Association Between Medications Acting on the Central Nervous System and Fall-Related Injuries in Community-Dwelling Older Adults: A New User Cohort Study

Abstract

Background

It is well established that individual medications that affect the central nervous system (CNS) increase falls risk in older adults. However, less is known about risks associated with taking multiple CNS-active medications.

Methods

Employing a new user design, we used data from the Adult Changes in Thought study, a prospective cohort of community-dwelling people aged 65 and older without dementia. We created a time-varying composite measure of CNS-active medication exposure from electronic pharmacy fill data and categorized into mutually exclusive categories: current (within prior 30 days), recent (31–90 days), past (91–365 days), or nonuse (no exposure in prior year). We calculated standardized daily dose and identified new initiation. Cox proportional hazards models examined the associations between exposures and the outcome of fall-related injury identified from health plan electronic databases.

Results

Two thousand five hundred ninety-five people had 624 fall-related injuries over 15,531 person-years of follow-up. Relative to nonuse, fall-related injury risk was significantly greater for current use of CNS-active medication (hazard ratio [HR] = 1.95; 95% CI = 1.57–2.42), but not for recent or past use. Among current users, increased risk was noted with all doses. Risk was increased for new initiation compared with no current use (HR = 2.81; 95% CI = 2.09–3.78). Post hoc analyses revealed that risk was especially elevated with new initiation of opioids.

Conclusions

We found that current use, especially new initiation, of CNS-active medications was associated with fall-related injury in community-dwelling older adults. Increased risk was noted with all dose categories. Risk was particularly increased with new initiation of opioids.

Falls and fall-related injuries constitute a critical, and growing, public health problem (1). One in three community-dwelling adults aged 65 and older, and one in two aged 80 and older, sustains a fall each year (2,3). About half of all falls result in an injury (4). Treatment of injurious falls is expensive, costing approximately $50 billion annually (5). Furthermore, the rate of deaths from falls among U.S. adults aged 65 and older increased 31% between 2007 and 2016 (6). Fall prevention efforts are vital to reduce the associated morbidity and mortality.

Multiple risk factors contribute to falls, many of which are modifiable (7,8). Medications affecting the central nervous system (CNS) may increase falls risk because they may cause sedation, dizziness, and cognitive impairment. Older adults often take multiple CNS-active medications, with 25% taking two or more different classes (9). Many observational studies have evaluated the use of a single class of CNS-active medications (eg, anticholinergics, antidepressants, benzodiazepines, opioids, and sedative-hypnotics) and have reported an increased risk of falls and fall-related injuries in older adults using these individual classes (10–14). Furthermore, risk for fall-related injuries associated with a single class of CNS-active medications appears greatest when the medication is first initiated (10,12,14–16). Given the prevalence of CNS polypharmacy in older adults, it is more clinically relevant to focus on falls risk associated with CNS-active medication exposure, taking into account dose (17–21), rather than on single medication classes. The two primary methods for summarizing composite dosage intensity measures for CNS medication exposure are the Drug Burden Index (DBI) and a measure that calculates the standardized daily dose (SDD) for medications (22). Both measures have been associated with increased risk for falls in older adults (18–21,23). For example, the DBI has been associated with risk for falls or fall-related injury in people living in the community (20,21,23) or residential aged care facilities (19) as well as during hospitalization (24).

An important gap that exists in studies conducted in community samples is that overall CNS-active medication exposure is often not measured at the time of the fall. Thus, it is not entirely clear that the person was exposed to the medication at the time of the fall (9,18,20,21). Furthermore, most studies focused on prevalent users of CNS-active medications, and thus risk associated with new initiation was not examined. To the best of our knowledge, only one study of community-dwelling older adults examined a composite measure of CNS-active medication exposure, incorporating dose, immediately preceding the fall, however this study did not examine new initiation of medications (17). To this end, we examined the risk of fall-related injury associated with a composite measure of CNS-active medication exposure that incorporates dose, as well as timing of use relative to the outcome, in community-dwelling older adults.

