates of the frequency and number of diarrhoeal deaths attributable to rotavirus among Indian children younger than 5 years by combining direct estimates of diarrhoea-related mortality from a nationally representative study of deaths2 with the fraction of diarrhoeal hospitalizations in selected Indian urban centres that were caused by rotavirus.11 We also calculate the number of deaths preventable in India by a national rotavirus vaccination programme by using the estimated annual number of rotavirus-associated deaths in India and rotavirus vaccine efficacy data from clinical trials in two developing countries.

Methods

We ascertained the number of deaths from diarrhoea among children during 2001–2003 on the basis of findings from the Million Death Study.2,12,13 The Million Death Study is an ongoing project conducted with the Registrar General of India and collaborators to determine causes of death in India. The study derives its data from a nationally representative sample of 6.3 million people in 1.1 million households within the Sample Registration System, a large, routine demographic survey performed by the Registrar General of India in all 28 Indian states and seven union territories that has been the primary system for the collection of Indian fertility and mortality data since 1971. Sample units are randomly selected to ensure that they are representative of the population at the state level. Enrolment in the system is voluntary and its confidentiality and consent procedures are defined as part of the Registration of Births and Deaths Act of 1969. Additional details about the Sample Registration System are available elsewhere.12

A dual-recording system was used in which births and deaths were continuously enumerated by a resident part-time enumerator and a retrospective survey was conducted independently every six months by a full-time supervisor. When a death occurred, the cause was assessed using the Routine, Reliable, Representative and Re-sampled Household Investigation of Mortality with Medical Evaluation, a standard verbal autopsy questionnaire.13,14 Two independent physicians reviewed the field reports and assigned a single cause of death by use of International classification of diseases, 10th revision codes.15 Deaths were defined as due to diarrhoea if they received a final code of A00–A09. Differences in coding were resolved by anonymous reconciliation of initial codes and, if needed, a third, senior physician adjudicated the final cause of death. We characterized diarrhoeal deaths by sex, age (< 1 month, 1–2 months, 3–11 months, 12–23 months and 24–59 months) and region (central, east, north, north-east, south and west). Sex-, age- and region-specific live births and deaths were corrected proportionally by use of methods described elsewhere2 to reflect the United Nations estimates of live births and deaths in India in 2005.

To estimate the number of deaths attributable to rotavirus infection, the annual number of diarrhoeal deaths by region was multiplied by the regional age-specific proportion of confirmed rotavirus infections among children hospitalized with diarrhoea. For example, the percentage of diarrhoea-associated hospitalizations due to rotavirus infection among boys aged 12–23 months in the state of Punjab was multiplied by the estimated number of diarrhoeal deaths among boys aged 12–23 months in the north region. Rotavirus data were obtained from the Indian Rotavirus Strain Surveillance Network (IRSSN) and covered admissions to 10 hospitals in 7 cities from December 2005 to November 2007.16 Overall, faecal samples from 1405 (39%) of 3580 children hospitalized with diarrhoea during this period tested positive for rotavirus by means of a commercial enzyme immunoassay.16 No data were available from the IRSSN for the central region, so we used an age-specific average of proportions from the other regions. In sensitivity analyses, we also applied the IRSSN age-specific rotavirus proportions from the regions with the highest and lowest rotavirus proportions (the north-east and west regions, respectively) to the central region.

Most diarrhoeal deaths in the Million Death Study occurred outside hospital (90%; 2734 of 3053) and in rural areas (92%; 2798 of 3053).2 Because no laboratory-confirmed data on the etiology of diarrhoea were available for children who died outside hospital, we assumed that, of these deaths, the proportion due to rotavirus infection reflected the proportion of deaths due to laboratory-confirmed rotavirus infection among children hospitalized with diarrhoeal disease. The proportion of diarrhoea cases attributable to rotavirus infection increases with the severity diarrhoea, so we also assumed that diarrhoea resulting in death in rural settings with limited access to health care is similar in severity to the diarrhoea that would be treated in hospitals in settings with more accessible health care.

To estimate the number of deaths preventable by a national rotavirus vaccination programme, we multiplied the estimated annual number of rotavirus-associated deaths in India by the values of vaccine efficacy from field studies of the monovalent rotavirus vaccine (Rotarix; GlaxoSmithKline, London, United Kingdom of Great Britain and Northern Ireland) in Malawi (efficacy, 49%) and the pentavalent bovine reassortant vaccine (Rotateq; Merck, Whitehouse Station, United States of America) in Bangladesh (efficacy, 43%).17,18 On-time vaccination and 100% coverage were assumed. Because infants aged < 3 months are too young to be fully vaccinated, we applied vaccine efficacy data only to children aged 3–59 months. We also calculated the number of children who would need to be vaccinated to save one life on the basis of the following method described by Kelly et al.: number needed to vaccinate = 1/([annual event rate attributed to disease among unvaccinated individuals] × [vaccine efficacy for prevention of the event]).19 All statistical analysis was performed using Stata SE version 10 (StataCorp. LP, College Station, United States of America).

