Antidementia Medication Use in Nursing Home Residents

Abstract

BACKGROUND:

Antidementia medication can provide symptomatic improvements in patients with Alzheimer’s disease, but there is a lack of consensus guidance on when to start and stop treatment in the nursing home setting.

METHODS:

We describe utilization patterns of cholinesterase inhibitors (ChEI) and memantine for 350,197 newly admitted NH residents with dementia between 2011 and 2018.

RESULTS:

Overall, pre-admission use of antidementia medications declined from 2011 to 2018 (ChEIs: 44.5% to 36.9%; memantine: 27.4% to 23.2%). Older age, use of a feeding tube, and greater functional dependency were associated with lower odds of ChEI initiation. Coronary artery disease, parenteral nutrition, severe aggressive behaviors, severe cognitive impairment, and high functional dependency were associated with discontinuation of ChEIs. Comparison of clinical factors related to anti-dementia drug treatment changes from pre to post NH admission in 2011 and 2018 revealed a change toward lower likelihood of initiation of treatment among residents with more functional dependency and those with indicators of more complex illness as well as a change toward higher likelihood of discontinuation in residents having two or more hospital stays.

CONCLUSIONS:

These prescribing trends highlight the need for additional research on the effects of initiating and discontinuing antidementia medications in the NH to provide clear guidance for clinicians when making treatment decisions for individual residents.

INTRODUCTION

Two classes of antidementia medication have been shown to provide modest symptomatic improvements in people with Alzheimer’s disease (AD).^1–4 The two classes with Food and Drug Administration (FDA) marketing approval for AD are the cholinesterase inhibitors (ChEIs) donepezil, galantamine, and rivastigmine (indicated for mild, moderate, and severe disease), and the N-methyl-D-aspartate (NMDA) antagonist memantine (indicated for moderate to severe disease).

Nursing homes (NHs) are the primary care setting for many people with advanced AD; yet, contemporary information regarding the use of ChEIs or memantine in NHs is limited. For NH residents with dementia, past studies conducted using data from the early 2000s found that one-third were prescribed ChEIs, one-sixth memantine, and few were prescribed both.^5–7

In clinical practice, many residents on ChEIs or memantine at the time of NH admission discontinue treatment shortly after entry (regardless of dementia severity).^6,8 Discontinuation of long-term antidementia drug treatment may be warranted for patients experiencing adverse events and for patients who have end-stage disease. Discontinuation of ChEIs has been associated with a reduced likelihood of falls and fractures in NHs,⁹ and may not lead to worsening aggressive behavior;¹⁰ however, discontinuation has been also been associated with worsening of behavior and functioning in some cases.^11–16 While end of life decisions for people with very advanced dementia often include discontinuance of antidementia drugs as futile interventions,^17,18 some practice guidelines¹⁹ and expert reviews²⁰ have recommended that antidementia drugs should not be stopped just because dementia severity increases or because a person is admitted to a NH.²¹

Evidence to guide how long to continue treatment with ChEIs or memantine in the NH setting is limited, however, by the short duration of follow-up in randomized controlled clinical trials in primarily outpatient cohorts.^1,2,4 Thus, clinical practice guidelines have been unclear and conflicting in their recommendations regarding what criteria should prompt discontinuation in the NH setting.^18,22–24

To inform our understanding of evolving NH prescribing practices for patients with AD related dementia, our objective was to describe and compare utilization patterns of ChEIs and memantine before and after NH admission. Secondly, we examined these prescription patterns over time. Lastly, we examined resident characteristics associated with initiation and discontinuation of each medication class shortly after NH admission. To provide a comprehensive view of recent antidementia drug usage in NHs, we provide an overall assessment of drug usage during the epoch of 2011–2018 as well as annual usage comparisons.

