Poor Adherence to Medications May Be Associated with Falls

Abstract

Background.

Poor medication adherence is associated with negative health outcomes. We investigated whether poor medication adherence increases the rate of falls as part of Maintenance of Balance, Independent Living, Intellect, and Zest in the Elderly of Boston (MOBILIZE Boston), a prospective, community-based cohort recruited for the purpose of studying novel risk factors for falls.

Methods.

A total of 246 men and 408 women (mean age, 78 years) were followed for the occurrence of falls (median follow-up, 1.8 years). Adherence was assessed by the Morisky scale based on the following four questions: whether an individual ever forgets, is careless at times, stops taking medications when feels better, or stops taking medications when feels worse. Low adherence was defined as a “yes” answer to one or more questions. High adherence was defined as a “no” answer to every question.

Results.

Forty-eight percent of subjects were classified as having low medication adherence. The rate of falls in the low adherence group was 1.1 falls/person-year (95% confidence interval [CI]: 1.0–1.3) compared with 0.7 falls/person-year (95% CI: 0.6–0.8) in the high adherence group. After adjusting for age, sex, race/ethnicity, education, alcohol use, cognitive measures, functional status, depression, and number of medications, low medication adherence was associated with a 50% increased rate of falls compared with high medication adherence (rate ratio = 1.5, 95% CI: 1.2–1.9; p < .001).

Conclusions.

Low medication adherence may be associated with an increased rate of falls among older adults. Future studies should confirm this association and explore whether interventions to improve medication adherence might decrease the frequency of falls and other serious health-related outcomes.

FALLS are one of the most common health concerns facing elderly persons. About one third of community-dwelling persons over the age of 65 years and nearly one half of institutionalized persons fall each year (1,2). Although most falls result in no injury, 31% of falls result in an injury requiring medical attention or restriction of activities(3). Additionally falls are the leading cause of fatal and nonfatal injuries for persons aged 65 years and older (4). Even among persons not experiencing a fall-related injury, falls are associated with functional decline, social withdrawal, depression, and an increased use of medical services (4,5).

Risk factors for falls among community dwellers include advancing age, dementia, Parkinson's disease, depression, stroke, osteoarthritis, previous falls, dizziness, and visual and balance impairments (6–8). Psychotropic medications and polypharmacy (use of more than three to four medications) have also been associated with an increased risk of falls (9,10).

Previous studies have found an association between poor medication adherence and early functional and cognitive decline in elderly persons (11,12). Some evidence suggests that poor medication adherence may contribute to this decline directly as poor adherence with use of lipid-lowering agents, bronchodilators, antipsychotics, and antiepileptic medications is associated with increased hospitalization and resource utilization among community dwellers treated for associated illnesses (13–16).

To our knowledge, no study has reported on the relationship between medication adherence and falls. We therefore investigated whether poor medication adherence increases the risk of falls as part of the Maintenance of Balance, Independent Living, Intellect, and Zest in the Elderly of Boston (MOBILIZE Boston) Study, a community-based cohort of seniors recruited for the purpose of studying novel risk factors for falls.

METHODS

Study Participants

Subjects included participants of MOBILIZE Boston, a prospective cohort study comprised of a sample of community-dwelling elders in the greater Boston area (17,18). Eligibility criteria included: (a) age 70 years and above, (b) residence within 5-mile radius of Hebrew Rehabilitation Center, (c) ability to hear and speak English language, (d) lack of significant cognitive impairment (Mini-Mental Status Examination [MMSE] ≥ 18), and (e) ability to ambulate 20 feet without the assistance of another person.

A total of 4,303 individuals were identified from town lists as being 70 years and older. Using doorstep recruitment, 6% (n = 280) refused to participate, 5% (n =217) were unable to be contacted, and one third were ineligible (n = 1,440). In a second stage of telephone screens, 57% (n = 1,360) refused to participate, 3% (n = 72) were unable to be contacted, and 6% (n = 133) were ineligible. After a final home interview stage, 5% (n = 44) refused to participate and 1% (n = 8) were ineligible. An additional 16 persons aged 64–69 years, who were spouses or living with participants, were added to the cohort, for a total of 765 study participants.