This study was approved by the institutional review board of the Postgraduate Institute of Medical Education and Research (Chandigarh, India), the Indian Council of Medical Research, the Indian Health Ministry’s Screening Committee and the institutional review board of St Michael’s Hospital (Toronto, Canada).

Results

In the Million Death Study, 3053 (13.2%) of 23 152 deaths among children younger than 5 years were due to diarrhoea (Table 1).2 This corresponds to approximately 334 000 diarrhoeal deaths nationally during 2005, or 1 in 82 Indian children dying from diarrhoea before the age of 5 years. On the basis of IRSSN data, rotavirus was estimated to cause approximately 34% (113 000; 99% confidence interval, CI: 86 000–155 000) of all diarrhoeal deaths in this age group. Together, these findings yielded an estimated mortality rate of 4.14 (99% CI: 3.14–5.68) deaths per 1000 live births during 2005 and suggest that approximately 1 in 242 children will die from rotavirus infection before their fifth birthday.

The estimated rotavirus-associated mortality rate varied by region, age and sex during 2005 (Appendix A, available at: http://www.cghr.org/wordpress/wp-content/uploads/Rotavirus-BWHO_Web-Table-1_20120103.doc). The rate among girls (4.89 deaths per 1000 live births; 99% CI: 3.75–6.79) was 42% higher than among boys (3.45 deaths per 1000 live births; 99% CI: 2.58–4.66) (Table 2). Overall, about 64 000 girls (99% CI: 49 000–89 000) and 49 000 boys (99% CI: 37 000–66 000) died from rotavirus infection. At every age group beyond the neonatal period, the ratio of the rotavirus-associated mortality rate among girls to the rate among boys exceeded 1.30 deaths per 1000 live births and peaked at 1.70 deaths per 1000 live births between ages 24 and 59 months. Approximately 50% of rotavirus-associated deaths (54 700 of 113 000) occurred in the first year of life and about 75% (85 400 of 113 000) occurred in the first two years of life (Fig. 1).

Fig. 1. Estimated overall number of diarrhoeal deaths and rotavirus-attributable diarrhoeal deaths among Indian children younger than 5 years, by age and sex, during 2005
Fig. 1. Estimated overall number of diarrhoeal deaths and rotavirus-attributable diarrhoeal deaths among Indian children younger than 5 years, by age and sex, during 2005
Note: The mean percentage of deaths attributable to rotavirus among boys and girls from the Indian Rotavirus Strain Surveillance Network was 26% for ages 0–2 months, 43% for ages 3–11 months, 44% for ages 12–23 months and 25% for ages 24–59 months. Values were calculated using data from the Million Death Study (collected during 2001–2003) and the Indian Rotavirus Strain Surveillance Network (collected during 2005–2007) that were corrected on the basis of United Nations estimates of live births (14 180 792 boys and 13 130 363 girls) and deaths in India during 2005.

Most rotavirus-associated deaths during 2005 occurred in the central region (56 400) and east region (28 900) (Fig. 2). The mortality rate from rotavirus-associated gastroenteritis among children younger than five years ranged from a low of 1.64 deaths per 1000 live births in the west region to a high of 5.49 deaths per 1000 live births in the central region. The age distribution of deaths from rotavirus infection was similar for boys across regions but girls in the north and south regions died at somewhat earlier ages than those in other regions (data not shown). Rotavirus-associated deaths were not distributed evenly across the country (Fig. 3). More than half (64 400) of deaths were estimated to occur in three states: Uttar Pradesh (35 700 deaths), Bihar (17 800) and Madhya Pradesh (10 900). The sensitivity analysis involving use of varying proportions of rotavirus-associated deaths for the central region did not markedly change the national totals.

