FormalPara Key Points

Our research suggests that residents that initiate pimavanserin may have a lower risk for falls or fractures compared with residents treated with other atypical antipsychotics in long-term care or nursing home (LTC/NH) settings.

The results showed that pimavanserin patients had a 37% decrease in risk of all-cause falls/fractures compared with other atypical antipsychotics and a 41% decrease when compared with quetiapine only.

1 Background

Falls and fractures are one of the major causes of morbidity and mortality among older residents in long-term care (LTC) and nursing home (NH) facilities in the USA [1]. The agency for health care research and quality reports that annual rate of falls among the 1.6 million nursing facility residents is at least 50%; with 1 in 3 of those falling repeatedly in a year [2]. It is estimated that approximately 10–25% of those with falls have a serious injury such as fractures [3, 4]. Falls and fractures are known to be drivers that reduce NH residents’ functioning and quality of life, increase NH staff burden, and contribute to a growing NH crisis in the nation [5].

One of the main causes of falls among older residents in LTC/NH facilities is the increasing prevalence of chronic diseases such as Parkinson’s disease (PD) [7, 8]. Some estimates suggest that approximately 5–7% of LTC/NH residents have PD, with the numbers expected to rise in the aging population [6]. Among residents with PD, nonmotor symptoms related to psychosis (i.e., hallucinations and delusions) and cognitive disorders are not only risk factors for premature NH placement and mortality, but they have also been implicated as risk factors for falls and fractures [8]. Emerging research suggests that risk of falls and fractures is known to be higher among residents with PD psychosis (PDP) compared with those without psychosis [7, 8].

Although residents with PD who are not on PD-specific maintenance treatments are more likely to fall, residents who are adequately treated may also be prone to falls due to adverse effects brought on by antipsychotics or other medications that may impair cognition (e.g., narcotics, benzodiazepines, hypnotics or sleeping agents, anticholinergics, muscle relaxants) [9]. AGS Beers criteria and other studies suggest that off-label atypical antipsychotics commonly prescribed to treat neuropsychiatric symptoms such as hallucinations and delusions may also elevate the risk of falls and fractures, potentially due to syncope, sedation, slowed reflexes, loss of balance, and impaired psychomotor function [10, 11]. To date, few studies have examined the risks of falls and fractures among residents treated with different atypical antipsychotics (AAPs) in the LTC/NH setting.

Currently, pimavanserin is the only Food and Drug Administration (FDA) approved AAPs for treating hallucinations and delusions associated with PDP. Previously published research suggested that other AAPs (e.g., quetiapine, risperidone, etc.) continue to be prescribed off-label among residents with PDP, especially in LTC/NH settings [12]. Since reducing falls and fractures remains an important clinical goal in managing older residents in LTC/NH settings, research comparing the risk of falls or fractures among PDP residents treated with different AAPs in such settings is needed. Therefore, a database analysis of Medicare beneficiaries in LTC/NH settings was conducted to examine the risk of falls or fractures among PDP residents treated with pimavanserin (PIM) versus (i) other AAPs [quetiapine (QUE), risperidone (RIS), olanzapine (OLA), aripiprazole (ARI)] and (ii) QUE only.

2 Methods

2.1 Data Source and Study Population

A retrospective analysis of data from the Center for Medicare and Medicaid Services (CMS) 100% Medicare fee-for-service (FFS) claims of PDP residents in long-term care/nursing home (LTC/NH) facilities was conducted. The Medicare program is the primary public health insurer for all US residents aged 65 years and above, as well as selected individuals under 65 years with disabilities and end-stage renal disease. As of 2019, the 100% Medicare sample represented all 63 million US Medicare beneficiaries, encompassing claims for medical services from inpatient (Part A) and outpatient (Part B), and pharmacy services (Part D). It is important to note that Medicare parts A, B, and D cover hospital insurance, medical insurance, and pharmacy insurance, respectively. Medicare FFS is the traditional Medicare program, where healthcare providers are reimbursed based on a fee schedule for each service provided to the beneficiary, determined by the Medicare annual fee schedule.

