Third National Family Health Survey in India: Issues, Problems and Prospects

NATIONAL FAMILY HEALTH SURVEY-3Economic & Political Weekly EPW november 29, 200833Third National Family Health Survey in India: Issues, Problems and ProspectsS Irudaya Rajan, K S JamesThe three rounds of the National Family Health Survey have generated vast amounts of data, which unfortunately have been subject to only limited critical examination by Indian research scholars, though the opposite is the case with scholars outside India. The NFHS-3, which was conducted in 2005-06, covered many more areas than the previous surveys and collected information in new and sensitive areas like sexual behaviour. However, there are questions about the quality of data thrown up byNFHS-3. Information on some indicators such as fertility and infant mortality remains of reasonably good quality, but the data on nutrition, immunisation, and genderviolence is suspect. There have been three of these very large surveys since 1992-93, and it is perhaps time to reflect on the experience so far and plan for the next survey a decade after NFHS-3 which would be five years after the 2011 Census.We are grateful to the Population Foundation of India, especially its executive director, A R Nanda, for funding that helped to organise a meeting on 26 July 2008 at the Centre for Development Studies, Thiruvananthapuram, to discuss the papers in this special issue. We also thank K N Nair, director of CDS, for all help rendered in organising the meeting.In the untimely demise of P N Mari Bhat, former director, International Institute for Population Sciences, Mumbai, we have lost a highly competent expert who could scrutinise and evaluate the data collected in the National Family Health Survey. He contributed papers to the special issues of this journal on NFHS-1 and NFHS-2. This special issue is dedicated to his memory.S Irudaya Rajan (rajan@cds.ac.in) is at the Centre for Development Studies, Thiruvananthapuram. K S James (james@isec.ac.in) is at the Population Research Centre, Institute for Social and Economic Change, Bangalore.One of the major objectives of the National Family Health Survey (NFHS) is to produce high quality information on demographic and health outcomes and to provide inputs for policy formulation, both at the national and state levels. The quality of data collected in this massive exercise is of utmost importance as it can lead to fallacies in understanding the emerg-ing and changing demographic and health scenarios. With this in mind, we evaluate the reliability of the third NFHS (2005-06) data using various qualitative and quantitative checks. TheNFHS-1 (1992-93) was considered a landmark event which yielded a wealth of demographic and health data for the country as a whole for the first time (Visaria and Irudaya Rajan 1999). Unlike theNFHS-1 andNFHS-2 (1998-99), the NFHS-3 included un-married women in the reproductive age group 15-49 as well as men. In addition, the enquiry was widened both in terms of coverage and content, using biomarkers for measuring HIV/AIDS and introducing several new questions pertaining to sexual life, marriage and family relations.In our overview of the NFHS-2 (Irudaya Rajan and James 2004), we concluded that there were sufficient grounds to suspect the quality of the data, compared to that of the NFHS-1, though it wasexpected that with each successive round, the quality of the information would improve. In addition, we argued strongly in favour of including men in the NFHS (Irudaya Rajan and James 2004; Roy 2004). For the first time, a separate module for men in the age group 15-54 was included in the NFHS-3. This provides a unique opportunity to evaluate the quality of the survey by com-paring the responses of couples within a household.It is always challenging to obtain reliable demographic data in India, where there are very high levels of illiteracy and social bar-riers that exclude many communities. In an earlier analysis, we argued that the quality of data has a direct link with the educa-tional status of respondents (James and Irudaya Rajan 2004). With social silence on several issues, particularly sexuality, it is often difficult to elicit responses to sensitive questions in a large-scale survey in India. However, the NFHS-3 decided to incorporate several such questions, both for men and women (International Institute for Population Sciences and Macro International 2007). This made the NFHS-3 questionnaire significantly different, bulkier and perhaps more sensitive than those of the earlier surveys. Complete privacy was a necessary condition to administering it.Serious underestimation is expected in the case of several sensitive questions in large surveys, and Lekha Subaiya’s paper elsewhereinthisissuebrings out the gender and class dimen-sions in reporting about pre-marital sex in India. For women,

