Patterns of Benzodiazepines and Hypnotic Drugs Dispensing after Hip Fracture among Older US Adults Previously on the Drugs

Abstract

Background

Benzodiazepines (BZD) and nonbenzodiazepine receptor agonist hypnotics (Z-drugs) strongly increase the risk for hip fractures and should be avoided among older adults with fall or fracture history. We assessed BZD and Z-drugs refill patterns after discharge for hip fracture among older US adults and explored factors associated with refilling.

Methods

We conducted a retrospective cohort study of US Medicare fee-for-service beneficiaries (20% random sample) aged ≥66 years, hospitalized for hip fracture, and discharged home or to a skilled nursing facility. Eligible patients had ≥365 days of Medicare Parts A, B, and D coverage and filled a BZD (2013–2019) or Z-drug (2008–2019) ≤90 days before hospitalization. We estimated cumulative incidences of refilling the same/increased dose (continuation) and reduced dose (reduction) and their corresponding 95% confidence intervals (CI) within 180 days after discharge, stratified by drug class. We examined demographic factors, comorbidities, and geriatric conditions associated with continuation versus deprescribing (reduction or no refill).

Results

We included 21,123 eligible patients [11,465 (54%) on short/intermediate-acting BZD, 3,473 (16%) on long-acting BZD, and 6,185 (29%) on Z-drugs at baseline]. The median age was 82 years, 81% were female, and 94% were non-Hispanic White. The continuation within 180 days post discharge was 54.9% (95% CI: 54.0%, 55.8%) for short/intermediate-acting BZD, 58.4% (56.7%, 60.1%) for long-acting BZD, and 56.0% (54.8%, 57.1%) for Z-drugs; 7.2%–16.3% refilled with dose reductions. Younger patients or those with anxiety were at higher risk of continuation.

Conclusions

A large proportion of US older adults on BZD or Z-drugs before hospitalization for a hip fracture continued the drugs within 180 days post hospital discharge. This highlights the need for comprehensive medication review and deprescribing intervention during transitions of care in this high-risk population.

INTRODUCTION

Hip fracture has a substantial impact on survival and quality of life for older adults and poses a significant economic burden on the healthcare system.^1,2 It is estimated that more than half a million US Medicare beneficiaries had a hip fracture between 2008 and 2018.³ Patients who experienced a hip fracture are more likely to have a recurrent fall during the recovery process, leading to higher mortality.^4,5 Therefore, the prevention of recurrent falls is crucial for older adults who have had a hip fracture.

Benzodiazepines (BZD) and nonbenzodiazepine receptor agonist hypnotics (Z-drugs) are known to be strongly associated with hip fractures among older adults as these drugs increase sedation, slow reaction times, disrupt balance and gait, and impair vision.^6,7 Although BZD are indicated for anxiety, seizure, and alcohol withdrawal management, and Z-drugs are short-term solutions for insomnia, the American Geriatrics Society (AGS)’s Beers and the European Screening Tool of Older Persons’ Prescriptions (STOPP) criteria recommend avoiding BZD and Z-drugs among older adults with a history of falls or fracture when the potential benefits outweigh harms.^8–10 However, the prevalence of BZD and Z-drugs usage in this population ranges from 12% to as high as 20%.^11–13

Hospitalizations due to hip fractures provide an opportunity for medication review and appropriate deprescribing.^11,14,15 If there is a continued indication for BZD or Z-drugs use, providers should switch to a safer alternative. When no safer alternative is available, dose reduction should be considered alongside a medication review to assess the risks and benefits for the patient. This assessment should include shared decision-making with patients, caregivers or family members, primary care providers, and other members from the patient care team. If there is no longer a clear indication for BZD or Z-drugs use, tapering strategies paired with nonpharmacological support strategies (e.g., cognitive behavioral therapy) and close monitoring for withdrawal symptoms should be pursued.^16–19 However, deprescribing among older adults in post-acute care transitions to home is often challenging due to poor care coordination between the hospital-based team and primary care providers in the community or nursing home, limited time and prioritization, or inadequate education and communication with patients and their caregivers.^20,21

While a hospitalization for a hip fracture provides an opportunity to decrease inappropriate prescribing, little is known about refilling or deprescribing patterns of BZD and Z-drugs after hospitalization.^22–24 To address this knowledge gap, we described refilling patterns after hospital discharge for a hip fracture among prevalent BZD and Z-drugs (BZDZ drugs) users and examined patient characteristics associated with refilling these drugs.

