Putting Quality Metrics in Context: A Novel Index Approach to Measuring Inpatient Utilization

Abstract

Research ObjectiveTo determine the preliminary feasibility and reliability of a novel health plan‐level quality index reflecting multiple aspects of inpatient utilization currently measured by separate quality metrics.Study DesignAnalysis of three Healthcare Effectiveness Data and Information Set (HEDIS) measures: Acute Hospital Utilization (AHU), Hospitalizations for Potentially Preventable Conditions (HPC), and Plan All‐Cause Readmissions (PCR). Proportion of plans able to report these measures were examined within and between measurement years. Individual measure performance was scored on the observed‐to‐expected (O/E) ratio: O/E = 1 “as expected” (score of 0), O/E < 1 “better than expected” (score of +1), and O/E > 1 “worse than expected” (score of ‐1). Measure scores were summed to create an index (‐3 to +3). Performance variation was examined. Index unidimensionality and reliability were evaluated using coefficient omega (total and hierarchical) and explained common variance (ECV).Population StudiedMedicare Advantage (MA) plans reporting HEDIS in measurement years (MY) 2017 (505 plans) and 2018 (525 plans). To align reporting strata across the three metrics, analysis was restricted to beneficiaries age 65 and older.Principal FindingsA majority of plans were able to report measures in both MY2017 and MY2018 (AHU: 67.0%, HPC: 65.7%, PCR: 60.7%). In MY2018, 71.4% of plans reported all three measures. The strongest correlation in O/E ratios was observed between AHU and HPC (Pearson: 0.29, Spearman: 0.64), the weakest between AHU and PCR (Pearson: 0.19, Spearman: 0.17). A moderate correlation was observed between PCR and HPC (Pearson: 0.31, Spearman: 0.36). “Better than expected” (+1) was the most common measure‐level score on AHU (48.0%) and HPC (51.4%). “As expected” (0) was the most common score (58.5%) on PCR.The most common index score was +2 (25.1%), and mean score was 0.3. HPC performance scores were most likely to contribute to an increased index score (52.4% of cases), and AHU performance was most likely to contribute to a decreased index score (42.1% of cases). PCR performance was as likely to increase (20.5%) as decrease (21.1%) the index score. When the original O/E ratio for each measure was retained, moderate‐to‐good reliability was observed. Omega hierarchical approached the minimum 0.50 threshold (0.49), and ECV met the 0.85 threshold for unidimensionality. Omega hierarchical decreased significantly (0.33), and ECV no longer achieved the 0.85 threshold when the composite was scored as an ordinal index (‐3 to +3).ConclusionsA composite inpatient utilization index is feasible and demonstrates meaningful variation. Index‐style scoring is useful for visualization and identification of trends; however, composite reliability is achieved only when original O/E ratios are maintained. Alternative scoring approaches should be examined.Implications for Policy or PracticeInpatient utilization is a primary driver of health care costs and critical target for quality improvement. Current health plan accountability metrics focus on different components of utilization separately and may miss the relationships between them. Interventions to improve one metric may overlap into another, while over‐focusing on one aspect in isolation may lead to perverse incentives (eg, keeping total hospitalizations high to make readmission rates appear low). The proposed composite approach provides a more complete representation of inpatient care, improving transparency and accountability.