The MPIRICA Quality Score is designed to be an objective comparison of facilities and physicians, so it is essential that the quality measurements upon which the score is based are themselves objective. Based on currently available data that are reported by hospitals, MPIRICA uses the following as the key measurements critical for evaluating quality:
While the data on readmissions and patient mortality inside a hospital require no subjective interpretation, the data on complications are not as clearly documented. This is because recording complications requires individual judgement, and coders may differ in their interpretation of events. As a result, complications may be recorded differently across facilities (or even by different coders within the same facility). Furthermore, even if codings were consistent, the codes themselves do not always separate a mild complication (like a simple infection) from a serious one that requires more medical attention (like sepsis).
For all these reasons, MPIRICA uses prolonged risk-adjusted length of stay to identify serious complications. Medical records have been reviewed and compared to cases using this approach to confirm that prolonged risk-adjusted length of stay is a valid way to determine when a severe complication occurred. In addition, this methodology has been published in respected peer reviewed journals. The prolonged risk adjusted length of stay measure is not only an effective way of associating cases with severe complications and their higher costs, but also shows strong association to adverse events after hospital discharge (e.g. complications and readmissions).
Patient risk adjustment is an essential component of the methodology used by MPIRICA because it is critical for achieving apples-to-apples comparisons between hospitals or physicians. For example, a hospital that typically handles the most complex cases or the sickest patients should not be penalized for the greater risks inherent to their more difficult patient population.
Over 500 risk factors, created by physicians, are used to reflect the health status of patients. These risk factors include patients’ demographic, diagnostic, and procedural information. Models are created and hand crafted by physicians to ensure they reflect medical reality. This process permits the valid comparison of performance across hospitals or physicians independent of the severity of their cases.
MPIRICA relies on models that predict outcomes at either the hospital or physician level based upon the prior health conditions of their patients. The models are built using data reported by hospitals nationwide for each procedure scored. MPIRICA uses the most recent three years of available data (or four years for surgeons) and scores are updated as new data becomes available. The data is released annually and updated as frequently as quarterly.
As the data from healthcare databases are complex, the analysis takes into account a variety of ways data can be missing, incorrect or otherwise defective. Data integrity tests and exclusion criteria are applied to problematic data. For example, we know cancer is not a complication resulting from a procedure, so if cancer is coded as a complication, the coding is corrected to reflect the cancer as a pre-existing condition.
The resulting predictive model takes a set of patients and their associated health status as inputs and produces predictions of adverse outcomes. The predicted values for all the patients of a specific hospital or physician can then be added together and compared to the sum of the observed outcomes of that provider.
Based on their patient mix, each hospital or physician will have a predicted outcome from the model for each of the measures we track. The predicted outcome is what the provider should have achieved for a procedure given their patient mix, and is compared against the observed outcome of what actually occurred. The relationships between observed and predicted outcomes are used to calculate the MPIRICA Quality Score.
The MPIRICA Quality Score is the sum of points awarded after comparing hospital or physician observed and predicted outcomes. To calculate a score, we established base points that can be earned for each broader measure, and apply rules for how points can be earned. Finally, we align the relative performance compared to the average, and the components are summed to produce a final score.
For a more in depth look at the statistical methodology, please send a message to email@example.com to request the MPIRICA Methodology Whitepaper.