Hi Edwin

That's a good question.

You're right that the largest effect is still the individual surgeon skill (how well I do the procedure, conditional on me electing to do it) and the surgeon's judgment (how well I made the decision to elect to do it)

But if surgeons don't roam across multiple hospitals, even if we accept your premise of just surgeons 'doing the surgery', then reporting hospital metrics of surgical quality would still give consumers, referring physicians and payors a good first order approximation of individual surgeon quality.

However, I think that there's enough work to suggest that there's an independent effect of hospital on surgical outcomes, and another, independent interaction effect between surgeon and hospital, that both affect surgical outcomes.

In other words, in a thought experiment in which you get, say, a cardiac bypass procedure with the same surgeon at two different hospitals, you would likely experience different outcomes.

You can think of this as a combination of two effects, both of which have been demonstrated pretty well.

First, there's a longitudinal effect: the procedure is just one of a series of care steps from referral, workup, prep, ICU care, step-down care, discharge planning and community org support, post-acute care, follow-up care which all differ by hospital.

Second, there's a cross-sectional effect: the surgeon, while clearly still the master of the ship, needs to have good working relationships with the anesthesiologist, the trainees, the nurses, the ICU fellows etc, and ideally good overlapping shared experience bases, and good mutual experience with the technological idiosyncrasies at hand in that particular hospital. Across different hospitals, all this varies even if the surgeon is the same person.

Best

Marco

From the viewpoint of a methodologist, outcome is impacted by various sources of variance, amongst them: patients' characteristics, skill level of surgeon, contextual variables of the hospital (e.g. availabilty of certain diagnostic procedures). Though there is a large body of literature on this topic, we've found only 25 articles dealing with the impact of hospitals and physicians simultaneously. And only 1 study did this using the adequate "multi-level-modeling" by random coefficient models:

(Wen, H. C., Tang, C. H., Lin, H. C., Tsai, C. S., Chen, C. S., & Li, C. Y. (2006). Association between surgeon and hospital volume in coronary artery bypass graft surgery outcomes: a population-based study. [Research Support, Non-U.S. Gov't]. Ann Thorac Surg, 81(3), 835-842. )

My conclusion for hospital specific quality reports would be to report hospital metrics adjusted for patient variables, but using multilevel models disentangling hospital, patient, and physician effects during internal quality management efforts.

Thank you both for your thoughtful comments, and references. While volume of practice may be a coarse proxy for surgeon skill, I feel it is a poor one, since a surgeon with mediocre skills may only be repeating those same behaviors over and over and not really developing their skills. And so even if volume is high, skill could still be only mediocre. The NEJM article cited in the report in my question is the only one I have seen reporting procedure skills, not merely the number of times they have exhibited them, and for that, it is groundbreaking, and frankly, disruptive. Perhaps by "this topic" Dr. Frick, you are referring to "surgeon characteristics", which have been studied, including: surgeon specialty, age, years of experience, and of course, volume, and others. But surgeon skill is what this topic is really about, and it is hard to characterize and measure, though the researchers at Univ. Michigan where the NEJM study was done, have done a credible job in assessing just that. And for that they have broken new ground. Take a look. It's at http://www.nejm.org/doi/full/10.1056/NEJMsa1300625

Again, I thank you for your interest and contributions to this topic.

Hi Dr Frick

With your permission, let me add the following two articles of mine which do deal with surgeons and hospitals simultaneously.

The first looks at individual surgeon and hospital volume (both recent and at various lags, as well as cumulatively) as predictors of surgical outcomes (http://www.ncbi.nlm.nih.gov/pubmed/19732169).

The second looks additionally at the interaction effect of surgeon*hospital in a true hierarchical 4 level multilevel model (http://www.ncbi.nlm.nih.gov/pubmed/20849555).

Edwin, let me also point to a little study of mine (flawed design and imperfect data but still...) which looked at the impact of surgeon breaks of a calendar month immediately preceding a patient's cardiac surgery. In line with a "manual skills" argument, my point estimate results were consistent with a small deterioration in such skills (http://www.ncbi.nlm.nih.gov/pubmed/23773602).

Best

Marco

Thanks!