METHODS

Data sources

Our study used Minimum Data Set (MDS) 3.0 linked with Medicare administrative data from the Master Beneficiary Summary file (MBSF), Medicare Part A and Medicare Part D for calendar years 2011–2018. MDS assessments are performed at admission, quarterly, annually, and upon a significant change in the resident’s status. The assessments are mandatory for all residents of Medicare- and Medicaid-certified NHs which represent >95% of all NHs in the United States. Medicare Part A records are part of the Medicare Provider Analysis and Review file, which contains inpatient claims submitted by hospitals or skilled nursing facilities. The Medicare Part D Event and Drug Characteristics files contain information on drug dispensing and characteristics, respectively. This study was approved by the UMass Chan Medical School Institutional Review Board.

Population

Newly admitted residents within the date ranges of 04/01/2011 to 09/30/2018 (allowing 3 months of data before and after admission for all residents) with a diagnosis of AD or other dementia disorder were eligible. Diagnoses of AD or other dementia were ascertained from ICD-9 and ICD-10 codes²⁵ on Medicare Part A claims and MDS assessments (Supplemental Table 1), as well as from MDS active diagnosis fields for these conditions (MDS items: I4200, I4800). Our study population included residents with uninterrupted enrollment in Medicare fee-for-service plans (Parts A, B, and D) 90 days before and 30 days after their initial admission date. Excluded residents were those who were in a coma (MDS item B0100), enrolled in hospice (MDS items O0100K1 and O0100K2), or who died within 30 days of NH admission. People entering NHs for a skilled nursing facility stay (i.e., their admission assessment indicated the assessment was used for a prospective payment system assessment: MDS item A0310B not equal to ‘99’) were also excluded because these resident’s medications are bundled into the per diem payment (and not observable) rather than in Part D claims. Figure 1 summarizes the application of inclusion and exclusion criteria.

Figure 1.

Population selection diagram

Antidementia Medications

In the 90 days preceding (“Pre-NH” period) and following (“Post-NH” period) admission, we examined the use of two classes of antidementia medications (ChEIs and NMDA receptor antagonists) for reasons discussed in the introduction. The ChEIs include donepezil, galantamine, and rivastigmine. Memantine is the only approved NMDA receptor antagonist medication in this class. For each medication class, residents were classified for pattern of use as non-users (no prescriptions in the Pre-NH and Post-NH period), discontinuers (at least one prescription in the Pre-NH period and no prescriptions in the Post-NH period), initiators (no prescriptions in the Pre-NH period and at least one prescription in the Post-NH period), and continuous users (at least one prescription in each of the Pre-NH and Post-NH periods) of each medication based on the presence or absence of at least one Part D claim before and one claim after admission.

Covariates

We examined demographic, clinical, and functional characteristics as covariates. Demographic characteristics were drawn from the MBSF and included age, sex, race/ethnicity, and Medicaid enrollment. Information on diagnoses was drawn from both the MDS and Medicare Part A claims. Physical comorbidities included diabetes, heart failure, cardiac arrhythmias, stroke, coronary artery disease, renal impairment, Parkinson’s disease, liver cirrhosis, and malnutrition. Mental health conditions included depression, bipolar disorder, and schizophrenia. Other clinical characteristics drawn from the MDS that were assessed upon admission to the NHs were recent fall and fracture history, and special nutritional needs (parenteral or feeding tube, mechanically altered food).

Polypharmacy was calculated based on the number of unique Part D claims for distinct generic drug names (excluding the exposure medication) and categorized as: using ≤5 medications, using 6–10 medications, and using >10 medications. Functioning in activities of daily living (ADLs) were measured using the long form of the ADL score (range 0–28, with higher scores indicative of greater ADL limitations) and categorized into 0–7 (low functional dependency), 8–14 (mild functional dependency), 15–21 (moderate functional dependency), and 22–28 (high functional dependency).²⁶ Cognitive impairment was measured using the Cognitive Function Scale (CFS) which uses MDS information from both staff assessment and resident reported information to categorize residents as either no cognitive impairment, mild, moderate, or severe impairment.²⁷ The MDS 3.0 Aggressive Behavior Scale (ABS) (range 0–12, categorized into 0 (none), 1–4 (moderate), 5–8 (severe), and 9–12 (very severe) was used to quantify the severity of behavioral symptoms, such as physical aggression towards others, verbal aggression towards others, rejection of evaluation or treatments, and behaviors classified as other.²⁸

