Next Steps in Measuring Clinical Quality Joe V. Selby, MD Division of Research Kaiser Permanente Northern California
Differences in Clinical Quality – Diabetes Care * HEDIS Health Plan Summary Data
What We Know What We Measure The Chasm in Clinical Quality Assessment Quantitative effects of many process measures and of differences in outcomes on survival and on non-fatal complications in populations – from clinical trials Processes of care not known to be related to outcomes or effectiveness Semi-quantitative outcomes (Hb A1c >9.5%, LDL-C < 100) that hide more effectiveness differences than they reveal
Population Rates for Simple Process Measures do NOT Consistently Reflect Clinical Benefit in those Populations Point #1
PacifiCare Texas Kaiser Permanente No. California Pacific Health Research Institute U. Michigan Indiana U. UCLA UMDNJ CDC Centers for Disease Control - Sponsor and Data Coordinating Center Translating Research Into Action for Diabetes A multi-center cohort study of diabetes in managed care settings
10 health plans (n=500 to 2000 per plan) 67 physician groups with > 50 members in sampling frame The TRIAD Sampling Scheme (Sampling scheme: Aimed for equal numbers from each physician group within health plan, so from per physician group)
TRIAD Data ( ) Patient Surveys (telephone or mailed) – 11,928 respondents Chart Reviews – 8,757 patients Medical Director Surveys – health plan and provider group directors
Four Measures of Disease Management Intensity – from Health Plan and Provider Group Director Surveys Use of diabetes registries Use of clinician reminders Performance feedback to physicians Diabetes care management: Guideline use Patient reminders, Patient education Use of care/case managers
Provider Group Performance Difference (%) (80 th – 20 th Percentile of Dis Mgmt Intensity) PROCESS MEASURES Care Management Performance Feedback Diabetes Registry MD Reminders Hb A1c Test LDL-C Test Retinal Exam Urine Albumin Foot Exam Aspirin Advised adjusted for patient age, sex, race, education/income, diabetes treatment and duration, comorbidities, SF-12 (PCS), health plan disease mgmt intensity
Provider Group Performance Differences (80 th – 20 th Percentile of Dis Mgmt Intensity) INTERMEDIATE OUTCOMES Care Management Performance Feedback Diabetes Registry MD Reminders Hb A1c (%) Syst. Blood Pressure (mmHg) LDL- cholesterol (mg/dL) adjusted for patient age, sex, race, education/income, diabetes treatment and duration, comorbidities, health plan disease mgmt intensity
Moreover, Provider Group intensity of disease management also unrelated to the appropriateness* of treatment for each condition Provider Group Quality Scores based on process measures were unrelated to provider group levels of control of blood pressure, LDL-C or Hb A1c *Proportion in control or on appropriately aggressive pharmacotherapy
Point #2 Even if we measure evidence-based processes or outcomes, the potpourri of indicators within and across diseases dont readily yield a measure of overall clinical benefit
Differences in Clinical Quality (hypothetical) based on evidence-based processes/ outcomes
How Do We Quantify the Net Benefit? Each of these differences represents a predictable change in expected survival and complications (i.e., each measures a clinical benefit ) But practical questions remain: Which is more important, the difference in Hb A1c levels or the difference in BP control? Should plans, providers work to improve multiple measures modestly, or drive one indicator toward the optimal for all patients? We need a composite, quantitative measure of net clinical benefit that can be compared across plans, provider groups, systems.
Quality-adjusted life-year
The QALY A common metric for measuring clinical quality (both survival and quality of life) Across interventions (using aspirin, BP lowering) Across perspectives (patient, provider, purchaser) Across diseases (diabetes, CHF, CAD, asthma) Across activities (e.g., chronic disease care, prevention)
Where Do QALYs Come From?
Creating a Quantitative Metric for Diabetes Natural History Model Systolic Blood Pressure Hemoglobin A1C Aspirin Use LDL- Cholesterol Expected Survival & Complicatons Adjusted Life-expectancy Risk Adjusters
Potential Advantages of Model Expresses quality in familiar metric – life expectancy Requires clinical trial evidence clearly evidence-based Allows exploration to explain differences, which emphasizes population importance of various indicators
Potential Disadvantages/Questions Will require extensive explanation and transparency of the model to gain acceptance New evidence will have to be incorporated over time, potentially altering metrics across years Because it takes a population or public health perspective, will not capture quality of care well for rare conditions (because prevalence too small) Questions of whether and how to adjust for case-mix differences between population will have to be addressed