Introduction

Estimates of measles-related deaths are critical for monitoring progress towards global measles elimination goals [1], [2]. However, absence of complete and reliable mortality data, particularly from countries with high disease burden, has been a major challenge in measuring the measles mortality reduction goals. In many developing countries, deaths from measles are not routinely reported and measles cases are substantially under-reported. Measles case fatality rates (CFRs) are a key input in models used by the World Health Organization (WHO) to estimate measles related deaths [1], [3]. Hence, generating reliable estimates of measles CFR in a variety of settings and at different periods of time at the national and sub-national levels is important for monitoring the impact of national immunization programmes.

India has had low coverage with measles-containing vaccine (MCV), 74% among children aged between 12–23 months in 2008 [4], and was one of the last countries to provide a second dose of MCV. Critical inputs to the estimation of measles mortality in India are accurate estimates of CFR. According to a review of 25 studies published during 1970–2008 from 12 states, the median case fatality ratio was 1.63% [5]. New estimates are needed from additional states, especially those with low MCV coverage, and in the period after the introduction of a second dose of measles vaccine and the recently-concluded measles catch-up immunization campaign (MCUP) [6].

Bihar, one of the north Indian states, is consistently having lower rates of childhood vaccination. MCV1 coverage was only 58% in 2008 [4], increasing to 76% in 2010–11 according to the Annual Health Survey [7]. MCUP were conducted in the state during December 2010–January 2011 in five districts and during November-December 2011 in another 15 districts, out of the total 38 districts, followed six months later by the introduction of a second dose of MCV into the routine immunization schedule.

Since August 2011, WHO-National Polio Surveillance Project (NPSP) – primarily established for polio surveillance – has been supporting measles surveillance in the state. To the best of our knowledge, no information is available about measles CFR in Bihar. In order to better define measles CFRs in India and specifically in Bihar and identify risk factors associated with mortality, we conducted a retrospective cohort study in areas affected by measles outbreaks.

Materials and Methods

Sampling Frame and Sampling Procedure

During 1 October 2011 to 30 April 2012, measles surveillance system in Bihar was notified of 136 suspected outbreaks (triggers), 47 of which were further investigated. Based on the laboratory criteria mentioned above, 31 outbreaks were due to measles, 2 were due to rubella and 14 were negative for both measles and rubella. In four of the 31 measles outbreaks, one serum sample each was also positive for rubella. All 31 measles outbreaks occurred in 11 districts that had not yet conducted the MCUP campaign. During this period no outbreaks were reported from the seven other districts that had not yet held the MCUP campaign or from the 20 districts that had already conducted MCUP campaigns during 2010–2011.

For the study, 50% of outbreaks were selected from the 8 districts reporting >1 outbreak using probability proportional to size of the outbreak, and all outbreaks from 3 districts reporting a single outbreak. Each laboratory confirmed outbreak was considered as a cluster. All tolas (section of the village)/urban localities surveyed by the district health authorities during the outbreak investigation/response were considered as the geographical boundary of the selected cluster and all households in these tolas/urban localities as the cluster size (Table 1). In total, we selected 18 outbreaks in 11 districts out of the total 31 measles outbreaks. Of the 90 sera collected for laboratory confirmation from these 18 outbreaks, 68 were positive for IgM antibodies against measles. The remaining 22 sera were tested for IgM antibodies against rubella, of which 2 samples, one each from 2 outbreaks, were positive.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Laboratory confirmed measles outbreaks detected during October 2011 and April 2012 and sampling details.

https://doi.org/10.1371/journal.pone.0096668.t001

Results

We surveyed 9350 households covering a population of 54,297 from 87 village tolas/urban localities from 18 outbreaks in 11 districts. Of these, 1384 (14.8%) households were locked while 16 (0.2%) refused to participate (Fig. 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Flowchart showing measles case ascertainment, Bihar, India, 2011–12.

https://doi.org/10.1371/journal.pone.0096668.g001

Measles Deaths and Case Fatality Rate

Of the 3,670 measles cases in 16 outbreaks surveyed, 28 died within 30 days of rash onset with an overall adjusted CFR of 0.78% (95% CI: 0.47–1.30). Fourteen cases each died during the first and second week of rash onset. As compared to the surveillance system, we identified 14 additional deaths (50% of the total measles deaths); four of whom had died after the investigation of the outbreak. Measles CFR in different districts is presented in Table S1 in File S1.

