For what it's worth, a review of PubMed shows about half-a-dozen studies looking at the supposed "July phenomenon", and none validate increased morbidity/mortality.
Not in the NBER study:
In Cohort Turnover and Productivity: The July Phenomenon in Teaching Hospitals (NBER Working Paper No. 11182), authors Robert Huckman and Jason Barro investigate a third form of turnover, the extreme, though not uncommon, scenario that they term cohort turnover.
Using data on all patient admissions from a large, multi-state sample of American hospitals over a five-year period, the authors find that both minor and major teaching hospitals experience a significant increase in resource utilization -- measured by average length of stay (LOS) -- immediately following the July turnover, and that the effect appears to last for several months. They also find that teaching hospitals with medium teaching intensity experience a significant increase in patient mortality over the same period. The confluence of increased resource utilization and increased mortality (in other words, decreased quality) during the July-August period implies that this cohort turnover reduces medical productivity.
Nevertheless, those hospitals with the highest teaching intensities (the greatest reliance on residents for the provision of care) seem to avoid the disruption of the July phenomenon with respect to average mortality rates. The authors' preliminary evidence suggests that higher supervision levels play a role in mitigating the impact of the July turnover in major teaching facilities.
The magnitude of the estimated effects is substantial and appears to last for roughly six months.
The average LOS for the average, major teaching hospital increases by roughly 2 percent following the July turnover and remains between 1 percent and 2 percent higher throughout the final six months of the calendar year. Similarly, the average, major teaching hospital experiences an
increase in risk-adjusted mortality of roughly 4 percent in the July-August period. This effect also remains at levels between 2 percent and 4 percent for the last six months of the calendar year. For the average major teaching hospital, this translates into between 7.8 and 13.8 "accelerated" deaths (that is, deaths that occur earlier than they would have in the absence of the July turnover) per year. Based on a total of roughly 200 major teaching hospitals in the United States, the July phenomenon is thus associated with roughly 1,500 to 2,750 accelerated deaths per year in the United States. The authors do not estimate the social cost of this increase in mortality.