Effect of Care Rehabilitation on Medical Expenses,Care Costs,and Total Costs of Elderly Individuals with Parkinson's Disease

Abstract

The fast-growing prevalence of Parkinson's disease (PD) creates a heavy burden for society and the health care system. Although different ways to mitigate the economic burden of PD have been discussed in the literature, including several effective treatments, few studies have paid attention to the effect of care rehabilitation (CR) on PD costs over a long-term care period. This study tracked medical expenses, care costs, and total costs of elderly individuals with PD for 3 years based on medical claims data merged with long-term care insurance (LTCI) claims data, and determined whether CR reduced PD costs. Using a retrospective, longitudinal cohort design, 3950 elderly individuals with PD who received LTCI services from April 2014 to March 2017 in Fukuoka Prefecture, Japan were followed. PD costs were compared between the CR group and the non-CR group, and a hierarchical linear model was used to examine whether CR was associated with medical expenses, care costs, and total costs. The mean value of total costs in fiscal years 2014, 2015, and 2016 were ¥3,124,944 (US$29,504), ¥3,328,398 (US$31,425), and ¥3,615,892 (US$34,140), respectively. In a hierarchical linear model, CR alone was not associated with medical expenses and care costs; additionally, CR had a positive association with higher total costs. However, the interaction term between CR and baseline care needs level significantly reduced care costs and total costs. That indicates that if older PD patients with higher care needs level receive CR, their care costs and total costs will be reduced. Further research is needed to clarify how CR reduces these patients' costs.

Introduction

Parkinson's disease (PD) is a chronic, progressive neurodegenerative disease characterized by a decline in motor and non-motor function, and it is frequently associated with various complications.^1,2 The number of patients with PD is increasing because of population aging in developed countries; consequently, vast social and medical resources are required, leading to a huge overall economic burden.^3
–5 Many studies have analyzed the direct and indirect costs of PD to estimate its economic burden.^6

–9 One of the conclusions reached is that hospitalization and long-term care costs were the major expenditures in PD direct costs.^5,10
–12 From another point of view, several studies suggested that treatments slowing progression and controlling symptoms could reduce the growing economic burden of PD.^13,14

The efficacy of rehabilitation therapies in the treatment of PD has long been established.^15,16 Currently, PD guidelines in many countries, including Japan, recommended rehabilitation on the same grade as surgery and medications.^17
–19 Interestingly, some studies have pointed out that continuous rehabilitation may delay worsening of the severity of PD.^20
–22 However, no studies have investigated whether continuous rehabilitation therapies affect the costs of PD.

In Japan, patients with chronic diseases such as PD who need long-term care services can receive care rehabilitation (CR) provided by the long-term care insurance (LTCI) system. CR emphasizes living independently and actively socializing by improving physical function, and is characterized by long-term implementation, as opposed to acute rehabilitation provided by the medical insurance system.²³ Nevertheless, research that evaluates the quality of LTCI services has only just begun in Japan, and evidence of the effect of CR is scarce.^24
–26 This study thus aimed to track medical expenses, care costs, and total costs of elderly individuals with PD for 3 years and determine whether CR reduced their medical expenses, care costs, and total costs.

Methods

Data source

Medical claims data and LTCI claims data were obtained from Latter-Stage Elderly Healthcare Insurance (LSEHI) in Fukuoka Prefecture. The population in Fukuoka Prefecture was 5,130,000 (4% of Japan's total population) in 2019, with about 694,000 (13.5%) aged 75 years or older.²⁷ LSEHI is public insurance in Japan for older residents aged 75 years or older, and those aged 65–74 years with a specified disorder. Medical claims data record all medical services, disease diagnosis information, and medical expenses of health care visits. LTCI claims data consist of long-term care services information and payment from LTCI-certified residents aged 65 years or older, and users aged 40 years or older with a specified disorder.²⁸

There is an eligibility process to receive LTCI services in Japan. An official assessment program classifies the care needs level depending on the physical and mental status of applicants and decides on the corresponding care services. There are 7 eligibility care needs levels. The 2 lowest levels are classified as “Assistance required” (Levels 1 and 2), in which users often are living independently and require few care services; the remaining 5 levels are classified as “Care required” (Levels 3, 4, and 5), meaning more care services are needed as the level increases.²⁹

Study design

A retrospective, longitudinal cohort study design was used. Participants were elderly individuals aged 75 years or older and diagnosed with PD between April 2013 and March 2014. The follow-up period was from April 2014 to March 2017.

Participants

First, medical claims data were used to identify individuals diagnosed with PD (International Classification of Diseases, Tenth Revision [ICD-10] code: G20) between April 2013 and March 2014, and who had been prescribed antiparkinsonian agents (medicine category code: 116) at least once during this period. Individuals who lost their insurance status because of death or moving during the follow-up period were excluded.

Next, medical claims data and LTCI claims data were connected to identify patients with PD who used LTCI services every year during the follow-up period, and their care needs level and care costs were examined. Because the assessment program for the care needs level in the Japanese LTCI system reevaluates patients every 6 months, the first care needs level was chosen as the criterion if patients had 2 different care needs level in the same year.

Additionally, individuals who had used CR before the follow-up start day (April 1, 2014) were excluded from the study, as the definition was CR newly starting as of April 1, 2014. Those aged younger than 75 years also were excluded.

