BACKGROUND: Understanding resource utilization patterns among high-cost patients may inform cost reduction strategies. OBJECTIVE: To identify patterns of high-cost healthcare utilization and associated clinical diagnoses and to quantify the significance of hot-spotters among high-cost users. DESIGN: Retrospective analysis of high-cost patients in 2012 using data from electronic medical records, internal cost accounting, and the Centers for Medicare and Medicaid Services. K-medoids cluster analysis was performed on utilization measures of the highest-cost decile of patients. Clusters were compared using clinical diagnoses. We defined Bhotspotters^as those in the highest-cost decile with ≥4 hospitalizations or ED visits during the study period. PARTICIPANTS AND EXPOSURE: A total of 14,855 Medicare Fee-for-service beneficiaries identified by the Medicare Quality Resource and Use Report as having received 100 % of inpatient care and ≥90 % of primary care services at Cleveland Clinic Health System (CCHS) in Northeast Ohio. The highest-cost decile was selected from this population. MAIN MEASURES: Healthcare utilization and diagnoses. KEY RESULTS: The highest-cost decile of patients (n = 1486) accounted for 60 % of total costs. We identified five patient clusters: BAmbulatory,^with 0 admissions; BSurgical,^with a median of 2 surgeries; BCritically Ill,^with a median of 4 ICU days; BFrequent Care,^with a median of 2 admissions, 3 ED visits, and 29 outpatient visits; and BMixed Utilization,^with 1 median admission and 1 ED visit. Cancer diagnoses were prevalent in the Ambulatory group, care complications in the Surgical group, cardiac diseases in the Critically Ill group, and psychiatric disorders in the Frequent Care group. Most hot-spotters (55 %) were in the Bfrequent care^clus-ter. Overall, hot-spotters represented 9 % of the high-cost population and accounted for 19 % of their overall costs. CONCLUSIONS: High-cost patients are heterogeneous; most are not so-called Bhot-spotters^with frequent admissions. Effective interventions to reduce costs will require a more multi-faceted approach to the high-cost population.

Hospital systems, payers, and regulators have focused on reducing length of stay (LOS) and early readmission, with uncertain benefit. Interpretable machine learning (ML) may assist in transparently identifying the risk of important outcomes. We conducted a retrospective cohort study of hospitalizations at a tertiary academic medical center and its branches from January 2011 to May 2018. A consecutive sample of all hospitalizations in the study period were included. Algorithms were trained on medical, sociodemographic, and institutional variables to predict readmission, length of stay (LOS), and death within 48-72 h. Prediction performance was measured by area under the receiver operator characteristic curve (AUC), Brier score loss (BSL), which measures how well predicted probability matches observed probability, and other metrics. Interpretations were generated using multiple feature extraction algorithms. The study cohort included 1,485,880 hospitalizations for 708,089 unique patients (median age of 59 years, first and third quartiles (QI) [39, 73]; 55.6% female; 71% white). There were 211,022 30-day readmissions for an overall readmission rate of 14% (for patients ≥65 years: 16%). Median LOS, including observation and labor and delivery patients, was 2.94 days (QI [1.67, 5.34]), or, if these patients are excluded, 3.71 days (QI [2.15, 6.51]). Predictive performance was as follows: 30-day readmission (AUC 0.76/BSL 0.11); LOS > 5 days (AUC 0.84/BSL 0.15); death within 48-72 h (AUC 0.91/BSL 0.001). Explanatory diagrams showed factors that impacted each prediction.

BACKGROUND: Rates of preventable admissions will soon be publicly reported and used in calculating performance-based payments. The current method of assessing preventable admissions, the Agency of Healthcare Research and Quality (AHRQ) Preventable Quality Indicators (PQI) rate, is drawn from claims data and was originally designed to assess population-level access to care. OBJECTIVE: To identify the prevalence and causes of preventable admissions by attending physician review and to compare its performance with the PQI tool in identifying preventable admissions. DESIGN: Cross-sectional survey. SETTING: General medicine service at an academic medical center. PARTICIPANTS: Consecutive inpatient admissions from December 1-15, 2013. MAIN MEASURES: Survey of inpatient attending physicians regarding the preventability of the admissions, primary contributing factors and feasibility of prevention. For the same patients, the PQI tool was applied to determine the claims-derived preventable admission rate. KEY RESULTS: Physicians rated all 322 admissions and classified 122 (38 %) as preventable, of which 31 (25 %) were readmissions. Readmissions were more likely to be rated preventable than other admissions (49 % vs. 35 %, p = 0.04). Application of the AHRQ PQI methodology identified 75 (23 %) preventable admissions. Thirty-one admissions (10 %) were classified as preventable by both methods, and the majority of admissions considered preventable by the AHRQ PQI method (44/78) were not considered preventable by physician assessment (K = 0.04). Of the preventable admissions, physicians assigned patient factors in 54 (44 %), clinician factors in 36 (30 %) and system factors in 32 (26 %). CONCLUSIONS: A large proportion of admissions to a general medicine service appeared preventable, but AHRQ's PQI tool was unable to identify these admissions.Before initiation of the PQI rate for use in pay-forperformance programs, further study is warranted.KEY WORDS: preventable admissions; quality indicators; value-based purchasing.

This study describes a new method to directly quantify the cost of defensive medicine. Defensive medicine appears to have minimal impact on primary care costs.