For this study, eligible sample of residents in LTC/NH setting within the PDP population were identified using a combination of parts A, B, and D claims from 1 January 2013 to 31 December 2019. The PDP population included residents identified from ≥ 1 ICD-9 and ICD-10 diagnostic claim of 332.0 and G20, respectively, for PD with a concurrent psychosis diagnosis (occurrence of ≥ 1 psychosis or psychotic disorder diagnostic claim: F06.0, F06.2, F22, F23, F28, F29, H53.16, R44.0, R44.1, R44.2, R44.3) (see Supplementary Table 1). This analysis was conducted in accordance with the CMS data use agreement that was established after New England Institutional Review Board review and approval.

2.2 Study Design and Population

Residents with PDP in the LTC/NH setting who were prescribed and maintained continuously for at least 6 months on PIM monotherapy, other AAPs monotherapy, or QUE-only monotherapy continuously (post-index) between January 2014 and December 2018 formed the final study population. In addition to the comparative cohort of PIM versus “other AAPs” cohort, a “QUE only” cohort was used as a second comparator group as a secondary analysis. This analysis was conducted in compliance with HIPAA under a CMS data use agreement that was established pursuant to a New England Institutional Review Board review and approval. The index date among the eligible sample was identified as the date of first prescription for PIM, other AAPs, or QUE only within LTC/NH settings. Only residents without prior use of any AAPs or PIM for at least 6 months prior to the index date were included (Supplementary Fig. 1). All residents with PDP with a pre-index diagnosis of psychosis, secondary parkinsonism, delirium, other psychotic disorders, alcohol/drug-induced psychosis, schizophrenia, paranoia, or personality disorders were also excluded from the study population. A list of diagnosis codes used for the inclusion/exclusion criteria are given in Supplementary Table 1. Study participant selection process is provided in Fig. 1.

Fig. 1.
figure 1

Study Population Selection. PD Parkinson’s disease, PDP Parkinson’s disease psychosis, AAPs atypical antipsychotics, PIM pimavanserin. *Diagnosis of secondary parkinsonism, delirium, other psychotic disorder, alcohol/drug-induced psychosis, schizophrenia, paranoia, or personality disorders. †Residents treated with other AAPs (n = 6,182) i.e., risperidone (n = 668), olanzapine (n = 316), aripiprazole (n = 188), and quetiapine (n = 5010)

Full size image

2.3 Covariates

Pre-index baseline characteristics including age, sex, race or ethnicity, region, clinical comorbidities, and coexisting insomnia or dementia status were examined during 6 months prior to the index date. Clinical comorbidities were evaluated using the Elixhauser comorbidity indexes during the pre-index period.

2.4 Outcomes

2.4.1 Falls, Fractures, and Falls/Fractures

Falls were identified from ICD-9 or ICD-10 diagnostic claims for at least any one of the following during the 6-month follow-up: falls on the same level, falls from different level, or unspecified falls for any other reason. Fractures were defined using ICD-9 or ICD-10 diagnostic claims for hip fractures, pelvic, or femur fractures. Additionally, a composite outcome of either falls or fractures were also analyzed. The specific ICD-9 or ICD-10 diagnostic codes used to define these outcomes are explained in detail in Supplementary Table 2.

2.5 Propensity Score Matching (PSM)

Residents initiating PIM versus other AAPs or versus QUE only were propensity score matched using a 1:1 ratio in a pairwise fashion to create a balanced sample between the cohorts. Propensity scores were calculated using multivariable logistic regression on resident age, sex, race, region, and 27 of 31 Elixhauser comorbidity characteristics. Four Elixhauser comorbidities such as psychosis, human immunodeficiency virus (HIV), alcohol abuse, and substance abuse were not used in propensity score matching. These 31 variables were chosen as covariates in the propensity score matching as they were hypothesized to be potential confounders based on clinical and statistical plausibility to be associated with the falls/fracture outcomes. Two other concurrent conditions, insomnia and dementia, were used as controlling variables in the regression analysis of outcomes. Residents with psychosis in the pre-index (i.e., baseline) were excluded in this analysis and data for residents with HIV, alcohol abuse, and substance use may be suppressed by CMS to accommodate resident confidentiality and would not have allowed an appropriate method of matching [13,14,15,16]. A greedy nearest neighbor matching algorithm was used for matching as well as other matching details are described elsewhere [17]. Covariate balance was assessed using standardized mean differences (SMDs) of < 0.1 between PIM and other AAPs beneficiaries for each of the comorbidities. All missing data were excluded before matching and the final, matched sample had no missing data. The same methodology was repeated for the secondary analysis of PIM versus QUE.