Measures

The outcome variables were medical expenses, care costs, and total costs for 3 years from April 2014 to March 2017. Medical expenses comprised outpatient, inpatient, and medication fees.

The independent variable was CR use, with individuals who had used CR at least once in the follow-up period being defined as “1,” and those who never used it as “0.” CR included commuting rehabilitation and home-visit rehabilitation, and was defined by the following codes: (1) CR in home visit nursing care category (service codes and Item numbers: 131501 to 131555) and (2) CR in home visit rehabilitation category (service codes: 14, 16, 64, 66, 86, 87).

Covariates were divided into time level and patient level. Time-level variables included follow-up year (2014, 2015, and 2016), annual hospitalization, and annual outpatient rehabilitation, which changed over time. Annual inpatient rehabilitation also was considered to be a time-level variable; however, it was not included in the models because of its high correlation with annual hospitalization.

Patient-level variables included sex, age, Charlson comorbidity index (CCI), deterioration of care needs level, and baseline care needs level. Following 3 studies in the literature,^30
–32 a categorical variable for CCI was generated as follows: Category 0 (CCI: 0), Category 1 (CCI: 1–2), Category 2 (CCI: 3–4), and Category 3 (CCI: ≥5). To assess if there was care needs level deterioration, the baseline care needs level was subtracted from the level in the third year (2016). This was assigned a “1” if there was deterioration or “0” if not. Baseline care needs levels were classified into 3 categories: Category 0 (Assistance required: level 1–2), Category 1 (Care required: level 3–4), and Category 2 (Care required: level 5–7).

Statistical analysis

Descriptive statistics were used to summarize participants' characteristics. Student and Welch's t tests were applied for continuous variables (age, medical expenses, care costs, and total costs), and Pearson chi-square tests were performed for categorical variables (major PD comorbidity, CCI, deterioration of care needs level, baseline care needs level, hospitalization, inpatient rehabilitation, and outpatient rehabilitation).

To investigate the effect of CR on medical expenses, care costs, and total costs, the research team first confirmed the data distribution. Because medical expenses, care costs, and total costs did not have a Gaussian distribution but showed a heavy right tail, the team conducted a logarithmic transformation. Next, a hierarchical linear model (HLM) was used for the longitudinal time-series data because repeated measures for 3 years are nested in patient characteristics (random effect). That is, the team forecast the effect of CR on medical expenses, care costs, and total costs of elderly individuals with PD with hierarchical data in which time-variant observations (eg, hospitalization every year) for 3 years are measured at Level 1 (Time level) and time-invariant observations (eg, sex, age) at Level 2 (Patient level). This was done because PD is a chronic and progressive disease, so its cost will change over time, and it is greatly affected by PD patients with different characteristics.

Compared to standard pooling regression models, HLM cannot only observe changes of different time, but it also incorporates patient characteristics at a higher level to manage different intercepts and slopes for each patient. HLM (also called a random effects model, multilevel model, or mixed model) is applicable for place-based hierarchical data or temporal hierarchical data.³³ The generic pooling regression model assumes that residuals are independent and normally distributed, and that there are not further correlations (ie, dependence). However, in hierarchical data, the measurements are not independent in each level 2, because repeated measures at level 1 (eg, every year) are nested in level 2 (eg, sex and age). Therefore, these correlations are regarded as unexplained residual variance and divided into 2 (higher level variance between higher level entities and lower level variance within these entities) in HLM. The HLM can be defined as follows.³⁴

Level 1 $Y_{i j} = β_{o i} + β_{1} X_{1 i j} + β_{2} Z_{i} + e_{i j}$

Level 2 $β_{0 i} = β_{0} + μ_{i}$

Where $Y_{i j}$ is the dependent variable; $β_{o i}$ is the mean intercept for higher level entity i; $β_{0}$ is the mean intercept; $μ_{i}$ is the higher level residual for higher level entity i (random effect) $μ_{i} \sim N (0, σ_{μ}^{2})$ ; $X_{1 i j}$ is a covariate that is measured at the lower level entity j for higher level entity i; $β_{1}$ is the coefficient of $X_{1 i j}$ ; Z_i is a covariate that is measured at the higher level entity i; $β_{2}$ is the coefficient of Z_i ; and $e_{i j}$ is the lower level residual for lower level entity j, $e_{i j} \sim N (0, σ_{e}^{2})$ .

The research team assumed fixed effects and random intercept effects in this study. $Y_{i j}$ denotes the cost on time j of patient i. $μ_{i}$ refers to the deviation between $β_{0}$ (mean intercept of all patients) and $β_{o i}$ (mean intercept of patient i), and denotes the residual term of random effects. $X_{1 i j}$ is the time-variant covariate on time j of patient i with coefficient $β_{1}$ . Z_i is the time-invariant covariate and refers to characteristics of patient i with coefficient $β_{2}$ . $e_{i j}$ is the residual term in the model.

In addition, the team considered the independent variable (CR) as a patient-level factor in this model, because whether someone uses CR depends on patient characteristics, such as baseline care needs level or willingness to improve daily functioning.

Finally, for medical expenses, care costs, and total costs, the team constructed 4 models, respectively. The first model (Null model) examined only the variance of the intercept random effect for outcome variables. The second model (Model 1) expanded the Null model by introducing time-level covariates (follow-up year, annual hospitalization, and annual outpatient rehabilitation). The third model (Model 2) added patient-level variables (sex, age, CCI category, deterioration of care needs level, and baseline care needs level). The full model (Model 3) investigated the regression coefficients and change in variance explained by adding CR.