METHODS

Database

We conducted a retrospective cohort study using data from a 20% random sample of fee-for-service Medicare beneficiaries between 2007 and 2019. The database includes longitudinal claims for inpatient and outpatient services, prescription drugs, enrollment and demographics until death, disenrollment, or end of data availability (December 31, 2019). The study was approved by the University of North Carolina at Chapel Hill Institutional Review Board (IRB #21–0971).

Cohort

We assembled a cohort of patients aged 66 years or older who were hospitalized for hip fracture between January 1, 2008, and June 30, 2019 (only the first eligible hospitalization was selected if patients had multiple hospitalizations during the study period). Patients were selected if they had a Current Procedural Terminology (CPT) or International Classification of Diseases, Ninth/Tenth Revision, Clinical Modification (ICD-9/10-CM) procedure code for repair, revision, and/or reconstruction of the pelvis and hip joint, and ICD-9/10-CM diagnosis code for hip fracture as their principal diagnosis. We required patients to have 365 days of continuous enrollment in Medicare Parts A (inpatient), B (outpatient), and D (prescription) with no enrollment in a health maintenance organization prior to the hospital admission and to be discharged home or to a skilled nursing facility. Patients with an ICD 9/10 code indicating a motor vehicle crash were excluded from the study to focus on fragility/osteoporotic hip fracture. Codes used to define hip fracture and motor crash are reported in Supplementary Table S1a and Table S1b, respectively.^25–27 We also required patients to have a prescription for an oral BZDZ drug dispensed ≤90 days before admission. We used the five-level Anatomical Therapeutic Chemical (ATC) classification system to identify all BZDZ drugs listed in the Beers criteria (Table S2) and matched these medications to National Drug Codes (NDCs). We then used the NDCs to identify BZDZ drugs dispense in the claims. If patients had multiple BZDZ drugs dispensed before the admission, the most recent drug prior to admission was used to define the baseline BZDZ drug. Since Medicare Part D did not cover BZD until 2013, we excluded eligible patients who were on BZD before 2013.²⁸ A schematic diagram illustrating the study design is in Figure S1.

Outcomes

The primary outcomes were the cumulative incidences of refilling the same/increased dose (continuation) or a reduced dose (reduction) within 180 days after hospital discharge, stratified by baseline BZDZ drug ‘class’ (Figure S1). We defined baseline ‘class’ as 1) short/intermediate-acting BZD, 2) long-acting BZD, or 3) Z-drugs. We followed patients from discharge to the first refilling (categorized as either a same/increased or a reduced dose), death (competing event), disenrollment in Medicare Part D (drop-out censoring event), or the end of the study period (December 31, 2019, administrative censoring), whichever occurred first. Dose comparison was based on the daily diazepam milligram (DDM) equivalent (mg/day) change between the baseline BZDZ fill and the first refill using the formula DDM = ([dose conversion factor] × [the medication strength] × [the prescription quantity]) ÷ [number of days supply], where the dose conversion factor for each drug is in Table S3.²⁹ We selected the discharge date as the index date to eliminate the impact of hospitalization on the assessment of refilling, as patients with longer stays may get the prescription during the hospitalization and delay refilling after discharge. We treated death as a competing event because refilling prescription drugs could not occur after death.

Since skilled nursing facility stays following discharge may preclude patients from refilling at a pharmacy, we also assess the refilling patterns after 90 days post hospital discharge among patients who had continuous Medicare Part D coverage, excluding those who died or lost coverage during the first 90 days after discharge (Figure S1). Refilling patterns were assessed by comparing the baseline BZDZ drug and the first refill after 90 days post-discharge (with a 90-day grace period) using the DDM equivalent for dose comparison. We categorized refilling patterns as: 1) refilling the same/increased dose, 2) refilling a reduced dose, 3) no refilling of BZDZ drugs and switching to an alternative (e.g., no refill of lorazepam but initiated doxepin after discharge), or 4) no refilling of BZDZ drugs and no alternatives. A list of BZDZ drugs and their alternatives for each indication is reported in Table S2.^10,30,31