Method

Design, Study Setting, and Participants

We employed a new user design using data from the Adult Changes in Thought (ACT) study, a population-based prospective cohort study conducted within Kaiser Permanente Washington (KPWA), an integrated health-care delivery system in the northwest United States. Study methods have been described in detail elsewhere (25). Participants aged 65 years and older without dementia were randomly sampled from Seattle-area members. Participants were enrolled during three waves: the original cohort between 1994 and 1996 (n = 2,581), the expansion cohort between 2000 and 2003 (n = 811), and continuous enrollment beginning in 2004. Participants were assessed at study entry and at 2-year intervals to evaluate cognitive function and collect information about demographic characteristics, medical history, health behaviors, and functional measures. ACT biennial follow-up visits with collection of these measures ended once a participant developed dementia, thus participants were censored at this time in analyses. As of April 30, 2014, ACT had enrolled 4,682 people across the three waves who met our current study’s initial eligibility criteria of having at least 1 year of health plan membership before start of follow-up and no evidence of a fall-related injury in that year. From this sample, we performed the primary analysis focused on overall CNS-active medication exposure, excluding those participants with evidence of CNS-active medication use in the year before study entry (n = 2,043) and resulting in an eligible sample of 2,639, of which 2,595 had complete covariates. The protocol was approved by the KPWA institutional review board; participants provided written informed consent.

Assessment of Fall-Related Injuries

The primary outcome was fall-related injury from health plan electronic databases as defined by Tinetti and colleagues (26) and modified to include outpatient events. We expect to have complete capture of health care utilization (ie, medically attended falls) as KPWA provides insurance coverage and health care to members and receives claims for outside care. We defined fall-related injury based on emergency department, inpatient, and outpatient claims for visits with the International Classification of Diseases, Ninth Revision (ICD-9) injury codes for nonpathological skull, facial, cervical, clavicle, humeral, forearm, pelvic, hip, fibula, tibia, or ankle fractures (800.00–806.19, 807.0–807.2, 808.0–808.9, 810.00–814.19, 818.0–825.1, or 827.0–829.1), brain injury (852.00–852.39), or dislocation of the hip, knee, shoulder, or jaw (830.0–832.19, 835.00–835.13, or 836.30–836.60) provided that there was no E code for motor vehicle accidents. We also included inpatient and emergency department visits that were accompanied by a fall-related E code (8800–8889) even when one of the listed ICD-9 injury codes was not present.

CNS-Active Medication Exposure

We ascertained CNS-active medication exposure from computerized pharmacy fill data that included drug name, strength, route of administration, date dispensed, and amount dispensed. We included the following CNS-active medication classes: benzodiazepine receptor agonists (benzodiazepines and Z-drugs) and other sedatives, antidepressants, opioids, antipsychotics, skeletal muscle relaxants, and anticholinergics. A complete list of included CNS-active medications is provided in Supplementary Table S1. The basis for combining these medication classes comes from prior research suggesting similar risk profiles (10–14).

At start of follow-up, all included participants were required to be “nonusers” of CNS-active medications. We then constructed a time-varying exposure measure for each participant for each day during study follow-up based on any prescription fills for CNS-active medications during that time, taking into account adherence and early refills (Supplementary Methods) (27). The exposure period of primary interest was current use defined as an episode of medication use that overlapped with the 30-day window before the fall (index date). We examined the following three exposure contrasts (Supplementary Methods):

• 1) Current, recent, past compared with nonuse: we categorized our time-varying CNS-active medication exposure measure into mutually exclusive categories of current, recent (run-out date occurred within 31–90 days of index), past (run out date within 91–365 days of index), or nonuse (no exposure within prior year).

• 2) Dose–response (among current users) compared with nonuse: we created a time-varying measure of standardized daily dose (SDD) based on previous methods (18,22,28) using the minimum effective geriatric daily dose (29). The SDD approach of calculating a composite measure of CNS exposure has been associated with a dose–response relationship with falls (18,30), although there is no evidence that it is a better predictor of falls compared with the DBI. Current users were categorized as <1 SDD (low), 1 to <2 SDD (medium), or ≥2 SDD (high).

• 3) Type of current use (new initiation or prevalent) compared with no current use.

Covariates

Based on a literature review, we selected covariates that may be potential confounders of the association between CNS-active medication use and falls, including demographic and health-related characteristics (7,18,20). Information about covariates came from standardized questionnaires administered at each ACT study visit and from health plan electronic databases. Demographic characteristics included age at study entry, sex, and years of education (categorized as beyond high school or not). Body mass index (BMI) was calculated from measured weight in kilograms divided by height in meters squared. Participants were asked about smoking and self-rated health (fair/poor vs good/very good/excellent). We ascertained presence of several comorbidities, including medication-treated hypertension and diabetes mellitus (computerized pharmacy data), urinary incontinence (diagnosis codes), and self-reported Parkinson’s disease, osteoarthritis, stroke, heart failure, and coronary heart disease (eg, prior myocardial infarction, coronary artery bypass graft, angioplasty, or angina). Depressive symptoms were assessed using the short version of the Center for Epidemiologic Studies Depression scale. We created a binary variable for depressive symptoms (≤10 points vs > 10 points), considering a score >10 consistent with high depressive symptoms (31). We defined slow gait speed as completing a 10 foot timed walk at a speed of <0.6 m/s (32). Participants who were referred for diagnostic cognitive evaluation (eg, due to scoring <86 on the Cognitive Abilities Screening Instrument [CASI]) but were determined not to meet the criteria for a dementia diagnosis were defined as having impaired cognition (33). Poor vision was defined based on self-reported eyesight interview questions or inability to complete CASI test due to poor eyesight.

