Predictors of CNS-Active Medication Use and Polypharmacy Among Homebound Older Adults With Depression

The rates of outpatient care visits in which older adults were prescribed antidepressant, anxiolytic, sedative-hypnotic, or analgesic medications have risen over the past decade (1). Especially notable has been the increase in polypharmacy for medications active in the central nervous system (CNS), defined by the Beers Criteria as three or more CNS-active medications (antipsychotics, benzodiazepines, nonbenzodiazepine benzodiazepine receptor agonist hypnotics, selective serotonin reuptake inhibitors [SSRIs], tricyclic antidepressants, and opioids) (1). Opioid prescribing rates were declining overall in recent years (2); however, benzodiazepine and continuing (vs. new) antidepressant prescriptions continued to increase in primary care visits (3, 4).

Although CNS-active medications are prescribed and used for symptom relief, risks (e.g., development of frailty, hepatotoxicity, falls, fractures, delirium) associated with their long-term use, their misuse, and polypharmacy are well-documented (5, 6). Despite extensive research on polypharmacy and its adverse effects on older adults in general, little research has focused on increasing numbers of homebound older adults who, along with chronic illness and disability, have depression and other psychiatric comorbid conditions at significantly higher rates and tend to be more socioeconomically disadvantaged than nonhomebound peers (7).

The purposes of this study were to examine patterns of CNS polypharmacy among homebound older adults enrolled in a clinical trial of short-term psychosocial treatment for depression, associations between the participants’ use of CNS polypharmacy with their depressive symptoms and pain ratings, and associations among use of different medication groups. Our hypotheses were, first, that CNS polypharmacy would be associated with more severe depressive symptoms and higher self-reported pain intensity and, second, that antidepressant, benzodiazepine, and opioid analgesic medication intake would be significantly associated with one another. Covariates were sociodemographic characteristics, number of chronic illnesses, emergency department (ED) visits, disability, and perceived social support. The findings of this study provide a valuable insight into one of the most disadvantaged groups of older adults with respect to CNS-active medication use and polypharmacy.

Methods

Participants were 277 homebound individuals with depression (age ≥50) enrolled in a randomized controlled trial evaluating the clinical effectiveness of short-term depression treatments with the use of video conferences (Choi et al., 2019, unpublished; ClinicalTrials.gov identifier: NCT02600754). They were referred to the study by case managers of a large, home-delivered meals program in Central Texas. Inclusion criteria were moderate-to-severe depressive symptoms and willingness to participate. Exclusion criteria were high suicide risk, probable dementia, bipolar disorder, psychotic disorder, and substance abuse. Written informed consent, approved by the authors’ university institutional review boards, was obtained from each participant after the study procedures had been fully explained. Following informed consent, participants were administered the baseline assessment at their own home by trained assessors. To examine the study questions, we used baseline data, collected between February 2016 and April 2019.

Participants’ current intake, dose, and intake frequency of CNS-active medications (antipsychotics, benzodiazepines, nonbenzodiazepine benzodiazepine receptor agonist hypnotics, SSRIs, tricyclic antidepressants, and opioids) were collected from their medication containers. Data on other classes of antidepressant medications, sedative-hypnotic or sleep medications, and nonsteroidal anti-inflammatory drugs (NSAIDs) were collected in the same manner. CNS polypharmacy referred to taking three or more CNS-active medications.

Depressive symptoms were assessed with the 24-item Hamilton Depression Rating Scale (HAMD)—the GRID-HAMD-21 structured interview guide (8) augmented with three additional items (hopelessness, helplessness, and worthlessness) by Moberg et al. (9). A HAMD score ≥15 (moderately severe to severe symptom severity) was the study inclusion criterion. Participants self-reported intensity of bodily pain “during the past 6 months” on scale ranging from 0 (no pain) to 10 (pain as bad as it could be), along with the pain frequency and interference with daily activities.

Sociodemographic variables included age, gender, race-ethnicity, education, and financial situation. The number of chronic illnesses included up to nine chronic medical conditions that have ever been diagnosed by a health care professional (arthritis; diabetes; hypertension; heart disease; stroke; emphysema, chronic bronchitis, and other lung problems; kidney disease; liver disease; and cancer). An ED visit was coded 1 (vs. 0) if the participant visited an ED at least once in the past 90 days. Disability (in the past 30 days) was assessed with the 12-item World Health Organization Disability Assessment Schedule 2.0 (10) to determine the activity limitations and activity-participation restrictions (0=none; 4=extreme/cannot do), irrespective of medical diagnosis. Perceived social support was assessed with the 12-item Multidimensional Scale of Perceived Social Support (11) on a 7-point scale (1=very strongly disagree; 7=very strongly agree), with higher scores indicating higher perceived support. Living arrangement, household income, impairments in activities of daily living (0–6), impairments in instrumental activities of daily living (0–6), and overnight hospitalization within the past 90 days are reported for descriptive purposes only.

All analyses were performed with Stata 15-MP (StataCorp, College Station, Texas). We used descriptive statistics to examine participants’ characteristics, specifics of each CNS-active medication group, and other psychotropic and analgesic medication intake. To test the first hypothesis (associations of CNS polypharmacy with depression severity and pain intensity), we fit a multinomial logistic regression model with CNS-active medication use (0, 1–2, and ≥3) as the dependent variable. To test the second hypothesis (associations among use of different medication groups), we fit two logistic regression models with antidepressant and opioid use, respectively, as dependent variables. In multivariable models, we excluded 22 participants undergoing cancer treatment. Multinomial and binary logistic regression results are presented as relative risk ratios (RRRs) with 95% confidence intervals (CIs), and adjusted odds ratios (AORs) with 95% CIs, respectively. Statistical significance was set at p<0.05.