Information on participants was collected during a baseline home visit and clinical assessment conducted at Hebrew Rehabilitation Center (2005–2007). Individuals who were not taking medications (n = 25) or had incomplete information on medication adherence (n = 86) were excluded from these analyses. This study was approved by the Institutional Review Board of Hebrew SeniorLife.

Falls

Falls, defined as unintentionally coming to rest on the floor or lower surface, were ascertained by self-report using falls calendar postcards (19) and telephone interviews. Falls calendars were provided to participants at the baseline visit and returned by mail monthly (October 2005 to October 2008). Any subject reporting a fall was asked to complete a telephone questionnaire to provide details regarding the circumstances of the fall. Syncopal falls were excluded. Individuals who failed to complete a fall calendar (<20% each month) were contacted by telephone to ascertain falls. Less than 2% of subjects were missing falls information after telephone interviews.

Medication Use

At the baseline home visit, participants were asked to provide interviewers all over-the-counter and prescription medications taken in the past 2 weeks. The Iowa Drug Information Service was used to categorize medications by pharmacological therapeutic class (20).

The total number of medications was ascertained from a count of all over-the-counter and prescription medications. We excluded topicals, herbals, vitamins, and minerals. We retained all other classes provided that participants used these medications during the 2 weeks before the baseline visit. Psychotropic medications were defined as use of any antidepressant, benzodiazepine, or antipsychotic.

Medication Adherence

Medication adherence was ascertained using the 4-item Morisky scale (Table 3; 21). Low adherence was defined as a “yes” answer to at least one of the four questions. High adherence was defined as a “no” answer to every question. In secondary analyses, we considered low medication adherence as a “yes” answer to each of the four items separately.

Table 3.

Adjusted^* RR and 95% CI for Associations Between Specific Items Used to Characterize Low Medication Adherence and Falls in the MOBILIZE Boston Study

Item used to characterize low adherence to medications	RR*	95% CI
Do you ever forget to take your medications? (n = 270)	1.5	1.2–1.8
Are you careless at times about taking your medications? (n = 87)	1.1	0.8–1.5
When you feel better do you sometimes stop taking your medications? (n = 42)	1.1	0.7–1.8
Sometimes if you feel worse when you take your medication, do you stop taking it? (n = 65)	0.9	0.7–1.4

Notes: 95% CI = 95% confidence intervals; ADL, Activities of daily living; RR = rate ratios.

^*Adjusted for age, race, education, alcohol use, functional status, executive function, depression, and total number of medications.

Other Characteristics of Interest

We considered the following characteristics at the baseline visit that have been associated with an increased risk of falls in at least one study (6–8,10,22,23) and may influence medication adherence: older age, female sex, white race, higher education, alcohol use, low functional status, decreased physical activity, cognitive impairment, poor executive function, poor visual acuity, depression, increased number of comorbidities, urinary incontinence, polypharmacy, and use of psychotropic medications.

Race/ethnicity, education level, and alcohol use were determined by self-report. Functional status was determined by the modified Katz Activities of Daily Living (ADL) scale (24). We classified subjects into two categories: (a) no difficulty in performing any of the five ADLs, and (b) difficulty or inability to perform one or more ADL independently. Physical activity was measured by the Physical Activity Scale for the Elderly (PASE), a self-reported weighted score of 10 commonly performed activities, with higher numbers indicating greater physical activity (25).

Cognition was measured using the MMSE, with cognitive impairment defined as an MMSE score of less than 24 (26). Executive function was measured by time to complete Trails B test (seconds) (27). Visual acuity was measured in both eyes with corrective lenses using the Snellen eye chart, with poor acuity defined as best-corrected vision less than 40/100. Depression was ascertained using the Revised Center for Epidemiologic Studies Depression (CESD-R) instrument with severe depression defined as a CESD-R score greater than 60 (28). We assessed burden of comorbidities using the modified Self-Administered Co-morbidity questionnaire (29), which determines the presence or absence of 12 major medical conditions by self-report (score 0–12).