Several studies have reported that severity of PD or care needs level at baseline may affect subsequent outcomes.^22,25 Therefore, the team confirmed the effect of CR and baseline care needs level on the outcome variable “total costs” using a 2-way analysis of variance (ANOVA) before model.

In all instances, P values were 2-tailed, and the level of significance was set at P < .05. Statistical analyses were performed using Stata Statistical Software: Release 14 (StataCorp LP, College Station, TX).

Ethical issues statement

The claims data used in this study were anonymized. Ethics approval was obtained from the Institutional Review Board of Kyushu University (Clinical Bioethics Committee of the Graduate School of Medical Sciences, Kyushu University).

Results

Participants' sociodemographic characteristics are shown in Table 1. Of the 3950 participants, 498 used CR and 3452 did not. Woman accounted for almost three quarters of the total sample. The proportion of woman in the non-CR group was significantly higher as compared to the CR group. Mean age was 82.3 years. The non-CR group was significantly older than the CR group. Except for dementia and cancer, there were no differences between the CR and non-CR groups in major PD comorbidity. The CR group had a significantly higher rate of cancer and lower rate of dementia than the non-CR group. Regarding CCI, Categories 1 and 2 accounted for about 70% of the total, and there was no significant difference between groups. The mean number of CR months was 15.7 (Table 1).

Table 1.

Characteristics of Study Participants

	Overall (n = 3950)	CR group (n = 498)	Non-CR group (n = 3452)	P
Woman (%)	2950(74.7%)	334(67.1%)	2,616(75.8%)	<0.01
Age				<0.01
Mean (SD)	82.3(4.7)	81.5(4.3)	82.5(4.8)
Cerebrovascular disease (%)	2353(59.6%)	290(58.2%)	2063(59.8%)	0.52
Dementia (%)	1618(41.0%)	169(33.9%)	1449(42.0%)	<0.01
Heart failure (%)	1541(39.0%)	195(39.2%)	1346(39.0%)	0.94
Cancer (%)	470(11.9%)	73(14.7%)	397(11.5%)	0.04
Renal disease (%)	305(7.7%)	39(7.8%)	266(7.7%)	0.92
CCI category (%)				0.11
Category 0	314(8.0%)	51(10.2%)	263(7.6%)
Category 1	1575(39.9%)	182(36.6%)	1393(40.4%)
Category 2	1323(33.5%)	165(33.1%)	1158(33.6%)
Category 3	738(18.7%)	100(20.1%)	638(18.5%)
Number of CR months
Mean (SD)	NA	15.7(10.5)	NA
Deterioration of care needs level (%)	1529(38.7%)	292(58.6%)	1237(35.8%)	<0.01
Baseline Care needs level (%)				<0.01
Category 0	967(24.5%)	209(42.0%)	758(22.0%)
Category 1	1542(39.0%)	182(36.6%)	1360(39.4%)
Category 2	1441(36.5%)	107(21.5%)	1334(38.6%)
Hospitalization (%)
In 2014	1549(39.2%)	252(50.6%)	1297(37.6%)	<0.01
In 2015	1437(36.4%)	247(49.6%)	1190(34.5%)	<0.01
In 2016	1603(40.6%)	251(50.4%)	1352(39.2%)	<0.01
Inpatient rehabilitation (%)
In 2014	1349(34.2%)	256(51.4%)	1093(31.7%)	<0.01
In 2015	1167(29.5%)	222(44.6%)	945(27.4%)	<0.01
In 2016	1289(32.6%)	210(42.2%)	1079(31.3%)	<0.01
Outpatient rehabilitation (%)
In 2014	60(1.5%)	15(3.0%)	45(1.3%)	<0.01
In 2015	45(1.1%)	9(1.8%)	36(1.0%)	0.13
In 2016	31(0.8%)	5(1.0%)	26(0.8%)	0.55
Medical expenses, ¥
Mean (SD)
In 2014	1,389,640(1,511,555)	1,755,786(1,707,077)	1,336,818(1,474,005)	<0.01
In 2015	1,330,752(1,425,740)	1,739,320(1,664,260)	1,271,811(1,378,266)	<0.01
In 2016	1,461,913(1,648,125)	1,816,124(1,802,687)	1,410,814(1,618,509)	<0.01
Care costs, ¥
Mean (SD)
In 2014	1,735,305(1,279,889)	976,806(999,061)	1,844,729(1,278,969)	<0.01
In 2015	1,997,645(1,279,945)	1,419,077(1,125,789)	2,081,112(1,279,402)	<0.01
In 2016	2,153,979(1,267,420)	1,733,672(1,195,160)	2,214,614(1,266,205)	<0.01
Total costs, ¥
Mean (SD)
In 2014	3,124,944(1,773,800)	2,732,592(1,940,549)	3,181,547(1,741,453)	<0.01
In 2015	3,328,398(1,688,501)	3,158,396(1,802,554)	3,352,923(1,670,254)	0.02
In 2016	3,615,892(1,756,827)	3,549,796(1,888,262)	3,625,428(1,737,129)	0.40

CCI, Charlson comorbidity index; CR, care rehabilitation; NA, not available; SD, standard deviation.