Statistical Analyses

We used multivariable Cox regression models, with time on study as the time scale, to estimate hazard ratios (HRs) and 95% confidence intervals (CI) for the association between CNS-active medication use and fall-related injury. We estimated separate models examining our three exposure contrasts of interest: (i) current, recent, and past (reference group was nonuse); (ii) dose–response relationship among current users (low, medium, and high dose; reference group was no current use); and 3) new-initiation among current users (new initiation, prevalent use; reference group was no current use). For all analyses, participants were followed until the earliest of fall-related injury or a censoring event (ie, health plan or ACT disenrollment, dementia onset, death, or end of the study follow-up period). We calculated HR and 95% CI estimates from a minimally adjusted model that only included age (natural cubic splines with six knots) and sex, as well as a primary model that included additional adjustment for ACT enrollment wave (original, expansion, continuous enrollment), BMI, self-rated health, hypertension, diabetes, stroke, urinary incontinence, osteoarthritis, coronary artery disease, congestive heart failure, Parkinson’s disease, slow gait speed, depression, impaired cognition (ie, not dementia), and poor vision. All covariates, other than sex and cohort, were treated as time-varying. We performed a complete-case analysis, excluding observation time when participants were missing covariate information from all model estimates (excluded 44 participants and a total of 531 person-years). We assessed proportional hazards using Schoenfeld residuals (34).

Interactions and Sensitivity Analyses

We assessed effect modification by using interaction terms to estimate separate HRs for CNS-active medication use according to sex, gait speed (binary), and age. We also performed prespecified sensitivity analyses to assess robustness of our primary findings. We considered models that did not adjust for cognition and gait speed as these variables could be in the causal pathway. We also examined only fall-related injuries that resulted in a visit to the emergency department or an inpatient stay.

Post hoc Analyses

We performed additional analyses to better understand the contributions of the individual medication classes that comprised our overall CNS-active medication exposure and how these could have impacted overall dose–response results. We summarized the prevalence of exposure and the distributions of dose categories across individual medication classes and performed post hoc analyses estimating the association between use of specific classes and fall-related injury risk (Supplementary Methods). We created new user cohorts for each specific medication class to maximize sample size and enhance external validity. Based on these investigations, particularly with regard to the strong association observed with opioids (presented later), we repeated our primary dose (SDD) analysis setting the minimum effective geriatric daily dose of morphine to 10 mg (rather than 30 mg) to align with a recently published study (30) and to assess the impact on our dose findings. The literature is less clear on the minimum effective geriatric daily dose for opioids.

All analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC) and R, version 3.3 (R Foundation for Statistical Computing, Vienna, Austria).

Results

Characteristics of participants at study entry are given in Table 1. The median age of participants was 73 years, 54% were female, and 70% had some education beyond high school. The most common medication classes were antidepressants, opioids, bladder anticholinergics, and benzodiazepines and other sedative hypnotics (Supplementary Table S2). Among 2,595 participants included in the complete-case analyses, there were 624 fall-related injuries over 15,531 total person-years of follow-up. A description of falls according to method of ascertainment and visit encounter type (ED, hospital, or outpatient) is provided in Supplementary Table S3. The most common serious injuries included 203 lower limb fractures (including 80 hip fractures), 146 upper limb fractures, 83 major joint dislocations, 72 rib fractures, and 33 pelvic fractures.

Table 1.