Statistical Analysis

Descriptive statistics summarized the characteristics of the study population according to categories of medication use (antidementia drug users and non-users). For continuous variables, the median, 25^th and 75^th percentiles were calculated. We calculated the frequencies and percentages for all categorical variables.

The prevalence of medication use was then calculated separately for the pre-admission and post-admission period for each class of medication during 2011–2018. Prevalence of medication use over time according to the four utilization patterns previously described was examined for each study year. Cochran-Armitage tests with two-sided statistical significance levels of 0.05 were performed for linear trend in use of each medication class. We also fit multivariable logistic regression models to compare the odds of being a pre-admission user of each drug class based on the year of admission to the NH while adjusting for covariates.

In the overall (2011–2018) population, a series of four multivariable models were developed to identify factors associated with change in the use of each of the two medication classes under study, i.e., ChEIs, memantine. Specifically, among non-users of a medication class during the pre-admission period, logistic regression models estimated odds ratios for initiation. Similarly, among users of a medication class prior to admission, logistic regression models estimated odds ratios for discontinuation. Supplemental analyses were performed within the subgroup with a specific diagnosis of AD and separately within the earliest and latest years of data available (2011 and 2018). All covariates were considered as potential predictors, along with pre-admission use of the other medication classes.

RESULTS

Among 350,197 residents admitted to U.S. NHs during April 1, 2011 to September 30, 2018, the average age of the study population was 85 years, and 70.4% of the population were women. Characteristics of the residents before NH admission according to drug use are provided in Table 1. Non-users were slightly older than users, had more comorbidities and pre-admission fractures, and a lower prevalence of depression. Polypharmacy was less prevalent in non-users. An admitting diagnosis of AD occurred in 37.6% of residents using ChEIs and 41.0% of those using memantine, compared to 19.3% of non-users. Compared to 54.6% of non-users, 62.9% of ChEI users and 68.3% of memantine users had moderate or severe cognitive impairment. Prevalence of severe impairment was lowest among ChEI users. Prevalence of ADL impairment and aggressive behaviors were comparable in both drug user groups and greater than that seen in non-users.

Table 1.

Characteristics of ChEI, memantine, and non-users before NH admission among residents with dementia (2011 to 2018: N=350,197)

	ChEI Users (n= 143780)	Memantine Users (n=87,985)	Non-users (n= 171387)
Median age in years (25th, 75th)	84 (78,94)	84 (78,88)	86 (80,97)
Women (%)	71.0	71.1	70.6
Race/Ethnicity (%)
Non-Hispanic White	84.8	85.4	83.9
Black	9.5	8.8	9.0
Hispanic	2.5	2.6	2.3
American Indian/Alaska Native	0.4	0.4	0.8
Asian/Pacific Islander	1.7	1.7	2.3
Other	0.8	0.8	0.9
Married (%)	27.5	29.5	23.2
Dual Eligibility (%)	50.3	50.2	49.9
Alzheimer diagnosis (%)	37.6	41.0	19.3
Comorbidities (%)
Parkinson’s disease	2.2	1.9	2.3
Atrial fibrillation or other dysrhythmia	6.0	5.6	9.9
Diabetes mellitus	26.2	25.1	28.0
End stage renal disease	8.4	7.9	12.1
Cirrhosis	0.2	0.2	0.4
Heart failure	12.4	11.4	18.0
Stroke	8.6	8.2	12.2
Coronary disease	20.2	19.6	22.2
Malnutrition	2.0	2.0	3.0
Mental health conditions
Depression	46.1	47.3	37.2
Bipolar	2.5	2.4	1.67
Schizophrenia	2.5	2.5	1.3
Polypharmacy (%)
6–10 medications	40.1	40.9	39.7
> 10 medications	34.6	35.3	29.3
Fall history (%)
Fall since admission	10.5	10.8	10.1
History of fracture 6 months prior to or since admission	7.8	7.3	10.0
Median ADL score* (25th,75th)	17 (11,19)	17 (11,20)	18 (13,20)
Cognitive Function Scale** (%)
Intact - mildly impaired	37.0	30.8	45.3
Moderately impaired	54.5	58.9	44.5
Severely impaired	8.4	10.4	10.1
Aggressive behavior scale*** (%)
Moderate	20.4	20.9	17.2
Severe	2.2	2.4	1.4
Very severe	0.5	0.5	0.3

