Risk Adjustment
Getting to know your risk adjustment: What it is, why we do it, and how it's calculated.
What is Risk Adjustment?
Specific to the Merit-Based Incentive Payment System (MIPS), risk adjustment works to "level the playing field" between providers who see higher populations of high-risk patients in comparison to their peers who see more easily progressed patients. Risk Adjustment generates more accurate representations for Taxpayer Identification Number (TINs) by accounting for risk factors that exist prior to the Episode of Care.
Every individual, regardless of the risk, should have access to affordable, inclusive care. Risk adjustment makes sure that providers who care for these higher-risk patients do not have a financial penalty for doing so.
What is the clinical value of Risk Adjustment?
At Keet, we facilitate our Nation's move towards value-based care by measuring our provider's clinical performance through the use of Patient Reported Outcome Measures. As a QCDR we are equipped with the knowledge of patient populations and historical outcomes which allows us to determine the likelihood of patient improvement. We call this our Risk Adjustment Model. In the application of this, we are able to compare the outcomes of otherwise inequal patient populations, allowing for a fair comparison of providers.
Value-based care is not only measured by how a provider performs in relation to all other providers, but additionally how a provider performs against his/herself. In competition with self, providers can ask, “given the patient populations that I typically see, are my patients progressing more than expected?” Risk adjustment offers insight into how well a provider is doing while providing a baseline expectation of the number of patients a provider is likely to have progress vs. failure to progress.
How are patients Risk Adjusted?
The risk adjustment process assigns probability values for failure to progress (FTP) to a given patient based on various factors. The FTP, or progression adjusted rate, is the probability that a patient will not meet or exceed a Minimal Clinically Important Difference (MCID), the threshold value, or the minimum score change between the initial and final PRO, used to determine progress. For DHI, the risk adjustment process assigns probability values for percentage of patients who have progressed rather than those that have failed to progress.
During the episode of care setup, certain fields are informed by the therapist which correlates to the below logistic model.
Once the patient has completed the first survey, the patient’s risk adjusted outcome is calculated. The resulting adjustment for IROMS produces a probability of “failure to progress” between 0.0 and 1.0.
The mean sum of log odds for each patient represents the patient’s risk adjustment probability for FTP. Probabilities for each measure and predictors for that measure are loaded into a table that informs the risk adjusted value to each patient. In layman's terms, the likelihood of a patient progressing during care is assigned based on the complicating factors applied during risk adjustment.
For the HM7 (DHI) measure, patients who achieved a MCID in vestibular dysfunction, as measured via the validated Dizziness Handicap Inventory or equivalent instrument, to indicate functional improvement greater than zero and a Risk Adjusted Functional Status Change Residual Score for the dizziness handicap successfully calculated with an MCID score that is greater than zero from their initial visits and just prior to or at their discharge visits from the PT/OT practice.
In order to maximize the value of a predicted patient outcome, it is important to understand the risk adjusted variables that account for variance in the predicted outcome. If you know a patient has negative predictors (age, surgical status, SDOH, etc..) this should ideally inform the creation of the patient’s care plan. For example, evidence-based care plans for Surgical Knee patients can be leveraged, but they should be individualized based on individual patient characteristics, to include variables associated with risk adjustment.
The Technical Components
The ROMS risk adjustment methodology uses a logistic regression with multiple indicators (predictors) X_i, to predict a binary outcome, Y. In this case, whether a patient is likely to progress, or not. The predictors used to define the logistic model include:
- age
- admit function score
- admit pain score
- duration of symptoms
- surgery response
- gender
- insurance response
- While insurance type itself is not submitted directly to CMS for MIPS, it is a value used for risk adjustment which may influence the failure to progress rate that we do submit to CMS.
The DHI/HM7 measure is adjusted to patient characteristics known to be associated with functional status and quality of life outcomes (risk adjusted) and used as a performance measure at the patient level, at the individual clinician, and at the clinic level to assess quality.
So how do we use Risk Adjustment to monitor patients?
In short, you don’t. Risk adjustment levels the playing field, but it should not be used to inform care decisions.