Characteristics at Study Entry (n = 2,639)

Characteristic	N	%
Age, median (IQR)	73 (70, 79)
Female	1,436	54.4
Race, nonwhite	282	10.7
Some education beyond high school	1,834	69.5
Body mass index, median (IQR)	27 (24, 30)
Current smoker	131	5.0
Fair or poor self-rated health	305	11.6
Health conditions
Treated hypertension*	1,255	47.6
Treated diabetes†	198	7.5
Stroke‡	65	2.5
Urinary incontinence§	75	2.8
Osteoarthritis‡	936	35.5
Coronary artery disease||	425	16.1
Congestive heart failure‡	79	3.0
Parkinson’s disease‡	11	0.4
Slow gait speed (<0.6 m/s)	163	6.2
Depression (CES-D score ≥ 10)	165	6.4
Impaired cognition¶	137	5.2
Poor vision	355	13.4

Note: Missing information includes: race n = 4; body mass index n = 20; self-rated health n = 6; gait speed n = 15; depression n = 50; smoking n = 5. IQR = interquartile range; CES-D = Center for Epidemiologic Studies Depression (CES-D).

*Two or more fills in computerized pharmacy data for antihypertensive medications within a 12-month period in the 3 years before ACT study entry.

^†Two or more fills in computerized pharmacy data for insulin or oral diabetic medications within a 12-month period in the 3 years before ACT study entry.

^‡Self-report.

^§Codes 788.30–788.39 from the International Classification of Diseases, Ninth Revision in 3 years before ACT study entry.

^||Self-reported history of heart attack, angina, angioplasty, or coronary artery bypass surgery.

^¶Score on the Cognitive Abilities Screening Instrument (CASI) of <86 or if the CASI was missing or invalid but the participant was referred for full diagnostic evaluation.

In multivariable analyses, current use of any CNS-active medication was associated with a greater risk of fall-related injuries (HR = 1.95; 95% CI = 1.57–2.42) compared with nonuse (Table 2). Risk was not increased for recent (HR = 1.07; 95% CI = 0.73–1.57) or past use (HR = 1.15; 95% CI = 0.91–1.45). Among current users, elevated risk was observed for all dose levels, although a dose–response relationship was not observed (test for trend; p = .610). The adjusted HR was 1.73 (95% CI = 1.20–2.48) for use of low dose, 2.60 (95% CI = 1.93–3.51) for medium dose, and 1.56 (95% CI = 1.10–2.22) for high dose, compared with no current use.

Table 2.

CNS-Active Medication Use and Risk of Fall-Related Injury

CNS Medication Use	Primary Analysis					Sensitivity Analysis†,‡
	Person-Years	Fall-Related Injury	Fall-Related Injury Rate (per 100 person-years)	Adjusted Model* ,†
				HR	95% CI	HR	95% CI
None	10,986	386	3.5	1.00	Reference	1.00	Reference
Past	2,162	88	4.1	1.15	(0.91–1.45)	1.19	(0.87–1.64)
Recent	754	29	3.8	1.07	(0.73–1.57)	1.08	(0.65–1.81)
Current use	1,629	121	7.4	1.95	(1.57–2.42)	1.86	(1.38–2.50)
According to SDD
<1	476	33	6.9	1.73	(1.20–2.48)	1.17	(0.66–2.07)
1–<2	531	52	9.8	2.60	(1.93–3.51)	2.72	(1.83–4.04)
≥2	622	36	5.8	1.56	(1.10–2.22)	1.72	(1.10–2.67)

Note: SDD = standardized daily dose; HR = hazard ratio; CI = confidence interval.

*Results for primary outcome of fall-related injury resulting in ED or outpatient visit or hospitalization.

^†Adjusted for age, sex, ACT study cohort, and time varying measures of body mass index, self-rated health, hypertension, diabetes, stroke, urinary incontinence, osteoarthritis, coronary artery disease, congestive heart failure, Parkinson’s disease, gait speed, depressive symptoms, impaired cognition, and poor vision.

^‡Adjusted HR for the outcome of fall-related injury resulting in ED visit or hospitalization.

Among current users, risk of fall-related injury was highest for those who had newly initiated a CNS-active medication. Relative to no current use, new initiation was associated with an increased risk for fall-related injury (HR = 2.81; 95% CI = 2.09–3.78) as was prevalent use (HR = 1.52; 95% CI = 1.18–1.97; Table 3). Comparing new initiation to prevalent use yielded a HR of 1.85 (95% CI = 1.28–2.67). Among people with new initiation, increased risk was observed for all dose levels, however high dose was not statistically significant. A dose–response relationship was not observed (p = .655).

Table 3.

