Angelyn Fairchild and Dr. Joseph Price, Economics Department
Purpose
Patients trust that their medical provider will help them receive the best care possible even when this may involve referring them to another provider. However, transferring patients to other providers represents a transfer of income between the health care providers, thus creating a strong profit incentive for physicians to retain patients in their care. In order to examine how physicians respond to competitive threats to their income, we exploit changes in insurance laws that increased competition in the market for labor and delivery by expanding policies to cover midwives as well as physicians.
The changes in insurance law represent a random change in the market structure for labor and delivery services. This change is empirically useful because it affects the decision making process of women and health care providers without reference to their underlying characteristics. We hypothesize that, all things being equal, more patients would choose midwives because of the decrease in the relative cost. This should lead to fewer medically intensive births, as midwives are not licensed to perform certain medical procedures. However, we expect that physicians may react to the increase in midwives by enacting policies that would tend to decrease the midwives’ ability to practice.
Research Approach and Preliminary Result
Our hypothesis rests on three key interactions. First, we establish the change in insurance policy as a random variable that impacts the competitive structure between physicians, midwives, and patients but that does not otherwise directly impact the medical needs of the patients. We find that active insurance policies are associated with a 12% increase in the proportion of births delivered by midwives. However, we find a troubling association between the mothers’ health characteristics and the insurance status of the state. We are currently troubleshooting this issue and believe that problems within the data are creating statistically significant results across all variables.
Second, we use the change in legislation as an instrument that tests whether states with insurance see changes in their birth outcomes. We find that the insurance change is associated with an increase in the number of repeat caesarians and a decrease in the number of vaginal births, services only a doctor can perform. However, the change is also associated with a decrease in the number of mothers who have induced labor and who have monitoring equipment in place during the birth, an indication of increased midwife activity. These conflicting results seem to indicate that there is a competing effect associated with the change in insurance—physicians may be performing more intensive procedures that would block midwives from practicing while at the same time more women use midwives who tend to use fewer technology-intensive methods.
It is important to note that these results could also reflect a local average treatment effect— specifically, both physicians and midwives might be changing their behavior in reference to a specific population of women who are healthy enough to use either provider and who are more elastic to the change in insurance. This effect would be greater if we assume that physicians have enough information about their patients to change their behavior only towards patients in danger of switching. If physicians cannot differentiate, then we would expect the changes in insurance policy to affect the patient pool generally as physicians hedge their bets against competition.
In order to more fully answer this question, we need to differentiate between patients who would have used a physician but who switched to using a midwife, and then compare those patients to similar patients who remained with their physician. We are interested in recovering the difference between these groups over time because this will help us to tease apart the competing effects, where midwives do fewer c-sections but doctors do more in reaction to midwives. The biggest challenge in this approach has been effectively isolating the “switchers”—patients who would have used a physician but who changed their behavior based on insurance policy. We have been trying to establish propensity scores that calculate the likelihood that a woman use a midwife based on her pre-existing characteristics, then use those scores to predict caesarean birth rates. This method shows that women whose health characteristics make them more likely to use a midwife are much less likely to have a first caesarean birth and slightly less likely to have a repeat caesarean birth. However, this does not fully separate the competing effects; we are currently working to devise a better method of using propensity scores to differentiate between groups.
Challenges and Future Research Needs
In order to bring this research to publishable quality, we need to establish a better mechanism for isolating the relevant patient group affected by the policy change. Our improved methodology will need to more precisely differentiate between groups of patients. It will also need to more accurately. There is not really a tried and trusted method for doing this, so it is going to require more creativity to devise a workable model. It has also been challenging to isolate the competing effects of the insurance policy on physician and midwife behavior. Their behaviors are only shown implicitly through patient outcomes, so determining which shifts are really associated with behavioral changes is difficult. Fortunately, there is still time left for us to fine tune our methodology and hopefully overcome some of these problems.