We examined the efficiency of the health system in improving maternal health outcomes, i.e. increasing institutional delivery (desirable output) and reducing the maternal mortality ratio (undesirable output). We used a nonradial, nonoriented Slack based measure data envelopment analysis model to assess the efficiency of the health system across 19 states of India between 2016 and 2020. The overall average efficiency score suggested potential for a 30% improvement in the health system efficiency in India, with only six states emerging as efficient. Interestingly, few states that were endowed with more than adequate health resources were inefficient. We find that resource constraint states such as Bihar and Jharkhand have emerged as efficient. To substantiate the use of undesirable adjusted model instead of the transformed model, we used the Wilcoxon ranked-signed test to estimate the statistical significance. We conclude that incorporating undesirable outputs while measuring efficiency is imperative. Our findings suggest that states that are inefficient with excess capacity should prioritize the efficient utilization of resources. Additionally, adequate resources should be allocated to resource constraints and efficient states such that they achieve better health outcomes. Furthermore, the diversification of health resources across levels of care could be a suggestive measure for some states.