Results

On average, participants were age 68; most were female, were Hispanic or non-Hispanic black, lacked a bachelor’s degree, had income ≤$25,000, and reported financial difficulty (cannot make ends meet or just about manage to get by) (see online supplement). They had, on average, four chronic illnesses, moderately severe disability, and moderate social support; moreover, almost one third had one or more ED visit in the preceding 90 days.

Of the participants, 77% (N=214) were taking psychotropic or analgesic medication, including antidepressants (51%, N=142), benzodiazepines (22%, N=62), sedative-hypnotic or sleep medications (21%, N=57), and opioids (42%, N=115) (see online supplement). Although the median lengths of medication use were 3–4 years, there were long-term (32–37 years) users. As expected from the study inclusion criteria, there was no association between any antidepressant use and depression severity; however, opioid use was associated with significantly higher pain intensity, frequency, and interference. Most (91%) antidepressant users (N=129), 80% of benzodiazepine and sedative-hypnotic or sleep medications users (N=95), and 76% of opioid users (N=87) reported that they used their medications always or mostly as prescribed. No association was found between assessment years (2016–2019) and any medication use rate, although opioid use rates were higher among those assessed in 2016 than in those assessed between 2017 and 2019 (50% [N=48] vs. 37% [N=67], p=0.056).

In addition, 65% (N=180) used one or more CNS-active medications (see online supplement). Among those with any CNS-active medication use, 46% (N=82), 34% (N=62), and 64% (N=115) were using SSRIs, benzodiazepines, and opioids, respectively; moreover, 16% (N=45) engaged in CNS polypharmacy. Among those with CNS polypharmacy, 69% (N=31), 69% (N=31), and 89% (N=40) were using SSRIs, benzodiazepines, and opioids, respectively.

Table 1 shows that higher pain intensity ratings were associated with use of CNS-active medications (1–2 CNS-active medications, RRR=1.14, 95% CI=1.01–1.28; ≥3 CNS-active medications, RRR=1.20, 95% CI=1.01–1.45) compared with no CNS-active medication use. Depression severity was not a significant factor. Only financial difficulty was a significant covariate and was associated with lower risk of CNS-active medication use.

Sedative-hypnotic or sleep medication users (AOR=5.04, 95% CI=2.11–12.05) and NSAID users (AOR=2.32, 95% CI=1.25–4.32) were more likely to use antidepressants (see online supplement). Antidepressant users (AOR=1.99, 95% CI=1.06–3.74) and benzodiazepine users (AOR=3.36, 95% CI=1.65–6.84) were more likely to use opioids. Of the covariates, financial difficulty was associated with lower odds of antidepressant use; moreover, the number of chronic illnesses and higher pain intensity ratings were associated with greater odds of opioid use.

Discussion

This study found CNS polypharmacy among one-sixth of the study participants, which is higher than the 1% in all outpatient visits by those age ≥65 between 2004 and 2013 (1). Likely causes for the higher rate are that low-income homebound older adults with depression are more likely than older adults in general to have complex medical and psychological conditions, visit the ED, receive prescriptions from multiple health care providers, and have lower health literacy (12, 13), which is a barrier to their questioning health care providers about polypharmacy. The high rate of CNS polypharmacy, coupled with potential unintentional misuse due to their lack of understanding, are especially concerning for the safety of homebound older adults, given their multiple chronic illnesses and associated frailty.

Homebound older adults with depression need effective depression treatments because untreated depression will further aggravate their physical and functional health, social isolation, and quality of life. However, pharmacotherapy for these older adults requires special attention to certain safety issues associated with antidepressant use (e.g., increased bleeding risk, hyponatremia, decreased bone density, and falls) (14) because consequences (side effects and adverse outcomes) may be more severe, given their frailty and polypharmacy.

The high odds of benzodiazepine use by opioid users among these homebound older adults are also worrisome, given the safety issues related to poisoning from intentional or unintentional overdose and greater mortality associated with their concomitant use (15). In line with a national downward trend in opioid prescribing in recent years, opioid use rates among the study participants appear to have declined to a certain extent but not to a significant degree. Higher pain intensity ratings among opioid users may signal opioid-induced hyperalgesia among long-term opioid users. When possible, nonpharmacological, psychotherapeutic treatment of late-life depression, anxiety, sleep problems, and pain should be used to provide these older adults with more effective self-management strategies.

This study had several limitations. First, given the geographically limited, relatively small, and largely racial-ethnic minority sample, the findings may not be generalizable to other low-income homebound older adults. Second, self-reported data (e.g., medication adherence, disability, social support) may have been affected by social disability and recall biases. Third, the cross-sectional data could show only association, not causation. Thus, it was not clear whether depressive symptoms and pain intensity ratings could have been worse for some participants without antidepressant and analgesic medications.

Conclusions

Although polypharmacy for these older adults may be necessary to manage their multiple conditions, medication reviews and education about each medication are warranted, especially for long-term users of benzodiazepines, opioids, and NSAIDs. Health care providers should also be aware of CNS polypharmacy problems among these vulnerable older adults and take steps to minimize coprescriptions of multiple medication groups and to monitor toxicity and efficacy among polydrug users. Finally, given the strong evidence base of psychotherapeutic treatments for depression, anxiety, insomnia or other sleep problems, and chronic pain, improving access to these treatment options among low-income homebound older adults should be a national goal in a rapidly aging society.