Health care resource utilization patterns among patients with Parkinson’s disease psychosis: analysis of Medicare beneficiaries treated with pimavanserin or other-atypical antipsychotics

Abstract

Background

Pimavanserin (PIM) is the only FDA-approved atypical antipsychotic (AAP) for hallucinations and delusions associated with Parkinson’s disease psychosis (PDP). Comparative real-world analyses demonstrating its benefits are needed.

Objectives

To evaluate health care resource utilization (HCRU) outcomes among PDP patients treated with PIM vs. other-AAPs.

Methods

Retrospective cohort analysis of Parts A, B, and D claims from 100% Medicare sample from 01 January 2013–31 December 2019 was conducted. PDP Patients initiating (i.e. index date) continuous monotherapy (PIM vs. other-AAPs) for ≥12-months during 01 January 2014–31 December 2018 without 12-months pre-index AAP use were selected after 1:1 propensity score matching (PSM) on 31 variables (sex, race, region, age, and 27 Elixhauser comorbidities). HCRU outcomes included: annual all-cause and psychiatric hospitalization (short-term stay, long-term stay, and SNF-stay [skilled nursing facility]) rates, annual all-cause and psychiatric-ER visit rates, mean per-patient-per-year (PPPY) hospitalizations, and average length of stay (ALOS). PIM and other-AAPs were compared using generalized linear models (GLM) controlled for demographic characteristics, comorbidities, coexisting-dementia, and coexisting insomnia.

Results

Of 12,164 PDP patients, 48.41% (n = 5,889) were female, and mean age was 77 (±8.14) years. Among 1:1 matched patients (n = 842 in each), 37.8% (n = 319) on PIM vs. 49.8% (n = 420) on other-AAPs (p < .05) reported ≥1 all-cause hospitalizations, respectively. Specifically, short-term and SNF-stay among PIM patients vs. other-AAPs were: 34% (n = 286) vs. 46.2% (n = 389) and 20.2% (n = 170) vs. 31.8% (n = 267) (p < .05), respectively. Similarly, 9.6% (n = 81) of PIM vs. 14.6% (n = 123) of other-AAPs patients had ≥1 psychiatric hospitalization (p < .05). Furthermore, ≥1 all-cause and psychiatric ER visit among PIM vs. other-AAPs were 61.6% (n = 519) vs. 69.4% (n = 584) and 5.2% (n = 43) vs. 10.2% (n = 86) (p < .05), respectively. PIM also had significantly lower ALOS, and mean PPPY short-term hospitalization and SNF-stays.

Conclusions

In this analysis of PDP patients, PIM monotherapy resulted in nearly 12% and 7% lower all-cause hospitalizations and ER visits vs. other-AAPs.

Keywords:

• Parkinson’s disease
• psychosis
• pimavanserin
• atypical antipsychotics
• HCRU
• health care resource utilization
• hospitalizations

JEL CLASSIFICATION CODES:

• C55
• C5
• C
• D87
• D8
• D
• H51
• H5
• H
• J14
• J

View correction statement:

Correction

Introduction

Approximately, one million people in the US are anticipated to have Parkinson’s disease (PD), a chronic and progressive neurological condition; its estimated prevalence is anticipated to rise to 1.6 million by 2037.^1,2 Approximately, 25%–40% of PD patients are known to develop Parkinson’s disease psychosis (PDP), a condition characterized by symptoms of hallucinations and delusions over its progressively degenerative disease course.^1,3 While the pathophysiological basis for PDP is unclear, both exogenous and endogenous factors may play a role in the incidence of PDP. It is believed that the complex interplay between neurotransmitters including dopamine, acetylcholine, and serotonin may play a role in the occurrence of psychosis among these patients, potentially by contributing to dopaminergic activity in the pathways to limbic system.^4–7 Additionally, it is also hypothesized that patients with PD may have an elevated risk for PDP due to deposition of Lewy bodies which result in abnormalities in dopamine, serotonin, and glutamate neurotransmission causing neuronal destruction throughout various parts of the brain.² Literature suggests that symptoms of hallucinations and delusions associated with PDP can resemble other conditions such as delirium, schizophrenia or other psychiatric disorders, infections, metabolic abnormalities, and medications.^4,8,9

PDP is known to compound the underlying burden of PD, often resulting in greater hospitalizations, ER visits, and accelerated nursing home placement.^10,11 In addition, PDP imposes a significant burden to caregivers and family members.¹¹ In a Medicare survey of claims data from 2000 to 2010, patients with PDP had significantly higher resource utilization; incurring approximately 116%, 68%, and 219% higher LTC costs, inpatient costs, and nursing facility costs, respectively, when compared to PD patients without psychosis.¹⁰ Additionally, in a study of PDP patients in Olmsted County, Minnesota, patients with PDP had a 14% (HR = 1.14, p = .005) greater risk for hospital admission compared to PD patients without psychosis.¹² In another study of Medicare patients, psychosis among PD patients was associated with a more than three-fold increased risk of custodial care and a nearly one-third increased risk of death.¹³ In addition to increased health care utilization, PDP markedly impacts the quality of life of patients and their caregivers.^14,15

