Tuesday, July 14, 2015

Big data is not big enough

Today ProPublica released its “Surgeon Scorecard” touting it as the best way to pick the right surgeon.

It took me less than a minute to discover some interesting omissions from the application.

For laparoscopic cholecystectomy, the only general surgery procedure listed, the app omits approximately one-third of the hospitals in my state including two where I have practiced.

It looks like the problem is that using Medicare fee-for-service data does not yield enough surgeons performing 20 or more cases in some categories such as laparoscopic cholecystectomy for the five years included in the database.

At one of the biggest hospitals in my state, apparently only one surgeon performed 20 laparoscopic cholecystectomies on fee-for-service Medicare patients in the five years studied; 23 other surgeons were listed as having performed fewer than 20 laparoscopic cholecystectomies on patients in the target population. I don’t see how patients who want to use that hospital for their gallbladder surgery will benefit from the Surgeon Scorecard.

In general, the complication rate for laparoscopic cholecystectomy is low, but I think I understand why ProPublica chose that procedure to review. They needed to select a procedure that was done frequently enough to yield a sufficient number of cases for analysis. Unfortunately, because of the limitations of the Medicare fee-for-service data and the low complication rate of the procedure, the Surgeon Scorecard is useless for anyone looking to compare general surgeons.

Similar problems with the scorecard may be in play for prostate surgery. Again, the procedure was chosen because of its high frequency, but in quickly looking through some searches in that area, I note that a number of urologists I know also did not perform 20 cases on fee-for-service Medicare patients.

Perhaps the next iteration of the scorecard will utilize a data set that contains enough patient and surgeon records to make a meaningful comparison.

Until then, general surgeons can relax. They will not have to explain away their complications but will simply have to explain why they aren’t listed in the Surgeon Scorecard.


DrSoup said...

Using only the inpatient cases for an operation most of us do as outpatient skews the data. I can understand their saying comp rates were same whether in or out patient, but clearly left us that strive to keep patients out, save costs, out of the analysis. Also prob skews up the overall complication rate, since the inpatients are prob sicker & more prone to complications. Most of my inpatient Medicare gallbladders come thru the ER being ruled out for MIs.

Would almost argue that inpatient Medicare gallbladders are, almost by definition, not the elective operation they were looking to study.

Skeptical Scalpel said...

I agree completely. One of the authors of an article that accompanied the Surgeon Scorecard release has confirmed to me that the laparoscopic cholecystectomy data only involves inpatients. That not only skews the data it also eliminates the vast majority of laparoscopic cholecystectomies from consideration. This is why so few general surgeons are listed. As I said, the scorecard is useless for comparing general surgeons.

Windriven said...

Points well made. How then is the medical consumer to evaluate and select a surgeon? The physician community is often perceived, and rightly I think, to be slow to address even egregious shortcomings among its members. If the community is unable or unwilling to adequately police itself, it should at a minimum work to establish objective metrics for medical consumers to employ. It is to my mind an issue of informed consent.

Skeptical Scalpel said...

Windriven, that is a good question with no clear answer. Administrative (billing) databases such as the one used in the Surgeon Scorecard are inadequate for comparing clinical outcomes. A whole new clinical database needs to be set up. Problems include who will do this monumental task, how will data be submitted and analyzed and who will pay for it.

Proponents of the Surgeon Scorecard say it is better than nothing. I disagree. Bad data may be worse than no data at all.

Windriven said...

"Bad data may be worse than no data at all."

Indeed, the hamartia of many meta analyses. Garbage in, garbage out. Bad data isn't better than nothing, at best it is only as good as nothing.

Skeptical Scalpel said...

And at worst, it is worse than nothing. For example, Some of the surgeons in the scorecard may erroneously be labeled as having excessive complications and the converse is also true. Some who fall in the low rate of complications may actually be bad surgeons.

Maggie Mahar said...

Skeptical scalpel-- the fact that the information is incomplete (many general surgeons not included) does not make it wrong.
There is enough data to show several very important things: a) while some surgeons account for very few complications, others are at the other end of the bell curve and probably shouldn't be operating.
Moreover, even within a single, brand-name (or marquee) hospital you see wide variation. Any surgical nurse knows this (though she won't tell you unless you are a relative. Nurses are afraid of being fired or sued.

Skeptical Scalpel said...

The info is not merely incomplete. The methodology is flawed. Here are links to 2 of the many posts that have dissected the scorecard: https://twitter.com/buckeyesurgeon/status/622888610140811264 and https://ejsmdblog.wordpress.com/2015/07/20/propublicas-surgeon-scorecard-call-for-peer-review/

Post a Comment

Note: Only a member of this blog may post a comment.