Research Article: Understanding intersections of social determinants of maternal healthcare utilization in Uttar Pradesh, India

Date Published: October 4, 2018

Publisher: Public Library of Science

Author(s): Arnab Dey, Katherine Hay, Bilal Afroz, Dharmendra Chandurkar, Kultar Singh, Nabamallika Dehingia, Anita Raj, Jay G. Silverman, Olalekan Uthman.

http://doi.org/10.1371/journal.pone.0204810

Abstract

To explore intersections of social determinants of maternal healthcare utilization using the Classification and Regression Trees (CART) algorithm which is a machine-learning method used to construct prediction models.

Institutional review board approval for this study was granted from Public Health Service—Ethical Review Board (PHS-ERB) and from the Health Ministry Screening Committee (HMSC) facilitated by Indian Council for Medical Research (ICMR). IRB review and approval for the current analyses was obtained from University of California, San Diego. Cross-sectional data were collected from women with children aged 0–11 months (n = 5,565) from rural households in 25 districts of Uttar Pradesh, India. Participants were surveyed on maternal healthcare utilization including registration of pregnancy (model-1), receipt of antenatal care (ANC) during pregnancy (model-2), and delivery at health facilities (model -3). Social determinants of health including wealth, social group, literacy, religion, and early age at marriage were captured during the survey. The Classification and Regression Tree (CART) algorithm was used to explore intersections of social determinants of healthcare utilization.

CART analyses highlight the intersections, particularly of wealth and literacy, in maternal healthcare utilization in Uttar Pradesh. Model-1 documents that women who are poorer, illiterate and Muslim are less likely to have their pregnancies registered (71.4% vs. 86.0% in the overall sample). Model-2 documents that poorer, illiterate women had the lowest ANC coverage (37.7% vs 45% in the overall sample). Model-3, developed for deliveries at health facilities, highlighted that illiterate and poor women have the lowest representation among facility deliveries (59.6% vs. 69% in the overall sample).

This paper explores the interactions between determinants of maternal healthcare utilization indicators. The findings in this paper highlights that the interaction of wealth and literacy can play a very strong role in accentuating or diminishing healthcare utilization among women. The study also reveals that religion and women’s age at marriage also interact with wealth and literacy to create substantial disparities in utilization. The study provides insights into the effect of intersections of determinants, and highlights the importance of using a more nuanced understanding of the impact of co-occurring forms of marginalization to effectively tackle inequities in healthcare utilization.

Partial Text

The Sustainable Development Goals for 2030 continue the focus on maternal health, pushing beyond the targets set by the Millennium Development Goals for 2015. The SDGs call for reducing maternal mortality across the globe to less than 70 per 100,000 live births by the year 2030. This goal also highlights the immense disparities between developing and developed regions in their contributions to the global maternal mortality ratio [1]. Globally, 99% of all maternal deaths are estimated to occur in developing regions [2]. India, with a Maternal Mortality Ratio of 167 [3] contributed nearly 15 percent of global maternal deaths in 2015 [2]. In addition to huge disparities existing between developed and developing regions, maternal mortality is disproportionately high among socially and economically marginalized communities within regions[4, 5].

Institutional review board approval for this study was granted from Public Health Service—Ethical Review Board (PHS-ERB) and from the Health Ministry Screening Committee (HMSC) facilitated by Indian Council for Medical Research (ICMR). IRB review and approval for the current analyses was obtained from University of California, San Diego.

Participants included 5,565 women with children aged 0–11 months. The mean age of participants was 26.4 years (SD = 4.4) and their mean age at first marriage was 18.1 years (SD = 2.4). Approximately half of participants were illiterate (53.8%) and the majority belonged to either the SC/ST (29.3%) or the Other Backward Classes (OBC) category (53.5%). Overall, 16.2% of the participants were Muslims and almost half of the participants had 3 or more births at the time of the interview (45.9%). These proportions closely correspond to the Census of India 2011 figures for rural Uttar Pradesh. According to the Census, 54.9% of rural women in Uttar Pradesh were illiterate, 23.6% of rural women belonged to the SC/ST social groups and 19.3% of the women in the state were Muslims [24, 25].

In this paper, we have explored the use of decision trees in quantifying intersections of determinants of inequities in healthcare utilization. This paper contributes to the global discourse on interactions between determinants of inequities and demonstrates the utility of an additional empirical tool to examine intersections quantitatively. The Classification and Regression Tree (CART) analysis used in the paper provides methodological advantages over the traditional approaches that have been used by researchers exploring healthcare utilization patterns among different segments of populations. Research on population segments vulnerable to low healthcare utilization has primarily used bivariate analysis of dependent variables across individual strata (e.g. caste, religion etc.) and logistic regression models assessing associations between outcomes of interest and key social determinants while controlling for other factors. Both of these methods have serious limitations in providing a deeper understanding of the intersections of socio-economic determinants of healthcare utilization. While bivariate analysis does not allow for consideration of multiple co-occurring determinants, regression models fail to specify population segments most at risk, information critical for program and policy guidance [22]. The use of CART mitigates some of these challenges by allowing consideration of multiple forms of inequity to segment populations into clearly defined sub-groups that differ based on a specified outcome, information with clear relevance for programs that intend to reduce inequities in healthcare utilization or similar resource.

The models identify household wealth and literacy as the most prominent determinants whose intersections affect healthcare utilization for all the three indicators. Religion and age at marriage have also been highlighted as determinants leading to crucial intersections for utilization.

One of the limitations of the CART model is that it does not provide an estimate of the relative strength of the determinants in an interaction. For example, questions like whether wealth is more potent than literacy in determining utilization remain unanswered through the use of this models.

Despite the limitations, our analysis provides important insights into intersections of key social determinants in predicting healthcare utilization. The paper reveals the role of interactions between wealth and literacy in creating substantial disparities in healthcare utilization among women living in rural Uttar Pradesh. We also found that religion and age at marriage also interact with wealth and literacy to exacerbate disparities. The understanding of such interactions is still evolving and further research across multiple populations and geographies is required to determine its broader application. Future research should assess the utility of exploring intersections as an approach to understand other developmental issues (e.g., completion of education). Finally, although it remains to be seen how these insights may be successfully utilized from a programmatic perspective, programmatic responses to the intersections of multiple social determinants and persistent health inequities are clearly warranted.

Institutional review board approval for this study was granted from Public Health Service—Ethical Review Board (PHS-ERB) and from the Health Ministry Screening Committee (HMSC) facilitated by Indian Council for Medical Research (ICMR). IRB review and approval for the current analyses was obtained from University of California, San Diego.

 

Source:

http://doi.org/10.1371/journal.pone.0204810

 

0 0 vote
Article Rating
Subscribe
Notify of
guest
0 Comments
Inline Feedbacks
View all comments