Pimavanserin (PIM), an atypical antipsychotic (AAP) was approved by US Food and Drug Administration (FDA) in 2016 for the treatment of delusions and hallucinations associated with PDP.¹⁶ PIM is a selective 5-HT_2A inverse agonist/antagonist that also binds to a lesser extent to 5-HT_2c receptors, but has no binding affinity for dopaminergic, adrenergic, muscarinic, or histaminergic receptors.⁴ While PIM remains the only FDA-approved PDP treatment to date, other antipsychotics (i.e. risperidone, olanzapine, etc.) are used off-label to treat PDP despite an unfavorable benefit-risk profile related to increased risk of stroke, mortality, weight gain, and extrapyramidal symptoms, among others.^17,18 Movement Disorder Society (MDS) commissioned evidence-based review of treatments for non-motor symptoms, published in 2019, suggest that clozapine and quetiapine (QUE) may be “possibly useful” for treating psychotic symptoms despite lack of demonstrable Class I evidence. Although not FDA approved and used off-label, clozapine and quetiapine have demonstrated moderate, yet inconsistent, effects in improving hallucinations and delusions associated with PDP. In contrast, the recommendations indicate that PIM is clinically useful for treating psychosis symptoms among patients with PDP.¹⁹ PIM has demonstrated effectiveness in improving psychotic symptoms of PDP, is well-tolerated and displays no worsening of motor symptoms. The safety and efficacy of PIM was demonstrated in a six-week pivotal phase 3, randomized, double-blind, placebo-controlled clinical trial.^8,9 Evidence-based reviews by expert panels of the MDS and the American Geriatrics Society also have recommended the use of PIM for the treatment of PDP symptoms.^4,20

While PIM’s effects on symptoms of hallucinations and delusions have been established from clinical trials, real-world assessment of health care resource utilization (HCRU) associated with use of PIM are needed, particularly in comparison to other off-label AAPs. Therefore, the primary objective of this analysis was to examine HCRU patterns such as hospitalizations and emergency room (ER) visits among PDP patients treated with PIM versus off-label AAPs.

Materials and methods

Study design

A retrospective cohort analysis of PDP patients newly initiating PIM or other-AAPs was conducted using Parts A, B, and D claims data from the Centers for Medicare and Medicaid Services (CMS) 100% Medicare sample of fee-for-service (FFS) beneficiaries. 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’s review and approval.

Patient population

Patients with at least one International Classification of Disease, 9th version (ICD-9, 332.0) or 10th version (ICD-10, G20) for PD along with occurrence of one or more psychosis or psychotic disorder diagnostic claim (Supplementary Table 1) from January 2013 to December 2019 were identified. Since adjudicated claims data generally have a lag of nearly a year to be made available for research; Parts A, B, and D claims from the 100% Medicare sample was available only till 31 December 2019 at the time of this analysis.

Of these patients, the eligible study sample included treatment-naive PDP patients on PIM or other-AAP continuous monotherapy (i.e. quetiapine, risperidone, olanzapine, and aripiprazole) between January 2014 to December 2018. Date of first prescription for PIM or “other-AAPs” formed the index date for the two study groups (Figure 1). Only patients without prior use of any AAP for 12 months (i.e. pre-index) and ≥12 months post-index follow-up were included. All PDP patients with a pre-index diagnosis of psychosis, secondary parkinsonism due to any reasons, delirium, other psychotic disorders, alcohol/drug-induced psychosis, schizophrenia, paranoia, or personality disorders were also excluded from the study population. List of diagnostic codes used to describe study population selection are listed in Supplementary Table 1. Details describing patient selection and patient disposition chart is provided in Figure 2.

Figure 1. Study schema. Abbreviations. PDP, Parkinson's Disease Psychosis; AAPs, atypical antipsychotics.

Figure 2. Patient selection. Abbreviations. PD, Parkinson's Disease; PDP, Parkinson's Disease Psychosis; PIM, pimavanserin; AAP, atypical anti-psychotics. *Diagnosis of secondary parkinsonism, delirium, other psychotic disorder, alcohol/drug-induced psychosis, schizophrenia, paranoia, or personality disorders. †Other-AAPs (Matched to PIM) group included quetiapine (n = 639), aripiprazole (n = 34), olanzapine (n = 61), and risperidone (n = 108).

Study outcomes

Patient characteristics were described (i.e. 12-month pre-index) in terms of baseline demographics, clinical comorbidities, concomitant movement disorder medication (i.e. levodopa, carbidopa, levodopa/carbidopa) use, and coexisting insomnia or dementia status. Baseline demographic characteristics included age, sex, race or ethnicity, region that were identified and evaluated prior to index date. Clinical comorbidities were evaluated using the Elixhauser comorbidity index score as well as individual comorbidities during pre-index period. Concomitant movement disorder medication use patterns, and breakdown of AAPs in the other-AAP cohort were described for 12 months follow-up. HCRU outcomes related to ≥1 total inpatient hospital admissions and ER visits for all-cause and psychiatric-related causes were analyzed. All-cause and psychiatric-related (i.e. ≥1) inpatient hospital admissions were also analyzed broken down by hospital-stay type based on provider/facility characteristics. It should be noted that in Medicare claims, inpatient hospital admissions are defined based on provider type/facility characteristics (i.e. facilities characterized by allowable length of stay) and reported as either short-term stay (ST-stay), long-term care stay (LTC-stay) or skilled nursing facility stay (SNF-stay). ST-stays are hospitalizations in a facility/hospital that provides care to patients with needs that require an acute or critical setting following surgery, sudden sickness, injury, or flare-up of a chronic sickness while LTC-stays are hospitalizations in certified long-term acute care hospitals (LTACHs) who, on average, may stay more than 25 days. Patients typically are transferred from intensive care units (ICUs) to LTACHs. On the other hand, SNF-stays are typically hospitalizations that are longer than LTC-stays and may house patients for up to 100 days. In our analysis, long-term care admission type (LTCA) was examined as a composite of any LTC-stay or SNF-stay, not including ST-stays.