Users may be receiving more than one class of medication

^*The ADL(activities of daily living) score summarizes independence in seven ADLs with each ranging from a minimum of 0 (independent) to 4 (dependent) for a score range of 0–28

^**The Cognitive Function Scale draws information from two sources on the Minimum Data Set: 1) the five-item Brief Interview for Mental Status (BIMS) screener that assesses temporal orientation and recall, and 2) the Cognitive Performance Score (CPS), which is only fully complete if the resident who are deemed inappropriate or unable to complete the BIMS. The Cognitive Function Scale has four levels (intact, mildly, moderately, and severely impaired); we collapsed intact and mild impairment because the BIMS and CPS group these levels.

^***The Aggressive Behavior Scale is a summary of four MDS items including verbally abusive, physically abusive, socialy inappropriate or disruptive behavior, and resisting care. Each behavior is coded from 0 (not exhibited) to 3 (occurred daily) over the past seven days so that the summary score ranges from 0 to 12. The categories presented correspond to 1–4 (moderate), 5–8 (severe), and 9–12 (very severe).

Annual comparisons for prevalence of medication use pre- and post NH admission are provided in Figure 2. Patterns of use over time show a progressive decline in use of ChEIs and memantine from 2011 to 2018. Usage following NH admission was lower in every year for ChEIs but not for memantine. Trends for the prevalence of pre- and post-NH admission use of ChEIs and memantine gradually showed significant decline over years from 2011 to 2018 with some fluctuations (p-values <0.001 for both ChEIs and memantine at pre- and post-NH admission); the prevalence of pre-NH admission use of ChEIs and memantine decreased from 44.5% to 36.9% and 27.4% to 23.2%, respectively, during 2011 to 2018; the prevalence of post-NH admission use of ChEI’s and memantine dropped from 43.2% to 35.0% and 27.2% to 22.5% respectively, during 2011 to 2018. Adjusting for changes in resident characteristics over time, residents admitted to the NH in 2018 had reduced odds of being pre-NH users of ChEIs (adjusted Odds Ratio (aOR): 0.79 (95% confidence interval (CI): 0.77–0.82), and memantine (aOR: 0.94 (95% CI: 0.91–0.97) compared with 2011. Annual comparisons for prevalence of medication use pre and post NH admission for those residents specifically diagnosed to have AD are provided in Supplemental Figure 1. The trends are very similar, though use following NH admission was lower in every year for memantine as well as for ChEIs.

Figure 2.

Trends in the prevalence of pre- and post-nursing home admission use of ChEIs and memantine by year among residents with dementia (repeated cross-sections: 2011–2018)

Abbreviations: ChEI (cholinesterase inhibitors)

*p <0.001 for Cochran-Armitage test for linear trends in pre- and post-admission prevalence of ChEI and memantine use

Details showing the change in the medications of interest that were made between 2011 and 2018 are provided in Table 2. Overall, pre-admission use of antidementia medications declined from 2011 to 2018 (ChEIs: 44.5% to 36.9%; memantine: 27.4% to 23.2%). Comparable details regarding patterns of change in medications for residents diagnosed specifically as AD are provided in Supplemental Table 2. Overall use of antidementia drugs was higher for those diagnosed with AD. Before NH admission, 54.0% of the population used ChEIs (continuers + discontinuers), and 37.0% of them used memantine. The prevalence of ChEI use decreased mildly after NH admission (54.0% to 50.5%; continuers + initiators) as well as for memantine (37.0% to 35.7%).