CFR decreased with age, from 2.3% in infants to 1.1% in children aged 1–4 y to 0.42% in children aged > = 5 y. The CFR was significantly higher cases who did not receive vitamin A during the illness, those who belonged to families whose head was illiterate, had an income below the poverty line or belonged to SC/ST (Table 2). CFR tended to be higher among single cases or primary cases in households compared to secondary cases, though the difference was not statistically significant. Among the under-five as well as older children, CFR was significantly higher among children belonging to SC/ST compared to G/OBC (Table 3). Considering only measles cases who did not receive vitamin A during outbreak response, the adjusted CFR was significantly higher among children belonging to SC/ST as compared to G/OBC (Table 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Measles case fatality rate by selected variables, in 16 outbreaks, Bihar, India, October 2011 to April 2012.

https://doi.org/10.1371/journal.pone.0096668.t002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Measles case fatality rate among SC/ST and G/OBC cases stratified by age group and vitamin A administration, Bihar, India.

https://doi.org/10.1371/journal.pone.0096668.t003

The median interval between the occurrence of the first case in these outbreaks and the notification of the outbreak was 49 days (IQR: 20–67 days), the interval between notification and outbreak response was 17 days (IQR: 9–24 days), and the median interval between the onset of first case and initiation of outbreak response was 70 days (IQR = 39–82 days). Measles cases occurring after the outbreak investigation had significantly lower coverage of vitamin A as compared to those occurring before (31.7% vs 38.7%, p = 0.01).

On multivariate analysis, independent risk factors associated with higher odds of death included age <5 years, SC/ST status, and non-administration of vitamin A. Outbreaks with longer interval between the rash onset of first case and the notification of outbreak also had higher deaths (Table 4). Though the CFR was higher among measles cases belonging to SC/ST and who did not receive vitamin A, a zero-cell value in table 3 (zero deaths among the SC/ST cases who received vitamin A) prevented the estimation of odds ratio for the interaction term for caste and vitamin A treatment.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Multiple logistic regression analysis of the risk factors associated with measles death during 16 outbreaks in Bihar, India, October 2011 to April 2012.

https://doi.org/10.1371/journal.pone.0096668.t004

Discussion

This community-based study of measles deaths in 16 outbreaks in Bihar, India found that the overall CFR was 0.78%, but varied from 2.3% in children less than one year to 1.1% in children 1–4 years of age to 0.4% in children five years and older. Measles mortality was also higher among children who belonged to SC/ST, those families whose heads were illiterate, and among children who were the first or only case in a household. Most deaths occurred among measles cases who did not receive vitamin A. The CFR observed in our study was lower than the median CFR of 1.63% reported in a review based on 25 studies done earlier in India [5]. Comparisons with earlier studies is difficult as 8 used prospective case ascertainment, all in the pre-vaccination era, and though the remaining studies used house-to-house surveys for cases, most surveys were done right at the end of an outbreak or in periods of “endemic disease”, with some cases likely to be surveyed before 1 month had elapsed since rash onset. Most previous surveys in India did not clearly describe use of household lists and lists of cases in order to maximize the ascertainment of deaths. The CFR in Bihar was comparable to that found in Nepal, where the CFR was 1.1% (95% CI: 0.5–2.3) overall, 3.5% in cases <1 year of age, 2% in cases 1–4 years, and 0.2 in cases 5–14 years of age [2]. Methods for ascertainment of measles cases and death in the current study were similar to the study conducted in Nepal. The findings of our survey have implications in further reducing the measles CFR in Bihar as well as in India. CFR estimates will also improve the estimates of measles deaths in India.

During 2000–2011, the global estimates of measles mortality decreased by more than three-quarters, however in the south-east Asian region the decrease was only by half, with India alone accounting for 36% of the total deaths in 2011 [13], [14]. The measles mortality model developed by the World Health Organization uses vaccination coverage estimates and surveillance data to estimate the number of measles cases by country. The estimated age distribution of cases and case fatality rates are then used to calculate the number of deaths [13], [15]. The findings of estimated number of measles deaths from India should change with the completion in 2013 of the MCUP in 14 states with MCV1 coverage <80% and introduction of a second dose of measles vaccine in routine in all states. Having updated estimates of measles CFR in high-burden countries such as India will be critical for accurately estimating national, regional and global mortality.