All-cause ER visits was defined as an ER admission for any diagnosis and psychiatric-related ER visits was defined as an ER admission for any one of the psychotic disorders described in Supplementary Table 1.

Propensity score matching (PSM)

To create a balanced sample on measured characteristics, patients initiating PIM or other-AAPs were propensity score-matched in a 1:1 ratio. Propensity scores were calculated using multivariate logistic regression on patient age, sex, race, region, and 27 of the Elixhauser comorbidity characteristics. Four (psychosis, HIV, alcohol-abuse and substance-abuse) of the 31 Elixhauser comorbidities were not used in propensity score matching. Patients with psychosis with pre-index were excluded from this analysis and therefore not applicable for matching. Additionally, data for patients with HIV, alcohol-abuse and substance use may be suppressed by CMS to accommodate patient confidentiality and would not have allowed an appropriate method of matching.^21–24 These four comorbidities were not used in propensity score matching since patients with psychosis in the pre-index (i.e. baseline) were excluded in this analysis. An 8:1 Digit Match, a greedy nearest neighbor matching algorithm was used for matching. The algorithm first matches PIM cases to other-AAPs on eight digits of the propensity score. Among the unmatched, PIM are then matched to other-AAPs on seven digits of the propensity score, sequentially proceeding to the lowest digit (one digit) match on propensity score, until no more matches can be made.²⁵ Covariate balance were assessed using standardized mean differences (SMDs) value of <0.1 between PIM and other-AAP beneficiaries. All missing data were excluded prior to matching and the final matched sample had no missing data.

Statistical analysis

All analyses were conducted using SAS Enterprise Server via the CMS Virtual Research Data Center. Baseline patient demographics and clinical characteristics such as sex, race/ethnicity, region, Elixhauser comorbidities, other comorbidity status (i.e. dementia or insomnia) and baseline movement disorder medication use were described using frequencies and proportions before and after propensity score matching. Additionally, proportion of patients on quetiapine, risperidone, olanzapine, and aripiprazole among other-AAPs before and after matching were examined. Descriptive statistics were reported as frequencies and percentages for categorical variables; 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. Percentage of ≥1 all-cause and psychiatric-related inpatient hospital admission stays (including ST-stay, LTC-stay, or SNF-stay), ≥1 all-cause and psychiatric-related long-term care (LTCA) admissions (i.e. defined as a composite of LTC-stay or SNF-stay) were estimated as proportions. The percentage of ≥1 ER visits and mean per-patient-per-year (PPPY) ER visits for both all-cause and psychiatric-related reasons were also reported. Length of inpatient stay, mean and median length of stay (LOS) for inpatient hospital admissions, overall and by stay type, were also assessed.

Additionally, differences in all-cause and psychiatric admissions, including ST-stay, LTC-stay, and SNF-stay between PIM and other-AAP cohorts were analyzed using logistic regression from a class of generalized linear models (GLM) with binomial logit link while controlling for demographic characteristics, comorbidities, baseline dementia, and baseline insomnia. Finally, standardized mean PPPY rates of all-cause and psychiatric-hospital admissions and ER visits were assessed. Unless otherwise specified, all p values were set to a threshold of p < .05.

Results

There were 12,164 eligible PDP patients identified; approximately 48.41% (n = 5,889) were female and mean age was 77.8 (±8.14) years. Figure 2 shows the attrition of the study population and Tables 1 and 2 outline the demographic and baseline characteristics of eligible PDP patients with at least one year of PIM or other-AAP continuous monotherapy before and after matching. Of 12,164 patients, 842 and 8,810 patients were treated with PIM and other-AAPs, respectively. Prior to matching, 80.8%, 10.4%, 5.1%, and 3.8% of patients were treated with quetiapine, risperidone, olanzapine, and aripiprazole, respectively, in the other-AAP cohort. After 1:1 propensity score matching, gender, mean, and median age were similar in both PIM (n = 842) and other-AAP (n = 842) cohorts. However, the PIM cohort had lower rates of patients with coexisting insomnia or dementia and higher comorbidity scores compared to the other-AAP cohort. Demographic and other baseline characteristics for PIM and other-AAPs in the matched sample are described in Tables 1 and 2. After matching, there were 75.89% (n = 639), 12.83% (n = 108), 7.24% (n = 61), and 4.04% (n = 34) patients treated with quetiapine, risperidone olanzapine, and aripiprazole, respectively, in the other-AAP cohort.