Compared to baseline, more than 38% of participants had deteriorated care needs level in the third year; this rate was significantly higher in the CR group than in the non-CR group. As for the baseline care needs level, Categories 1 and 2 accounted for about 70% of the total. Compared with the non-CR group, the CR group had a significantly higher rate of Category 0 and lower rates of Categories 1 and 2 (Table 1).

In general, the rate of hospitalization over 3 years ranged from 35% to 40%. The CR group showed significantly more hospitalization and inpatient rehabilitation than the non-CR group. As for outpatient rehabilitation, the overall utilization rate was very low, but still relatively higher in the CR group than in the non-CR group (Table 1).

Mean medical expenses are shown in Table 1. Although the 2 groups (CR and non-CR) also showed the same changes as a whole, there were significantly higher medical expenses in the CR group than in the non-CR group. Moreover, mean care costs increased annually and exceeded medical expenses. The CR group had significantly lower care costs than the non-CR group. The total costs increased from 2014 to 2015, and from 2015 to 2016. Although the total costs of the CR group were lower than those of the non-CR group, there was no significant difference in the third year (Table 1).

ANOVA results revealed significant main effects of CR (F ₁, ₁₁₈₄₄ = 6.20; P = .01) and baseline care needs level (F ₂, 11844 = 515.19; P < .01), and a significant effect of the interaction (F ₂, ₁₁₈₄₄ = 22.72; P < .01) on total costs. Figure 1 shows the interactive effect. Compared with the non-CR group, whose total costs increased rapidly, the CR group increased slowly from baseline care needs level onward.

FIG. 1.

The interactive effect between care rehabilitation and care needs level. CR, care rehabilitation; JP¥, Japanese yen.

Table 2 shows the results for medical expenses for the 4 models. In each model, the variance of random intercept effect for medical expenses was significant. Model 1, which introduced time-level covariates, explained 25.5% of the variance of random effects (PCV = 0.255), the largest amount among the models; followed by Model 2, which added patient-level covariates; and Model 3, which added CR and explained the smallest variance amount. As shown in Model 3, CR and the interaction term between CR and baseline care needs level had no significant effect on medical expenses. Time-level covariates were all significantly associated with medical expenses. In particular, annual hospitalization had the most significant impact (exp(β) = 1.33, 95% CI 1.30–1.36).

Table 2.

Hierarchical Linear Model for Medical Expenses

	Null model			Model 1			Model 2			Model 3
	exp(β)	95% CI	P	exp(β)	95% CI	P	exp(β)	95% CI	P	exp(β)	95% CI	P
Intercept	13.64	13.61 – 13.67	<0.01	80.36	52.91 – 107.82	<0.01	82.47	55.03 – 109.91	<0.01	82.35	54.91 – 109.79	<0.01
Time-Level
Follow-up year				−0.03	−0.05 – −0.02	<0.01	−0.03	−0.05 – −0.02	<0.01	−0.03	−0.05 – −0.02	<0.01
Hospitalization
No				Ref.			Ref.			Ref.
Yes				1.35	1.32 – 1.38	<0.01	1.33	1.30 – 1.36	<0.01	1.33	1.30 – 1.36	<0.01
Outpatient rehabilitation
No				Ref.			Ref.			Ref.
Yes				0.22	0.07 – 0.38	0.01	0.17	0.02 – 0.33	0.03	0.18	0.02 – 0.33	0.03
Patient-Level
Age							−0.02	−0.03 – −0.02	<0.01	−0.02	−0.03 – −0.02	<0.01
Sex
Man							Ref.			Ref.
Woman							−0.04	−0.10 – 0.01	0.09	−0.04	−0.09 – 0.01	0.10
CCI Category
Category 0							Ref.			Ref.
Category 1							0.04	−0.05 – 0.12	0.40	0.04	−0.05 – 0.12	0.38
Category 2							0.23	0.14 – 0.32	<0.01	0.23	0.14 – 0.32	<0.01
Category 3							0.38	0.29 – 0.48	<0.01	0.38	0.28 – 0.47	<0.01
Deterioration of care needs level
No							Ref.			Ref.
Yes							0.04	−0.01 – 0.09	0.10	0.03	−0.01 – 0.08	0.15
Baseline care needs level
Category 0							Ref.			Ref.
Category 1							−0.17	−0.23 – −0.11	<0.01	−0.15	−0.21 – −0.08	<0.01
Category 2							−0.24	−0.30 – −0.18	<0.01	−0.24	−0.30 – −0.17	<0.01
CR
No										Ref.
Yes										0.09	−0.02 – 0.20	0.10
CR × Baseline care needs level
Non-CR or										Ref.
Category 0
CR × Category 1										−0.12	−0.27 – 0.03	0.13
CR × Category 2										0.10	−0.07 – 0.28	0.26
Random Effect
Variation (SE)	0.550(0.018)	0.52 – 0.59		0.410(0.012)	0.39 – 0.44		0.371(0.012)	0.35 – 0.39		0.369(0.012)	0.35 – 0.39
ICC	0.436			0.521			0.496			0.495
PCV (%)				0.255			0.071			0.004
Log-likelihood	−17031.6			−13785.3			−13631.4			−13626.2
AIC	34069.2			27582.7			27290.8			27286.4
BIC	34091.3			27626.9			27394.0			27411.7

AIC, Akaike information criterion; BIC, Bayesian information criterion; CCI, Charlson comorbidity index; CI, confidence interval; CR, care rehabilitation; ICC, interclass correlation; PCV, proportional change in variance; Ref., reference; SE, standard error.

exp(β): exponential coefficient, percent increase in the response for a 1-unit increase in the independent variable.