2.6 Statistical Analysis

Baseline resident demographics (e.g., sex, age) and clinical characteristics (e.g., Elixhauser comorbidities and other comorbidities such as dementia or insomnia) among residents in matched PIM versus other AAPs (or QUE-only) cohort were described using frequencies and proportions after propensity score matching. Additionally, proportion of residents on specific AAPs such as QUE, RIS, OLA, and ARI in the other AAPs cohort before and after matching were examined. Descriptive statistics were reported as frequencies and percentages for categorical variables, and mean, median, and range for continuous variables. Chi-square tests (categorical measures), t-tests, and Wilcoxon rank sum tests (continuous measures) were used to describe differences in outcomes associated with PIM versus other AAPs.

Risk of falls/fractures among PIM versus other AAPs and PIM versus QUE residents (the secondary analysis) were compared using log binomial regression models, controlling for age, gender, region, race, coexisting insomnia, or coexisting dementia. Relative risk (RR) ratios and 95% confidence intervals are reported for 6-months follow-up. Unless otherwise specified, the statistical significance was set to a threshold of p < 0.05. All analysis were conducted using SAS Enterprise Server via the CMS Virtual Research Data Center.

3 Results

Of the 7187 residents who initiated continuous monotherapy of AAPs with 6-month follow-up, 14% (n = 1005) were prescribed PIM, 86% (n = 6182) were prescribed other AAPs [QUE, 70% (n = 5010); RIS, 9.29% (n = 668); OLA, 4.40% (n = 316); ARI, 2.62% (n = 188)]. The mean (SD) age for the study cohort was 78.8 (± 7.75) years and 47.59% (n = 3420) were female. The mean age across the two cohorts (PIM, other AAPs) was similar and rates of males and females were comparable. Resident characteristics for PIM versus other AAPs cohort in a matched cohort are described in Table 1, while the clinical comorbidity profile for PIM versus other AAPs cohort after matching is described in Table 2. Prior to matching, rates of Elixhauser comorbidities between PIM and the other AAPs cohorts were significantly different; however, after 1:1 matching, rates between the cohorts were largely balanced. The proportion of patients with insomnia or dementia were significantly higher in the other AAP or QUE cohorts.

Table 1 Resident Demographics, PIM versus Other AAPs (Matched to PIM) Cohort
Full size table
Table 2 Baseline Comorbidities, Matched PIM versus Other AAPs
Full size table

Figure 2 shows the risk of falls or fractures for PIM versus other AAPs. Among the matched cohort, the risk for fractures only were 1.39% (n = 14) for PIM compared with 2.09% (n = 21) for other AAPs. PIM residents also experienced lower risk of falls only [4.58% (n = 46)] compared with other AAPs residents [7.66% (n = 77)] [RR = 0.60 (0.42, 0.85), p < 0.05]. Risk for falls or fractures among PIM residents was 5.67% (n = 57) versus 9.05% (n = 91) for other AAPs [RR = 0.63 (0.46, 0.86), p < 0.05] (see Fig. 2 and Table 3).

Fig. 2
figure 2

Risk of Falls/Fractures: PIM versus Other AAPs. AAPs atypical antipsychotics, NS not significant, PIM pimavanserin

Full size image
Table 3 Risks of falls only, fractures only, and falls or fractures among PIM versus other AAP-treated residents at 6-months follow-up
Full size table

A secondary analysis of the matched PIM and QUE cohorts (Fig. 3) showed a nonsignificant difference in fracture risk between the cohorts. PIM residents had a 1.39% fracture risk, while those who received QUE had a 1.89% risk (p-value: ns). Risk for falls only were significantly lower for PIM residents (4.6%) compared with 8.26% for QUE [RR = 0.55 (0.39, 0.79), p < 0.05]. The proportion of residents with falls or fractures was lower in the PIM group (5.67% versus 9.55%) compared with the QUE group [RR = 0.59 (0.43, 0.81), p < 0.05] (see Fig. 3 and Table 3).