Table 1. Baseline patient demographics among pre-matched and post-matched pimavanserin and other-atypical antipsychotics.

	PIM (n = 842; 8.7%)	Other-AAPs (n = 8,810; 91.3%)	p Value	SMD	PIM (n = 842; 50%)	Other-AAPs (n = 842; 50%)	p Value	SMD
Age in years
Mean (SD)	77.37 (7.20)	78.12 (7.89)	.0043	0.100	77.36 (7.20)	77.52 (7.22)	.65	0.022
Median (IQR)	77 (73, 82)	78 (73, 84)			77 (73, 82)	78 (73, 82)
Minimum, maximum	53, 98	41, 106			53, 98	45, 98
<65, n (%)	24 (2.85%)	339 (3.85%)			24 (2.85%)	26 (3.09%)
65–70, n (%)	124 (14.73%)	1,065 (12.09%)			124 (14.73%)	109 (12.95%)
71–75, n (%)	188 (22.33%)	1,788 (20.30%)			188 (22.33%)	190 (22.57%)
76–80, n (%)	238 (28.27%)	2,190 (24.86%)			238 (28.27%)	238 (28.27%)
>80, n (%)	268 (31.83%)	3,428 (38.91)			268 (31.83%)	279 (33.14%)
Gender, n (%)			.63	0.018			.5909	0.029
Female	396 (47.03%)	4,224 (47.95%)			396 (47.03%)	384 (45.61%)
Male	446 (52.97%)	4,585 (52.05%)			446 (52.97%)	458 (54.39%)
Race, n (%)
Unknown	11 (1.31%)	73 (0.83%)	.1538	0.047	11 (1.31%)	^a	.0062	0.142
White	758 (90.02%)	7,801 (88.55%)	.1964	0.048	758 (90.02%)	760 (90.26%)	.8701	0.008
Black	28 (3.33%)	406 (4.61%)	.0861	0.066	28 (3.33%)	32 (3.8%)	.5990	0.026
Other	17 (2.02%)	127 (1.44%)	.1867	0.044	17 (2.02%)	15 (1.78%)	.7211	0.017
Asian	14 (1.66%)	185 (2.10%)	.3937	0.032	14 (1.66%)	16 (1.9%)	.7125	0.018
Hispanic	^a	187 (2.12%)	.0384	0.084	^a	^a	.1138	0.077
American native	^a	31 (0.35%)	.2386	0.035	^a	^a	.2179	0.08
Geographic region, n (%)
Midwest	193 (22.92%)	2,149 (24.39%)	.3414	0.035	193 (22.92%)	112 (13.3%)	<.0001	0.252
Northeast	165 (19.60%)	1,796 (20.39%)	.5864	0.020	165 (19.6%)	405 (48.1%)	<.0001	0.632
South	326 (38.72%)	3,472 (39.41%)	.6944	0.014	326 (38.72%)	255 (30.29%)	.0003	0.178
West	158 (18.76%)	1,393 (15.81%)	.0258	0.078	158 (18.76%)	70 (8.31%)	<.0001	0.309
Elixhauser comorbidity score
Median (IQR)	4 (2, 6)	4 (3, 6)			4 (2, 6)	3 (2, 5)
Mean (SD)	4.12 (2.47)	4.71 (2.86)	<.0001		4.12 (2.47)	3.79 (2.55)	.0070
0	304 (36.10%)	1,722 (19.55%)			304 (36.10%)	217 (25.77%)
1	59 (7.01%)	648 (7.36%)			59 (7.01%)	116 (13.78%)
2–4	285 (33.85%)	3,293 (37.38%)			285 (33.85%)	311 (36.94%)
4–9	175 (20.78%)	2,643 (30%)			175 (20.78%)	175 (20.78%)
≥9	19 (2.26%)	504 (5.72%)			19 (2.26%)	23 (2.73%)
Movement disorder medication, n (%)
Carbidopa	^a	60 (0.68%)	.7759		^a	^a	–
Levodopa	^a	^a	^a		^a	^a	–
Carbidopa/levodopa	737 (87.52%)	6,481 (73.56%)	<.0001		737 (87.52%)	616 (73.16%)	<.0001
Other comorbidities, n (%)
Dementia	600 (71.26%)	7,476 (84.86%)	<.0001		600 (71.26%)	716 (85.04%)	<.0001
Insomnia	246 (29.22%)	3,291 (37.36%)	<.0001		246 (29.22%)	308 (36.58%)	.0015

Abbreviations. AAPs, atypical antipsychotics; SD, standard deviations; IQR, interquartile range; PIM, pimavanserin; SMD, standardized mean differences.

^aCells with sample size of <11 are suppressed per CMS requirements.

Table 2. Baseline patient comorbidities among pre-matched and post-matched pimavanserin and other-atypical antipsychotics.