Table 2:

Patterns of cholinesterase inhibitors (ChEIs) and memantine use in the 90 days pre- and 90 days post-nursing home admission in 2011 and 2018 among residents with dementia.

	Year 2011 (N=41,642)		Year 2018 (N=39,379)
	ChEIs	Memantine	ChEIs	Memantine
Prevalent users,* %	38.9	24.0	31.9	20.2
Discontinued users,* %	5.7	3.4	5.0	3.1
Initiators,* %	4.3	3.2	3.1	2.1
Non-users,* %	51.2	69.4	60.1	74.4

^*We defined continuing users as residents with at least one Part D claim in the 90 days before and 90 days after nursing home admission; discontinued users as those with a claim in the 90 days before but not after nursing home admission; initiators as a claim in the 90 days after and not before nursing home admission; and non-user as those without any claim in the 90 days before or 90 days after nursing home admission.

We performed multivariable logistic regression to identify clinical factors that were associated with ChEI and memantine initiation or discontinuation (Supplemental Table 3). Older age (oldest versus youngest quintile: aOR: 0.49, 95% CI: 0.46–0.53), use of a feeding tube (aOR: 0.41, 95% CI: 0.21–0.80), and greater dependency in ADLs (full dependency versus low: aOR: 0.41, 95% CI: 0.38–0.45) were associated with lower odds of ChEI initiation. Characteristics associated with discontinuation of ChEIs included coronary artery disease (aOR: 1.40, 95% CI: 1.00–1.96), receipt of parenteral nutrition (aOR: 2.64, 95% CI: 0.99–7.05), severe (aOR 1.64, 95% CI: 1.51–1.79) and very severe aggressive behaviors (aOR: 1.67, 95% CI: 1.40–2.00), severe cognitive impairment (aOR: 1.72, 95% CI: 1.61–1.84), and high functional dependency in ADLs (aOR: 1.72,95% CI: 1.61–1.83). The direction and magnitude of associations for memantine initiation and discontinuation were generally similar to those for ChEIs.

Comparable details regarding clinical factors that were associated with ChEI and memantine initiation or discontinuation for residents diagnosed specifically as AD are provided in Supplemental Table 4. Again, older age and greater dependency in ADLs were associated with lower odds of ChEI initiation. Hispanic ethnicity was associated with higher odds of initiating either drug, and severe aggression was associated with higher odds of initiating memantine. Characteristics of residents diagnosed with AD associated with higher odds of discontinuation of both drug classes included severe aggressive behaviors, severe cognitive impairment, high functional dependency, and 2 or more hospitalizations.

We then performed multivariable logistic regression separately in 2011 and 2018 to examine changes over time in associations of clinical characteristics with (“Post-NH” period) initiation (Supplemental Table 5) or with discontinuation (Supplemental Table 6). The aOR of initiating memantine in the oldest quintile rose from 0.59 (95% CI: 0.47–0.73) in 2011 to 0.68 (95% CI: 0.54–0.87) in 2018, while the odds remained relatively consistent in the three oldest quintiles for ChEI. The aOR of initiating either class of drugs for residents with moderate functional dependency dropped to a similar degree. The aOR of initiating ChEI in residents who had a skilled nursing facility stay dropped from 1.30 (95% CI: 1.11–1.52) in 2011 to 1.11 (95% CI: 0.93–1.33) in 2018. Compared with non-Hispanic white residents, black residents appeared more likely to initiate ChEIs in 2011 (aOR: 1.16; 95% CI: 0.98–1.37) and in 2018 (aOR: 1.23: 95% CI: 1.00–1.51). The aOR of initiating ChEI in residents receiving polypharmacy (>10 medications) dropped from 0.81 (95% CI: 0.72–0.91) in 2011 to 0.69 (95% CI: 0.60–0.80) in 2018.