NATIONAL FAMILY HEALTH SURVEY-3november 29, 2008 EPW Economic & Political Weekly34pre-maritalsexhas serious implications because they mostly lack the power of self-determination and are reluctant to speak about it. This is also true in the case of domestic violence (see Leela Visaria’s paper in this issue). Even empowerment of women does not seem to help significantly either in encouraging report-ing of violence or providing effective safeguards. Regardlessof socio-economic background, women seem to accept the right of men to discipline them and often justify physical violence in certain circumstances. Interestingly enough, the publication of provisionalNFHS-3 data through fact sheets in 2006 and the entire report in 2007 has brought out several paradoxes in the country (International Institute for Population Sciences and Macro International 2007). Bose (2006) has expressed concern over the paradox of declining fertility and increasing malnutrition. Though the Indian economy is growing at around 8% a year, the NFHS-3 indicates worsening nutritional levels. It also reports a decline in child immunisation despite significant increases in the budget allocation for health in the last few years, particularly the reproductive and child health programme (see the paper by Srinivasan and Mohanty in this issue) and the National Rural Health Mission (Srinivasan et al 2007). This has come as a rude surprise to health administrators, and it is vital to assess the quality of information collected through theNFHS-3. Demographers use several indicators to assess the accuracy of data collected in censuses and sample surveys. Different methods have been developed to measure the extent of errors in the age data and its impact on estimates of fertility and child mortality indicators (Shryock and Siegal 1976). The NFHS-3 made a serious effort to overcome anticipated problems and obtain quality informationinthese areas. We shall therefore discuss the quality issue using other approaches. Size Does MatterAs stated earlier, the NFHS-3 broadened its area of enquiry substantially. It also ventured into collecting data on issues that are often considered too sensitive to include in large surveys. So, there was significant increase in the number of questions administered per household. A rough estimate of the number of ques-tions administered in theNFHS-3 to women in the age group 15-49 is more than 450 – almost double that in the NFHS-2. Table 1 presents the number of questions administered in the three rounds of the NFHS.As evident from Table 1, the new areas of enquiry in theNFHS-3 were marriage and cohabitation, contacts with health personnel, the sexual life of never and ever-married women and the utilisation of the Integrated Child Development Scheme (ICDS). The number of questions went up considerably in sections like antenatal and postnatal care, immunisation, nutrition, household relations and HIV/AIDS. The printed version of the NFHS-3 questionnaire ran into about 130 pages (inclusive of household, women and men). Did the size of the questionnaire lead to a deterioration in the quality of data? The inverse relationship between length of a questionnaire and the quality of responses is often an accepted fact although no empirical evidence exists to verify this (Bogen 1996).As we have no direct empirical evidence, we have calculated the time taken to administer the women’s questionnaire and used it as a proxy to assess the quality of the data. Table 2 presents the average time taken to interview women in three selected states in theNFHS-3, NFHS-2 andNFHS-1.The average time taken for administering the women’s sched-ule in the NFHS-3 was around 70 minutes, 21 and 28 minutes higher than in theNFHS-2 andNFHS-1 respectively. As already pointed out, an increase in the length and sensitivity of the questionnaire meant that the time necessary to administer it was more. Bogen (1996) found that the increased length of a questionnaire adds to the bur-den on respondents and pushes more of them over the threshold beyond which they will no longer cooperate or provide good responses.This is also true of the research investigators.It is interesting to note that there is wide variation in the time spent for the survey in different states across India. While it was as high as 86 minutes in Tamil Nadu, it was almost half that long in Haryana. We extensively reviewed con-sulting organisations and their efficiency while evaluating the NFHS-2 data (Irudaya Rajan and James 2004). The quality of investigators, for instance, has a direct link with the quality of the information gathered from the field. The amount of honorarium inNFHS-3 and its mode of payment to the investigators by different agencies are not clear to us. This is Table 1: Number of Questions Canvassed in the Three Rounds of NFHSAreas of Enquiry No of Questions NFHS-1NFHS-2NFHS-3Women’s questionnaire respondent’sbackground 56 24 18 Reproduction 32 34 42 Marriage and cohabitation na na 18 Contraception 54 50 39 Contacts with health personnel na na 18 Quality of care na 19 na Antenatal and postnatal care 47 52 82 Immunisation, health and women’snutrition 51 47 83 Utilisation of ICDS na na 15 Sexuallife na na 20 Fertilitypreference 18 10 23 Husband’s background and work 21 22 29 Household relations (status of women) na 7 29 HIV/AIDS na 6 38Total 279 271 454Number of questions in household questionnaire 79 64 68Number of questions in men’s questionnaire na na 216Grand Total 358 335 738Compiled by the authors; (1) na= Not administered. (2) Numbers are only approximate and indicative. All questions were not asked to all women because the answer to loop questions determined whether the following question would be asked. At the same time, there were several questions like information on birth history that were repeated based on the number of children born and surviving.Table 2: Average Time Taken (in Minutes) to Administer the Women’s Schedule in Selected States, Three Rounds of NFHS State NFHS-3NFHS-2NFHS-1DifferenceDifference (1) (2) (3) (1)-(2) (2)-(3)Least time taken to interview women Haryana 45.19 34.4 36.24 10.79 -1.84 Punjab 50.85 36.86 30.35 13.99 6.51 HimachalPradesh 57.76 36.79 45.19 20.97 -8.40Longest time taken to interview women Bihar 76.79 39.5 48.05 37.29 -8.55 Orissa 76.79 46.99 39.42 29.8 7.57 TamilNadu 86.06 71.15 65.91 14.91 5.24India 68.91 48.5441.94 20.376.60Estimated by the authors; Only women having at least one child less than three years of age were selected for the antenatal care and immunisation section and only one currently married woman from each household (for the household relation part) was selected for computation of time.

).

73.2 72.1 96.5 94.4 78.5 779 64 2 60.1 NFHS 2 NFHS 3

Radhakrishna, R and C Ravi (2004): “Malnutrition in India: Trends and Determinants”, Economic & Political Weekly, Vol 39, No 7, pp 671-76.

Bose, Ashish (2006): “Falling Fertility and Rising Anemia?”, Economic & Political Weekly, Vol 41, No 37, 16 September, pp 3924-36.

Fred Arnold, Praveen Nangia and Umesh Kapil (2004): “ Vol 39, No 7, pp 664-70.