Comorbidities, n (%)	PIM	Other-AAPs	p Value	SMD	PIM	Other-AAP	p Value	SMD
	(n = 842; 8.7%)	(n = 8,810; 91.3%)			(n = 842; 50%)	(n = 842; 50%)
Congestive heart failure	59 (7.01%)	1,084 (12.3%)	<.0001	0.180	59 (7.01%)	57 (6.77%)	.9233	0.009
Cardiac arrhythmia	114 (13.54%)	1,919 (21.78%)	<.0001	0.217	114 (13.54%)	113 (13.42%)	>.9	0.003
Valvular disease	50 (5.94%)	711 (8.07%)	.0334	0.084	50 (5.94%)	61 (7.24%)	.3261	0.053
Pulmonary circulation disorder	17 (2.02%)	272 (3.09%)	.1027	0.068	17 (2.02%)	20 (2.38%)	.7351	0.024
Peripheral vascular disease	116 (13.78%)	1,553 (17.63%)	.0055	0.106	116 (13.78%)	131 (15.56%)	.3349	0.050
Hypertension uncomplicated	350 (41.57%)	5,076 (57.62%)	<.0001	0.325	350 (41.57%)	350 (41.57%)	>.9	0.000
Hypertension complicated	70 (8.31%)	1,029 (11.68%)	.0039	0.112	70 (8.31%)	66 (7.84%)	.7885	0.017
Paralysis	^a	167 (1.90%)	.0095	0.118	^a	15 (1.78%)	.0429	0.110
Other neurological disorders	534 (63.42%)	6,354 (72.12%)	<.0001	0.187	534 (63.42%)	566 (67.22%)	.1125	0.080
Chronic pulmonary disease	53 (6.29%)	1,228 (13.94%)	<.0001	0.256	53 (6.29%)	51 (6.06%)	.9194	0.010
Diabetes uncomplicated	108 (12.83%)	1,820 (20.66%)	<.0001	0.211	108 (12.83%)	105 (12.47%)	.8834	0.011
Diabetes complicated	71 (8.43%)	953 (10.82%)	.0367	0.081	71 (8.43%)	63 (7.48%)	.5285	0.035
Hypothyroidism	102 (12.11%)	1,614 (18.32%)	<.0001	0.173	102 (12.11%)	128 (15.20%)	.076	0.090
Renal failure	68 (8.08%)	1,111 (12.61%)	.0001	0.149	68 (8.08%)	79 (9.38%)	.388	0.046
Liver disease	^a	139 (1.58%)	.0077	0.125	^a	^a	–	0.022
Peptic ulcer disease	^a	76 (0.86%)	.1745	0.065	^a	^a	–	0.018
Lymphoma	^a	73 (0.83%)	.8638	0.013	^a	^a	–	0.026
Metastatic cancer	^a	54 (0.61%)	>.9	0.002	^a	^a	–	0.000
Solid tumors without metastasis	37 (4.39%)	645 (7.32%)	.0019	0.125	37 (4.39%)	56 (6.65%)	.0548	0.099
Rheumatoid arthritis	16 (1.90%)	398 (4.52%)	.0004	0.149	16 (1.90%)	16 (1.90%)	>.9	0.000
Coagulopathy	27 (3.21%)	330 (3.75%)	.4862	0.029	27 (3.21%)	26 (3.09%)	>.9	0.007
Obesity	26 (3.09%)	469 (5.32%)	.0063	0.112	26 (3.09%)	31 (3.68%)	.5899	0.033
Weight loss	41 (4.87%)	738 (8.38%)	.0004	0.141	41 (4.87%)	68 (8.08%)	.01	0.131
Fluid and electrolyte disorders	102 (12.11%)	1687 (19.15%)	<.0001	0.195	102 (12.11%)	104 (12.35%)	.9407	0.007
Blood loss anemia	^a	106 (1.20%)	.157	0.065	^a	^a	–	0.000
Deficiency anemia	42 (4.99%)	698 (7.92%)	.0027	0.120	42 (4.99%)	56 (6.65%)	.176	0.071
Depression	187 (22.21%)	3072 (34.87%)	<.0001	0.283	187 (22.21%)	184 (21.85%)	.9064	0.009

Abbreviations. AAPs, atypical antipsychotics; PIM, pimavanserin; SMD, standardized mean differences.

^aCells with sample size of <11 are suppressed per CMS requirements.

All comparative HCRU results between PIM vs. other-AAPs are described in Tables 3 and 4. PIM patients had significantly lower all-cause hospital admission rates compared to patients on other-AAPs (37.8% vs. 49.8%, p < .05). More specifically, all-cause ST-stays (34% vs. 46.2%, p < .05) and SNF-stays (20.2% vs. 31.8%, p < .05) were lower among PIM versus other-AAPs cohorts, respectively. Similar results were reported for psychiatric-hospital admissions (Figure 3) with fewer PIM patients having psychiatric admissions compared with other-AAP patients (9.6% vs. 14.6%, p < .05). More specifically, psychiatric-related ST-stays were reported among 5.4% of PIM versus 9.6% of other-AAPs patients (p < .05). Finally, LTCA (defined as a composite of LTC-stay or SNF-stay) was found to be 10% lower among PIM patients (23.2% vs. 34.6%; p < .05) compared to other-AAPs. Adjusted regression results controlled for age, race, region, comorbidities, baseline dementia, and insomnia indicate that all HCRU differences were significant (p < .05).