Table 3 shows the results for care costs. The variance of random intercept effect for care costs was significant in all models. Model 2, which introduced patient-level covariates, explained 50.5% of the variance of random effect for care costs, the highest in the 4 models (PCV = 0.505), followed by Model 3, adding CR. Model 1 added time-level covariates and did not explain any variance. In Model 3, although CR was not significant on care costs, the interaction terms, “Category 1 × CR” (exp(β) = -0.41, 95% CI −0.56 – −0.27) and “Category 2 × CR” (exp(β) = -0.34, 95% CI −0.51 – −0.17) were significant. All covariates were significantly associated with care costs.

Table 3.

Hierarchical Linear Model for Care Costs

	Null model			Model 1			Model 2			Model 3
	exp(β)	95% CI	P	exp(β)	95% CI	P	exp(β)	95% CI	P	exp(β)	95% CI	P
Intercept	14.13	14.10 – 14.16	<0.01	−348.24	−370.55 – −325.94	<0.01	−351.74	−374.04 – −329.44	<0.01	−351.71	−374.02 – −329.41	<0.01
Time Level
Follow-up year				0.18	0.17 – 0.19	<0.01	0.18	0.17 – 0.19	<0.01	0.18	0.17 – 0.19	<0.01
Hospitalization
No				Ref.			Ref.			Ref.
Yes				−0.17	−0.19 – −0.14	<0.01	−0.18	−0.20 – −0.15	<0.01	−0.17	−0.19 – −0.15	<0.01
Outpatient rehabilitation
No				Ref.			Ref.			Ref.
Yes				−0.38	−0.53 – −0.24	<0.01	−0.32	−0.46 – −0.19	<0.01	−0.31	−0.45 – −0.18	<0.01
Patient Level
Age							0.02	0.01 – 0.02	<0.01	0.02	0.01 – 0.02	<0.01
Sex
Man							Ref.			Ref.
Woman							0.25	0.20 – 0.30	<0.01	0.24	0.19 – 0.29	<0.01
CCI Category
Category 0							Ref.			Ref.
Category 1							0.18	0.09 – 0.26	<0.01	0.17	0.08 – 0.25	<0.01
Category 2							0.22	0.14 – 0.31	<0.01	0.22	0.13 – 0.30	<0.01
Category 3							0.20	0.11 – 0.30	<0.01	0.20	0.11 – 0.29	<0.01
Deterioration of care needs level
No							Ref.			Ref.
Yes							0.17	0.12 – 0.21	<0.01	0.18	0.13 – 0.22	<0.01
Baseline care needs level
Category 0							Ref.			Ref.
Category 1							0.96	0.91 – 1.02	<0.01	1.02	0.96 – 1.08	<0.01
Category 2							1.56	1.50 – 1.62	<0.01	1.60	1.54 – 1.66	<0.01
CR
No										Ref.
Yes										0.05	−0.05 – 0.16	0.31
CR × Baseline care needs level
Non-CR or										Ref.
Category 0
CR × Category 1										−0.41	−0.56 – −0.27	<0.01
CR × Category 2										−0.34	−0.51 – −0.17	<0.01
Random Effect
Variation (SE)	0.774(0.020)	0.74 – 0.81		0.774(0.019)	0.74 – 0.82		0.383(0.011)	0.36 – 0.40		0.376(0.010)	0.36 – 0.40
ICC	0.727			0.755			0.604			0.560
PCV (%)				0.000			0.505			0.009
Log-likelihood	−13814.2			−13237.1			−12024.7			−11993.5
AIC	27634.4			26486.2			24077.4			24021.0
BIC	27656.6			26530.5			24180.7			24146.5

exp(β): exponential coefficient, percent increase in the response for a 1-unit increase in the independent variable.

Table 4 shows the results for total costs. The variance of random intercept effect for total costs was significant in all models. Model 2, which added patient-level variables, explained 45.7% of the variance of random effect for total costs, the highest among all models (PCV = 0.457), followed by Model 1 with time-level variables. Model 3 added CR and explained the smallest amount of variance. CR was significantly associated with higher total costs (exp(β) = 0.12, 95% CI 0.07 – 0.18). However, the interaction terms “Category 1 × CR” and “Category 2 × CR” were significantly associated with lower total costs. Follow-up year, annual hospitalization, woman, CCI category, deterioration of care needs level, and baseline care needs level were significantly associated with total costs.

Table 4.