Covariates

Baseline demographic characteristics included age at admission (66–74, 75–84, ≥85 years old), race and ethnicity (Asian/Pacific Islander, Hispanic/Latino, non-Hispanic Black, non-Hispanic White, other race or ethnicity), sex (male or female), and Medicare low-income subsidy (yes or no). Baseline clinical characteristics were assessed using claims in the 365 days prior to the hospital admission (look-back period) and included: Faurot Frailty Index (FFI) [low probability of frailty (FFI<0.05), low to medium probability of frailty (0.05≤FFI<0.1), medium probability of frailty (0.1≤FFI<0.2), medium to high probability of frailty (0.2≤FFI<0.4), and high probability of frailty (FFI≥0.4)], Gagne comorbidity index (−2–1, 2–3, 4–5, >5), and comorbidities (dementia, chronic pain, and depression).^32–36 BZD indications included insomnia, anxiety, seizure disorder, spasticity, and alcohol withdrawal, and Z-drugs are indicated for insomnia.^8,18 Given that the indication for a dispensed prescription is not recorded in claims data, we identified these conditions using ICD 9/10 diagnosis code in the 365-day look-back period (Table S1b).^36–39 Patients with multiple conditions were counted in each relevant condition. Co-medications were assessed using claims in the 90 days prior to hospitalization and included antidepressants, opioids, and polypharmacy. Polypharmacy was defined as having 5 or more drugs (including BZDZ drugs, by generic drug name) dispensed in the 90 days before hospital admission.

Statistical analysis

We used the Aalen Johansen function⁴⁰ to calculate the outcome-specific cumulative incidences of refilling the same/increased (continuation) and reduced dose (reduction) and their corresponding 95% confidence intervals (CI) within 180 days after discharge, stratified by baseline BZDZ drug class. We repeated this analysis in patients stratified by different conditions (insomnia, anxiety, seizure, and spasticity). We performed a series of sensitivity analyses comparing continuation and reduction within 180 days post-discharge under various assumptions. In the first, we restricted to long-time users of BZDZ drugs, defined as patients who had an additional dispensing of the same baseline BZDZ drug from 91 to 365 days before hospital admission. The second restricted to recent users, defined as those who had the baseline BZDZ drug dispensed ≤30 days before admission. The third restricted to concurrent users of opioids or antidepressants with dispensations in the 90 days before admission, since evidence suggests that these patients have higher risk of fall or fracture than the monotherapy users and should reduce the use of other medications if one is a must use. ^10,36,43 Finally, we restricted to non-pathological hip fracture patients (excluding those with an ICD 9/10 code containing 733, M80, or M84).

To assess factors associated with continuation versus deprescribing (reduction or no refill), we estimated the ratios of cumulative incidence of continuation across different levels of baseline characteristics and 95% CIs at 180 days post-discharge. ^41,42 The ratios (CIR) were estimated from a multivariable log-linear model (including each baseline characteristic, with log link and Poisson distribution), treating death and loss of Medicare Part D coverage as competing events.

For refilling patterns after 90 days post hospital discharge, we calculated the proportions and 95% CIs of patients who refilled the same/increased dose, reduced dose, or had no BZDZ refill with or without initiating an alternative in patients who were alive and had continuous Medicare coverage at 90 days after discharge, separately, by baseline BZDZ drug class and drug. We also reported the number of deaths during 91–180 days after discharge, by baseline BZDZ drug class and drug. We assessed trends in refilling the same/increased dose by year of index hospitalization. All statistical analyses were performed using SAS, version 9.4 (SAS Institute, Cary, NC).

RESULTS

Characteristics of study population

The final cohort included 21,123 eligible patients (Figure S2). At baseline, 11,465 (54%) patients were on short/intermediate-acting BZD, 3,473 (16%) were on long-acting BZD, and 6,185 (29%) were on Z-drugs. The median age was 82 years old, 81% were female, 94% were non-Hispanic White, 42% were eligible for the Medicare low-income subsidy, and the median length of hospital stay was 5 days. More than one third (39%) had more than five comorbidities and 50% had high predicted frailty probability. More than half (58%) co-prescribed antidepressants, 47% co-prescribed opioids, and 89% were classified as polypharmacy. Lorazepam (26%), Clonazepam (11%), and Zolpidem (27%) were the most prescribed short/intermediate-acting BZD, long-acting BZD, and Z-drugs at baseline, respectively. (Table 1)

Table 1.

