Data from multiple sources for policy, planning, and actions: How valid it is?

Availability of data is foremost and crucial for planning, implementation, monitoring, and evaluation of an intervention in every sector; public health being no exception. However, the data generated should be complete, up to date, and representative of situation. No matter how much powerful data appear to be, it can play many tricks while collection to analysis due to myths and traps that lie within, which are commonly referred as data fallacies. This may ultimately lead to draw incorrect conclusions and make poor decisions.[1] In order to get a true and powerful insights, availability of data from multiple sources is desirable as more complex and diverse the datasets are, more surprising, and potent insights it may produce. Additional data sources also increase the scope of “informed actions” fueling top-line and bottom-line growth. Data generated from multiple sources through a triangulation exercise help in getting the true picture. Globally, countries with high human development index ranking have robust demographic data systems for centuries. Such data help them address many socioeconomic problems in a decentralized way. High-quality, open, transparent, and uncensored demographic data are essential for creating an accountable system. It helps measure the success of policies and programs adopted by the governments.[2] Interestingly, India has multiple sources of demographic data such as census, sample registration system (SRS), National Family Health Survey (NFHS), and civil registration system (CRS); all have been crucial for planning and evaluation; however, none of them is comprehensive enough to address all the three components of population dynamics, namely, births, deaths, and migration. Census, a huge, resource-intensive exercise, undertaken decennially provides a limited data on births and no information on deaths and immigration.[2] Thanks to Corona, census due for 2021 has been postponed to 2023–2024. While the census and CRS are based on true and complete reporting of events, SRS and NFHS gather the information from a small subsample selected through robust sampling techniques at periodic intervals. Often, we face a situation where the similar type of data collected from multiple agencies/sources provides different values. Reasons for this variability can be attributed to sampling methodology, sample size, and agency involved in data collection. This leads to a question whether it is justified to spend scarce resources to generate multiple data for same outcome and later which of these multiple data is to be relied upon. HOW FAR IS OUR CIVIL REGISTRATION SYSTEM RELIABLE? CRS, popularly known as birth and death registration system, is the recording of vital events such as birth, death, stillbirth, and marriage under the statutory provisions on a continuous and permanent basis. It started as a voluntary reporting in the middle of the 19th century (1834) for British India, but was made compulsory in the entire India under the provisions of the Birth and Death Registration act, 1969.[3] Undoubtedly, CRS is on improvement for the past few decades due to (a) data digitization and (b) registration of birth/death provides a legal identity/access to the rights of a citizen including entitlements and social benefits provided by the governmental agencies. Till recently, public health personals were critical of CRS as it is plagued with (a) inaccurate/incomplete enumeration of deaths as well as births and (b) inadequate medical certification of deaths limiting the ability to infer the cause of death.[4] Based on CRS (2019), the level of registration of births and deaths was 92.7% and 92.0%, respectively. Globally, the WHO reported some 15 million deaths due to Corona in 2020 and 2021, including 4.7 million in India.[5] As per this report, India accounted for the highest number of deaths due to the Corona, and this number given by the WHO was almost 10 times high to the number of corona deaths officially reported by India. While some countries accepted this disparity gracefully, in India, all governmental agencies stood up collectively to defend our CRS which reported Corona-related deaths between 4 and 5 lakhs. CRS according to them is very old robust system and working perfectly well; hence, the claim of the WHO cannot be accepted. However, the popular perceptions and media reports do acknowledge the under reporting of mortality. According to the SRS-GOI in 2020, some 8.1 million deaths were occurred, and 99.9% of these were registered in CRS. In fact, the CRS 2020 also acknowledges the registered death as 8.12 million.[6] This claim has two parts about (a) mortality and (b) registration completion; however, it has only one. If we believe that only 8.1 million deaths were occurred in 2020, then data in the 2020 CRS report confirm that essentially all these deaths were registered. Hence, the question is whether the estimate of 8.1 million annual deaths is correct or not?