Hierarchical Linear Model for Total Costs

	Null model			Model 1			Model 2			Model 3
	exp(β)	95% CI	P	exp(β)	95% CI	P	exp(β)	95% CI	P	exp(β)	95% CI	P
Intercept	14.87	14.85 – 14.88	<0.01	−168.04	−181.20 – −154.88	<0.01	−169.21	−182.37 – −156.04	<0.01	−169.22	−182.38 – −156.06	<0.01
Time Level
Follow-up year				0.09	0.08 – 0.10	<0.01	0.09	0.08 – 0.10	<0.01	0.09	0.08 – 0.10	<0.01
Hospitalization
No				Ref.			Ref.			Ref.
Yes				0.45	0.44 – 0.47	<0.01	0.45	0.44 – 0.46	<0.01	0.45	0.44 – 0.47	<0.01
Outpatient rehabilitation
No				Ref.			Ref.			Ref.
Yes				−0.02	−0.11 – 0.06	0.60	0.01	−0.07 – 0.09	0.84	0.01	−0.07 – 0.09	0.77
Patient Level
Age							0.00	0.00 – 0.01	0.49	0.00	−0.00 – 0.00	0.63
Sex
Man							Ref.			Ref.
Woman							0.15	0.12 – 0.17	<0.01	0.14	0.12 – 0.17	<0.01
CCI Category
Category 0							Ref.			Ref.
Category 1							0.14	0.09 – 0.19	<0.01	0.14	0.09 – 0.18	<0.01
Category 2							0.22	0.17 – 0.26	<0.01	0.21	0.17 – 0.26	<0.01
Category 3							0.28	0.23 – 0.33	<0.01	0.28	0.23 – 0.33	<0.01
Deterioration of care needs level
No							Ref.			Ref.
Yes							0.16	0.14 – 0.19	<0.01	0.17	0.14 – 0.19	<0.01
Baseline care needs level
Category 0							Ref.			Ref.
Category 1							0.44	0.41 – 0.48	<0.01	0.49	0.46 – 0.52	<0.01
Category 2							0.82	0.79 – 0.86	<0.01	0.86	0.83 – 0.90	<0.01
CR
No										Ref.
Yes										0.12	0.07 – 0.18	<0.01
CR × Baseline care needs level
Non-CR or										Ref.
Category 0
CR × Category 1										−0.27	−0.36 – −0.19	<0.01
CR × Category 2										−0.27	−0.37 – −0.18	<0.01
Random Effect
Variation (SE)	0.243(0.007)	0.23 – 0.26		0.226(0.006)	0.21 – 0.24		0.115(0.003)	0.11 – 0.12		0.113(0.003)	0.11 – 0.12
ICC	0.645			0.720			0.567			0.562
PCV (%)				0.070			0.457			0.008
Log-likelihood	−8591.6			−6663.8			−5537.9			−5508.8
AIC	17189.2			13339.5			11103.8			11051.7
BIC	17211.4			13383.8			11207.1			11177.1

exp(β): exponential coefficient, percent increase in the response for a 1-unit increase in the independent variable.

Discussion

This study tracked the medical expenses and care costs of elderly individuals with PD for 3 years. The total costs increased from ¥3,124,944 (US$29,504) in the first year, to ¥3,328,398 (US$31,425) in the second year, to ¥3,615,892 (US$34,140) in the third year. Some Western studies suggested an average PD cost per person of $20,000–$40,000 (approximately 2–4 million JP Yen),^4,35 and the present study showed a similar trend in Japan to that seen in America and Europe. Additionally, care costs were 1–1.5 times the medical expenses every year and accounted for more than half of total costs. This indicates that, in the cost of PD, care costs exceed medical expenses, accounting for a considerable proportion of the total costs, and continued to increase with deterioration of symptoms.

CR and the interaction between CR and baseline care needs level did not affect medical expenses. It was originally expected that continuous CR for elderly patients could improve their physical function and reduce the risk of falling or complications; thus, hospitalization and medical expenses would decrease.³⁶ Unfortunately, that scenario was not confirmed in this study.

Although CR did not reduce care costs, the interaction between CR and baseline care needs level was associated with reduced care costs. Compared with non-CR or CR in Category 0 at baseline, CR in Category 1 and Category 2 significantly reduced care costs. In particular, this was most noticeable in Category 1. In general, the upper limit of publicly available LTCI service increases with care needs level deterioration, which leads to a rise in care costs. As shown in Table 3, care costs increased with higher care needs level at baseline or care needs level deterioration. However, as noted in previous studies, exercise consistently shows benefits to motor function and daily living and thus may reduce long-term care service needs.^21,22 Therefore, the research team considers that patients with a higher care needs level can improve their function, even if slightly, through continuous rehabilitation; the effect reflected in cost reduction is greater than in patients with a lower care needs level. In this study, the cost reduction effect of CR in Category 1 was particularly remarkable.

Finally, CR significantly increased total costs. However, as with care costs, when considering the interaction term, the total costs decreased. Compared with non-CR in Category 0, CR in Categories 1 and 2 significantly reduced the total costs. That is clearly shown in Figure 1, in which an interactive effect is demonstrated. At a lower care needs level (Category 0), there is a large gap in total costs between CR and non-CR; however, with the increase in care needs level, both became closer, indicating that CR restrained the increase in costs. As with care costs, CR significantly reduced PD total costs, especially for patients with a higher care needs level.

Another strength of this study is that it confirmed the existence of individual heterogeneity as a random effect in independent variables using a HLM. Compared with annual hospitalization and baseline care needs level, CR has an even smaller but substantial effect on PD cost.