Baseline characteristics of eligible patients between January 1, 2008, and June 30, 2019

Characteristics	n	%
Total	21,123	100%
Age, years, median (IQR)	82 (75–88)
66–74	4,692	22%
75–84	7,894	37%
≥85	8,537	41%
Sex
Male	4,075	19%
Female	17,048	81%
Race/ethnicity
Non-Hispanic White	19,779	94%
Non-Hispanic Black	471	2.2%
Hispanic/Latino	373	1.8%
Asian/Pacific Islander	236	1.1%
American Indian/Alaska Native or other races	200	0.9%
Unknown	64	0.3%
Medicare low-income subsidy	8,868	42%
Length of Hospital stay, days, median (IQR)	5 (4–6)
Gagne comorbidity index
(-2)-1	1,607	7.6%
2–3	5,019	24%
4–5	6,239	30%
>5	8,258	39%
Conditions*
Anxiety	12,378	59%
Insomnia	6,011	28%
Seizure	1,568	7.4%
Alcohol withdrawal	106	0.5%
Spasticity	9,257	44%
Depression	11,178	53%
Chronic Pain	11,297	53%
Dementia	9,618	46%
Co-medications†
Antidepressants	12,203	58%
Opioids	10,003	47%
Polypharmacy	18,898	89%
Predicted frailty index
Low probability of frailty <5%	1,001	4.7%
Low to medium 5-<10%	1,950	9.2%
Medium 10-<20%	3,106	15%
Medium to high 20-<40%	4,579	22%
high ≥40%	10,487	50%
Most recent prescription before hospital admission‡
Short/intermediate-acting BZD	11,465	54%
Lorazepam	5,527	26%
Alprazolam	4,920	23%
Temazepam	933	4.4%
Triazolam	35	0.2%
Oxazepam	37	0.2%
Others (Flurazepam/ Estazolam)	13	0.1%
Long-acting BZD	3,473	16%
Clonazepam	2,323	11%
Diazepam	991	4.7%
Clorazepate	98	0.5%
Chlordiazepoxide	61	0.3%
Z-drugs	6,185	29%
Zolpidem	5,682	27%
Eszopiclone	414	2.0%
Zaleplon	89	0.4%

^*Conditions were identified by diagnosis code ≤365 days before admission.

^†Co-medications were assessed ≤90 days before admission. Polypharmacy represents patients who had 5 or more drugs dispensing (based on different generic name, including BZD and Z-drugs) in the 90 days before admission.

^‡Based on the latest dispensed drug ≤90 days before admission. BZD cohorts included people from 2013 to 2019, while Z-drug cohort included people from 2007 to 2019.

IQR, interquartile range; BZD, benzodiazepines

Cumulative incidences of refilling the same/increased dose (continuation) and reduced dose (reduction) post discharge

Among the total population, the cumulative incidences of refilling the same/increased dose (referred to hereafter as ‘continuation’) and reduced dose (referred to hereafter as ‘reduction’) at 180 days post discharge was 55% and 14% for short/intermediate-acting BZD, 58% and 16% for long-acting BZD, and 56% and 7.2% for Z-drugs, respectively (Table 2). The slope of the continuation curve (hazard) was highest in the first 60 days and lower during the 60 to 180 days post discharge, while the slope of the reduction curve was highest in the first 120 days and declined until 180 days post discharge (Figure 1a). For patients on short/intermediate-acting BZD at baseline, those with anxiety had higher continuation (57%), while those with seizure had lower continuation (52%) compared to other conditions. For patients on long-acting BZD at baseline, those with anxiety had higher continuation (59%) and those with insomnia or seizure had higher reduction (20%). For patients on Z-drugs at baseline, those with anxiety had higher continuation (59%), while those with seizure had lower continuation (53%). (Table 2, Figure 1b–1e) Recent users and long-time users had higher continuation for all baseline BZDZ drug classes, while concurrent antidepressants and opioids users had higher reduction among patients on short/intermediate-acting BZD at baseline. (Table 2, Figure S3–S4)

Table 2.