[7] It may be noted that the CRS 2020 reports approximately 6 lakhs more deaths and 6 lakhs less births in 2020 than 2019 without any explanation. The SRS is an annual exercise done in randomly sampled units and provides mortality estimates. During 2020, SRS estimated the crude death rate (CDR) as 6.0 per 1000, and with India's population around 1.35 billion, this gives 8.1 million deaths. The percentage of registration of mortality is also available with NFHS-5, where a question was asked with surveyed families whether the death(s) in the family were registered or not. SRS does underreport the mortality, and whenever excess deaths are reported, the government system defends this by assuming more registration of events (reaching to 99.9% in 2020). Bihar and Uttar Pradesh, where the CRS is undoubtedly weak, account for a quarter of the national population. According to the CDR estimates as per SRS, annually 2.1 million deaths are expected. According to this CRS report, Bihar and UP together reported 1.3 million death registrations in 2019. Hence, as per the SRS itself, around 800,000 deaths went unregistered in UP and Bihar. When this exercise was extended to all the states, 1.2 million death registrations were found missing in the pre-Corona period of 2019. Mortality registration cannot be above 95% nationally when the data indicate such many unregistered deaths.[7] This entire exercise was possible due to the availability of data from SRS and NFHS-5. It is worth knowing that it was for the first time that the NFHS-5 recorded the proportion of registered deaths and found this as 70.8%.[8] INFANT MORTALITY RATE IN GUJARAT – SAMPLE REGISTRATION SYSTEM VERSUS CIVIL REGISTRATION SYSTEM As per the SRS, the infant mortality rate (IMR) in Gujarat in 2020 was 23 per 1000 live births. For the same year, as per the CRS, there were 12,119 infant deaths in 1,103,241 births – giving an IMR of 11 per 1000 live births. It is worth knowing that the IMR for 2019–1020 in Gujarat as per the NFHS-5 was 31.2 per 1000 live births; separately being 24.1 for urban and 35.5 for rural areas [Table 1]. Mute question is which one is to be relied on as the difference is too much between IMR of 31.2 (NFHS-5) and 23 (SRS) or even 11 (CRS). That the truth will be somewhere in between, but one thing is for sure that CRS underreports mortality. Not only that, CRS will also miss the births as well, especially for early neonatal mortality. Finally, we would say that despite all the defense raised for the CRS, if our CRS is as good as claimed, why should we have the SRS at all. SRS is needed to make some projections only when CRS is not reliable. The same disparity exists when we look at the India's IMR for the same period with the same sources, i.e., NFHS-5. Interpretation of this disparity becomes difficult as values at 95% CI are not given for IMR derived from the NFHS-5. Further when we extend this comparison with values derived in CRS (registered live births and infant deaths), the difference becomes more obvious. Low values of IMR derived with CRS prove again the underreporting of vital events and question its validity. Rather than calling it very robust, it highlights the need to improve the CRS making it complete and more inclusive. A similar disparity may be seen in case of maternal mortality ratio as derived from SRS and compared with CRS.Table 1: Infant mortality rate/1000 live births for India and Gujarat[8 , 9 , 10]DATA TRIANGULATION INSIGHTS Small and local exercise always highlights the disparity even more. In a study from 51 villages of three primary health centers from the tribal belt of South Gujarat,[11] information was gathered for 1 year from four sources, namely, health-care system, Integrated Child Development Services (ICDS) scheme, CRS, and investigator himself. A total of 48 infant deaths were recorded by investigator against 2, 10, and 8 infant deaths reported by the CRS, health system, and ICDS, respectively. Such a wide variation in reported events from the same area for the same period by multiple agencies creates a question that which data should be relied on for data-based planning, management, and monitoring of programs. OVERALL SEX RATIO The NFHS-5 data (2019–2021) indicate that the All-India sex ratio crossed 1000 for the first time and is 1020 females per 1000 males.[12] As per the NFHS-4 (2015–2016), this ratio was 991. It seems highly improbable to gain 29 points in <5 years. If the data are further segregated to states and union territories, the change seems impossible to occur. For example, a small territory of Lakshadweep islands gained 165 points, and a state Kerala, which already had a favorable sex ratio of 1049, gained another 72 points all within <5 years. SEX RATIO AT BIRTH The sex ratio at birth as per the NFHS-5 is 929 females per 1000 males, quite comparable with the sex ratio of 922.9 females per 1000 males at birth (CRS, 2020).[6] Furthermore, except for Arunachal Pradesh and Ladakh, in the same report, it was everywhere not

Abstract licence: CC BY-NC-SA

2 more references, whose wording the paper’s licence keeps private.

A person counts once a day per paper. We keep no cookie and no address: a reader is a hash that changes every day. Any month under five reads shows no number.