Initiation of CNS-Active Medication and Risk of Fall-Related Injury

CNS Medication Use	Primary Analysis					Sensitivity Analysis†,‡
	Person-Years	Fall-Related Injury	Fall-Related Injury Rate (per 100 person-years)	Adjusted* ,†
				HR	95% CI	HR	95% CI
No current use	13,902	503	3.6	1.00	Reference	1.00	Reference
Prevalent use	1,160	72	6.2	1.52	(1.18–1.97)	1.46	(1.04–2.05)
New initiation	468	49	10.5	2.81	(2.09–3.78)	2.68	(1.78–4.04)
According to SDD
<1	184	17	9.2	2.47	(1.52–4.01)	2.15	(1.06–4.36)
1–<2	160	24	15.0	4.04	(2.67–6.10)	3.80	(2.16–6.68)
≥2	125	8	6.4	1.75	(0.87–3.52)	1.96	(0.80–4.77)

Note: SDD = standardized daily dose; HR = hazard ratio; CI = confidence interval.

*Results for primary outcome of fall-related injury resulting in ED or outpatient visit or hospitalization.

^†Adjusted for age, sex, ACT study cohort, and time varying measures of: body mass index, self-rated health, hypertension, diabetes, stroke, urinary incontinence, osteoarthritis, coronary artery disease, congestive heart failure, Parkinson’s disease, gait speed, depressive symptoms, impaired cognition, and poor vision.

^‡Adjusted HR for the outcome of fall-related injury resulting in ED visit or hospitalization.

Interactions and Sensitivity Analyses

We did not detect interactions with sex (p for interaction = .064), gait speed (p for interaction = .413), and age (p for interaction = .314), when examining current, recent, or past use compared with no use. Estimates were appreciably unchanged in models that did not adjust for cognition and gait speed (data not shown). We obtained a mostly similar pattern of results when we examined only fall-related injuries that resulted in a visit to the emergency department or hospital (Tables 2 and 3).

Post hoc Analyses

Relative to no current use, new initiation of opioids (HR = 3.25; 95% CI = 2.44–4.31) was associated with an increased risk for fall-related injury as was prevalent current use (HR = 1.79; 95% CI = 1.32–2.42). An elevated risk was noted for new initiation of benzodiazepines (HR = 1.68; 95% CI = 0.90–3.13) and anticholinergics (HR = 1.26; 95% CI = 0.77–2.04), but the findings were not statistically significant. New initiation of antidepressants (HR = 1.12; 95% CI = 0.53–2.37) was not associated with fall-related injury (Supplementary Table S4).

We found differences in the distribution of dose (low, medium, high) among the different classes of CNS-active medications (data not presented). For example, the distribution of current users of opioids was shifted toward the low dose category (<1 SDD) while for antidepressants more than half of use was in the high dose category (>2 SDD). When repeating our primary dose (SDD) analysis setting the minimum effective dose of morphine to 10 mg (rather than 30 mg) to align with a recently published study (30), we found a dose–response relationship for current users compared with nonuse (test for trend; p = .017): low dose (HR = 1.22, 95% CI = 0.73–2.06); medium dose (HR = 1.78, 95% CI = 1.23–2.59); and high dose (HR = 2.32; 95% CI = 1.80–3.01).

Discussion

In this longitudinal cohort study of community-dwelling older adults, we found that people with current CNS-active medication exposure had a greater risk for fall-related injury compared with no exposure. The risk was elevated with all dose levels; we did not find a dose–response relationship. Furthermore, we found that risk was greatest for people with new initiation of a CNS-active medication. Our findings were robust to sensitivity analyses. Post hoc analyses revealed that risk was particularly elevated with new initiation of opioids.

Although prior studies have examined risk of falls associated with composite measures of CNS-active medications (17–21), our study is novel partly because we examined CNS-active medication exposure around the time of the fall-related injury in community-dwelling older adults and also took into account whether the medication was newly initiated. Although no published studies are directly comparable, the most similar study to ours is a retrospective cohort study that examined the association between a time-varying measure of psychoactive medication burden, based on the defined daily dose, and hospitalization for falls among 73,690 Australian veterans 65 years and older (17). The authors reported that increased number and doses of psychoactive medications were associated with an increased risk of hospitalization for falls. For example, the highest risk was found in those taking three or more defined daily doses per day (adjusted IRR = 4.26, 95% CI = 2.75–6.58) compared with no use. The list of psychoactive medications in this study and ours was similar, although Pratt and colleagues included drugs for seizures and migraines, whereas we did not. Pratt and colleagues did not assess fall risk associated with new initiation of CNS-active medications. Although not assessing exposure at the time of the fall as was done in the study by Pratt and colleagues another well-executed population-based study conducted in New Zealand (n = 70,553) used a time-varying measure of the DBI, updated in 90-day intervals, and reported a dose–response relationship with hip fracture (23). The estimated subhazard ratio was 1.52 (1.28–1.81) for those with DBI > 3 compared with those with DBI = 0 (23).

