Incremental health care resource utilization and costs associated among patients with Parkinson's disease psychosis and incident dementia: An analysis of medicare beneficiaries

Real‐world evidence examining the incremental health care resource use (HCRU) and cost burden of incident dementia among patients with Parkinson's disease psychosis (PDP) are needed within the United States (US).

Conclusions: Patients with PDP þ D experience significantly higher all-cause and psychiatric-related IP hospitalizations, including ST-stays, LT stays, and SNF stays.
They also had 54% greater mean PPPY IP hospitalization costs versus PDP patients.

K E Y W O R D S
dementia, health care resource utilization, hospitalizations, medicare, Parkinson's disease psychosis

Key points
� The results of this retrospective analysis of Medicare beneficiaries with Parkinson's disease psychosis (PDP) show that nearly 4 in 5 patients with PDP have incident concurrent dementia (PDP þ D).
� PDP þ D patients had significantly higher rates of hospitalizations for both all-cause and psychiatric-related reasons.Particularly they had more frequent short-term stays, longterm stays, and skilled nursing facility stays.
� PDP þ D patients also had incurred 54% higher costs for inpatient hospitalization compared to PDP patients without dementia.
� These findings highlight the additional burden imposed by dementia on PDP patients and the potential need to manage dementia symptoms as well.

| BACKGROUND
A hallmark characteristic of Parkinson's disease psychosis (PDP) is the spectrum of neuropsychiatric symptoms such as hallucinations and delusions. 1Other neuropsychiatric symptoms among patients with Parkinson's disease (PD) include cognitive impairment; with research suggesting that nearly 1 in 4 patients with PD exhibit mild cognitive impairment at diagnosis and almost 80% with PD potentially progressing to dementia within 2 decades. 2 It is estimated that the prevalence of psychosis occurs in over 50% of patients with PD during their lifetime. 3However, the prevalence rate of psychosis may increase to 70% among PD patients who also experience co-existing dementia. 4P with dementia (PDP þ D) is characterized by the coexistence of psychosis and dementia symptoms in individuals with PD.Additionally, research studies suggest that the constellation of PDP symptoms such as hallucinations and delusions may aggravate cognitive deficits, potentially worsening the severity of dementia.2,4 Interestingly, the co-existing nature of both psychosis and dementia among patients with PDP suggests that the development of one symptom may often signal the advent of the other.5 Despite the existing literature on the association between psychosis and dementia in PD patients, systematic investigations specifically focused on the new occurrence of dementia in patients already diagnosed with PDP are lacking.Furthermore, no definitive estimates are available to understand the incidence of dementia among patients with PDP.
The co-occurrence of psychosis and dementia in PD is known to have a significant deleterious impact on the quality of life for both patients and their caregivers. 6It is also associated with increased morbidity, mortality, increased caregiver burden and nursing home placement. 4,7,8However, there is limited research examining the incremental economic burden due to increased health care resource utilization (HCRU) and associated costs of incident dementia burden among patients with PDP (PDP þ D).
Given the current gaps in PDP literature and incident occurrence of dementia, a database analysis of Medicare beneficiaries within the United States (US) diagnosed with PDP was conducted to examine: (i) the incidence of dementia among patients diagnosed with PDP, (ii) compare HCRU outcomes such as all-cause inpatient (IP) and psychrelated hospitalizations and Emergency Room (ER) visits among PDP patients with and without dementia occurrence after diagnosis of PDP, (iii) examine the difference in costs of all-cause IP hospitalization, both total and by type of IP stay, were examined as additional outcomes to determine whether similar trends were observed for costs among PDP þ D patients compared to PDP.S1.The study sample of PDP þ D and PDP-only were derived from this selected patient population.

| Study sample
The study sample included PDP patients with no incident dementia (PDP) and PDP patients with incident dementia (PDP þ D).These patients were identified during the study time period 01/01/14 to 12/ 31/18.List of diagnosis codes used for the inclusion/exclusion criteria and patient selection attrition diagram of patients with PDP and PDP þ D are listed in supplementary Table S1 and Figure 1, respectively.This analysis was conducted in accordance with Health Insurance Portability and Accountability Act (HIPAA) compliance pursuant to a New England Review Board ethics committee waiver and CMS data use agreement.

| Covariates
Baseline covariates such as age, sex, race or ethnicity, region, clinical comorbidities, and coexisting insomnia were examined during the pre-index period.Clinical and individual comorbidities were also evaluated on data prior to the pre-index date.A list for these codes a described in Supplementary Table S2.