Statistical Analysis

We compared differences in baseline characteristics between subjects with high and low adherence to medication using a Wilcoxon rank sum test for quantitative variables and a chi-square test for categorical variables. The annualized rate of falls was calculated as the total number of falls per person-year of follow-up. We used negative binomial regression models to estimate the association of medication adherence (low vs high) with the rate of falls because subjects had unequal follow-up time and number of falls was unequally distributed in the cohort (ie, most people had no or a small number of falls, whereas a few people had a large number of falls). We expressed the association of low medication adherence and the rate of falls as rate ratios (RR) with 95% confidence intervals (CI) (30).

In multivariable analyses, we decided a priori to adjust for age, sex, number of medications, and executive function. Other characteristics associated with the RR of falls at p ≤ .1 in the bivariate analyses were considered for inclusion in the multivariable analysis. We then performed a backwards stepwise regression by sequentially removing characteristics that were the least significant as independent risk factors for falls in the multivariable model. Interactions between medication adherence and number of medications and medication adherence and alcohol use were considered because we hypothesized that medication adherence might impact the risk of falls differently in these groups. Interaction terms were not significant, and thus, they were not considered further. The adequacy of the multivariate model presented here was assessed using a goodness-of-fit test (p = .23).

RESULTS

Subjects included 654 community-dwelling elders (246 men and 408 women). Mean age of participants was 78 years. During a median follow-up of 1.8 years (range, 0.6 months to 3.2 years), 376 subjects (141 men and 235 women) experienced a total of 1,052 falls.

Forty-one percent of subjects reported ever forgetting to take medications, and 6–13% reported sometimes being careless or not taking their medications when either feeling better or when they believed their medications made them feel worse. Overall, 48% of subjects were classified as having low medication adherence.

Subjects characterized as having low medication adherence were more likely to be non-white, have less than a high school education, report severe depression, urinary incontinence, more comorbidities, and to take more medications compared with subjects classified as having high adherence (Table 1). Although not statistically significant, subjects classified as having low medication adherence tended to have worse functional status, cognitive impairment, executive function, and visual acuity, whereas subjects classified as having high medication adherence were more likely to consume more than three to four alcoholic drinks/week.

Table 1.

Baseline Characteristics and Follow-up of Participants in the MOBILIZE Boston Study, by Classification of Low and High Medication Adherence

Characteristic, n (%) unless otherwise specified	Low adherence, (N = 314)	High adherence, (N = 340)	p Value
Age*	78 ± 5.3	79 ± 5.4	.02
Male	116 (37)	130 (38)	.73
Race (white)	231 (74)	278 (82)	.01
Consumes alcohol 3–4 days/week	70(22)	93 (27)	.14
Greater than high school education	1188 (61)	234 (69)	.02
Functional status			.09
No difficulty with any of the five ADLs	230 (73)	268 (79)
Difficulty or inability to perform one or more ADL	84 (27)	72 (21)
Physical activity
PASE score†	100.0 (53–138)	95.7 (56–145)	.27
Cognitive impairment‡	64 (20)	58 (17)	.28
Executive function
Trails B time (s)†	120 (85–193)	118 (81–185)	.20
Poor visual acuity§	30 (10)	19 (6)	.06
Severe depression‖	73 (23)	51 (15)	.01
Co-morbidity index*	3.3 ± 1.7	3.0 ± 1.5	.02
Urinary incontinence	142 (45)	122 (36)	.01
Total number of medications			.02
1–4 medications	96 (31)	96 (28)
5–7 medications	107 (34)	150 (44)
>7 medications	111 (35)	94 (28)
Psychotropic medication use	73 (23)	70 (21)	.41
Total number of falls	627	425	<.001
Person-years of follow-up	585	614	.46
Number of falls per subject			<.001
0	113 (36)	165 (49)
1	62 (20)	79 (23)
≥2	139 (44)	96 (28)

Notes: ADL = Activities of daily living; PASE = Physical Activity Scale for the Elderly.