Fig. 3
figure 3

Risk of Falls/Fractures: PIM versus QUE. NS not significant, PIM pimavanserin, QUE quetiapine

Full size image

4 Discussion

While published research before the advent of pimavanserin implicated AAPs medication use with an increased risk of falls and fractures among older nursing home residents, not many studies have examined the association of pimavanserin and other AAPs use for falls or fractures among residents with PDP in nursing home residents [18, 19]. The current analysis adds to the body of published literature comparing the risk of fall/fracture outcomes among PDP residents treated with PIM versus other AAPs and PIM versus QUE in an LTC/NH setting.

Previous research had also suggested that all AAPs may have higher, yet similar, rates of falls or fractures compared with nonusers [20]. Recently published claims data analysis of a population with PDP that have either commercial insurance or Medicare advantage coverage suggests that pimavanserin demonstrated no increase in the risk of falls or fractures compared with other AAPs [21]. While nursing home residents may have been included in the study, they were not separately discussed or analyzed. Our research suggests that residents that initiate PIM may have a lower risk for falls or fractures compared with residents treated with other AAPs in LTC/NH setting. A few plausible reasons may explain why PIM may have lower risks of falls or fall related injuries such as fractures. First, it is widely acknowledged that other AAPs such as olanzapine, risperidone, and quetiapine cause somnolence, postural hypotension, and motor and sensory instability, which may lead to falls and, consequently, fractures or other injuries. On the other hand, patients treated with pimavanserin in placebo-controlled studies had low rates of orthostatic hypotension or sedation-related adverse events [22]. While these are not causal in nature, low rates of orthostatic hypotension or sedation may be associated with lower rates of falls/fractures observed with PIM. Second, it should be noted that in a phase 3, 6-week, double-blind randomized controlled trial (RCT), PIM significantly reduced the frequency and/or severity of hallucination and/or delusions associated with Parkinson’s disease psychosis versus placebo, without impairing motor function [23]. These effects may be especially critical in the context of treating residents with PDP given their motor impairment associated with PD symptoms without the potential for increasing falls and associated fractures.

While we examined a balanced sample of residents to the extent possible, it cannot be ruled out that residents with potential concomitant use of antidepressants, benzodiazepines, acetylcholinesterase inhibitors, memantine, selected antihypertensives, and other drugs with potentially higher risk of falls may also have played a role in the differential risk of falls or fractures that was observed in this analysis. It should be noted that in this analysis, the study did not control for prior falls/fractures since controlling for the same baseline variables in the causal pathway can hide true differences in follow-up outcomes and bias the results. Since this analysis assessed 6-month risk of falls and/or fractures among continuously enrolled patients with least 6-month continuous index monotherapy, patients who discontinue monotherapy, switched, or added any AAPs during the 6-month post-index period were excluded from analysis. Thus, the study had none that switched from one AAP medication to another or added another atypical antipsychotic during the follow-up period.