Limitations

This study had several limitations. First, because there was still a disparity between caregivers working in long-term care services in Japan, relying only on the claims data of Fukuoka Prefecture is prone to sample bias when evaluating the overall effect of CR. Second, because there were patients who could not be traced because of death or moving during follow-up, there is a possibility of selective bias. Third, CR was defined as individuals using CR during the follow-up period; thus, a dose-response relationship was not confirmed. Fourth, confounding factors may be present, as 2 comorbidities (dementia and cancer) were not adjusted for; although CCI was used. Fifth, the CR group had a significantly higher rate of Category 0 needs level at baseline and care needs level deterioration in the third year. Because some previous studies indicated that patients with lower (vs. higher) care needs level at baseline display significantly higher subsequent deterioration,^37
–39 it is possible that the difference in results was affected by the lack of unified baseline care needs level.

Conclusions

Based on current circumstances and the present results, this study concluded that continued CR use in elderly PD patients with a higher care needs level can reduce their care costs and total costs. Although the mechanism by which CR may reduce PD costs was not elucidated in this study, one cannot rule out the possibility that CR may provide long-term training or consultation for PD patients with a higher care needs level when they encounter a problem in daily life related to functional decline or deterioration. Thus, without using further care services, these patients may be more likely to have active and independent lives than those without CR support, which ultimately may reduce their total costs. Furthermore, the research team suggests that CR may not only alleviate the economic burden of PD but also can support and facilitate aging in place among patients with PD.

There were 162,000 patients with PD estimated in Japan according to a patient survey in 2017.⁴⁰ Moreover, the prevalence of PD in those aged 65 years or older will rise to about 10 times higher on average.¹⁹ Considering the upcoming demographic shifts in Japan, those aged 65 years or older will comprise one third of the population by 2030²⁸; thus, reducing the huge economic burden of PD in elderly individuals has undoubtedly become an important issue in the public health system.

Footnotes

Authors' Contributions

Ms. Liu: conception and design, acquisition of data, analysis and interpretation of data, drafting the manuscripts and statistical analysis. Dr. Babazono: conception and design, critical revision of the manuscript for important content, supervision, and administrative and material support. Ms. Kim: acquisition of data. Mr. Li: administrative and material support.

Acknowledgments

We thank the Fukuoka Prefecture Association of Latter-Stage Elderly Healthcare for the provision of health care claims data and long-term care insurance claims data. We also thank Editage for English language editing.

Author Disclosure Statement

The authors declare that there are no conflicts of interest.

Funding Information

No funding was received for this article.

References

George

, Ashok

. Parkinson's disease and its management part 1: disease entity, risk factors, pathophysiology, clinical presentation, and diagnosis. P T, 2015; 40:504–532.

Obeso

, Stamelou

, Goetz

, et al. Past, present, and future of Parkinson's disease: a special essay on the 200th Anniversary of the Shaking Palsy. Mov Disord, 2017; 32:1264–1310.

Dorsey

, Constantinescu

, Thompson

, et al. Projected number of people with Parkinson disease in the most populous nations, 2005 through 2030. Neurology, 2007; 68:384–386.

Rubenstein

, DeLeo

, Chrischilles

. Economic and health-related quality of life considerations of new therapies in Parkinson's disease. Pharmacoeconomics, 2001; 19:729–752.

The Lewin Group

. Economic burden and future impact of

Parkinson's disease final report

. https://www.michaeljfox.org/sites/default/files/media/document/2019%20Parkinson%27s%20Economic%20Burden%20Study%20-%20FINAL.pdf Accessed January 4, 2021 .

Findley

, Aujla

, Bain

, et al. Direct economic impact of Parkinson's disease: a research survey in the United Kingdom. Mov Disord, 2003; 18:1139–1145.

Lindgren

, Von Campenhausen

, Spottke

, Siebert

, Dodel

. Cost of Parkinson's disease in Europe. Eur J Neurol, 2005; 12(s1):68–73.

Martinez-Martin

, Macaulay

, Jalundhwala

, et al. The long-term direct and indirect economic burden among Parkinson's disease caregivers in the United States. Mov Disord, 2019; 34:236–245.

Yang

, Hamilton

, Kopil

, et al. Current and projected future economic burden of Parkinson's disease in the U.S. NPJ Parkinsons Dis, 2020; 6:15.

10.

Bhattacharjee

, Metzger

, Tworek

, Wei

, Pan

, Sambamoorthi

. Parkinson's disease and home healthcare use and expenditures among elderly Medicare beneficiaries. Parkinsons Dis, 2015; 2015:606810.

11.

Noyes

, Liu

, Li

, Holloway

, Dick

. Economic burden associated with Parkinson's disease on elderly Medicare beneficiaries. Mov Disord, 2006; 21:362–372.

12.

von Campenhausen

, Winter

, Rodrigues e Silva

, et al. Costs of illness and care in Parkinson's disease: an evaluation in six countries. Eur Neuropsychopharmacol, 2011; 21:180–191.

13.

Martinez-Martin

, Rodriguez-Blazquez

, Paz

, et al. Parkinson symptoms and health related quality of life as predictors of costs: a longitudinal observational study with linear mixed model analysis. PLoS One, 2015; 10:e0145310.

14.

Schenkman

, Wei Zhu

, Cutson

, Whetten-Goldstein

. Longitudinal evaluation of economic and physical impact of Parkinson's disease. Parkinsonism Relat Disord, 2001; 8:41–50.

15.

Abbruzzese

, Marchese

, Avanzino

, Pelosin

. Rehabilitation for Parkinson's disease: current outlook and future challenges. Parkinsonism Relat Disord, 2016; 22 Suppl 1:S60–S64.

