Physicians Aren’t Engaged in Performance Because Measure Results Aren’t Real

According to management guru Peter Drucker, “If you can’t measure it, you can’t fix it.” Quality measurement and reporting have been rooted in similar reasoning. The idea is that we find out what’s wrong, and then we launch programs to improve it.

That’s the linear route mapped out by Medicare starting with Meaningful Use, PQRS quality reporting, Value Modifier comparisons, and moving into current MACRA MIPS and APMs.

But physicians have known something for a while that others have been unwilling to accept: quality reporting measures don’t give you a foundation for improving outcomes. Why? Because performance measurement does not tell them what they need to know.

Performance Results for Outcome Measures Can Paint the Wrong Picture

When providers are launching improvement projects to gain Medicare MIPS dollars or APM savings, they see that something is missing from the current measurement system. That missing something is enough good information from measure results to show them where the problems lie and where outcomes must be improved.

CMS made compromises in implementing PQRS (that were transitioned to MIPS) to include outcome measures with clinical significance, but that don’t go far enough to achieve real value. This is especially true of typical outcome measures for primary care. Look at any intermediate outcome measure, such as blood pressure, and you will see that a once-a-year reading of blood pressure will not tell you anything about the patient’s control level.

In fact, blood pressure performance is a perfect case in point. A review of values across time, even of a single patient, reveals a remarkable variation—often as much as 60 points between high and low values that could have been taken a day or a week apart. The cause could be the method of taking blood pressure, devices or misrecording of data—but the effect is uniformly unreliable outcome data. That same individual patient, depending on whether the value chosen is 125 or 176, could be classified as either well-managed or high risk.

Understand that this happens over hundreds of patients for an individual physician, and thousands for a group. It means that results accumulated for both are inaccurate. Also know that this happens over many outcomes with repeated values, although blood pressure illustrates one of the worst cases of variation. Therefore, improvement activities based on patient outreach, shared goal setting, or even clinical interventions based on such data, is problematic, if it is based on quality measure reporting results.

Outcome data is also the most frequently missing data in retrieved EMR data files. Pain scores, disease staging and complications are among those most frequently absent from much structured EMR data. As a result, the final performance numbers are not complete and not reflective of real provider performance. Real performance is unknown.

Doing something with such performance data to improve outcomes is shaky, at best. Since data is missing or incorrect for individual patients, there aren’t any criteria to assign them correctly to higher risk groups for focused interventions.

Process Measures Rely on Data Capture and Do Not Represent Provider Excellence

Non-outcome, or “process” measures have generally been perceived by physicians as “administrative” measures and assigned to staff to complete. These are measures that simply report that a patient received some service: a diabetic foot exam, in the case of a patient of a primary care patient, or administration of certain drugs within a certain timeframe of a cardiac event. Physicians are almost never responsible for recording results for process measures. They also don’t always see them as legitimate quality measures, for good reason: these measures are usually less a result of performance than of data availability.

Process measures, which are the most common quality measures, usually assume the collection of various procedure codes to indicate the completion of the process defined by the protocol. Since procedure codes are included in claims data, the assumption is that these results are more available—and reliable.

This may be unfounded. Claims-based codes are not universally applicable to all process measures, requiring manual or other forms of retrieval to complete. All process measures are subject to data capture limitations that impair their performance results. Even claims procedure codes can be wrong or missing.

Also, a patient could report to her physician that an exam or test required by a quality measure was performed by a previous provider. If this data is not verified and captured in the record, the current physician will have a lower performance score. The performance reflects how the state-of-art of data is facility or group-centric, and not patient-centric.

Data That Will Facilitate Physician Engagement—and Propel Performance Improvement

Physicians want to know that their patients have done well and are getting better. Information about patient status must be presented in a way that is valuable for learning and improvement.

Make no mistake about this: once-a-year outcome values can only be interpreted as scores, and this is how they are used not only by CMS and health plans, but also by physicians’ own health systems. Measure results are invariably used to show physicians how they compare to others, regardless of whether that comparison is valid.

With a new anti-regulatory environment in Washington, and MIPS in its first year of implementation, proposals are already percolating to make it streamlined and simpler. Unfortunately, the focus appears to be on easier reporting alone without achieving better value.

But both goals can be achieved: focused performance measurement, yet smaller external reporting requirements. That happens by measuring and reporting improvement efforts instead of ineffective quality metrics.

Performance improvement must come from a review of current patients and their outcomes, but it requires a more focused emphasis to be visualized by physicians. Since data is almost always coming from EMRs, especially for the large groups that will be participating in these endeavors, there is little burden to actual data collection beyond data integrity efforts.

Outcomes and not Process Measures. Process measures should be monitored internally to ensure the use of clinical protocols and evidence-based medicine, but unless they are associated with targeted outcomes or problems, should be separate from performance improvement.
Measurement over time. Since improvement in outcomes should be the real focus, continuous values (blood pressure, BMI, HbA1c) can only be properly evaluated through trends.
Measurement of interventions. Performance improvement by necessity consists of one or more interventions. This could include patient education, shared decision-making, changes in medication, and similar initiatives. Physicians need to be able to see what is producing results beyond anecdotal information by their patients and peers.
Selected representative patient results for learning and investigation. Physicians need both aggregate and individual patient results, but we cannot overwhelm them with data. They should see outcome variations for small numbers of patients who are improving, or not, and those who may have frequent variations.
Data integrity issues in patient information. As in the blood pressure example, physicians should see samples of such widely variant values in their patients, because this likely represents a process that is not working and must be corrected by practice to provide better care. They should also be exposed to conditions where outcome information is missing in the data, so that they can support efforts to improve.
Risk adjusted patients and performance and comparisons with peers. Physicians want to know how they are comparing with peers, but the underlying data point should be fair. Comparison of the aggregate improvement trend for any outcome and comparison of a physician’s trend with the organizational trend are valuable information. But that performance data should have been risk adjusted and should allow or require physician feedback on the results.
Per patient and episode costs. Providers should always negotiate for patient-identified claims data with payers, so that they can use per-patient cost reports and episodic care costs. The aggregate costs should be able to identify in- and out-of-network costs, referral volumes and types, diagnostic tests including imaging, emergency services, and facility costs including inpatient and outpatient services.

The years of focus on reporting measures have discouraged many physicians from performance measurement. Fortunately, MIPS and APMs, along with many health plan initiatives, reflect movement toward performance improvement in both quality and cost.

Providers can adjust the fundamentals of performance measurement and get better data to begin improvements. By evaluating the issues revealed through outcomes and cost, they can launch better future efforts—with physician support.

Founded as ICLOPS in 2002, Roji Health Intelligence guides health care systems, providers and patients on the path to better health through Solutions that help providers improve their value and succeed in Risk. Roji Health Intelligence is a CMS Qualified Clinical Data Registry.

Image Credit: Illustration from Ajā’ib al-makhlūqāt wa-gharā’ib al-mawjūdāt (Marvels of Things Created and Miraculous Aspects of Things Existing) by al-Qazwīnī, Islamic Medical Manuscripts from the National Library of Medicine.