In Bihar, by April 2012, no outbreaks had been detected by the surveillance system in the 20 districts of Bihar where MCUP had been conducted during 2010 and 2011 (Measles surveillance in Bihar, unpublished data) indicating a potential early impact of the MCUP. The coverage of MCUP in Bihar using administrative data was 93% (NPSP, unpublished data). A significant decline in the number of measles following MCUP was also reported in the states of Maharashtra, Rajasthan and Gujarat (NPSP, unpublished data). These states reported 3188 measles cases one year before the MCUP compared to 69 cases reported within one year of MCUP (NPSP, unpublished data). In the 11 study districts of Bihar where MCUP had not been conducted by 2012, implementation of enhanced measles surveillance with support from WHO India-NPSP probably has still improved the CFR through the detection and laboratory confirmation of suspected outbreaks as well as their investigation and appropriate response including administration of vitamin A. MCUP in the remaining 18 districts of Bihar was conducted during Jan 2013.

The lower overall CFR observed in Bihar might be related to the older age of measles cases. Though CFR was higher among under-five children, over half of the measles cases were aged > = 5 years and the CFR was lowest (0.41%) in this age group. The declining CFR with age observed in our study is consistent with the studies conducted in other countries as well as community based studies in India [5], [11]. The higher proportion of cases among older children reflects the improvement in MCV1 coverage. MCV1 coverage in Bihar has increased from 40.4% during 2005–6 [16], to 76% during 2010–2011 [7].

The association with vitamin A and lower measles mortality found in Bihar is consistent with reports from several studies conducted in other countries [1], [17]. The association between measles mortality and vitamin A was modified by caste, with higher CFR among those not receiving vitamin A in SC/ST. SC/ST is one of the most disadvantaged socio-economic groups recognized by the Constitution of India. In Bihar, there were significant differences between the measles cases belonging to SC/ST and G/OBC with respect to their socio-economic status as well as age distribution. These factors could partly explain the differences in CFR. It is therefore necessary to improve coverage with two doses of measles vaccine in these communities and during measles outbreaks prioritize them for vaccination activities and for vitamin A administration.

Our survey identified nearly twice the number of measles cases as compared to the cases identified by the surveillance system, with a reporting fraction of 51%. Overall, <5% measles cases received treatment from public health facilities which is likely to affect the sensitivity of the surveillance system and may be the reason for delayed notification of outbreaks [18]. The majority of measles cases in these outbreaks sought treatment from traditional healers. Similar health seeking behaviour was also reported by other studies in India [19], [20]. In Bihar, the interval between the occurrence of the first case and the notification of outbreak was long, with investigation and response beginning after a median interval of two weeks.

Late notification of outbreaks resulted in a lower overall coverage of vitamin A administration to measles cases and thereby higher case fatality. Our survey detected 50% more measles deaths than the surveillance system and 29% (4/14) of these cases had died after the investigation of the outbreak. Measles cases occurring after the investigation of the outbreak had lower coverage of vitamin A. These findings suggest poor adherence to vitamin A treatment of measles cases occurring after the investigation of outbreak.

Several studies have reported milder infection, reduced complications and lower CFR among measles vaccinated children [11], [21], [22]. It has been suggested that higher measles vaccination coverage lower mortality rates directly through reduced disease incidence and indirectly by increasing the median age at infection, resulting in more cases in older age groups with lower mortality rates, or by decreasing incidence, resulting in a lower intensity of transmission [11]. These studies often relied on records of vaccination status before the outbreak [21] or had higher proportions of cases vaccinated [2], [23]. No difference in CFR between vaccinated and unvaccinated children was found in our study, with only 27% of cases with a history of vaccination. Assuming 85% efficacy of the single dose of measles vaccine, 27% measles cases with history of vaccination translates to MCV1 coverage of 71% in outbreak areas [24], which was comparable to the MCV1 coverage of 78% reported in the Annual Health survey – 2010–11 (7). In settings such as Bihar, with low retention of immunization cards (15% vaccinated cases had immunization card in our survey), multiple round of vaccinations, and high proportion of illiterate families, collecting reliable information about vaccination, especially from measles cases who died remains a challenge.