Cumulative incidence of refilling BZDZ with same/increased dose or reduced dose at 180 days after discharge among total and different patient subgroups, stratified by baseline drug class

Population	N	Short/ intermediate-acting BZD, CIF (95% CI), %		Long-acting BZD, CIF, (95% CI), %		Z-drugs, CIF (95% CI), %
Refill dose		Same/increased	Reduced	Same/increased	Reduced	Same/increased	Reduced
Total	21,123	54.9 (54.0, 55.8)	13.8 (13.2, 14.4)	58.4 (56.7, 60.1)	16.3 (15.1, 17.5)	56.0 (54.8, 57.1)	7.2 (6.6, 7.8)
Conditions *
Insomnia	6,011	54.9 (52.9, 56.9)	15.9 (14.6, 17.3)	57.8 (54.5, 61.3)	20.0 (17.3, 23.2)	57.7 (55.8, 59.6)	7.9 (6.9, 9.1)
Anxiety	12,378	56.5 (55.4, 57.6)	14.9 (14.2, 15.6)	59.3 (57.3, 61.4)	17.8 (16.5, 19.2)	59.1 (57.2, 61.0)	8.2 (7.0, 9.6)
Seizure	1,568	51.6 (48.8, 54.5)	16.0 (13.7, 18.6)	57.6 (52.5, 63.2)	20.0 (16.2, 24.5)	53.0 (48.7, 57.6)	7.8 (5.4,11.2)
Spasticity	9,257	53.6 (52.3, 54.9)	14.9 (13.9, 16.0)	57.2 (55.0, 59.6)	17.0 (15.3, 18.9)	56.1 (54.1, 58.2)	7.4 (6.3, 8.7)
Subgroups
Recent user	14,323	62.0 (61.0, 63.0)	13.5 (12.7, 14.4)	64.6 (62.9, 66.4)	15.9 (14.5, 17.6)	60.8 (59.3, 62.3)	7.5 (6.8, 8.3)
Long-time user	17,190	59.1 (58.2, 60.1)	14.2 (13.7, 14.9)	62.0 (60.5, 63.6)	17.1 (15.7, 18.7)	60.6 (59.4, 61.8)	8.0 (7.2, 8.9)
Concurrent user
Antidepressant	12,203	54.3 (53.2, 55.5)	15.6 (14.8, 16.5)	59.4 (57.3, 61.5)	17.8 (16.2, 19.5)	56.4 (54.8, 58.0)	7.7 (6.9, 8.7)
Opioids	10,003	57.1 (55.8, 58.5)	15.0 (14.1, 16.1)	58.3 (56.0, 60.7)	17.7 (15.9, 19.8)	57.5 (56.0, 59.0)	7.4 (6.6, 8.3)
Non-pathological fracture user	20,139	54.9 (54.1, 55.8)	13.8 (13.2, 14.5)	58.6 (57.2, 60.0)	16.4 (15.2, 17.7)	55.9 (54.7, 57.1)	7.2 (6.6, 7.9)

CIF, outcome-specific cumulative incidence of refilling was calculated from Aalan Johanssen estimator based on the first refill after discharge, treating death as a competing event. The first refill was stratified by refilling the same/increased dose or a reduced dose. If patients refilled a drug different from the baseline drug, dose change was compared using daily diazepam milligram (DDM) equivalent (mg/day, DDM = ([dose conversion factor] × [the medication strength] × [the prescription quantity]) ÷ [number of days supply].)

BZD cohorts included people from 2013 to 2019, while Z-drug cohort included people from 2007 to 2019.

^*Conditions were identified by diagnosis code ≤365 days before admission

BZDZ, benzodiazepines and z-drugs; CI, confidence interval; BZD, benzodiazepines

Figure 1. Cumulative incidence* of refilling BZDZ with same/increased dose or reduced dose within 180 days after discharge among all patients and patients restricted to different conditions, stratified by baseline drug class.

* Outcome-specific cumulative incidence of refilling was calculated from Aalan Johanssen estimator based on the first refill after discharge, treating death as a competing event. The first refill was stratified by refilling the same/increased dose or a reduced dose. If patients refilled a drug different from the baseline drug, dose change was compared using daily diazepam milligram (DDM) equivalent (mg/day, DDM = ([dose conversion factor] × [the medication strength] × [the prescription quantity]) ÷ [number of days supply].)

BZD cohorts included people from 2013 to 2019, while Z-drug cohort included people from 2007 to 2019.