Fig. 2. Estimated rotavirus-attributable diarrhoeal deaths and mortality rates among Indian children younger than 5 years, by age and sex, during 2005
Fig. 2. <b>Estimated rotavirus-attributable diarrhoeal deaths and mortality rates among Indian children younger than 5 years, by age and sex, during 2005</b>
CI, confidence interval.Note: Values were calculated using data from the Million Death Study (collected during 2001–2003) and the Indian Rotavirus Strain Surveillance Network (collected during 2005–2007) that were corrected on the basis of United Nations estimates of live births (14 180 792 boys and 13 130 363 girls) and deaths in India during 2005.Source of shapefile: GfK Geomarketing, Bruchsal, Germany.
Fig. 3. Estimated mortality rates and diarrhoeal deaths due to rotavirus infection among Indian children younger than 5 years, by state, during 2005
Fig. 3. Estimated mortality rates and diarrhoeal deaths due to rotavirus infection among Indian children younger than 5 years, by state, during 2005
CI, confidence interval; C, central region; N, north region; NE, north-east region; E, east region, S, south region; W, west region.Note: Data for the 15 states with the greatest number of estimated rotavirus deaths is shown and are arranged by estimated rotavirus mortality rates, with data for the remaining states grouped together under “Other states”. The size of the boxes denoting mortality rates is proportionate to the state’s total estimated rotavirus deaths. Values were calculated using data from the Million Death Study (collected during 2001–2003) and the Indian Rotavirus Strain Surveillance Network (collected during 2005–2007) that were corrected on the basis of United Nations estimates of live births (14 180 792 boys and 13 130 363 girls) and deaths in India during 2005.

Between 41 000 and 48 000 deaths among children aged 3–59 months (18 000–20 000 boys and 24 000–27 000 girls) could be avoided every year by administering one of the currently available rotavirus vaccines. These figures correspond to 3.6–4.2% of the 1 135 000 deaths in this group, with a greater benefit for girls (3.6–4.0% of 505 000 deaths among boys, compared with 3.8–4.3% of 630 000 deaths among girls) if vaccination coverage and timing for both sexes were equal. On the basis of our estimated rotavirus-associated mortality rates for boys and girls 3–59 months old (2.91 and 4.23 deaths per 1000 live births, respectively), 570 girls and 660 boys would need to be vaccinated against rotavirus to prevent 1 girl and 1 boy, respectively, from dying due to rotavirus infection.

Discussion

Approximately 113 000 children (99% CI: 86 000–155 000) younger than five years died from rotavirus infection in India in 2005, for a mortality rate of 4.14 (99% CI: 3.14–5.68) deaths per 1000 live births. This suggests that roughly 1 child in 242 will die from rotavirus infection by the age of 5 years. The first year of life was the period of highest risk for death due to rotavirus infection, a finding that underscores the need for on-time vaccination.20 At every age beyond the neonatal period, a marked sex-related disparity was apparent and more girls than boys died from rotavirus infection. The greatest discrepancy by sex and region existed between girls in the central region (mortality rate, 6.55 deaths per 1000 live births) and boys in the south region (mortality rate, 1.88 deaths per 1000 live births), with a mortality rate ratio of 3.5 for girls relative to boys.

Our mortality estimates, which are based on directly measured data on diarrhoea-associated deaths, are similar to previously published estimates of 122 0003 and 122 000–153 00010 rotavirus-associated deaths that were based on indirect national estimates of deaths attributable to diarrhoea among children younger than five years. Our estimate is higher than that of Jain et al.9 who multiplied the fraction of diarrhoeal deaths due to rotavirus infection by the total number of diarrhoeal deaths in India and concluded that rotavirus caused approximately 98 000 deaths during 1998. Unlike these earlier estimates, our results provide age-, sex- and region-specific patterns of rotavirus-associated mortality in India, which can be used to optimize the influence of rotavirus vaccination programmes.

Rotavirus vaccine is available in India through the private sector, although on a small scale. Therefore, the number of vaccinated children is likely to be exceedingly small, although no data on who is being vaccinated are available. Our data indicate that if rotavirus vaccine were delivered as part of a national programme, about 4% of all deaths among Indian children younger than five years could be prevented. This would make rotavirus vaccine a valuable component of India’s strategy to reduce childhood mortality.

One of the most important findings in this study is the sex-based difference in mortality. Increased overall and disease-specific mortality rates among Indian girls have previously been shown for children of different ages.2,21 Various factors associated with increased mortality rates have been documented to be more common among girls, including poor nutritional status, delayed or incomplete vaccination or none, limited access to health care and failure to seek medical attention, and decreased likelihood of receiving appropriate antibiotic therapy.20,2225 The study of these factors, especially about what is needed to minimize sex-based differences, must be a key component of interventions to reduce overall childhood mortality in India.

