This is a joint post with Bill Savedoff.

A new pay-for-performance approach to spur the 35 states of India to perform better in the health sector was recently announced. For the first time, central government funding to individual states under the country's 'flagship' health program, the National Rural Health Mission (NRHM), will depend on the state's performance. According to a Times of India news article, states that fail to perform on certain areas – primarily a more equitable distribution of doctors and nurses – will have their NRHM budget reduced, while states demonstrating performance on other areas, such as providing free generic drugs at public facilities, can earn additional outlays.

This new approach is encouraging and laudable because it gives states clear incentives to measure and improve their performance in providing public health care. But the design appears to focus primarily on changing the supply of services rather than creating incentives directed at improving the population’s health. The program also involves a fairly complicated set of sticks (punishment) and carrots (rewards). The states will be penalized if they don’t place enough doctors in districts with low health indices and rewarded for posting performance audits of health facilities online. But how will the quality of all this information be verified? And how will it even measure such things, listed in the article, as "responsiveness, transparency, and good accountability" and "efficient inter-sectoral convergence"?

Work at CGD on the concept of Cash on Delivery demonstrates the value of paying for results in ways that focus as much as possible on outcomes, include independent verification, and provide stronger incentives when they are simple and focused.  So why doesn't the Government of India provide financial incentives for increasing the coverage of key health services? The information for such a payment could come from the independent, robust Annual Health Survey, the world's largest demographic survey with 18 million individuals in 8 states conducted by the Registrar General of India and which records performance in the low-performing states of the country. The combination of focusing on coverage, using independent survey information, and focusing on a few important measures would solve critical problems that commonly occur in programs paying for results.

An ideal incentive program would pay states for progress on health status, such as reductions in child mortality or declining prevalence of malaria. But that can be difficult. By contrast, it is relatively easy to pay for improvements in delivering services that are closely related to good health outcomes, such as "Percentage of births attended by skilled personnel", "Percentage of mothers who sought antenatal care in at least four visits during their last pregnancy", and "Percentage of children under age 1 with full immunization", especially since such measures are collected from the Annual Health Survey which is now routine in the country’s poorer states. It will be a shame if the government fails to heed these lessons and take advantage of its own rich data on service coverage.

The authors thank Jenny Ottenhoff for helpful comments.