Figure 3. Percentage of patients with psychiatric-related hospitalizations. Abbreviations. PIM, Pimavanserin; AAPs, atypical antipsychotics; SNF, skilled nursing facility. *Groups differences significant (p = 0.05). **Sample size of <11 are suppressed per CMS requirements.

Table 3. Healthcare resource utilization and all-cause length of stay among post-matched pimavanserin and other-atypical antipsychotics.

	PIM (n = 842; 50%)	Other-AAPs (n = 842; 50%)	p Value
≥1 All-cause hospitalization, n (%)
Any hospitalization	319 (37.80%)	420 (49.80%)	<.0001
Short-term stay	286 (34%)	389 (46.20%)	<.0001
Long-term stay	52 (6.20%)	57 (6.80%)	.3828
SNF-stay	170 (20.20%)	267 (31.80%)	<.0001
Outpatient visit	777 (92.20%)	776 (92.20%)	>.9
Office visit	814 (96.60%)	783 (93%)	.0009
ER visit	519 (61.60%)	584 (69.40%)	.001
≥1 Psychiatric-related hospitalization, n (%)
Any hospitalization	81 (9.60%)	123 (14.60%)	<.0001
Short-term stay	46 (5.40%)	80 (9.60%)	.0022
Long-term stay	^a	^a	–
SNF-stay	40 (4.80%)	52 (6.20%)	.2382
Outpatient visit	154 (18.20%)	182 (21.60%)	.09
Office visit	494 (58.60%)	269 (32%)	<.0001
ER visit	43 (5.20%)	86 (10.20%)	.0001
All-cause LOS: long-term stay
n (%)	60 (7.60%)	74 (5.96%)
Mean (SD)	9.93	10.69	.56
Median (IQR)	10 (3, 15)	9 (4, 15)
Minimum, maximum	1, 27	1, 42
All-cause LOS: short-term stay
n (%)	493 (62.48%)	748 (60.23%)
Mean (SD)	5.43	6.48	.0024
Median (IQR)	4 (2, 6)	5 (3, 8)
Minimum, maximum	1, 66	1, 97
All-cause LOS: SNF
n (%)	236 (29.91%)	420 (33.82%)
Mean (SD)	36.27	41.96	.184
Median (IQR)	27 (14, 46)	27 (14, 44)
Minimum, maximum	1, 494	1, 494

Abbreviations. HCRU, health care resource utilization; AAP, atypical antipsychotic; SNF, skilled nursing facility; ER, emergency room, LOS, length of stay; AAP, atypical antipsychotic; SNF, skilled nursing facility; SD, standard deviations; IQR, interquartile range; PIM, pimavanserin.

^aCells with sample size of <11 are suppressed per CMS requirements.

Table 4. Mean all-cause per-patient per-year health care resource utilization among post-matched pimavanserin and other-atypical antipsychotic.

	PIM (n = 842; 50%)	Other-AAPs (n = 842; 50%)	p Value
All-cause: any hospitalization
Mean (SD)	0.94 (1.66)	1.48 (2.13)	<.0001
Median (IQR)	0 (0, 1)	0 (0, 2)
Minimum, maximum	0, 11	0, 14
All-cause: short-term stay
Mean (SD)	0.59 (1.09)	0.89 (1.35)	<.0001
Median (IQR)	0 (0, 1)	0 (0, 1)
Minimum, maximum	0, 8	0, 8
All-cause: long-term stay
Mean (SD)	0.07 (0.3)	0.09 (0.36)	.215
Median (IQR)	0 (0, 0)	0 (0, 0)
Minimum, maximum	0, 3	0, 3
All-cause: SNF
Mean (SD)	0.28 (0.66)	0.5 (0.9)	<.0001
Median (IQR)	0 (0, 0)	0 (0, 1)
Minimum, maximum	0, 6	0, 7
All-cause: ER visit
Mean (SD)	1.27 (1.57)	1.65 (2.56)	.0002
Median (IQR)	1 (0, 2)	1 (0, 2)
Minimum, maximum	0, 10	0, 55

Abbreviations. PPPY, per-patient per-year; HCRU, health care resource utilization; AAP, atypical antipsychotic; PIM, pimavanserin; SNF, skilled nursing facility; ER, emergency room; SD, standard deviations; IQR, interquartile range.

All-cause ER visits among PIM patients were lower compared to patients on other-AAPs (61.6% vs. 69.4%, p < .05) while psychiatric-related ER visits were reported among 5.2% of patients on PIM vs. 10.2% on other-AAPs (p < .05) (Figure 4). Additionally, standardized mean ST-stay PPPY admissions (0.59 [±1.0] vs. 0.89 [±1.35], p < .05) and SNF PPPY admissions (0.28 [±0.66] vs. 0.50 [±0.9], p < .05) were found to be significantly lower for PIM patients versus other-AAP patients, respectively (Table 4).

Figure 4. Percentage of patients with emergency room visits. Abbreviations. ER, emergency room; PIM, Pimavanserin; AAPs, atypical antipsychotics. *Groups differences significant (p = 0.05).

Average length of stay among hospitalized patients, reported in Table 3 were also significantly lower among PIM patients vs. other-AAP patients, largely driven by significant differences in ST-stays (5.43 ± 5.45 vs. 6.48 ± 6.65 days; p < .05) and numerical differences in SNF-stays (36.27 ± 44.2 vs. 41.96 ± 64.89; p = .18) between PIM vs. other-AAPs.

