Risky Business: How to Make Risk Adjustment Fair and Accurate for Quality Measurement

written by Dave Halpert | November 9, 2016

No two patients are the same. Some are much sicker than others. Obvious? Yes. But this is the fundamental challenge of quality measurement. As public scrutiny of physician performance intensifies via the CMS Physician Compare website, and as outcome comparisons become ever more crucial to Medicare and private insurance reimbursement strategies, finding a fair and accurate way to adjust for risk is critical.

Each patient population has its own challenges. Academic medical centers may get the most difficult cases, but solo practitioners see the entire community, and without the infrastructure available elsewhere. For health care to improve for all patients, we need meaningful quality measurement to demonstrate better outcomes (whether clinical or financial). These outcomes must be measurable, but cannot be arbitrary. To create meaningful benchmarks and compare providers against peers, we must level the playing field via risk adjusting for differences in patients cared for by different physicians.

While scoring providers has traditionally been limited to CMS, health plans and others who oversee quality initiatives, QCDRs—promoted as a reporting mechanism in the Quality Payment Program under MACRA—are also responsible for adjusting risk. Here are some inherent challenges for developing fair and appropriate methods, plus strategies to improve risk adjustment for both providers and QCDRs:

Challenges to Accurate Risk Adjustment
Practice-Level Challenges

An algorithm is only as good as its included measureable components. This is the greatest obstacle to accurately adjusting for risk. In other words, you can only adjust to account for the risks you can measure (or record). For example, an obese patient may be more likely to experience complications following surgery. However, if that patient was never coded with a diagnosis of obesity, the surgeon will be graded on the patient’s outcome using the same standards as a patient at ideal weight. This is a common example—as a QCDR that integrates non-billed clinical data as well as transactional data, we frequently see patients whose height and weight indicate obesity, but no coded diagnosis of obesity.

A misguided sense of efficiency can also exacerbate inaccuracy when calculating risk. Many factors can significantly impact outcomes (or already have, and indicate the patient will require more care), but are not coded, and are therefore unrecognized in risk-adjustment. Why aren’t they coded? If an “unspecified” diagnosis is reimbursed similarly to one that includes complications, it’s easy to leap to the conclusion that saving time on the coding is better for the practice. Unfortunately, this fails to recognize that additional specificity will ensure that providers are scored accurately, in a way that truly reflects the population.

Program-Level Challenges

There is another challenge to risk-adjustment, related to the care setting. Many of the established risk adjustment protocols were developed on the hospital side (e.g. 30 day mortality following an event or procedure, all-cause 30-day re-admission rate, per capita costs associated with a specific DRG.). However, the PQRS (formerly PQRI) has traditionally included quality metrics focused primarily on the ambulatory side. That was less of an issue when the program was focused exclusively on reporting, but with revenues tied to comparative provider performance through the Value Modifier, risk-adjustment is not widely utilized, and is far less defined. In PQRS and VM (and as will continue in the CMS Quality Payment Program), this means that the playing field is no longer level, and not everyone holds the high ground. The challenge is intensified by the introduction of care coordination metrics, which include measures of communication, accessibility and more—there are few established benchmarks for these areas.

PQRS measures (and the measures proposed under MIPS) include an abundance of process measures—measures of whether an action was performed. These are often adjusted by including responses that address reasons for not performing the action (a medical, patient or system-related reason for not doing so), so that providers are not unduly penalized (e.g. a female patient did not have mammography results documented and reviewed because she previously had a bi-lateral mastectomy).

However, since at least one outcomes measure is required for Quality reporting under CMS’s Final Rule, some providers are finding themselves in a challenging spot. For example, an existing outcome measure looks at patients who have hypertension and whether blood pressure is controlled. At present, there is no risk-adjustment applied to this measure—no accounting for socio-economic status, co-morbidity, race or other factor. Clearly, with such variation between populations, providers in one setting have the deck stacked against them compared to peers. To add to the challenge, there are far more process measures than outcome measures. The combination of a required outcome measure and a limited selection will force many providers to report an unfavorable measure—and accept the consequences.

How Can Risk Adjustment Improve?

Quality measurement is critical, but it must be reliable and trusted by all stakeholders. Providers and QCDRs must work together to determine accurate methodologies to account for risk. For these risk-adjustment strategies to be meaningful, trusted and (most importantly) tools for improving outcomes, each group has responsibilities:

For Providers:
  • Ensure that what goes out in claims reflects what you see in your office—coding should be accurate, complete and specific. Your patients may actually be “sicker,” but unless that is documented, you are the only one who knows. The Value Modifier actually offers providers an opportunity to earn a supplemental incentive if their patients are considered “high-risk” (75th percentile risk score and above). If you are not coding for complications and risk-factors, you are sabotaging yourself by lowering your patients’ adjusted risk, making it harder for you to compare favorably to peers when being scored on quality or cost.
  • Push for interoperability between your Health Information Technology vendors and your clinically integrated network, as well as for utilization data for their patients. Relying on patient responses is risky. You can’t be certain that an immunization was received, or that a patient received treatment (or didn’t) at another facility. Those with claims data may be able to accurately risk-adjust, but if you are not apprised of that patient’s history, treatment may deviate from what those adjusting risk may anticipate. Failing to treat a symptom or condition may have severe consequences. In other words, a patient may be adjusted more in your favor than you realize, but if you aren’t treating all conditions, you’ll still come out on the wrong end of a comparison.
For QCDRs:
  • If developing a risk-adjustment platform, look beyond what’s available in claims. Non-billed clinical data, such as height, weight, tobacco use and the number of medications a patient takes are all useful when determining which patients are more likely to incur high costs. Adjust accordingly.
  • Incorporate information gleaned from other sources. By bringing in out-of-network services, QCDRs can further level the playing field for providers by giving them the same information as health plans. Knowing that a patient has been hospitalized or has seen other providers for other conditions is critical to making informed clinical choices.
  • Listen to providers—what else may be considered? As we QCDRs continue to develop measures, we must also consider how to score providers accurately. There may be non-quantifiable data that providers may eventually wish to incorporate into customized risk scores, or to develop clinical registries to facilitate care delivery for an at-risk group of patients. Some other options:
    • Use hospital risk tools as “process” measures.
    • What tools have been helpful in the past? Is it possible to build off of an existing platform (e.g. readmission risk tool)?
    • What results can you obtain by comparing patients in varying levels of risk to utilization type and/or volume?

Scoring providers on patients’ outcomes is, by nature, an uneven process. With so many factors outside of a clinician’s control, many face challenges even in measuring processes, never mind outcomes. Even so, value-based care is here, and Medicare is going to be paying based on quality, rather than outcome. To remain competitive and to help identify at-risk patients, both providers and QCDRs must investigate new means of identifying and adjusting for risk. Specific documentation and multi-source data integration, combined with an open dialog will create a mechanism that not only adjusts for risk, but also facilitates better care for patients.

Founded in 2002, ICLOPS has pioneered data registry solutions for performance improvement in health care. Our industry experts provide comprehensive Solutions that help you both report and improve your performance. ICLOPS is a CMS Qualified Clinical Data Registry.

Image Credit: Unsplash