16.

Foster

, Bedekar

, Tickle-Degnen

. Systematic review of the effectiveness of occupational therapy-related interventions for people with Parkinson's disease. Am J Occup Ther, 2014; 68:39–49.

17.

Grimes

, Fitzpatrick

, Gordon

, et al. Canadian guideline for Parkinson disease. CMAJ, 2019; 191:E989–E1004.

18.

Keus

, Munneke

, Graziano

, et al. European physiotherapy guideline for Parkinson's disease. The Netherlands: KNGF/ParkinsonNet, 2014. https://www.parkinsonnet.nl/app/uploads/sites/3/2019/11/eu_guideline_parkinson_guideline_for_pt_s1.pdf Accessed October 6, 2020 .

19.

Takahashi

. Japanese guideline for Parkinson's disease. J Jpn Soc Internal Med, 2019; 108:1435–1441.

20.

Miller

, Mayol

, Moore

, Heron

, Nicholos

, Ragano

. Rate of progression in activity and participation outcomes in exercisers with Parkinson's disease: a five-year prospective longitudinal study. Parkinson's Dis, 2019; 2019:5679187.

21.

Oguh

, Eisenstein

, Kwasny

, Simuni

. Back to the basics: regular exercise matters in parkinson's disease: results from the National Parkinson Foundation QII registry study. Parkinsonism Relat Disord, 2014; 20:1221–1225.

22.

Rafferty

, Schmidt

, Luo

, et al. Regular exercise, quality of life, and mobility in Parkinson's disease: a longitudinal analysis of national parkinson foundation quality improvement initiative data. J Parkinsons Dis, 2017; 7:193–202.

23.

Ministry of Health Labour and Welfare (MHLW). Report for rehabilitation. 2017. www.mhlw.go.jp/file/05-Shingikai-12404000-Hokenkyoku-Iryouka/0000162530.pdf Accessed September 21, 2020 .

24.

Kato

, Tamiya

, Kashiwagi

, Sato

, Takahashi

. Relationship between home care service use and changes in the care needs level of Japanese elderly. BMC Geriatr, 2009; 9:58.

25.

Moriyama

, Tamiya

, Kawamura

, Mayers

, Noguchi

, Takahashi

. Effect of short-stay service use on stay-at-home duration for elderly with certified care needs: analysis of long-term care insurance claims data in Japan. PLoS One, 2018; 13:e0203112.

26.

Pedro

, Nanako

. Effective of in home and community based services on the functional status of elderly in the long term care insurance system in Japan. BMC Health Serv Res, 2012; 12:239.

27.

Fukuoka Prefecture. Report of the elderly population in Fukuoka prefecture. 2019; https://www.pref.fukuoka.lg.jp/contents/koreisyajinko20191001.html Accessed September 21, 2020 .

28.

Pedro

, Nanako

. Trends and factors in Japan's long-term care insurance system: Japan's 10-year experience. Dordrecht: Springer, 2014.

29.

Matsuda

. Health policy in Japan—current situation and future challenges. JMA J, 2019; 2:1–10.

30.

Mori

, Tamiya

, Jin

, et al. Estimated expenditures for hip fractures using merged healthcare insurance data for individuals aged >/ = 75 years and long-term care insurance claims data in Japan. Arch Osteoporos, 2018; 13:37.

31.

Jamal

, Babazono

, Li

, Yoshida

, Fujita

. Multilevel analysis of hemodialysis-associated infection among end-stage renal disease patients: results of a retrospective cohort study utilizing the insurance claim data of Fukuoka Prefecture, Japan. Medicine (Baltimore), 2020; 99:e19871.

32.

Murata

, Babazono

, Fukuda

. Effect of income on length of stay in a hospital or long-term care facility among older adults with dementia in Japan. Int J Geriatr Psychiatry, 2020; 35:302–311.

33.

Bell

, Jones

. Explaining fixed effects: random effects modeling of time-series cross-sectional and panel data. Political Sci Res Methods, 2014; 3:133–153.

34.

Charles

, Shayle

, John

. Generalized, linear, and mixed models. Wiley, 2008, pp. 1–424.

35.

Winter

, Balzer-Geldsetzer

, von Campenhausen

, et al. Trends in resource utilization for Parkinson's disease in Germany. J Neurol Sci, 2010; 294:18–22.

36.

Okunoye

, Kojima

, Marston

, Walters

, Schrag

. Factors associated with hospitalisation among people with Parkinson's disease—A systematic review and meta-analysis. Parkinsonism Relat Disord, 2020; 71:66–72.

37.

Jin

, Tamiya

, Jeon

, Kawamura

, Takahashi

, Noguchi

. Resident and facility characteristics associated with care-need level deterioration in long-term care welfare facilities in Japan. Geriatr Gerontol Int, 2018; 18:758–766.

38.

Lin

, Otsubo

, Imanaka

. The effects of dementia and long-term care services on the deterioration of care-needs levels of the elderly in Japan. Medicine (Baltimore), 2015; 94:e525.

39.

Masao

, Nanako

. The long-term care insurance system in Japan: past, present, and future. JMA J, 2019; 2:67–69.

40.

Ministry of Health Labour and Welfare (MHLW). Patients survey 2019. https://www.mhlw.go.jp/toukei/saikin/hw/kanja/10syoubyo/ Accessed September 21, 2020 .