BZDZ, benzodiazepines and z-drugs; BZD, benzodiazepines

Factors associated with continuation

Overall, patients aged 66–74 years, female, those with anxiety, or those who did not have dementia were at higher risk of continuing (Table 3, Figure S5–1a to S5–3c). Patients who were on opioids and short/intermediate-acting BZD at baseline (CIR=1.05 [1.00, 1.11]) or those identified as polypharmacy and on Z-drug (CIR=1.17 [1.03, 1.33]) at baseline were more likely to continue. Patients with lower Gagne comorbidity score (<2) and on short/intermediate-acting BZD or patients with lower frailty (<10%) and on Z-drug at baseline were more likely to continue, while no difference was detected across different levels of comorbidity score and frailty among patients who were on long-acting BZD. (Table 3)

Table 3.

Patient characteristics associated with cumulative incidence of refilling BZDZ with same/increased dose versus deprescribing at 180 days after discharge, stratified by baseline drug class

Characteristics	Short/intermediate- acting BZD CIR (95%CI)	Long-acting BZD CIR (95%CI)	Z-drug CIR (95%CI)
Age
66–74	1	1	1
75–84	0.97 (0.90, 1.04)	0.96 (0.86, 1.07)	0.88 (0.81, 0.96)
≥85	0.92 (0.85, 0.99)	0.97 (0.85, 1.10)	0.83 (0.75, 0.92)
Sex
Male	1	1	1
Female	1.07 (1.00, 1.14)	1.01 (0.90, 1.14)	1.12 (1.03, 1.22)
Race/ethnicity
Non-Hispanic White	1	1	1
Non-Hispanic Black	0.95 (0.79, 1.14)	0.87 (0.61, 1.23)	0.98 (0.79, 1.22)
Hispanic/Latino	1.05 (0.86, 1.28)	0.88 (0.59, 1.31)	1.01 (0.81, 1.27)
Asian/Pacific Islander	1.10 (0.84, 1.45)	0.67 (0.32, 1.41)	0.94 (0.73, 1.22)
American Indian/Alaska Native or Other races	0.97 (0.73, 1.28)	0.84 (0.56, 1.26)	1.13 (0.81, 1.58)
Medicare low-income subsidy	1.02 (0.97, 1.08)	1.07 (0.97, 1.17)	0.98 (0.91, 1.05)
Gagne comorbidity index
(−2)-1	1	1	1
2–3	0.95 (0.86, 1.05)	1.02 (0.85, 1.22)	0.96 (0.84, 1.08)
4–5	0.87 (0.78, 0.96)	0.99 (0.83, 1.19)	0.95 (0.83, 1.07)
>5	0.82 (0.74, 0.91)	0.94 (0.78, 1.13)	0.89 (0.78, 1.02)
Anxiety	1.10 (1.04, 1.17)	1.05 (0.95, 1.17)	1.09 (1.01, 1.18)
Insomnia	1.00 (0.94, 1.06)	1.00 (0.90, 1.11)	1.04 (0.97, 1.11)
Seizure	1.00 (0.91, 1.10)	1.00 (0.86, 1.17)	0.98 (0.85, 1.14)
Spasticity	0.99 (0.94, 1.05)	0.98 (0.89, 1.07)	1.00 (0.93, 1.08)
Depression	1.05 (0.99, 1.12)	1.05 (0.94, 1.16)	1.04 (0.96, 1.12)
Chronic Pain	0.97 (0.92, 1.02)	0.99 (0.90, 1.09)	0.97 (0.90, 1.04)
Dementia	0.82 (0.77, 0.87)	0.87 (0.78, 0.97)	0.92 (0.84, 1.00)
Antidepressants	0.99 (0.93, 1.05)	1.04 (0.93, 1.15)	0.99 (0.92, 1.07)
Opioids	1.05 (1.00, 1.11)	0.97 (0.89, 1.07)	1.03 (0.96, 1.11)
Polypharmacy	1.02 (0.94, 1.11)	1.03 (0.89, 1.20)	1.17 (1.03, 1.33)
Predicted frailty index
low probability of frailty <5%	1	1	1
low to medium 5-<10%	1.04 (0.91, 1.20)	1.05 (0.83, 1.33)	0.88 (0.76, 1.02)
medium 10-<20%	0.98 (0.85, 1.12)	1.00 (0.80, 1.26)	0.85 (0.74, 0.99)
medium to high 20-<40%	0.95 (0.83, 1.08)	1.03 (0.82, 1.30)	0.79 (0.68, 0.91)
high ≥40%	0.89 (0.78, 1.03)	0.97 (0.77, 1.23)	0.74 (0.63, 0.86)

