Gray B, Vandergrift J, Lipner RS. — American Board of Internal Medicine
Presented: AcademyHealth Annual Research Meeting, June 2014
Objective: To measure how well a patient's assessment of their physician's care quality predicts clinical (chart-based) measures of care quality.
Study Design: Quality of care measures (QMs) were drawn from patients of 1,299 internists who completed an American Board of Internal Medicine hypertension or diabetes practice improvement module (PIM) between 07/2011 and 11/2012 to fulfill their 10-year Maintenance of Certification requirement. PIM-QMs were drawn from patient survey responses and chart abstractions (25 per physician). Applying data drawn from charts, the following physician-level clinical QM composites were constructed by applying weights and a scoring algorithm from an expert panel of internists: (1) an overall chart-based composite, (2) a process of care composite and (3) an intermediate outcome composite. Patient-reported QMs included an overall assessment of care quality and a composite based on patient-reported self-care items. Patient-reported QMs were constructed to mirror typical measures of patient-centered care (proportion with excellent or very good ratings) and typical website physician star rating systems (1=Poor satisfaction; 2=Fair; 3=Good; 4=Very Good; 5=Excellent).
Methodology: Clinical QMs were applied as dependent measures in separate regressions with one or both patient-reported QMs as predictors. Because the PIM-QMs were bounded by one and zero, a binomial regression was used with a logit link to construct these estimates accounting for the specific PIM that was completed and whether an internist subspecialized. An IV methodology was used to account for the possibility that patients who were surveyed also had their charts audited because our goal was to measure whether patient-reported QMs drawn from one set of patients predicted clinical QMs drawn from another set of patients.
Population Studied: Mid-career board certified internists who routinely treat patients with diabetes or hypertension.
Principal Findings: Regardless of the manner in which patient-reported QMs were applied, they were clinically and statistically significant predictors of all the clinical QM composites (ps<.01). For example, a one out of five star rating increase in either the overall patient assessment of care quality or the self-care quality composite was associated with about an 11 percentage point improvement in the overall QM composite score. In terms of individual clinical QMs, this one star increase was associated with an 8 percentage point improvement in the share of a physician's patients with LDL under control (Ps< .001). Risk adjustment increased the magnitude of these results by about one percentage point.
Conclusion: QMs drawn from patients predict clinical QMs drawn from charts.
Implications for Policy or Practice: From a patient-centered care perspective, the study findings support the notion that, regardless of their intrinsic value, the patient voice is a valid predictor of arguably less subjective measures of care quality drawn from chart abstractions. For consumers, this study suggests that there is a potential for reliable patient ratings on the Web to provide valuable information regarding a provider's clinical quality of care. The take-home message for providers is that they should take seriously feedback from their patients not only because it is an important dimension of general care quality but also as a barometer of clinical care quality.
For more information about this presentation, please contact Research@abim.org.