Discussion

To date, few studies have evaluated the impact of PIM on HCRU outcomes such as hospital admissions and ER visits in usual care studies. To our knowledge, this is the first observational study that examines the association between PIM or other-AAP use and HCRU in a real-world setting. In this analysis of Medicare patients with PDP, those who initiate continuous monotherapy with PIM had significantly lower rates of all-cause and psychiatric hospitalization compared with those initiating continuous monotherapy with other-AAPs. Nearly 12% fewer all-cause admissions were reported in the PIM cohort, a potentially meaningful and relevant benefit given the high cost of hospitalizations in the US. For example, in an analysis of the 2017 data from the Health Care Utilization Project (HCUP) National Inpatient Sample (NIS), mean all-cause hospitalization cost per stay was over $11,000 (>$13,000 in 2021 inflation adjusted USD).^10,11

In this study, fewer number of all-cause SNF-stays and LTCA (defined as a composite of LTC-stay or SNF-stay) were also observed among patients with PIM vs. other-AAPs. These results suggest that PIM may confer real-world cost-savings due to lower nursing home stays and associated lower costs. For instance, in a published analysis of Medicare claims data from 2000-2010, the average annual all-cause cost for PDP patients in the LTC was $31,178¹⁰. Thus, the clinical and economic benefits of PIM due to fewer SNF-stays and LTCA-stays may be even more profound given the potentially higher annual LTC costs among PDP patients reported in the literature. Similar pattern of results were reported for ER visits, wherein PIM patients reported almost 8% lower all-cause ER visits compared to other-AAP patients. Consistent with all-cause results reported here, psychiatric-related hospitalizations (5% lower) and ER visits (5% lower) were also seen to be fewer with PIM compared to other-AAPs. These results suggest that future investigations should further examine the association between drivers of psychiatric episode occurrence such as frequency and severity of psychiatric symptoms (i.e., hallucinations and delusions) and rates of psychiatric hospitalizations and ER visits as well as hospital length of stays.

Since this is the first real-world analysis of HCRU outcomes among PDP patients treated with PIM vs. other comparators, a pooled cohort of patients on any one of the four commonly used off-label AAPs (quetiapine, risperidone, olanzapine, and aripiprazole) formed the comparative cohort. However, it should be noted that quetiapine is the most used off-label other-AAP in real-life settings. Not surprisingly, the final matched sample of other-AAP cohort in this analysis had nearly three in four quetiapine patients compared to the other three off-label AAPs. To the extent that quetiapine may drive the outcome results in other-AAP cohort, further analysis that include head-to-head comparisons between PIM and quetiapine are warranted in future investigations. While the patient selection time period was 2016–2018 for PIM, an expanded patient selection period of 2014–2018 was chosen to ensure adequate representation of all other-AAPs used off-label. Regardless, the only four off-label AAPs mentioned above had a reasonable sample size for inclusion into the pooled “other-AAP” cohort.

In this analysis, the researchers did not control for variables in the causal pathway, such as baseline HCRU, to avoid masking true differences in HCRU outcomes between the treatment cohorts.²⁶ However, literature also suggests that baseline HCRU is a well-known predictor of subsequent HCRU. Given this, it is possible that some of the differences may be attributable to baseline HCRU differences. Therefore, future studies should examine the role of baseline HCRU on post-index follow-up HCRU.

This study included only patients diagnosed with PDP to avoid confounding by indication among the eligible patient population; however, it cannot be ruled out that PDP patients with concurrent dementia or insomnia may be prescribed AAPs for treating insomnia or agitation associated dementia instead of being treated for PDP. In fact, the significantly higher proportion of dementia or insomnia in the pre-matched and post-matched other-AAP cohort support this possibility. Nevertheless, adjusted logistic regressions examining differences between PIM vs. other-AAPs suggest that PIM patients in this analysis were demonstrated to have fewer hospitalizations even after accounting for insomnia or dementia. It is also possible that other-AAPs, and more specifically quetiapine, may be inappropriately prescribed for insomnia to aid sleep rather than treating the symptoms of psychosis. Literature suggests that a mean dose of >100-150mg quetiapine may be needed to adequately control the symptoms of PDP. In this analysis, median daily dose of quetiapine was found to be 38 mg/day (IQR: 25,50mg).²⁷ Future investigations of PIM vs. other-AAP comparisons should also consider comparing PIM patients with quetiapine patients who are prescribed therapeutic doses.

In this analysis, the researchers have examined the association between monotherapy initiation with pimavanserin or other-AAP and post-index HCRU. While association between treatment initiation and outcomes post-treatment initiation can be examined more cleanly with monotherapy patients, real-world practice patterns suggest that a majority of patients may have adjunctive AAP therapy. In fact, our analysis also suggest that AAP monotherapy patients represents only about 44% (n = 9,652) of the overall patient AAP population (n = 21,557). These results suggest that future studies examining the HCRU among patients on adjunctive therapy may also be needed.