CIR, Cumulative incidence ratio was calculated from multivariate model that included each characteristic listed here using proc genmod with link=log and distribution=Poisson in SAS, where the outcome was the cumulative incidence of refilling a same/increased dose at 180 days after discharge (continue) and the reference was deprescribing=1-CIF (continue). Deprescribing included those who refilled a reduced dose and those who did not refill. Some of those who did not refill were people who dead (n=3,942) or lost Medicare part D coverage(n=241) during the 180 days after discharge, and they were treated as competing events in the model.

If patients refilled a drug different from the baseline drug, dose comparison was calculated using daily diazepam milligram (DDM) equivalent (mg/day, DDM = ([dose conversion factor] × [the medication strength] × [the prescription quantity]) ÷ [number of days supply].) BZD cohorts included people from 2013 to 2019, while Z-drug cohort included people from 2007 to 2019.

BZDZ, benzodiazepines and z-drugs; BZD, benzodiazepines; CI, confidence interval; CIF, cumulative incidence function

Overall refilling patterns and trends after 90 days post discharge across 2008–2019

Table S4 shows refilling patterns after 90 days post hospital discharge among patients who were alive and had continuous Medicare coverage (n=18,458). Among patients using short/intermediate-acting BZD at baseline, 49% refilled the same/increased dose, 15% reduced dose, and 36% had no refill (with 5.7% switching to an alternative). Among patients using long-acting BZD at baseline, 53% refilled the same/increased dose, 17% reduced dose, and 30% had no refill (with 3.8% switching to an alternative). Among patients using Z-drugs at baseline, 50% refilled the same/increased dose, 7.2% reduced dose, and 38% had no refill (with 7.0% prescribed an alternative). The proportion of individuals who refilled the same/increased dose increased from 2008–2019 for Z-drugs (p<0.001), while the proportion of individuals who refilled the same/increased dose remained the same for BZD from 2013–2019 (p=0.154, p=0.899). (Figure S6).

DISCUSSION

In this cohort of fee-for-service Medicare beneficiaries in the US who were hospitalized for a hip fracture between 2008–2019 with a prior dispensed BZDZ drug and discharged home or to a skilled nursing facility, we found a high BZDZ drug continuation and a few dose reductions after discharge. More than half of eligible patients on a BZDZ drug continued, and ~15% on BZD and 7.2% on Z-drugs reduced dose at 180 days post discharge. Patients who were younger or who had anxiety were at higher risk of continuing. After 90 days post discharge, nearly half of eligible patients either reduced dose or stopped the BZDZ with or without switching to an alternative (e.g., Doxepin, Mirtazapine listed in Supplementary Table S3). BZD were more likely to be deprescribed through dose reduction, while Z-drugs were more likely to be deprescribed by stopping the drug.

Our findings suggest that deprescribing BZDZ drugs during and immediately following hospitalization for a hip fracture remains uncommon and is similar across people with different conditions (insomnia, anxiety, seizure, and spasticity) identified by the diagnosis code. These may be missed opportunities to reduce the use of potentially inappropriate medications (PIMs) among vulnerable older adults. A scoping review summarizing interventions to decrease BZDZ drug use in hospitals found the combination of patient education, sleep protocols, and system changes such as electronic health record alerts could be effective in reducing BZD use.⁴⁴ In addition to reducing the use of PIMs, an international review highlighted the importance of a clear deprescribing protocol specifying the tapering and switching strategies for older adults to avoid unnecessary harm from withdrawal syndrome.¹⁷

Several studies have reported similar results among Medicare beneficiaries. Munson et al. described utilization patterns for drugs associated with fracture risk before and after a fragility fracture and found that only 21% of older adults prescribed hypnotics prior to the fracture discontinued hypnotic use during the 4 months after the fracture.¹¹ Pham Nguyen et al. assessed central nervous system-active PIMs among older adults with Parkinson’s disease before and after hospitalization for injury and found that 40% of participants discontinued at 90 days post discharge and 62% discontinued throughout one year.²³

