Vanderbilt and MIT Study Links Post-Acute Care Hospital Costs with Lower Survival Rates

It’s not yet entirely clear whether mortality for patients in this study was hastened by poor initial hospital care leading to the need for SNF care, or by poor quality SNF care.

“We can’t really attribute our finding to one or the other. If it’s the SNFs that are poor quality, that could explain why we’re finding that that type of spending is inefficient, in the sense that we’re spending more and getting worse outcomes. Or it could reflect poor quality on the inpatient side: the patients just aren’t ready to go home because they got poor quality while they were in the hospital, so they need to rely on SNFs for post-acute care,” said co-author John Graves, Ph.D., assistant professor of Health Policy and Medicine.

He said either interpretation supplies grounds for making post-discharge SNF utilization a hospital quality measure.

“We find that patients treated in hospitals that discharge to SNFs at rates above and beyond what their patient case mix would indicate achieve worse outcomes,” he said.

The authors state that the “adverse effect of outpatient spending is predominately driven by spending at skilled nursing facilities following hospitalization.”

The study was published online as a National Bureau of Economic Research working paper. Graves’ co-authors are two MIT economists affiliated with the NBER, Joseph Doyle, Ph.D., and Jonathan Gruber, Ph.D.

The study is a so-called “natural experiment.” After adjusting for acuity and other patient characteristics, some hospitals have significantly better patient outcomes than others, but isolating the causes of this variability is difficult.

Patients have a choice of hospitals, and the bias at play in this choice quashes any system-wide causal analysis of outcomes.

An experiment that randomly assigned similar patients to different hospitals could theoretically throw light on wasteful spending and causes of higher mortality in health care, but such an experiment is unlikely.

“Without random assignment you’re always concerned about the confounding influence of underlying unobservable characteristics that might explain why somebody gets more spent on them and why they have worse outcomes,” Graves said.

To circumvent this impediment, the team tracks one-year post-emergency mortality for patients who, by virtue of their initial emergency medical condition and recourse to emergency ambulance services, effectively wind up being randomly assigned to local hospitals. The authors find that outcomes documented for these patients resemble experimental results.

“We couldn’t run the actual randomized controlled trial but by testing how robust our results are under different assumptions, we’re trying to get at something approaching that,” Graves said.

The study tracked claims data from 1,582,421 Medicare patients who needed immediate hospital care for emergencies like hip fracture, femur fracture and intracerebral hemorrhage. All patient outcome comparisons are within single zip codes and every U.S. residential zip code is represented.

The authors profiled U.S. hospitals according to average Medicare spending for their patients over 90-day periods beginning at hospital admission, adjusted according to how sick the patients were.

The study found that patients assigned to hospitals with high 90-day spending overall (relative to patient acuity) had no better survival than patients assigned to hospitals with low 90-day spending, which “validates earlier findings from a slew of highly influential observational studies from researchers at Dartmouth,” Graves said.

“Our key finding is that, if you’re treated at a high intensity hospital, patients quasi-randomized to those kind of hospitals achieve much better outcomes in terms of survival at one year.

“But the hospitals that tend to rely a little bit more on post-acute care services and that discharge patients to skilled nursing facilities at greater rates than their case mix would indicate, patients treated in those types of hospitals achieve worse outcomes.”

The authors conclude with a vigorous recommendation for incorporating risk-adjusted SNF utilization as a routine hospital quality measure.