Our study was subject to at least five limitations. First, though we selected laboratory confirmed outbreaks for the survey, due to the retrospective nature of the study, it is possible that some measles cases might have been misclassified, as the case status was based on the symptoms reported by the parents/care-takers. Parents might have inaccurately recalled the occurrence of measles cases in the family as well as some risk factor information. In particular, parents of children dying of measles are more likely to report the risk factor information in a socially desirable direction in order to avoid criticism by the interviewer or out of guilt. We tried to reduce the recall bias by limiting the recall period for measles cases included in the survey to be the period 5–11 months before data collection, and using a local event calendar to aid parents recall. We also followed recommendations to first enumerate all the household members at the beginning of the recall period and use this list to identify who developed rash illness and died within 30 days of rash onset [10]. Interviewing mothers to collect information about the child’s illness and focusing the study on laboratory-confirmed outbreaks increases the positive predictive value of the clinical diagnosis of measles [10]. We used local terms and standardized clinical algorithm to ascertain the cases and deaths. Second, about 15% of the households were locked at the time of the visit and could not be surveyed. We collected information about measles deaths from the village head and the local village guide prior to the survey. No death was reported from the households that were locked. However, as the information about measles cases was not collected from these houses, we might have overestimated the CFR. Third, measles surveillance in the country is presently geared to capture larger outbreaks with at least 20 cases. Measles cases and deaths occurring in smaller outbreaks were not included in our sampling frame. Fourth, the confirmation of measles outbreaks is based on laboratory testing of only five sera samples. Although we excluded 2 outbreaks with co-circulation of rubella, rubella may have been co-circulating with measles in the outbreaks included in the analysis. However, requiring laboratory confirmation of all cases would be difficult to ensure and ethically challenging while at the same time ensuring complete case ascertainment and information on outcomes 1 month after rash onset [10].Lastly, the study results might not apply to other time periods or other Indian states as measles incidence and severity varies by season, epidemic cycle, geography and levels of population immunity to measles.

This case fatality rate survey followed expert guidelines in the choice of a retrospective survey to ascertain all cases and deaths due to measles in the affected areas of Bihar. The inclusion of only outbreaks with at least 20 cases provided a sampling frame of confirmed outbreaks but may have introduced some bias as smaller outbreaks would necessarily have been excluded. Any smaller outbreaks that occurred in Bihar by definition would have involved many fewer cases relative to the ones chosen and probably no deaths. Their exclusion should not have significantly biased the CFR estimates. Given the status of measles control in Bihar it is likely that small outbreaks would grow and coalesce with neighbouring outbreaks to reach a size where they would have been detected, investigated, and eligible for inclusion in this study.

In conclusion, the overall measles CFR in Bihar was found to be lower than other studies from India. The lower rate may represent fewer cases in these outbreaks in groups where CFR was higher, such as young children. Children belonging to SC/ST groups however still have high CFR. Delayed notification of outbreaks might also have led to higher mortality. More systematic communication efforts are needed to promote parents to ensure their children receive two doses of measles vaccine and bring children with measles to public health facilities for treatment of measles complications. Improvements in vaccination coverage should reduce the size and number of future outbreaks in Bihar and thereby reduce the number of measles deaths. Further reductions in measles case fatality could result from timely outbreak investigation and focusing vitamin A administration to SC/ST communities and to children <5 years of age. Although the study was based in 11 districts of Bihar, the recommendations for reducing measles case fatality are applicable to measles outbreaks response in other Indian states. Finally, similar surveys need to be conducted in other states and repeated periodically in order to generate reliable estimates of measles mortality and monitor progress in measles control in India.

Supporting Information

Acknowledgments

We gratefully acknowledge the state health officials, Govt of Bihar for their support and cooperation during the study. We would like to thank the Surveillance Medical Officers/Sub-Regional Team Leaders of WHO India–NPSP, Bihar for their valuable contribution in supervision of the field work. Thanks are also due to the Field Volunteers who conducted the data collection. The constructive feedback given by Drs Sanjay Mehendale and B.N. Murthy is gratefully acknowledged.