ORIGINAL RESEARCH Opioid Utilization and Opioid-Related Adverse Events in Nonsurgical Patients in US Hospitals Shoshana J. Herzig, MD, MPH1,2*, Michael B. Rothberg, MD, MPH3, Michael Cheung, MBA1, Long H. Ngo, PhD1,2, Edward R. Marcantonio, MD, SM1,2,4 1Division of General Medicine and Primary Care (Herzig, Cheung, Ngo, Marcantonio), Beth Israel Deaconess Medical Center, Boston, Massachu- setts; 2Harvard Medical School (Herzig, Ngo, Marcantonio), Boston, Massachusetts; 3Department of Internal Medicine, Medicine Institute, Cleveland Clinic, Cleveland, Ohio; 4Division of Gerontology (Marcantonio), Beth Israel Deaconess Medical Center, Boston, Massachusetts. BACKGROUND: Recent studies in the outpatient setting morphine equivalents. Opioid-prescribing rates ranged from have demonstrated high rates of opioid prescribing and 5% in the lowest-prescribing hospital to 72% in the highest- overdose-related deaths. Prescribing practices in hospital- prescribing hospital (mean, 51% 6 10%). After adjusting for ized patients are unexamined. patient characteristics, the adjusted opioid-prescribing rates ranged from 33% to 64% (mean, 50% 6 standard OBJECTIVE: To investigate patterns and predictors of deviation 4%). Among exposed, 0.60% experienced severe opioid utilization in nonsurgical admissions to US hospitals, opioid-related adverse events. Hospitals with higher opioid- variation in use, and the association between hospital-level prescribing rates had higher adjusted relative risk of a use and rates of severe opioid-related adverse events. severe opioid-related adverse event per patient exposed DESIGN, SETTING, AND PATIENTS: Adult nonsurgical (relative risk: 1.23 [1.14-1.33] for highest-prescribing com- admissions to 286 US hospitals. pared with lowest-prescribing quartile). MEASUREMENTS: Opioid exposure and severe opioid- CONCLUSIONS: The majority of hospitalized nonsurgical related adverse events during hospitalization, defined using patients were exposed to opioids, often at high doses. Hos- hospital charges and International Classification of Diseases, pitals that used opioids most frequently had increased Ninth Revision, Clinical Modification (ICD-9-CM) codes. adjusted risk of a severe opioid-related adverse event per RESULTS: Of 1.14 million admissions, opioids were used in patient exposed. Interventions to standardize and enhance 51%. The mean 6 standard deviation daily dose received in the safety of opioid prescribing in hospitalized patients oral morphine equivalents was 68 6 185 mg; 23% of should be investigated. Journal of Hospital Medicine exposed received a total daily dose of 100 mg oral 2014;9:73–81. VC 2013 Society of Hospital Medicine Recent reports have drawn attention to the high and to determine patterns and predictors of opioid utiliza- increasing rates of opioid prescribing and overdose- tion in nonsurgical admissions to US medical centers, related deaths in the United States.1–9 These studies hospital variation in use, and the association between have focused on community-based and emergency hospital-level use and the risk of opioid-related department prescribing, leaving prescribing practices in adverse events. We hypothesized that hospitals with the inpatient setting unexamined. Given that pain is a higher rates of opioid use would have an increased frequent complaint in hospitalized patients, and that risk of an opioid-related adverse event per patient the Joint Commission mandates assessing pain as a vital exposed. sign, hospitalization is potentially a time of heightened use of such medications and could significantly contrib- METHODS ute to nosocomial complications and subsequent outpa- 10 Setting and Patients tient use. Variation in prescribing practices, unrelated We conducted a retrospective cohort study using data to patient characteristics, could be a marker of inappro- from 286 US nonfederal acute-care facilities contribut- priate prescribing practices and poor quality of care. ing to the database maintained by Premier (Premier Using a large, nationally representative cohort of Healthcare Solutions, Inc., Charlotte, NC). This data- admissions from July 2009 to June 2010, we sought base, created to measure healthcare utilization and quality of care, is drawn from voluntarily participat- ing hospitals and contains data on approximately 1 in *Address for correspondence and reprint requests: Shoshana every 4 discharges nationwide.11 Participating hospi- J. Herzig, MD, Beth Israel Deaconess Medical Center, 1309 Beacon St, Brookline, MA 02446; Telephone: 617-754-1413; Fax: 617-754-1440; tals are similar in geographic distribution and metro- E-mail: [email protected] politan (urban/rural) status to hospitals nationwide, Additional Supporting Information may be found in the online version of although large, nonteaching hospitals are slightly this article. overrepresented in Premier. The database contains Received: July 16, 2013; Revised: September 26, 2013; Accepted: patient demographics, International Classification of September 30, 2013 2013 Society of Hospital Medicine DOI 10.1002/jhm.2102 Diseases, Ninth Revision, Clinical Modification (ICD- Published online in Wiley Online Library (Wileyonlinelibrary.com). 