Academic Articles

Average vs. Marginal Benefit of Medical Treatment

The weak relationship between aggregate spending and health outcomes is in stark contrast to evidence showing pronounced medical benefits for use of specific medical devices, procedures, or pharmaceuticals. For example, advances in the treatment of heart attacks reduced the one-year mortality rate for these patients by 5 percentage points between 1984 and 1991 (Cutler et al., 1998). The use of anti-retroviral drugs among HIV/AIDS patients is associated with approximately a 70 percent drop in mortality…

A paper by Evans and Garthwaite aim to see if the marginal benefit of additional time in the hospital for newborns improves the baby’s health.  A traditional OLS regression will likely show that a shorter postpartum hospital stay is correlated with better health.  However, this is because doctors release healthier babies from the hospital sooner than they do for sick babies; shorter postpartum hospital stays do not cause an improvement in health.  

To identify the causal affect of newborn initial hospital stay, the authors use an instrumental variables specification.  The instruments are a series of federal and state laws passed in the late 1990s increased considerably postpartum stays for newborns.  This lead to an increase in the length of postpartum hospital stays that is unrelated to the baby’s health.  

With the 2SLS, the authors find little average effect (LATE) of longer postpartum hospital stays on the probability of hospital readmission (a course measure of the baby’s health).  However, the authors also look at the effect of the laws on high risk babies.  This include babies born via C-section, or those with other significant risk factors.  For these high risk babies, longer hospital stays did decrease the probability of hospital readmission.  

Thus, while the average benefit of longer postpartum hospital stays may not be cost-effective for the average baby, it can be highly cost effective for high risk babies.

Leave a Reply

Your email address will not be published. Required fields are marked *