The aORs for discontinuing either class of drugs showed less evidence of change overall. The aOR of discontinuing memantine in residents with severe aggressive behavior rose from 1.95 (95% CI: 1.18–3.23) in 2011 to 2.13 (95% CI: 0.99–4.59) in 2018, while the aOR of discontinuing memantine in residents with severe cognitive impairment dropped from 1.56 (95% CI: 1.19–2.04) in 2011 to 1.37 (95% CI: 1.04–1.81) in 2018, although confidence intervals were wide. The aOR of discontinuing both classes of drug rose by about 0.1–0.2 in residents having 2 or more hospital stays. The aOR of discontinuing memantine in residents with a skilled nursing facility stay rose from 1.24 (95% CI: 1.02–1.51) in 2011 to 1.35 (95% CI 1.11–1.63) in 2018. Residents with atrial fibrillation or other dysrhythmia had a higher odds of discontinuing ChEIs than those without dysrhythmias in 2011 (aOR 1.22; 95% CI: 1.02–1.46) but not 2018 (aOR: 1.00; 95% CI: 0.83–1.20).

DISCUSSION

Overall, we found that in any given year less than half of all residents with dementia were treated with ChEI and less than one-third with memantine over the entire period from 2011–2018. A previous study using the 2004 National NH Survey reported that approximately 30% of 5,866 NH residents with dementia in U.S. NHs were being treated with ChEIs.⁵ In 2006, another retrospective cohort study examining 3,506 newly admitted NH residents in the U.S. with dementia reported that 40.1% received antidementia medications on admission, including 35.5% of the residents receiving ChEIs and 16.4% receiving memantine.⁶ In our study, overall ChEI use dropped from 43.2% in 2011 to 35.0% in 2018, and memantine use dropped from 27.2% to 22.3%, indicating a reversal of the upward trends seen between 2004 and 2011. Our data do not provide information on the etiology for this most recent reversal of prescribing practices for dementia patients inside and outside of the NH, and this trend deserves further investigation.

This recent decline in usage of antidementia medications may be in part related to conflicting results and interpretations of data from the clinical trials that were completed prior to 2011. In these clinical trials, ChEIs produced short term improvements in cognitive and daily living function^{1–4,29–31} as well as global functional status³² for NH residents^33–36 as well as community dwelling people with severe dementia.^37–40 Although practice guidelines^19,41–45 and expert reviews^17,20,21,46 have recommended use of ChEIs for treatment of mild to moderate dementia and memantine for moderate to severe dementia, the magnitude of the reported clinical effects of these drugs has been mixed, variably described as clinically important,⁴⁷ low,³² modest,⁴⁸ small^2,49 or possibly insignificant.⁴⁹ ChEIs are often poorly tolerated (with the possible exception of donepezil^32,50), as they may result in clinically significant cholinergic side effects that can lead to discontinuation for gastrointestinal distress, weight loss, and syncope related to bradycardia.⁴⁹ Despite these early adverse effects, antidementia drugs have been associated with reduced mortality.^{31,32,51–57}

Memantine has been shown to produce short term benefits in cognitive and daily living function, as well as global functional status, for NH residents with severe dementia,⁵⁸ as well as community dwelling residents with moderate to severe dementia,^59,60 but with minor side effects including headaches, somnolence, and constipation.^1–4 The magnitude of the reported beneficial effects has also been described as small⁴ or minimal;⁴⁹ optimal drug treatment may involve combination therapy with ChEIs to overcome this limitation.^4,61,62 Although long term placebo-controlled clinical trial data are not available for either class alone, one observational study of 382 community-dwelling AD patients found that combination therapy with a ChEI and memantine slowed cognitive and functional decline in AD compared with ChEI monotherapy and no treatment. These benefits had small-to-medium effect sizes that increased with time on treatment and were sustained for four years.^63,64

