A hospital compares outcomes of teaching and hospitalist teams

Dr. Chin's study looked at length of stay, readmissions, and costs among 3 types of inpatient medicine teams.

Every hospitalist knows that the day goes a little differently when you have trainees in tow, but a recent study at one academic medical center quantified the results of that variation.

Researchers compared length of stay, 30-day readmissions, and costs among 3 types of inpatient medicine teams: residents led by a nonhospitalist attending, residents led by a hospitalist, and a hospitalist with no residents. Hospitalists alone had the lowest 30-day readmission rates, according to the results, published in the August Journal of General Internal Medicine.

Photo by Regan Scott-Chin
Photo by Regan Scott-Chin.

Both kinds of teaching services had lower length of stay than hospitalists alone, and the hospitalist teaching services had lower cumulative costs of hospitalization (readmissions were included) than the teaching services led by nonhospitalists.

ACP Hospitalist recently spoke to lead author David L. Chin, PhD, a postdoctoral scholar at University of California, Davis, about the findings.

Q: What led you to study this issue?

A: I received a small pilot award to utilize the electronic health record data that we have at the institution [for research]. One of the key issues in quality is readmission measures. Residents' effect [on quality measures] is one of the [issues of interest] that came from discussion with the faculty here.

Q: How did the results compare to your expectations?

A: Naively, I assumed that residents and the training service would be able to potentially deliver the same quality, but perhaps in a less efficient way. I actually expected the readmission rates to be similar across all services. But, as our paper described, it's never quite that simple.

Q: What could explain resident services having shorter lengths of stay?

A: It's definitely plausible that residents have an interest in moving through their patient load and discharging patients. Not to say that they are more haphazard about it, but they have so many things to do...so they have an interest in discharging patients perhaps at a faster rate than the hospitalists.

Q: Can you draw conclusions about the quality of hospitalist care from your research?

A: There are dozens and dozens of papers that have been written about hospitalists and whether they provide better or worse quality of care. I think the hospitalists are in good shape. One of the unique aspects of our study is that we were able to separate the teaching services, so we were able to pretty convincingly compare how patients that were assigned to academic physicians ended up [versus] those taken care of by hospitalists with residents. There's definitely an argument for having hospitalists do the training in academic medical centers. Academic physicians have a vital role, at our institution in particular, but they're more the visitors in the hospital than the primary folks who are in charge of training residents. The hospitalists take on a very large role in training, at least with internal medicine.

Q: Do the results indicate any problems in the current academic hospital model?

A: I wouldn't go that far. I think the effect [we found] has more to do with the way that we measure quality than it does with an inherent problem in the structure or the design of the training process. In my view, the issue should be 30-day readmission as a metric, and whether it's truly a reflection of the type of care that's received in the hospital or if it's a measure that picks up other things like outpatient care [or] discharge planning, some of which are attributable to the hospital, but others maybe not so much.

Q: You found that the teaching services had both shorter length of stay and higher 30-day readmissions. Does that suggest that pushing down length of stay increases readmissions?

A: That's the topic of a paper I'm working on now. It's intuitive that as you decrease the length of stay, you'll have a greater number of readmissions, but as far as I know, no one's really definitively shown that. That was one of my initial questions with this project, but just for practical and technical reasons it got too complicated to drill down to that level. I have a bigger, fancier, more representative data set that I'm working on now, and that's soon to come.