This study has several limitations. First, the main potential source of uncertainty in our estimates involves possible misclassification bias attributable to the use of verbal autopsy to ascertain causes of death. However, for infants older than one month, the accuracy of diagnosis by verbal autopsy has been shown to be reasonably similar to that of hospital-reference diagnosis when diarrhoea is defined as frequent liquid, watery, loose or soft stools or dysentery: sensitivities for verbal autopsy ranged from 72% to 91% and specificities ranged from 57% to 78% in studies in Bangladesh, Nicaragua and Uganda.26 However, we caution that hospital-based studies are not an appropriate comparison for the large number of rural, unattended deaths.2 Second, because most diarrhoeal deaths recorded in the Million Death Study occurred outside hospital and/or in rural areas, it was not feasible to obtain samples from these individuals for microbiologic analysis. Consequently, our estimates are based on the assumption that, of the children who died from diarrhoeal disease outside hospital, the proportion of deaths due to rotavirus infection reflects the proportion of deaths due to laboratory-confirmed rotavirus infection among hospitalized children with severe diarrhoeal disease. Third, the largest absolute number of diarrhoeal deaths occurred in the central region, an area where the IRSSN does not collect rotavirus data. However, the use of various proportions from the national rotavirus surveillance study did not greatly alter our national totals (range, 106 000–114 000 deaths; data not shown), primarily because the large number of estimated rotavirus-attributable deaths in the central region is mainly a result of the underlying high diarrhoeal mortality rate in the region.2,27 Fourth, the Million Death Study data are from 2001–2003, the IRSSN data are from 2005–2007 and the rotavirus-associated mortality burden we estimated is projected onto India’s 2005 population. Because it is possible that the diarrhoeal mortality rate and the proportion of severe diarrhoea cases due to rotavirus infection have changed over the past 5–10 years, our estimates may not be accurate. However, although use of more recent data would have been ideal, no other more recent nationally representative source of comparable data exists. Some data suggest that there has not been any substantial change in at least some of the key factors driving diarrhoeal mortality; a review of India’s District Level Household Survey 3 showed that between 2002–2004 and 2007–2008, the percentage of children with diarrhoea who received oral rehydration solution (30.3% versus 34.2%) and sought medical treatment (73.2% versus 70.6%) changed only minimally.24 The survey also shows that less than one quarter of India’s population has access to piped water and that well under one half has access to relatively improved means of sanitation and human waste disposal.24 Data collection in the Million Death Study continues to the present date and we hope to analyse more recent data, including for time trends, in the future. Fifth, our calculation of avoidable deaths from rotavirus infection assumes on-time vaccination of the complete cohort of vaccine-eligible children. Unfortunately, overall vaccination coverage among children in India remains low and significant discrepancies in vaccination rates between boys and girls persist.20 In addition, it is likely that many of the children who received delayed vaccination or who are unvaccinated have other characteristics that increase their risk of dying from a diarrhoeal episode and thus we may have overestimated the benefits of the vaccine. Finally, our calculations relied on vaccine efficacy findings from Bangladesh and Malawi, rather than from India. Although it is reasonable to apply vaccine efficacy data from settings similar to India, assessment of the effectiveness of rotavirus vaccine after it is introduced will help refine our estimates of vaccine-preventable rotavirus deaths.

Findings from this study support the introduction of rotavirus vaccine in India, particularly in states such as Bihar and Uttar Pradesh, where rotavirus-associated mortality rates are high. Although a national rotavirus vaccination programme can play a key role in achieving the United Nations Millennium Development Goal of reducing child mortality,28 ongoing evaluation must be conducted to address major gaps in knowledge. The cost-effectiveness of a national rotavirus vaccination programme is among the most prominent gaps29 and the data presented here can serve as a basis for future analysis of this and other factors that can prepare policy-makers for decisions about how to optimally target health-care interventions and policies to reduce rotavirus infection and diarrhoeal deaths in India.


Acknowledgements

The Registrar General of India established the Sample Registration System in 1971, has continued it ever since and is collaborating with several of the authors on the ongoing Million Death Study. The opinions expressed in this paper are those of the authors and do not necessarily represent those of the Government of India, the United States National Institutes of Health or the United States Centers for Disease Control and Prevention. We thank Brendon Pezzack, Chinthanie Ramasundarahettige, Peter Rodriguez, Vicky Hsiao and Wilson Suraweera for data support. Much of this work was conducted when one of the authors (SKM) was a research fellow at the Centre for Global Health Research, Li Ka Shing Knowledge Institute, St Michael’s Hospital.

Funding:

This work was supported by the Fogarty International Centre of the United States National Institutes of Health (grant R01 TW05991–01), the Canadian Institute of Health Research (grant IEG-53506), the International Development Research Centre (grant 102172), the Li Ka Shing Knowledge Institute and Keenan Research Centre at St Michael's Hospital, University of Toronto, and the United States Fund for the United Nations Children’s Fund (subgrant 50140, via a grant from the Bill & Melinda Gates Foundation for the Child Health Epidemiology Reference Group). Authors received support from the Canada Research Chair programme (PJ) and from the Paediatric Scientist Development Program, the Sick Kids Foundation, the Paediatric Chairs of Canada and the March of Dimes (SKM). The funders had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript.

Competing interests:

None declared.

References

Share