At some point during the disease progression, many AD patients exhibit behaviors that include aggression, agitation, and psychosis. While ChEIs and memantine have been reported to ameliorate dementia-related behavioral symptoms^{29,61,65–68}, meta-analyses and systematic reviews later provided little support for a significant effect of these drugs on behavior, particularly with regard to the ChEIs.^{30–32,49,61,69}

Among newly admitted NH residents with dementia, we found that various kinds of medication regimen changes were made. Admission to a NH is a major change in point of care providing opportunities to identify inappropriate prescriptions, reassess patient care plans and optimize drug regimens,⁷⁰ so it was not surprising to see changes being made on admission. The number of residents newly initiating was largely counterbalanced by the number discontinuing each medication class.

We found that older age, use of a feeding tube, and greater dependency in ADLs were associated with lower odds of ChEI initiation, while coronary artery disease, receipt of parenteral nutrition, severe aggression, cognitive impairment, and high functional dependency in ADLs were associated with discontinuation of ChEIs. A prior study of NH residents with advanced dementia found that factors associated with discontinuation of ChEIs were new admission, older age, difficulty being understood, aggressive behavior, poor appetite, weight loss, mechanically altered diet, limited prognosis designation, hospitalization in 90 days prior, and northeastern region. Factors associated with less discontinuation included memantine use, use of strong anticholinergics, polypharmacy, rurality, and primary care prescriber vs geriatric specialist.⁷¹ Avoidance of this class of medication is particularly prudent for those with persisting alimentary problems given the adverse effects profile of ChEIs that includes nausea, weight loss, and anorexia.

Overall, the prevalence of severe cognitive impairment was 1.7% lower among ChEI than non-users. An earlier report also found that users tend to be younger and less impaired than those not receiving ChEIs.⁵ Another study from the U.S. found that use of ChEIs and memantine on admission was significantly greater in residents with mild to moderately severe dementia compared to those with advanced dementia.⁶ This pattern contrasts with a previous report on long term care facilities in Canada finding that residents with severe cognitive impairment were more likely to receive dementia pharmacotherapy.⁷² Given that ChEIs are FDA-approved for the broad range of mild to severe dementia, and memantine for moderate to severe dementia, future research should explore the rationale for these observed patterns and whether it is possible to identify those who are most likely to experience greater benefit than risk from antidementia medication in the NH population, particularly among those with more advanced disease.

In a prior report from 2011, after 3 months from NH admission, ChEI and memantine use decreased by about half in both treatment groups.⁶ The smaller decline in antidementia drug usage that we found from pre- to post-NH admission likely reflects more recent changes in prescribing culture and policies in the NH setting, and is similar to a previous report from NHs in Ontario Canada in 2018 finding that less than one-fifth of residents on a ChEI at admission discontinued use during the following year.⁷³

Probably the most important finding from our study is the declining pre-NH admission use of antidementia drugs between 2011 and 2018, a trend that is further reflected in the post-NH admission period. The characteristics of NH residents with dementia were relatively stable during this period of time, and the trends were not explained by changes in resident characteristics in our adjusted models. Comparison of clinical factors related to antidementia drug treatment changes following NH admission in 2011 and 2018 revealed higher rates of discontinuation among residents who had two or more hospitalizations and lower rates of initiation of treatment among residents with greater functional dependency and residents with indicators of more complex illness such as polypharmacy or recent skilled nursing facility admission. An earlier study of Medicare beneficiaries in 2009 also found that initiation of treatment with antidementia drugs was lower in residential care and in those with more comorbidities. Neurologists and psychiatrists were more likely and geriatricians less likely to initiate treatment compared to primary care physicians;⁷⁴ NH residents may have less access to neurologists and psychiatrists. The possibility that a shift in provider vs. family perceptions over time¹² might also underlie the patterns of use of antidementia drugs in the NH should be explored by future studies.

