Crossing the Quality Assessment Chasm: Aligning Measured and True Quality of Care (Pennsylvania)
Current assessments of quality of care for diabetics rank providers based on the proportion of their diabetic patients that achieve specified given values for hemoglobin A1c (HbA1c), blood pressure, and low density lipoprotein (LDL). This approach penalizes providers who care for patients who are objectively more difficult to control, despite reasonable efforts. The main goal of this project was to develop a model of expected level of control and incorporate this model into a quality measure that ranks providers on the degree to which their patients are doing better or worse than expected instead of than by overall levels. The project team developed models of diagnoses, lab results, vital signs, demographics, and visits and evaluated these models on their ability to consistently predict HbA1c, blood pressure, and LDL. Expected values were compared with actual values, and provider rankings were calculated on the degree to which providers did better or worse than expected under the various models.
The main objectives of this project were to:
- Evaluate structural and clinical issues that may affect the validity of comparisons made by providers using quality measures for diabetes. These included: the manner in which diabetes is defined; the way patients are linked to providers; and the concordance between use of diabetes medications and achieving thresholds for quality-of-care.
- Develop a quality measure for diabetes that accounts for patient heterogeneity in terms of baseline HbA1c and expected improvement in diabetes control, based on clinical parameters and other data available through the electronic medical record.
- Explore the Diabetes Control and Complications Trial and patient data to assess the impact of year-to-year variability in an individual's diabetes control on microvascular outcomes.
While traditional risk factors of age, gender, and socioeconomic status had some explanatory power in predicting HbA1c control, the team found that the prior degree of control was the most significant predictor of current degree of control. Yet with current quality measurements, providers must start with objectively difficult patients having poor control, and improve the HbA1c in those patients beyond expectations to rank high in their care of diabetics.
This study supported the notion that the current focus on quality of care for diabetics favors providers that start with panels where a majority of patients already have good control. While some clinical and demographic parameters influence the expectation of control, these factors are dominated by the prior level of control. Although the team was able to rank providers according to a quantitative score on the degree to which panels exceeded diabetes control expectations, the standard errors around those scores were large enough that all but the highest and lowest ranked providers were statistically indistinguishable. Part of this is related to the relatively small panel sizes for many of the providers. Future work will focus on expanding the analyses to additional clinical settings to validate the findings of this study.