Recent literature has drawn attention to the high and increasing rates of opioid prescribing and overdose‐related deaths in the United States. These studies have focused on community‐based and emergency department prescribing, leaving prescribing practices in the inpatient setting unexamined. Considerable variation in use, unrelated to patient characteristics, could be a marker of inappropriate prescribing practices and poor quality of care.


We studied a large cohort of adult admissions to the internal medicine service from July 2009 through June 2010 using Premier's Perspective database — the nation's largest inpatient drug utilization database, containing administrative data from more than 600 hospitals similar in composition to acute care hospitals nationwide. We defined opioid exposure as the presence of at least 1 charge for an opioid medication during the admission. We used a generalized estimating equation (GEE) with a log link to investigate the association between opioid use and patient characteristics, hospital characteristics, and U.S. census region, controlling for repeated patient admissions with an autoregressive correlation structure.


Our cohort included 1.14 million medicine admissions, spanning 288 hospitals (median age, 64 years; 46% men). Opioids were used in 49% of admissions. Morphine was the most commonly used opioid medication (20%), followed by hydrocodone (14%), and hydromorphone (13%). Opioids were administered in 39%, 50%, 53%, and 55% of admissions to hospitals in the Northeast, Midwest, South, and West, respectively. After adjustment for patient demographics, comorbidities, and hospital characteristics, opioid use was more common in patients who were female, aged 25– 64 (compared with those older and younger), white, and with nonprivate insurance. The strongest predictor of receipt of opioid medication was U.S. census region; compared with patients in the Northeast, the relative risk of opioid receipt for patients in the Midwest, South, and West was 1.26 (1.26–1.27), 1.35 (1.35–1.36), and 1.41 (1.40–1.42), respectively.


In this large pharmacoepidemiologic cohort, we found that opioids were used in almost half of hospitalized medical patients. Considerable geographic variation in opioid use was evident, even after controlling for patient and hospital characteristics. Increased attention should be paid to the role that inpatient opioid prescribing plays in the increased rates of chronic opioid use and overdose‐related deaths in the United States.