Limitations

Our study relied on observational data during our 90 day baseline period from Medicare claims and MDS admission assessments to identify clinical conditions (Part A and the MDS) and define medication exposures (Part D). Although these administrative data sources are routinely used for research, some misclassification of diagnoses may occur when diagnostic information recorded on billing data do not accurately represent an individual’s clinical status. Recognizing that the etiology of dementia is often unclear, mixed dementias are common, and that Alzheimer’s disease may be under diagnosed, we studied all residents with dementia. Claims-based algorithms using the codes for dementia that were used in this study have been found to achieve sensitivity and specificity of 57% and 92% (without searching MDS data), respectively, in the general Medicare population when applied to three years of data.⁷⁵ Like diagnoses, prescription fill patterns may not reflect actual medication ingestion. However, since medications are staff administered in NHs, non-adherence in the NH setting is unlikely.

Our study focused on medication use in the transition of care period encompassing three months before and three months after NH admission. With this approach we can be confident that individuals with no history of use and at least one claim in the post-admission period were initiators, and that those who had a dispensing in the pre-NH period and no use in the post-NH period discontinued the drug. Further examination of longitudinal medication use patterns after NH admission is warranted. We restricted our population to newly admitted residents who survived at least 30 days (90% of residents) to ensure adequate follow-up time to ascertain medication use status after admission; to the extent these individuals had a higher or lower prevalence of antidementia medication use versus those who survived, then the prevalence of pre-NH use may have been slightly higher or lower than what we’ve reported. While we cannot infer cause and effect for the observed association of severe aggression and cognitive impairment, and high functional dependency in ADLs, with discontinuation of ChEIs, these associations raise the question of whether deprescribing could result in behavioral and functional decline for some patients vs. whether appropriate deprescribing among those with advanced disease is being done when the drugs are unlikely to be providing any clinical benefit. Lastly, the overall declining use of both drugs classes over time could reflect more effective or prevalent use of non pharmacological therapies; however, we had no data on such therapies with which to make comparisons.

Conclusions

The relatively continuous and steep decline seen for ChEI usage from 2011 to 2018 is noteworthy. Given the reported benefits of antidementia drugs, including reduced functional decline and mortality, reported in clinical trials and observational studies, these prescribing trends point to the need for more detailed outcomes research. In order to develop improved clinical guidelines, further study is required to determine the optimum use of ChEIs and memantine in NH residents according to individual risk factors and target behaviors as well as the effects of withdrawing these medications on resident outcomes such as daily functioning and mortality in this frail population.

Supplementary Material

Supplemental material

Supplemental Figure 1: Trends in the prevalence of pre- and post-nursing home admission use of ChEIs and memantine by year among residents with Alzheimer’s disease (repeated cross-sections: 2011–2018)

Supplemental Table 1: Operational definitions of clinical conditions

Supplemental Table 2. Patterns of cholinesterase inhibitors (ChEIs) and memantine use among residents with Alzheimer’s disease in the 90 days pre- and 90 days post-nursing home admission in 2011 and 2018

Supplemental Table 3: Associations of nursing home resident characteristics with initiation and discontinuation of ChEIs and memantine among residents with dementia: Results of multivariable logistic regression models

Supplemental Table 4: Associations of nursing home resident characteristics with initiation and discontinuation of ChEIs and memantine among residents with Alzheimer’s disease: Results of multivariable logistic regression models

Supplemental Table 5: Associations of nursing home resident characteristics with initiation of ChEIs and memantine in 2011 and 2018 among residents with dementia: Results of multivariable logistic regression models

Supplemental Table 6: Associations of nursing home resident characteristics with discontinuation of ChEIs and memantine in 2011 and 2018 among residents with dementia: Results of multivariable logistic regression models