| HCRU outcomes and costs
HCRU outcomes related to rates of all-cause and psychiatric-related inpatient hospitalizations were assessed during the 12-months follow-up.While all-cause inpatient hospitalization was defined as a hospitalization at a healthcare facility for any reason, psychiatricrelated hospitalization was defined as an admission for any one of the psychotic symptoms or disorders in any of the 26 diagnosis positions in the Medicare database.The list of psychiatric symptoms and disorders are described in supplementary Table S1.
In Medicare claims, inpatient hospitalizations can be further classified into three different types of admissions based on the characteristics of the healthcare provider or facility and allowable length of stay within each place of service:1) short-term stays (ST-stay) are defined as hospitalizations that occur in hospitals where patients require an acute or critical setting following surgery, sudden sickness, or flare-up of a chronic sickness.Hospital stays are short-term, if they are less than 25 days; 2) long-term care stays (LT-stay), per Medicare data definition, are hospitalizations in certified long-term acute care hospitals who, in which would be expected to stay for at least over 25 days. 10Patients typically are transferred from intensive care units (ICUs) to long-term acute care; 3) skilled nursing facility stay (SNFstay) are hospitalizations that extend beyond the duration of LT-stays.
Patients in SNFs may receive care for up to 100 days, focusing on skilled nursing services and rehabilitation.
Additionally, all-cause and psychiatric-related ER visits were also assessed.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 S1 after  There were four comorbidities from the Elixhauser comorbidity list that were not used in the PSM since patients with psychosis in the pre-index (i.e., baseline) were excluded in this analysis.[14][15] The exclusion of patients with alcohol/drug-induced psychosis was based on pre-index diagnosis of alcohol/drug-induced psychosis using diagnostic codes only.Since, nearly almost all patients in both groups are treated with levodopa or other dopaminergic drugs, matching based on dopaminergic drug use was not necessary.A greedy matching algorithm using the nearest neighbour was used and details of the PSM procedure used for this analysis has been described elsewhere. 16Covariate balance between PDP and PDP þ D beneficiaries were assessed using standardized mean differences (SMDs) value of <0.1. 17Any missing data (n = 4179) on covariates were excluded prior to matching to ensure that the final, matched sample had the necessary covariates to conduct the matching (Figure 1).

| Statistical analysis
Baseline patient demographics and clinical characteristics such as sex, age, Elixhauser comorbidities, and other comorbidity status (i.e., dementia or insomnia) among patients in unmatched and matched PDP þ D versus PDP cohorts were described using frequencies and proportions before and after propensity score matching.Descriptive statistics were reported as frequencies and percentages for categorical variables, while mean, median, and range were used to represent 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 PDP þ D versus PDP patients.HCRU differences between PDP þ D versus PDP patients were evaluated using logistic regression.Differences in all-cause mean PPPY costs were examined using gamma log-link regression models.
All models were controlled for age, sex, race, region and 27 Elixhauser comorbidity characteristics.Unless otherwise specified, the statistical significance threshold of p < 0.05 was used.All analyses were conducted using SAS® Enterprise Guide Server version 7.15 via the CMS Virtual Research Data Center.This analysis was conducted in accordance with the CMS data use agreement that was established after New England Institutional Review Board review and approval.

| RESULTS
A total of 12,484 patients were eligible after our applying the study inclusion and exclusion criteria.Of these, 85.3% (n = 10,609) patients with PDP had a diagnosis for dementia in the post-PDP diagnosis period, while 14.7% (n = 1875) PDP patients did not have a diagnosis of dementia.The patient selection process, as well as the descriptive characteristics of both matched groups, are described in Figure 1 and Table 1, respectively.
Prior to PSM, the mean age and percentage of females was 76% and 47.4%, and 71 years and 52%, in the PDP þ D and PDP groups, respectively.Patients were propensity score-matched in a 1:1 ratio from the unmatched pool of PDP þ D and PDP groups, yielding 1855 PDP þ D and 1855 PDP patients.After PSM, SMD between both groups demonstrate they were balanced with respect to demographics and comorbidities (Tables 1 and 2).Mean age (72 years), gender (50% males), race, and comorbidity profile were similar for both groups.As is the case with all database studies using claims data intended for reimbursement purposes, the study has several limitations.

| CONCLUSION
Overall, results of this analysis of Medicare beneficiaries with PDP show that nearly 4 in 5 patients with PDP have incident concurrent dementia.These patients had significantly higher rates of all-cause and psychiatric-related hospitalizations, particularly in ST-stays stays, LT-stays stays, and SNF stays.They also demonstrated higher hospitalization costs compared to PDP patients without dementia occurrence.These results highlight the incremental burden of dementia imposed by these patients and the potential need to manage dementia symptoms as well.Further research is warranted to understand the underlying factors contributing to these differences and to implement strategies to optimize care.

