The Trouble with Good Intentions

While attending the National Quality Forum Annual Meeting this Tuesday, Farzad Mostashari MD unleashed a 29-part tweet storm that was like a quick summary of almost a decade of challenges in quality reporting. If you don’t know Dr. Mostashari, he is the former National Coordinator of Health IT and the founder of Aledade, a firm dedicated to implementing the promise of value-based care. I encourage folks to read Dr. Mostashari’s full conversation but I felt I had to embed it here and provide some context to some of his points.

The Context

Dr. Mostashari is referring to meetings of the Medicare Payment Advisory Commission commonly reffered to as MedPac. MedPac releases periodic reports advising Congress on the reception of CMS payment initiatives. In their most recent report in March, MedPac forcefully raised some objections to the way quality reporting is being implemented under the new CMS payment programs:

The Commission has repeatedly raised concerns with Medicare’s current clinician quality programs and resulting payment adjustments. First, the quality reporting requirements are confusing and burdensome to providers, and the link between performance and the resulting payment adjustment is unclear. Second, the quality reporting process does not allow creation of a national performance benchmark across the entire universe of clinicians. Third, many of the quality measures are not linked to outcomes of importance for the beneficiary. And fourth, the measures do not generally assess low-value care. March Report to Congress

Of these three points there is little debate. CMS has a bewildering array of quality measures that are measured in the program each with their own complexities and biases. Go ahead and try out the Quality Payment Program Measure Selector and try to decide which of the 271 measures works for you. More on this later.

What you may have missed though is that in accompanying meetings at MedPac, three senior staffers prosed a recommendation eliminating clinician quality reporting from CMS payment all together:

“the MIPS system is a very overbuild system that’s unlikely to be successful at identifying high-value clinicians. and ideas for fixing it include eliminating all measure reporting by clinicians and replacing it with a set of CMS-calculated outcome and patient experience measures.” Transcript March 2 Meeting pg 190 emphasis added

this refers to the following slide that Dr. Mostashari shared:

MedPac Slide

This is a remarkable proposal. It would effectively end quality-based payments for physicians who are not participating in Accountable Care Organizations. Though if the incentives to participate in an Accountable care organization were strong enough, would the effect of this change be dramatic? That is a big unknown question.

The Complexity

The major reason cited for this shift is the intense complexity of reporting quality measures using EHR clinical data. Here, Dr. Mostashari does an amazing job showing how it didn’t have to be this way but the system got completly carried away by complexity:

Imagine a world where through just careful straightforward documentation, quality would be automatically calcualted by the EHRs? yea that world was possible but never materialized. Dr. Mostashari cites one study he participated in.

There were a lot of reasons this world never came to pass summarized in the tweet storm. They include EHR vendors who could certify as compliant in calculating or implementing a measure through whatever workflow they wanted regardless of whether their users actually captured data that way. Thats why in above study and real life practice, the measures automatically calculated by the EHR always under perform until analysts (like myself for many years) start crawling through the database tracing how users actually enter the data not using the certified checkboxes and free text forms that were meant to provide “automatic calculation”.

One EHR that I worked with (I won’t name the vendor but see if you can guess) counted preventative screening based on text typed into a free text box within the clinical notes section of the EHR. If the provider typed “reviewed colonoscopy on 5/12/2014” with the correct spelling and date format, the automated reports would capture it. If the provider happened to use “2014-05-12” it wouldn’t count, if “reveiwed” was misspelled it wouldn’t count. It was a crazy way to do reporting.

The EHR vendors are not alone in making this reporting endlessly complex. Part of the blame goes to the measures themselves with amazingly complex branching logic that required multiple documentation types. Consider that documentation of smoking required documentation of frequency, type of tobacco, and cessation of either pharmaceutical prescription, education, or other.

Why its important

By now you are probably saying, “oh god we should eliminate this”. But quality reporting plays a critical part in the move to value-based care.

If your going to provide a network with a pool of money to care for a population of patients, you can make money by putting up walls preventing patients from getting the care they need. This was the impression of the HMOs that proliferated throughout the 1990s. Quality measures are the guard against that. They help ensure the identification of networks that really deliver higher quality care, that provide a better patient experience and justify receiving higher payments because of that success. Removing quality measurement or pulling so far back that it fails to provide this accountability would be detrimental to those aims.

At Algorex, we are working to make this process easier. In a forthcoming blog series, we will be sharing how groups can use open-source technologies to automatically calculate quality measures based on FHIR standards. It combines libraries from a range of initiatives. If you its becomes trivial to calculate the measure, it means groups can return focus to having workflows that don’t get in the way. Also, fast calculation gives groups the opportunity to actually do something about missed measures.

But faster calculation is only part of the solution, its clear we need some thought to try and eliminate measures that increase the complexity of process at the expense of outcomes. Let’s close on that and one last tweet from Tuesday.