Limitations

Randomized controlled clinical trials represent level I evidence while retrospective studies such as this represent level II evidence. Nonetheless, they are appropriate for informing the body of literature about specific outcomes such as HCRU in usual care settings. Results from this retrospective study are subject to limitations of any claims data analysis including potential miscoding, under-coding, or other issues that are characteristic of claims data used for billing and reimbursement purposes. While administrative claims are excellent measures to assess HCRU outcomes, they are unable to account for unobservable patient characteristics such socioeconomic status, psychosocial support, body mass index (BMI), smoking, facility characteristics, or other factors that may bias the results. Despite adopting PSM to ensure balanced cohorts, it is possible that residual confounding may exist. Of note, only 27 of the Elixhauser comorbidities (HIV, alcohol-abuse, substance-abuse, and psychoses were excluded) were used in the PSM; however, residual confounding may exist. After matching, PIM patients had higher mean and median Elixhauser comorbidity index scores and yet had a larger number of patients with no difference in majority of comorbidities compared to other-AAP cohort. We posit that since our PSM matching algorithm matched PIM and other-AAP patients based on individual 27 Elixhauser comorbidities instead of the comorbidity index score, it can’t be assumed that the cohorts would have a balanced match on the index score as well. Since this analysis included all PDP patients receiving PIM or other-AAPs regardless of setting (i.e. community or nursing home/LTC), it is possible that PIM patients may be either younger patients with fewer comorbidities or older patients with higher number of comorbidities at the time of treatment initiation. Additionally, it is possible that patients in the other-AAP cohort, specifically quetiapine, may potentially be prescribed inappropriately for treating other symptoms of dementia or insomnia despite ensuring that all the eligible patients have confirmed diagnosis of PD with conjunctive psychosis. Given, these potential for imbalances, future comparative analysis of PIM vs. other-AAPs may include patients of therapeutically equivalent dose of quetiapine or other comparators based on chlorpromazine equivalents or some form of defined therapeutic daily dose.

In this analysis, the other-AAP cohort was comprised of patients who initiate four of the commonly used other-AAPs due to small sample sizes of other-AAPs (e.g. clozapine, brexpiprazole, etc.). It is, therefore, possible that the HCRU results for other-AAPs may be different if other-AAPs such as clozapine that may have fewer extrapyramidal symptoms are included. As in any regression analysis, it is important not to control for variables in the causal pathway. Similarly, in comparing pimavanserin or other-AAPs, adjusting for baseline HCRU in the previous year in the same treatment cohort would have masked the differences between the treatment cohorts.²⁶ Therefore, this was not considered in the main analysis. Additionally, given the large standard deviations in HCRU outcomes, it is possible that inter-individual variability among patients may be skewed by heavy service users. Although dually eligible patients and patients with low-income subsidy may not be adequately represented in 100% CMS sample, they were not actively excluded in the analysis. Therefore, these results may be used to generalize these results to the broader population. Notwithstanding all the above limitations, this analysis represents a significant addition to the body of neuropsychiatric literature about the real-world benefits of PIM compared to other-AAPs.

Conclusions

This analysis comparing the effects of PIM vs. other-AAPs in PIM monotherapy was associated with nearly 12% lower all-cause hospitalizations and 7% lower all-cause ER visits vs. other-AAPs. Additionally, similar patterns were observed with psychiatric-related hospitalizations and ER visits. PPPY short-term and SNF-stays were also significantly lower for PIM monotherapy vs. other-AAPs. While these results are not intended to establish a causal relationship, they demonstrate the potential real-world association between HCRU patterns and PIM or other-AAP use among patients with PDP. Additional analysis examining PIM’s role in other outcomes related to delays in LTC admissions and other relevant outcomes are warranted.

Transparency

Declaration of funding

This study was financially sponsored by Acadia Pharmaceuticals

Declaration of financial/other relationships

KR is a current employee of Anlitiks Inc and SK is a former employee of Anlitiks Inc., a company that received funding from Acadia Pharmaceuticals to conduct this study. NR and DD are employees of Acadia Pharmaceuticals.

Reviewer disclosures

A reviewer on this manuscript has disclosed that they have received manuscript or speaker’s fees from Astellas, Eisai, Eli Lilly, Elsevier Japan, Janssen Pharmaceuticals, Kyowa Yakuhin, Lundbeck Japan, Meiji Seika Pharma, Mitsubishi Tanabe Pharma, MSD, Nihon Medi-Physics, Novartis, Otsuka Pharmaceutical, Shionogi, Shire, Sumitomo Pharma, Takeda Pharmaceutical, Tsumura, Viatris, Wiley Japan, and Yoshitomi Yakuhin, and research grants from Eisai, Mochida Pharmaceutical, Meiji Seika Pharma, Shionogi and Sumitomo Pharma. The other peer reviewers on this manuscript have received an honorarium from JME for their review work but have no other relevant financial relationships to disclose.

Previous presentations

Abstracts containing results from this analysis were presented in part at American Society of Clinical Psychopharmacology meeting, 2022 and Movement Disorder Society (MDS) meeting, 2022.

Acknowledgements

Formatting and submission of the manuscript to JME was conducted by Safiuddin Shoeb Syed, employee of Anlitiks Inc. Additional analysis to address reviewer comments were conducted by Daksha Gopal, employee of Anlitiks Inc.