In the US, falls and fall-related injuries such as fractures constitute approximately 6% of total Medicare spending annually [24]. Falls and fractures have been known to contribute to greater functional impairment and increased physical and mental disabilities. Additionally, they may result in premature deaths; in 2017, over 28,000 fall-related deaths were identified among the elderly in USA [25]. From 2004 to 2017, the rate of fatal fall injuries increased 35% [26]. The risk of mortality within 1 year after experiencing a hip fracture is estimated to be around 40% among residents of nursing homes [27]. Several studies including systematic literature reviews have reported the increased risk of falls and fall-related fractures among nursing home residents treated with psychotropic medications such as AAPs in the last few decades [7,8,9,10,11]. However, few studies have reported the fall risk differences between specific antipsychotics to help clinicians in making evidence-based prescribing decisions for NH residents. It is widely acknowledged that inappropriate use of AAPs among NH residents may be one of the culprits in higher rates of falls, fractures, cerebrovascular events such as stroke, and ultimately death [10, 11]. The falls prevention program instituted by CMS has implemented strategies to limit the use of AAPs through onerous documentation measures and gradual dose reduction aimed toward dose tapering, among others [28]. Our study primarily focused on falls and fractures to hip, femur, and pelvic regions as these were the most detrimental (fall-related) fractures in the elderly. Hip fractures in the elderly not only affects mobility but also result in high mortality [29]. Therefore, arms, wrist, fingers, or other bone-related fractures that may result from falls are not reported in this analysis. Previous research that utilized claims-based algorithm for identifying the incidence of fractures were based on claims data integrated with electronic medical records; our analysis relied on ICD-9/10 diagnostic codes for identification of falls or fractures since our data source included only claims. Notwithstanding the potential for underestimating falls in diagnostic claims, this analysis provides a valid measure of assessing the incidence of falls [30]. While much research is needed, we believe this is the first analysis to develop an evidence base to examine the fall-risk level differences among residents treated with specific AAPs such as PIM or QUE. These results may provide important insights regarding the clinical value of specific AAPs as it pertains to fall risk and help prescribers in evidence-based prescription decision-making. Future investigations examining the association between characteristics and risk of falls/fractures may be needed.

5 Limitations

As with any claims analysis, this research has a few limitations. First, sampling errors due to miscoding or under-coding of claims may have occurred in this analysis of claims data that are primarily used for billing and reimbursement purposes. To the extent the coding errors can be minimized, the use of a second confirmatory diagnostic claim to identify resident sample may have reduced potential sampling errors due to coding issues. Second, the study may have resulted in biased estimates due to the absence of variables such as medical/clinical characteristics, socio economic characteristics or other factors in claims data. Third, the PIM cohort had lower obesity, higher diabetes, and higher hypertension rates (Table 2) even after PSM, suggesting residual imbalance may exist. Fourth, baseline or prior risk of falls and fractures were not included as a covariate in the adjusted analysis. Fifth, while differential usage of various antihypertensives at differing dose ranges may play a contributing role in falls and fractures, granular details about the intensity of these agents were outside of scope of this research. Finally, our analysis could not control for underlying PD or PDP severity since severity cannot be defined in claims data. Our study primarily focused on fall and fractures to hip, femur, and pelvic regions as these were the most common fall-related fractures in the elderly. Since only major fractures were considered, this analysis might not capture the full picture of other fractures resulting from falls. Fractures of the wrist, fingers, or other bones that may result from falls could be underestimated. While our study examined the 6-month risk differences among matched cohort of LTC/NH residents receiving either PIM or other AAPs (or QUE) for PDP, residual confounding due to unobserved variables (e.g., medications, eyesight related problems, environmental hazards, etc.) that can increase the risk of falls or fractures cannot be ruled out. While this analysis included propensity score-matched cohorts, it is plausible that residual confounding may exist due to variables not included in the analysis. Notwithstanding this limitation, these results add to the body of literature in describing the potential lower risk of falls or fractures with PIM treatment versus other AAP and provide guidance about the choice of appropriate AAPs to treat PDPs to clinical and formulary decisionmakers.

It should be noted that the other AAPs including QUE are known to increase the risk of falls and consequently fall-related fractures due to several reasons including increased sedation, exacerbation of mild cognitive problems, and extrapyramidal symptoms, among others [31]. While this study did not examine the predictors of falls or fractures in this vulnerable population, future investigations examining the predictors of falls in PDP patients are warranted.

6 Conclusions

In this analysis of residents with PDP that have an LTC/NH stay from the 100% Medicare sample, PIM monotherapy resulted in nearly a 40% lower risk of falls only and falls/fractures compared with other AAPs or QUE. Risks of fractures only were no different between PIM versus other AAP or QUE residents. Overall, the lower risk of falls/fractures observed among PIM monotherapy residents in this analysis may be suggestive of a more favorable clinical benefit risk profile with PIM compared with other AAPs including QUE in LTC/NH facilities. Future studies examining other predictive risk factors for falls or fractures among LTC/NH residents receiving PIM versus other AAPs may provide additional insights and guidance to clinicians.