Optimizing use of health system generated poor quality data: evidence using MCTS/HMIS data in improving maternal and child health outcomes in lowest performing state of Bihar, India

Arup Das, International Institute for Population Sciences (IIPS)

India is undergoing a paradigm shift, focussing on evidence based approach. Ministry of Health (MoH) is generating individualised & indicator based aggregated data from Mother-Child-Tracking -System (MCTS) and Health-Management-Information-System (HMIS) respectively, however, the quality of data still remains a bottleneck in data use. This paper proposes a methodology for systematic review (step-by-step) of data quality. The methodology should be able to provide -solution to the problem of identifying the data quality issues, -specific remedial strategy and -suggest data driven management at all level of program implementation. The data is downloaded from the MoH portal for financial years 2010-2011 and 2011-12. The other data used for triangulation are DLHS-3 & NFHS-3. The steps are: -Framing hypothesis related to program response gaps -Identifying missing data and checking internal inconsistencies -Generating relevant indicators using MCTS/ HMIS and triangulation using DLHS/NFHS for external validation -Classification of results into “program response gaps” and data quality gaps” -Verifying results with the program managers and community workers through qualitative interactions The existing denominators are compared with the denominators generated using indirect estimations (UN Manual-X). In addition to the descriptive analysis, whipple’s index (WI) is computed for assessing quality of age data. Control chart techniques (for mean, range and SD) are used, assuming Poisson distribution, to identify the data points that are out of control. Both data source are historical, hence ARIMA and VAR techniques are used to assess predictability. The finding shows that the denominators used for target setting are inappropriate and can be improved using indirect techniques. There are inconsistencies between the two data sources. The control charts suggest that significant proportion of data points of several indicators are out of control. We have shown that LQAS can be used as a tool to identify whether the quality of data is under accepted level of margin.