^*Mean ± standard deviation.

^†Median (IQ range).

^‡Mini-Mental Status Examination score <24.

^§Defined as <40/100.

^‖Revised Center for Epidemiologic Studies Depression score <60.

Subjects characterized as having low medication adherence were more likely to experience two or more falls compared with subjects characterized as having high adherence (44% vs 28%; Table 1). The maximum number of falls experienced by a subject characterized as having high medication adherence was 18 compared with 12 in the low adherence group. The annualized fall rates in the low and high adherence groups were 1.1 (95% CI: 1.0–1.3) and 0.7 (95% CI: 0.6–0.8) falls/person-year, respectively.

The unadjusted RR for falls in the group with low medication adherence compared with the group with high medication adherence was 1.6 (95% CI: 1.3–1.9; p < .001; Table 2). In the adjusted model, the RR between the two groups changed little (RR = 1.5; 95% CI: 1.2–1.9; p < .001). We repeated the analysis excluding frequent fallers (defined as a fall rate that exceeded the 90th percentile for all subjects), and the results were similar. We additionally considered whether social support, as defined by the number of living children, affected the association, and the results were unchanged (results not shown).

Table 2.

Unadjusted and Adjusted^* RR and 95% CI for Associations Between Medication Adherence, Other Characteristics of Interest, and Falls in the MOBILIZE Boston Study

	Unadjusted association		Adjusted association
	RR	95% CI	RR	95% CI
Low medication adherence versus high medication adherence	1.6	1.3–1.9	1.5	1.2–1.9
Age (per 5 years)	1.1	1.0–1.2	1.1	1.0–1.2
Male	1.1	0.9–1.4	1.0	0.8–1.3
Race (white)	1.5	1.2–2.0	1.3	1.0–1.8
Greater than high school education	1.7	1.3–2.1	1.6	1.2–2.0
Consumes alcohol >3–4 days/week	1.2	1.0–1.6	1.3	1.0–1.6
Functional status*	1.6	1.3–2.1	1.7	1.3–2.2
Poor executive function†	1.0	1.0–1.0	1.0	1.0–1.0
Total number of medications
1–4 medications	Ref	Ref	Ref	Ref
5–7 medications	0.8	0.6–1.0	0.9	0.7–1.1
>7 medications	1.2	0.9–1.6	1.1	0.8–1.4

Notes: 95% CI = 95% confidence intervals; ADL, Activities of daily living; RR = rate ratios.

^*Difficulty or inability to perform one or more of the five ADLs versus no difficulty.

^†Per 10 s of time to complete trails B test.

When we considered low medication adherence as a “yes” response to each of the four separate items from the Morisky scale (Table 3), we found that low medication adherence was associated with an increased rate of falls for only the first of the four items, “Do you ever forget to take your medications?”

DISCUSSION

We found that poor medication adherence was associated with an increased rate of falls among community-dwelling elders. This association persisted after adjusting for other variables, including age, sex, cognitive function, and total number of medications.

There are several plausible explanations for our findings. First, subjects with low medication adherence may be more frail, and thus, more susceptible to falls. However, controlling for number of medications and comorbidities did not change our findings. Nonetheless, we recognize that there may be unmeasured or inadequately measured confounders, such as access to medical care, which could in part explain our findings.

Second, subjects with low adherence to medications may be less cautious and less likely to avoid situations associated with a high fall risk. We did not collect information on risk-taking behaviors, and thus, we are unable to test this hypothesis.

Finally, subjects with low adherence to medications may be at an increased risk of falls because they have limited physiological reserve as a result of their poor adherence. For example, failure to consistently take antihypertensive mediations may cause marked fluctuations in blood pressure, which could result in a transient lowered reserve for falls. On the other hand, failure to adhere with other classes of medications, such as sedatives, may actually decrease the risk of falls. We did not have information on adherence within specific drug classes, and thus we are unable to directly test this hypothesis. Nonetheless, when we repeated our analysis among psychotropic medication users, the results were unchanged.

