Sedative Prescriptions After Discharge Linked To Adverse Events In Older Adults

Background: Whether sedative prescriptions after hospital admissions are associated with poor patient outcomes is unknown. We sought to determine the incidence and risk of adverse events associated with sedatives in older adults within 30 days after hospital discharge.

Methods: We conducted a population-based cohort study involving older adults (age ≥ 66 yr) discharged alive from hospital in Ontario (2003 to 2023). We assessed the association of sedative prescriptions (benzodiazepines, antidepressant sedatives, or antipsychotics) filled within 7 days after discharge with falls (with or without fracture), emergency department (ED) visits, and hospital readmission. We used cause-specific proportional hazard regression for outcomes other than death and Cox proportional hazard models for death to assess the association between a sedative prescription filled after discharge and the outcomes. Because of an interaction with prehospital sedative prescription, we stratified results for prehospital sedative-naive and sedative-exposed status.

Results: Among 1 868 484 older adults (mean age 77 yr, 52.1% female), 13.2% filled a sedative prescription after discharge; of these patients, 31.0% were sedative naive before hospital admission. Falls occurred in 1.6% ( n = 30 626), ED visits in 21.3% ( n = 397 402), hospital readmissions in 12.4% ( n = 231 191), and death in 3.8% ( n = 70 661). The adjusted hazard ratio (HR) for all outcomes was increased among those who filled a sedative prescription after discharge (v. no filled prescription) and were sedative naive before hospital admission (fall: 1.20, 95% confidence interval [CI] 1.13 to 1.26; ED visit 1.20, 95% CI 1.19 to 1.22; hospital readmission: 1.20, 95% CI 1.17 to 1.22; and death: 1.78, 95% CI 1.73 to 1.83). For those who were exposed to sedatives before their hospital admission, there was an increased hazard of death (adjusted HR 1.08, 95% CI 1.05 to 1.11), but not for the other outcomes. For sedative-naive older adults, benzodiazepines were associated with an increased hazard of all outcomes, antipsychotics were associated with an increased hazard of falls and death, and antidepressant sedatives were associated with a decreased hazard of falls.

Interpretation: New sedative prescriptions filled within 7 days after discharge were associated with increased hazards of a fall, ED visit, hospital admission, and death within 30 days after discharge, with differences based on the sedative class. These findings have important implications for in-hospital medication review and falls-risk assessment for older adults in Canada.

Prescribing potentially inappropriate medications for older adults is discouraged, as medications with limited effectiveness in this population may be associated with adverse events, resulting in harm or requiring additional health care services. 1 – 4 Sedatives and antipsychotics are listed as potentially inappropriate medications because of their known association with adverse events. 1 – 4 Serious adverse events associated with their use in older adults living at home, in community settings, or in long-term care facilities include falls, fractures, delirium, cognitive impairment, and death. 5 – 9

In hospitals, sedatives and antipsychotics are frequently prescribed for patients to facilitate diagnostic or therapeutic interventions, as well as to manage sleep disturbances, anxiety, and delirium. 10 – 12 The initiation of sedative and antipsychotic prescriptions in hospital is generally intended for short-term use while the patient is in hospital. However, in population-based studies, 3% to 10% of older adults in North America, Europe, and Asia who were not prescribed these medications before hospital admission (i.e., prehospital sedative naive) filled a new benzodiazepine or antipsychotic prescription at or shortly after hospital discharge. 13 – 15 Although adverse events associated with the initiation of sedatives or antipsychotics (e.g., falls) have been well described in community-dwelling and long-term care older adult populations, 5 – 9 , 16 – 22 the risks to patients who continue new hospital-initiated sedative and antipsychotic prescriptions after hospital discharge are unknown, as are the risks associated with chronic home sedative or antipsychotic prescriptions after hospital discharge. The association with adverse events may differ between new use and chronic use because of physiologic changes (e.g., organ dysfunction and altered drug metabolism) or debilitation from an acute illness requiring hospital admission. Given the widespread use of these medications in hospital, it is important to investigate the associated risks with their use after hospital discharge to inform discharge prescribing practices. We sought to determine the incidence and risk of serious adverse events, specifically falls requiring medical attention, emergency department (ED) visits, hospital readmission, and death, associated with a new or continued sedative prescription among older adults during the first 30 days after hospital discharge.

We conducted a population-based cohort study using health care, prescription, and administrative data available at ICES ( www.ices.on.ca ) for the province of Ontario, Canada. ICES is an independent, nonprofit research institute whose legal status under Ontario’s health information privacy law allows it to collect and analyze health and demographic data without consent, for the purpose of evaluating and improving the health system. The province of Ontario publicly funds hospital and physician care for all residents (approximately 16 million), including outpatient prescriptions for those aged 65 years and older. We linked databases at the individual patient level using unique encoded identifiers at ICES (Appendix 1, Table E1, available at www.cmaj.ca/lookup/doi/10.1503/cmaj.251965/tab-related-content ). The missing rate for data used in this study was less than 0.5%.

Our analysis plan was registered with the Open Science Framework (Apr. 11, 2023). We followed the Strengthening the Reporting of Observational Studies in Epidemiology and Reporting of Studies Conducted Using Observational Routinely-Collected Health Data for Pharmacoepidemiology statements for study reporting (Appendix 1, Table E2).

