LAUSR

Learning Objectives: Some ICU admissions may be preventable. Yet, it is unclear whether greater use of primary care reduces rates of these admissions. In this study, we tested the relationship between primary care use and rates of potentially preventable ICU admissions, hypothesizing that areas with higher rates of primary care use would have lower rates of preventable ICU admissions. Methods: A cross-sectional, ecological study among fee-for-service U.S. Medicare beneficiaries admitted between January 2014 and September 2015. Preventable ICU admissions were defined using standard ICD-9-CM criteria from the Agency for Health Research and Quality’s Prevention Quality Indicators. Data on primary care use (beneficiaries with ≥ 1 annual primary care visit) and hospital referral regions (HRRs) were obtained from the Dartmouth Atlas. Patient-level data were obtained from the inpatient Medicare Provider and Analysis Review. After collapsing data into HRRs, we used linear regression to estimate relationships between primary care use and per capita rates of all-cause and preventable ICU admissions. Models accounted for clustering by HRR and were adjusted for demographics (age, sex, race/ethnicity, median income), comorbidity burden, provider availability (primary care, intensivists, and pulmonologists), other preventive health use, and year. Results: We identified 2,441,999 patients in 306 HRRs. HRRs varied in all-cause yearly ICU admission rates (26.8 per 1000 beneficiaries in quartile 1 versus 66.3 in quartile 4) and in preventable ICU admissions (2.3 per 1000 in quartile 1 versus 7.8 in quartile 4). After adjustment, each 10% increase in primary care utilization was associated with a 12.7 per 1000 (95% CI -20.0 to -5.4, p=0.001) decline in the rate of all cause ICU admissions and a 1.5 per 1000 (95% CI -2.7 to -0.37, p=0.010) decline in preventable ICU admissions. This effect was stronger in the subset of chronic (e.g., asthma) (-1.16, 95% CI -2.0 to -0.29, p=0.01) than acute (e.g., pneumonia) preventable admissions (-0.3, -0.6 to 0.0, p=0.037). Conclusions: Higher rates of primary care use were associated with lower rates of all-cause and potentially preventable ICU admissions within a region. This effect was most notable among preventable chronic ICU admissions. Further work using patientlevel primary care data is needed to confirm this association and examine potential mechanisms.