Q: We get a different readmission rate for our nursing center from several sources, so how do I tell which rate is right?

A: Currently, there is no standard method that is used by the Centers for Medicare & Medicaid Services (CMS) to calculate a publicly reported readmission rate for skilled nursing facilities (SNFs). However, CMS does calculate a readmission rate during the QIS survey that I discussed in some detail in my June 2011 column. This 30-day readmission rate, based on a random sample of 30 admissions, is valid for surveyors to use in Stage 1 to determine whether to conduct a Stage 2 investigation.
In Stage 2, surveyors review individual readmissions to determine when they were elective or appropriate, and also when care was inadequate given the individual resident’s risk for readmission. For calculating a raw (or unadjusted) readmission rate, the QIS methodology is the best.
While your unadjusted rate is important so that you know the percentage of your admissions that are readmitted to hospital in 30 days, this is not the “right rate” to determine if you are improving when conducting QAPI, or when comparing with other facilities. When readmission rates are used for comparing facilities (such as in public reporting) or over time, then adjusting for readmission risk is essential.
A “risk-adjusted rate” is necessary for these purposes. Our research team has been calculating risk-adjusted rates that are used by the Medicare Payment Advisory Commission (MedPAC) in its annual reports on SNF quality since 2004 (see March 2014 “MedPAC Report to the Congress”). The challenges in calculating these risk-adjusted rates can result in different methods and different rates, some of which are flawed.
Risk adjustment requires predicting the likelihood that each resident will be readmitted to determine an “expected rate” of readmission. The selection of resident characteristics for making this prediction is where the science comes in. We use a combination of comorbid diseases; functional measures; and measures of cognition, mental health, and selected conditions. We do not use services such as therapy, which are under the discretion of the facility to provide, or prior hospitalization, which would give facilities that hospitalized residents multiple times a distinct advantage. Inclusion of these for risk adjustment is not appropriate.
Depending on what is included in the model and how they are weighted, one can have very different risk-adjusted rates, even though the raw rate is the same. Using the risk models that CMS uses for the 30-day readmission measures for hospitals has serious limitations, because these were developed for hospital discharges with selected diagnoses that go to any setting, including home with no further care. These models do not fit discharges of frail SNF residents well, because discharges to SNFs represent only a small portion of all hospital discharges.
In summary, the raw rate is the right rate when you want to know the proportion of your admissions that get readmitted. However, a correctly calculated risk-adjusted rate is necessary for tracking your rates over time; comparing with other facilities; or in discussions with hospitals, physicians, or integrated care systems such as managed care or accountable care organizations.
Andy Kramer, MD, is a long term care researcher and professor of medicine who was instrumental in the design and development of the Quality Indicator Survey (QIS).