9-CM) codes, hospital demographics, and a date- An Official Publication of the Society of Hospital Medicine Journal of Hospital Medicine Vol 9 | No 2 | February 2014 73 Herzig et al | Opioids and Opioid-Related Adverse Events stamped log of all charges during the course of each yses by the Agency for Healthcare Research and Qual- hospitalization, including diagnostic tests, therapeutic ity (AHRQ).14,15 To avoid capturing adverse events treatments, and medications with dose and route of associated with outpatient use, we required the ICD-9- administration. The study was approved by the insti- CM code to be qualified as not present on admission tutional review board at Beth Israel Deaconess Medi- using the present on admission indicator required by cal Center and granted a waiver of informed consent. the Centers for Medicare and Medicaid Services for all We studied a cohort of all adult nonsurgical admis- discharge diagnosis codes since 2008.16 sions to participating hospitals from July 1, 2009, through June 30, 2010. We chose to study nonsurgical Covariates of Interest admissions, as patients undergoing surgical procedures We were interested in the relationship between both have a clear indication for, and almost always receive, patient and hospital characteristics and opioid expo- opioid pain medications. We defined a nonsurgical sure. Patient characteristics of interest included (1) admission as an admission in which there were no demographic variables, such as age, sex, race (self- charges for operating-room procedures (including reported by patients at the time of admission), marital labor and delivery) and the attending of record was status, and payer; (2) whether or not the patient spent not a surgeon. We excluded admissions with unknown any time in the intensive care unit (ICU); (3) comor- gender, since this is a key demographic variable, and bidities, identified via ICD-9-CM secondary diagnosis admissions with a length of stay greater than 365 codes and diagnosis-related groups using Healthcare days, as these admissions are not representative of the Cost and Utilization Project Comorbidity Software, typical admission to an acute-care hospital. At the version 3.7, based on the work of Elixhauser hospital level, we excluded hospitals contributing et al17,18; (4) primary ICD-9-CM discharge diagnosis <100 admissions owing to resultant lack of precision groupings, selected based on hypothesized associations in corresponding hospital prescribing rates, and hospi- with receipt of opioids, and based on the Clinical tals that did not prescribe the full range of opioid Classifications Software (CCS)—a diagnosis and pro- medications (these hospitals had charges for codeine cedure categorization scheme maintained by the only), as these facilities seemed likely to have unusual AHRQ, and defined in the Appendix19; (5) and non– limitations on prescribing or incomplete data capture. operating-room-based procedures potentially necessi- tating opioids during the admission, selected from the Opioid Exposure 50 most common ICD-9-CM procedure codes in our We defined opioid exposure as presence of 1 charge cohort and grouped as cardiovascular procedures for an opioid medication during the admission. (catheterization and insertion of vascular stents), gas- Opioid medications included morphine, hydrocodone, trointestinal procedures (upper and lower endoscopy), hydromorphone, oxycodone, fentanyl, meperidine, and mechanical ventilation, further defined in the methadone, codeine, tramadol, buprenorphine, levor- Appendix. Hospital characteristics of interest included phanol, oxymorphone, pentazocine, propoxyphene, number of beds, population served (urban vs rural), tapentadol, butorphanol, dezocine, and nalbuphine. teaching status, and US census region (Northeast, We grouped the last 9 into an “other” category owing Midwest, South, West). to infrequent use and/or differing characteristics from the main opioid drug types, such as synthetic, semi- Statistical Analysis synthetic, and partial agonist qualities. We calculated the percent of admissions with expo- sure to any opioid and the percent exposed to each Severe Opioid-Related Adverse Events opioid, along with the total number of different We defined severe opioid-related adverse events as opioid medications used during each admission. We either naloxone exposure or an opioid-related