There has been active discussion about Arvind Panagariya’s column on improving health care in India. Shanta pointed out that increasing the number of trained medics is unlikely to solve India’s health problems if these medics don’t show up for work. Arvind responded that his recommendation of increasing the number of MBBS was based on the idea that they would work as “private practitioners and not add to the rolls of absentee providers in the public sector”. This assumes that the main constraint on good quality medical advice is doctor’s knowledge. What do the data say?

A new paper summarizes work on the quality of care in low-income countries that I and my colleagues Jeffrey Hammer (Princeton) and Ken Leonard (University of Maryland) have been engaged in over the last 5 years. Our approach has been to try and decompose the quality of medical advice into two components—what doctors know and what doctors do. What doctors know—measured by testing doctors—represents the maximum care that a doctor could provide. What doctors do—measured by watching doctors—represents the care they actually provide to real patients. We call the first “competence” and the second “practice quality”.

Figure 1 shows the relationship between what doctors do when we were watching and what they said they would do when tested. The vertical axis shows the percentage of essential tasks (for two illnesses) that doctors completed when we were watching. The horizontal axis plots the percentage of essential tasks that doctors completed for the same cases when we tested them. If doctors practiced at the frontier of their knowledge, practice quality would be given by the 45-degree line. The gap between knowledge and practice is the difference between the 45-degree line and their actual performance—red for private doctors and green for public-sector doctors.

There are several noteworthy features. First, public sector doctors always do a lot less than their private sector counterparts—between 66 and 50 percent less depending on how competent they were to begin with. Second, even in the private sector, there is a large gap between knowledge and practice. Doctors who correctly covered 60 percent of essential tasks when tested ended up covering just over 30 percent when dealing with actual patients. Third, about the only “doctors” who are practicing at the frontier of their knowledge are those with very low competence—those who covered 20 percent or less of essential tasks when tested.

With these data, back-of-the-envelope calculations regarding the benefits of training are relatively straightforward. The effect of training on practice quality is the effect of training on competence multiplied by the effect of competence on practice quality. In Tanzania we find that two additional years of school and three additional years of medical school buys an increase of only 1 point in the percentage of essential tasks completed. Results are similar for other countries. Training may be a necessary condition, but it is far from sufficient—even for doctors in the private sector. The reason is simple: the average doctor-patient interaction, even in the private sector, lasts only 3 minutes. In the government sector, it’s just over 1 minute.

Why do doctors in low-income countries exert such low effort? One uncomfortable reason may be that systems of medical liability are weak. Because patients almost never sue doctors, one channel of accountability is effectively shut off. It’s an uncomfortable answer because, as the raging debate on medical liability in the U.S. shows, medical liability comes with a host of problems. Since studies like these have not been conducted in rich countries, we don’t know what the competence-effort decomposition would show—what we do know is that increasing effort among both private and public sector doctors is a first-order policy concern.