It is unclear why we did not observe a dose–response relationship. The studies that have examined CNS exposure around the time of a fall-related injury have used either the defined daily dose (17) or the SDD (30,35) methodology to summarize overall CNS-medication exposure. Two of these studies found a dose–response relationship (17,30) but the other did not (35). Studies that have used the DBI have found a dose–response relationship when using a time-varying exposure measure (updated in 90-day intervals) (23), but mixed findings were reported when the DBI was measured up to a year before the outcome (19,21). While the explanation for our finding is unclear, our post hoc analyses offer some insights into why we did not observe such a relationship. For example, this finding may be an artifact of the method used to compute the SDD and combining different medication classes. The SDD takes into account the minimum effective geriatric daily dose of each medication, which is not always clearly delineated in published literature as is the case with opioids. Using a lower minimum effective geriatric daily dose to calculate the SDD for opioids shifted the distribution of SDD values to higher exposure categories, and resulted in finding an overall CNS-active medication dose–response relationship. Although this does not detract from the main findings, it does suggest composite measures of CNS exposure (22) that take into consideration dose may be sensitive to the dose used to standardize across medications.

We found that new initiation of opioids increased the hazard for fall-related injury by 325% compared with no current opioid use, which confirms findings from other observational studies (15,16). Although an increased risk was observed for new initiation of benzodiazepines, this finding was not statistically significant given the wide confidence intervals in part because of the relatively low prevalence of benzodiazepine use. We did not find that new initiation of antidepressants was related to fall-related injury. Increased risk for falls has been previously reported for initiation of antidepressants (13) and benzodiazepines in older adults (13,14). No studies have specifically examined the risk of new initiation of anticholinergics, but in general the literature is surprisingly mixed regarding anticholinergics and fall risk in older adults (11,36,37). Taken together, the period of time immediately after initiation of a CNS-active medication—especially opioids—is high risk for fall-related injuries.

This study addresses several gaps. First, to the best of our knowledge, this is the only study examining overall CNS-active exposure employing a new user design which minimizes bias that can occur when including only people with prevalent use. Including prevalent users—people using CNS-active medications prior to study entry—can introduce selection bias especially for adverse outcomes (eg, falls) for which risk varies with time on therapy. Second, we measured CNS-active medication exposure in the period before the fall-related injury to establish a temporal relationship.

A few limitations should be considered. First, we used automated pharmacy data as our measure of medication exposure, and it is possible that people did not actually consume the medication or consumed the medication outside the timing we assumed for calculating episodes of use, which is especially relevant to medications that are used only as needed. Thus, we may have exposure misclassification, which would likely have biased our findings toward the null. Second, bias may occur because of residual confounding and confounding by indication. To minimize this, we adjusted for a variety of demographic and health-related characteristics, and measures of gait speed, cognition and depressive symptoms, variables not commonly available in studies relying solely on administrative claims data. A limitation that we share with other studies is that we were unable to adjust for quality of sleep, pain, or pain severity which may result in confounding by indication. Pain severity is associated with falls and is also an indication for some of the CNS-active medications, which is especially pertinent for the analysis focusing on new initiation of opioids (38). Third, our findings may not be generalizable to people with dementia as they were not included in our sample

In conclusion, we found that current exposure, especially new initiation, of CNS-active medications was associated with fall-related injury in community-dwelling older adults. We found that risk was increased for all dose levels. Initiation of opioids was particularly problematic for increasing fall-related injury. At the time of prescribing a new CNS-active medication for older adults, clinicians should consider the potential risk for increasing risk for falls. If the medication is necessary to optimize patient health, then clinicians are encouraged to educate older adults and their caregivers on these risks and closely monitor for falls upon initiation of a new CNS-active medication.

Funding

This work was supported by the National Institute on Aging (U01AG00678 to E.B.L. and R03AG042930 to S.D.); and the Branta Foundation (to S.D.).

Authors’ Contributions

All the authors contributed to study conception and design, acquisition, analysis, or interpretation of data; S.L.G. and R.W. drafted the article; all authors revised the article for critical intellectual content; R.W. conducted statistical analyses; and E.B.L. and P.C. obtained funding.

Conflict of Interest

None declared.