Aof 8 -
retrospective analysis of inpatient (Part A), outpatient (Part B), and prescription drug (Part D) claims from the 100% Medicare sample) from 01/01/14-12/31/18 (i.e., patient identification period) was conducted to compare PDP þ D patients versus PDP patients.The 100% Medicare sample represents data from the federal health insurance program that provides health insurance coverage for all US residents ≥65 years, select individuals with disabilities aged <65 years, and individuals with end-stage renal disease.As of 2019, there were 63 million US Medicare beneficiaries, all of whom were represented in the 100% sample and contains claims for all inpatient, outpatient, and pharmacy services incurred by all the covered beneficiaries.Our dataset, therefore, represents all patients ≥65 years of 2 RAJAGOPALAN ET AL. age with Parkinson's disease psychosis during 2013-2019 in the United States.It should be noted that Medicare Part A, B, and D covers hospital insurance (i.e., inpatient hospitalizations, skilled nursing, hospice, and home health care), medical insurance (i.e., doctor visits, outpatient care, preventive services, and durable medical equipment), and pharmacy insurance (i.e., prescription medication coverage through private insurance plans).Medicare Fee-for-Service (FFS) refers to the traditional indemnity Medicare program, which includes Part A and Part B, where healthcare providers are paid on a fee schedule for each service provided to the beneficiary, and reimbursement is based on fees established by Medicare.Rigorous data collection and validation processes are employed for the billing claims by Medicare to ensure data quality by performing oversight of healthcare providers and plans, conduct audits, and data validation checks to identify and correct errors in submitted claims data. 9PDP population included PD patients who were identified from ≥1 International Classification of Diseases (ICD-9) and ICD-10 diagnostic claim codes of 332.0 and G20, respectively.Further, a concurrent psychosis diagnosis (occurrence of ≥1 psychosis or psychotic disorder diagnostic claim of either: F06.0, F06.2, F22, F23, F28, F29, H53.16, R44.0, R44.1, R44.2, R44.3) was identified.The PDP date was defined as the index date and any patients from this PDP patient population with a diagnosis during 12-months pre-index of dementia, psychosis, secondary parkinsonism, delirium, other psychotic disorders, alcohol/drug-induced psychosis, schizophrenia, paranoia, or personality disorder were excluded.Those with co-existing dementia at baseline also were excluded.PDP þ D patients were those with Parkinson's disease psychosis (PDP) who received a diagnosis of dementia within 12-months of PDP during post-index.Dementia codes are defined in supplementary Table

11 2. 5 |
12-months follow-up.Mean patient per year (PPPY) total costs for all-cause IP hospitalization and by type of IP hospitalization stay were also examined.Total cost for each claim, were calculated by adding the inpatient Medicare claim payment amount plus the claim pass through amount paid for the days stayed from the inpatient files.Total payments made by Medicare for each cost claim were obtained from the 100% inpatient files data.All cost amounts were inflated to 2019 dollars using the Medical Consumer Price Index (CPI) as reported by the US Bureau of Labor Statistics.Propensity score matching Patients who were identified as PDP or PDP þ D were matched using a propensity score methodology by a 1:1 ratio to create a balanced sample on measured characteristics.Propensity scores were calculated using multivariate logistic regression on 31 variables (patient age, sex, race, region, and 27 Elixhauser comorbidity characteristics) RAJAGOPALAN ET AL.

First, any secondary
databases such as Medicare claims data, may contain coding errors, missed claims, bias introduced by omission of variables, and these should be considered as limitations to this type of data.Second, identification of psychosis was based on a diagnosis of psychosis-related hallucinations and delusions -so it is likely that PDP diagnosis is underestimated.Third, residual confounding may still exist, even though the study was adjusted for potential confounding issues through appropriate propensity score matching and covariate adjustment.Fourth, we did not adjust for the duration of time to incidence of psychosis post-PD diagnosis because most people with Parkinson's disease are generally diagnosed with psychosis after many years.Finally, it is important to acknowledge that diagnosis of dementia among patients with PDP may be difficult and, as such, result in underestimates of dementia diagnosis among these patients.Notwithstanding these limitations, this analysis using Medicare 100% sample is the most representative data source for the PDP population in the US.

Table 3
displays the rates of all-cause and psychiatric-related hospitalizations for ST-stay, LT-stays stays, SNF-stays among PDP þ D and PDP groups.Any all-cause inpatient hospitalizations Demographic characteristics of PDP patients with dementia (PDP þ D) or without dementia (PDP-only).
Abbreviations: IQR, interquartile range; PDP, Parkinson's disease psychosis; PDP þ D, Parkinson's disease psychosis with Dementia; SD, standard deviation; SMD, standardized mean differences.Abbreviations: IQR, interquartile range; PDP, Parkinson's disease psychosis; PDP þ D, Parkinson's disease psychosis with Dementia; SD, standard deviation; SMD, standardized mean differences.RAJAGOPALAN ET AL.Abbreviations: NS, not significant; PDP, Parkinson's disease psychosis; PDP þ D, Parkinson's disease psychosis with Dementia.F I G U R E 2 Average annual inpatient hospitalization costs per patient per year for all-cause inpatient hospitalizations in PDP patients with and without dementia.tions,as well as IP hospitalizations by type of stay, and psychiatric inpatients hospitalizations.In fact, PDP þ D patients had 30%-40% higher rates of inpatient hospitalizations.Additionally, examination of hospitalization costs showed that PDP þ D reported 54% greater mean PPPY IP hospitalization costs and two times greater mean PPPY SNF-stay costs compared to patients with PDP.Our findings suggest that patients with PDP þ D tend to experience a higher rate of all-cause and psychiatric-related hospitalizations, specifically STstays stays, LT-stays stays, and SNF stays compared to those with only PDP.