To our knowledge, no previous studies have examined the effect of poor medication adherence on the risk of falls. A Canadian study of 319 community dwellers found a nonstatistically significant increased risk of death, hospitalization, and emergency room visits among elders who were poorly adherent with prescription medications after controlling for age, functional status, cognitive status, comorbidities, and medication use (pooled Hazard Ratio = 1.24, 95% CI: 0.93–1.65) (31). No information was provided on the reason for medical service utilization in this study, and thus it is unknown whether falls contributed to the observed increase in adverse health outcomes.

Poor medication adherence may occur in as many as 40%–50% of community-dwelling elders (31,32). Previous research indicates that forgetfulness and lack of interest or knowledge are major barriers to taking medication (32). Medication expense, complicated medication regimens, and limited access to health care providers are also probably important factors that influence adherence. Educational strategies that involve patients, their family members, and health care providers have been shown to be beneficial in improving adherence among elderly persons (33). It remains unclear whether improving medication adherence could also result in improved health-related outcomes, including a decreased rate of falls. However, our findings are still important as poor medication adherence is common, feasible to ascertain, and may help to identify community dwellers at an increased risk of falls.

Our study has several limitations. First, medication adherence was determined by self-report rather than by a pill count or days-of-possession. We expect that subjects are likely to have overreported adherence, resulting in some misclassification. Assuming that overreporting occurred to the same extent in fallers and nonfallers, this type of misclassification should have biased our findings toward the null; yet, we were still able to find an effect of medication adherence on the rate of falls.

Second, we were unable to fully evaluate whether the reason for poor medication adherence may have affected our results. When we considered low medication adherence as a “yes” response to each of the four separate items on the Morisky scale, we found that only a “yes” response to the question “Do you ever forget to take your medications?” resulted in an increased rate of falls. Future investigations should consider whether the reason or magnitude of poor medication adherence affects the risk of falls.

Third, participation in the MOBILIZE Boston study was somewhat low. However, our study used the least restrictive eligibility criteria that still enabled us to carry out the study procedures (17). In contrast to many other studies in older adults (34), we placed no upper age limit on eligibility in order that our results be applicable to very elderly individuals at greatest risk of falls. Among participants, 86 individuals (11%) were missing complete information on medication adherence and excluded from our study. Individuals excluded did not differ from study participants with respect to characteristics including age and number of medications. Nonetheless, we performed a secondary analysis using imputation for subjects with incomplete information on medication adherence, and results were unchanged. Thus, the association between low medication adherence and increased risk of falls observed in this study is not likely explained by participation bias.

Finally, this study was conducted among a cohort of relatively high-functioning English-speaking community-dwelling elders. Certain characteristics, such as number of medications and poor executive function, were not associated with an increased rate of falls as we might have expected. This may be explained because our cohort was high functioning, or because we combined indoor and outdoor falls in the analysis. Low adherence to medications may be more prevalent among certain ethnic and racial groups (35) or among persons with greater functional impairment compared with individuals in our study, and poor adherence may possibly be a greater risk factor for falls in these persons.

In conclusion, older persons who report that they sometimes forget to take medications are at a greater risk of falls compared with those who report better adherence. Future studies should examine the association between poor adherence with specific drug classes and falls and between reasons for poor medication adherence and falls. It remains unclear whether strategies to improve medication adherence might decrease the frequency of falls, but clinicians should screen for poor medication adherence as it is common and predicts adverse health outcomes including falls.

FUNDING

This work was supported by the Hartford Geriatrics Health Outcomes Research Scholars Awards Program, the HRCA/Harvard Research Nursing Home Program Project funded by National Institutes of Health (grant number AG004390), and by an unrestricted grant from Pfizer, Inc.