We included all adults aged 66 years and older who were discharged alive from an acute care hospital admission between Apr. 1, 2003, and Aug. 31, 2023. We restricted cohort entry to age 66 years and older at the hospital discharge date (hospital discharge date = index date) to permit a 180-day look-back in the outpatient prescription claims database (Ontario Drug Benefit [ODB]) for prehospital prescription medication exposures. We excluded hospital admissions with missing or invalid provincial health card numbers, missing age or age older than 105 years at the index date, missing sex, hospital length of stay of 1 day or less, and hospital admissions with a prior admission in the 180 days before the admission date (as we could not confirm sedative exposure in ODB while a patient was in hospital). 23 Additionally, we excluded hospital admissions after which patients were discharged to inpatient facilities that do not submit prescription claims to ODB, such as facilities for palliative care, rehabilitation, or mental health. For patients with more than 1 hospital admission who met the inclusion criteria, we selected 1 admission at random for inclusion.

The primary exposure was any sedative prescription filled within 7 days after hospital discharge. This time window has been used in similar epidemiologic studies and supports the attribution of a filled prescription to the hospital encounter. 13 , 24 Our definition of a sedative prescription, based on market availability, prescribing practices, and the ODB formulary, was a benzodiazepine, an antidepressant sedative (specifically, mirtazapine and trazodone), or an antipsychotic, dispensed for any dose or duration (see Appendix 1, Table E3 for a full list of medications). 13 Nonbenzodiazepine Z-drug (e.g., zopiclone, zolpidem, zaleplon) sedative–hypnotics were not eligible for reimbursement under the ODB at any point in the study; therefore, we could not include them in our analysis.

The primary outcome was a fall, with or without a fracture, that resulted in medical attention within 30 days after hospital discharge (i.e., documented in the ED, outpatient setting [including hospital- or community-based ambulatory care services, day surgery, outpatient, and community-based clinic services], or hospital records). These have excellent positive predictive value (0.91 for falls; 0.81 to 0.96 for fractures, depending on fracture site). 25 Secondary outcomes were all-cause ED visits, hospital readmissions, and death within 30 days after hospital discharge. 25 , 26 The data sources and diagnostic codes used to identify each outcome have been previously described and validated (Appendix 1, Table E3). 17 , 25 , 27 – 29

We identified the following patient-level characteristics at the index date: age, sex, neighbourhood median income quintile, Deyo-adapted Charlson Comorbidity Index (categorized as 0, 1–2, ≥ 3), 30 John Hopkins Adjusted Clinical Group System frailty markers in version 10 (binary), 31 and the presence of conditions associated with falls in the published literature — specifically, dementia, Parkinson disease, osteoporosis, alcoholism, and history of falls that received medical attention in the last year. We identified in-hospital therapies or diagnoses, including surgery; acute kidney injury; receipt of renal replacement therapy; sepsis; stroke; intensive care unit (ICU) admission; billing for geriatric, palliative, or psychiatry services; length of hospital stay; hospital discharge destination; and year of hospital discharge. We assessed for other medication exposures associated with falls (i.e., anticonvulsants, antidepressants [other than mirtazapine and trazodone], antihypertensives, and opioids). 32 For each medication class, we assessed prescriptions filled within 180 days before hospital admission to define patients as either prehospital naive or exposed.

We compared characteristics of patients with and without a sedative prescription filled within 7 days after hospital discharge, stratified by prehospital sedative exposure, reporting standardized mean differences. We calculated the unadjusted incidence of each outcome of interest for the entire cohort and stratified by whether patients did or did not fill a sedative prescription within 7 days after hospital discharge. We used cumulative incidence function (CIF) curves to describe outcomes, with death as a competing risk, and the Gray test to evaluate differences in the incidence function between groups.

We entered the association between a sedative prescription filled within 7 days after hospital discharge and falls, ED visits, and hospital readmission in the model as time-dependent covariates and conducted the analysis using cause-specific proportional hazard regression, with death as a competing risk and adjustment for the covariates listed above (patient-level and in-hospital therapies or diagnoses). 33 We used a Cox proportional hazard model to examine the association between a sedative prescription filled within 7 days after hospital discharge and death. Follow-up time was constructed from the index date to the first event, death, or administrative censoring at 30 days. We expressed measures of association between exposure and outcome as adjusted hazard ratios (HRs) with 95% confidence intervals (CIs). A priori, we planned to test for an interaction with prehospital prescription exposure within 180 days before hospital admission. As the interaction was significant, we presented results stratified on prehospital exposure status. We tested for an interaction with year of discharge; however, this interaction was not significant and was therefore not included in the models. We also calculated the E-value to quantify the strength of association that an unmeasured confounder would need to have with both the exposure and the outcome to nullify any identified associations. 34

We conducted a sensitivity analysis by modifying the primary analysis to shorten the observation window to 14 days, as we anticipated that most adverse events would occur early after hospital discharge. We conducted additional sensitivity analyses by excluding patients who were readmitted or died within 7 days after hospital discharge and by excluding patients who had any billing for palliative care while admitted to hospital. It is possible that patients with a prehospital sedative prescription did not need to fill a prescription within 7 days after hospital discharge if they had an adequate supply available at home. We therefore conducted a post hoc sensitivity analysis to classify those who had filled a sedative prescription within 30 days before hospital admission as a patient who filled the prescription within 7 days after discharge, and set the time of prescription fill as the hospital discharge date.

We assessed the following subgroups to explore differences in the association between treatment and the primary outcome: age (≤ 75 or > 75 yr), sex, surgery, ICU admission, dementia, and frailty, 31 also stratifying by prehospital sedative exposure. Finally, we assessed the relationship between each sedative class and each outcome. We examined Schoenfeld residuals for proportionality assumptions (the assumptions were met).

We performed analyses using SAS Enterprise Guide version 7.15 (SAS Institute) and created figures with R version 1.4. All p values reported are 2-sided, with a threshold of less than 0.05 considered significant.

The research ethics boards of Sunnybrook Health Sciences Centre (no. 2688) and Sinai Health (no. 19–02327-C) approved the study protocol.