We estimated the BZDZ drugs continuation and dose reduction after hospitalization for a hip fracture, accounting for the impact of death as a competing event. Mortality is high post discharge for hip fractures compared with other hospitalizations, and not properly accounting for mortality can lead to overestimation of the proportion of patients who refill BZDZ drugs.^1,45 Another strength of our study was our categorization of refilling patterns into four different types and the use of daily diazepam milligram equivalent for dose comparison, which provides a comprehensive picture of medication continuation, dose reduction, switching, and discontinuation and allowed us to assess trends in refilling patterns over time. Given our study sample, our results are most generalizable to the fee-for-service Medicare population, which represents approximately half of all Medicare beneficiaries and predominantly consists of non-Hispanic White women with multiple comorbidities and high frailty.^46,47

Our study has several limitations. We relied on insurance claims to identify BZDZ refills. We assumed those who refilled would use the medications and those who did not refill stopped medication use. Both assumptions are plausible, but some misclassification of the outcome is expected with unclear overall direction of bias. Skilled nursing facility stays following inpatient visits may preclude patients from refilling at a pharmacy, so we may not observe any refilling right after discharge. However, the median number of days after which drug refills could be observed following discharge was 24 (Table S5), so we would likely see prescriptions dispensed after this time. We chose the first refill post discharge to assess refill patterns, which could miss any medication discontinuation, dose change, or switch thereafter. Since we did not assess the refill pattern of other drugs that could increase the risk of falls in older adults (e.g., anticholinergics or other central nervous system (CNS)-active medications), it is possible that for those in whom we did not see a change in their refill patterns, there were other CNS-active medication modifications made to their medication regimens to reduce fall risk (e.g., some patients who were on anticholinergics or opioids at the same time with BZDZ drug before admission may have discontinued anticholinergics or opioids but continued BZDZ drug after discharge). We used any diagnosis code observed within 365 days prior to admission to identify conditions that may indicate for BZDZ drug, which may not represent the indication at the time the baseline BZDZ drug was dispensed. We treated patients who lost Medicare Part D coverage after discharge as a competing event when assessing factors associated with continuation versus deprescribing, but this should bias our results minimally because the proportion of these patients was very low (<1%). Lastly, recent deprescribing practices may not be reflected in our results, since our study period ended in 2019, a year before the US Food and Drug Administration (FDA) posted a boxed warning update to improve the safe use of BZD.⁴⁸

In conclusion, we found that more than half of US older adults on BZDZ drug before hospitalization for a hip fracture continued the drug and a small proportion reduced the dose within 180 days after hospital discharge, and the proportions did not vary by patients with different conditions. Providers could leverage the opportunities during patients’ hospitalization to prioritize comprehensive medication reviews and adhere to deprescribing guidelines if the patient no longer has a clear indication for BZDZ drug use.¹⁷ Providers could optimize care for older adults hospitalized for hip fractures by developing and implementing targeted strategies, such as deprescribing initiatives and structured medication management protocols.^21,49,50 These approaches may allow careful BZDZ drug use when clinically necessary, and reduce medication-induced recurrent falls.

KEY POINTS BOX.

Key Points

• 2023 Beers criteria recommend avoiding the use of benzodiazepines (BZD) and nonbenzodiazepine receptor agonist hypnotics (Z-drugs) among older adults with a history of falls or fracture due to risk-benefit imbalance.

• Among 21,123 fee-for-service Medicare beneficiaries who were on BZD (2013–2019) or Z-drugs (2008–2019) before hospitalization for a hip fracture, more than half refilled the same/increased dose, and 14% on short/intermediate-acting BZD, 16% on long-acting BZD, and 7% on Z-drugs refilled with a reduced dose within 180 days post hospital discharge.

• People who were younger or had anxiety were at higher risk of refilling the BZD and Z-drugs with same/increased dose after discharge, and deprescribing interventions could be targeted to these groups to reduce inappropriate medications.

Why does this paper matter?

This paper provides comprehensive analyses of the refilling patterns among prior benzodiazepines or Z-drugs users after hospitalization for a hip fracture. The results suggest missed opportunities for comprehensive medication review and deprescribing potential inappropriate medications during or after discharge for patients hospitalized for a hip fracture.