Patients

A total of 124 consecutive patients with PD who were admitted to Kokura Rehabilitation Hospital between April 1, 2013, and March 31, 2019, were enrolled in this study. Due to emergency transfer, missing data, difficulty of evaluation, and death, finally, 61 patients with 96 consecutive admissions were selected for data analysis. Notably, some patients were admitted more than once during the study period; thus, repeated measurements obtained from one patient were correlated. A detailed flowchart of patient selection is depicted in Fig 1.

Fig 1

Flowchart for patient selection.

Ethical considerations

The study protocol conformed to the ethical guidelines of the Declaration of Helsinki and was approved by the ethics committee of Kokura Rehabilitation Hospital (No.10001) and Kurume University (No.19073). The ethics committee waived the requirement for informed consent because this was a retrospective study. Additionally, we provided information regarding the study to all patients concerned, while giving them an opt-out option, which allowed patients to withdraw from the study at any time. For the protection of patient privacy, patient data were collected with unlinkable anonymization and saved in a password-protected storage medium for research use only. The data set is available upon request to the first author of this paper.

Nutritional status

The nutritional status was diagnosed using the CONUT method [15]. The CONUT method is a method of scoring albumin (Alb), total lymphocyte count (TLC), and total cholesterol (T-cho) values. Scoring methods of these items are illustrated in Table 1.

Table 1

Scoring methods of Alb, TLC, and T-cho values.

Alb (g/dl)	≥3.50	3.00–3.49	2.50–2.99	<2.50
Score	0	2	4	6
TLC (/μL)	≥1600	1200–1599	800–1199	<800
Score	0	1	2	3
T-cho (mg/dl)	≥180	140–179	100–139	<100
Score	0	1	2	3

Alb, albumin; TLC, total lymphocyte count; T-cho, total cholesterol.

The CONUT score was computed using the following formula:

CONUT score = (Alb score) + (TLC score) + (T-cho score).

CONUT scores of 0–1, 2–4, 5–8, and 9–12 are considered to indicate normal nutritional status, light malnourishment, moderate malnourishment, and severe malnourishment, respectively [15].

Activities of daily living

ADLs were assessed using the FIM [16]. The FIM is one of the most common measures of ADL and includes 13 lower-order items regarding motor function and five lower-order items regarding cognitive function. Each item is graded on a scale from 1 (total assistance) to 7 (complete independence). The FIM scores on admission and at discharge were assessed by a team of physical therapists, occupational therapists, and nurses. The motor FIM score is the total score of the motor function, ranging from 13 to 91 points. The primary outcome was FIM gain, which was defined as the difference in the motor FIM scores at discharge and admission.

Data

The following variables were reviewed from the medical records: age, sex, number of days of hospitalization, route of hospitalization, Alb (g/dl), TLC (/μL), T-cho (mg/dL), BMI, PD severity, dysphagia, grip strength, adjusting anti-Parkinsonian drugs, weight on admission and discharge, and the FIM. The CONUT score was calculated from the Alb, TLC, and T-cho values. The weight loss rate was defined as the difference between the weight at discharge and admission, divided by the weight on admission. The Hoehn and Yahr stage (HY stage) [17] and the Food Intake LEVEL Scale [18] were also considered. The HY stage was used to classify patients with PD into two states, severe (HY stage 4–5) or not severe (HY stage 1–3), and the Food Intake LEVEL Scale was employed to classify dysphagia as “Yes” when it was ≤9 and “No” when it was 10.

Statistical analysis

Because randomization was not employed in the study design, study findings could be easily influenced by confounding factors. We identified confounding factors at the beginning, based on the data, as follows: First, variables associated with FIM gain at a significance level of 10% were selected. Then, among these variables, those that were found to have a significant association with the CONUT score, at a significance level of 10%, were identified as confounding factors.

A mixed-effects model was employed to deal with the correlated measurements from a single patient due to repeated admissions. Also, linear and piecewise linear models with one knot were considered, and they were combined with the mixed-effects model. The combined models were fitted to the data after adjusting for confounders, and in the case of piecewise linear model, the model was repeatedly fitted to the data by changing the values of the knot in the CONUT scores. The model that achieved the smallest Akaike’s Information Criteria was selected as the best model. Details of the models and model fitting process are described in the supporting information section (S1 File). Statistical analysis was performed using SAS software version 9.4 (SAS Institute Inc., Cary, NC, USA).

Baseline characteristics of the patients

Table 2 shows the baseline characteristics of the patients. The values in the table are presented as the median and interquartile range for numerical variables and as numbers and percentages (%) for categorical variables. Age, BMI, CONUT score, and grip strength values presented in the table were obtained on admission. The median age was 76.5 years. Of the 96 admissions, 53 (55.2%) were male, 53 (55.2%) patients with PD were categorized as severe, and 55 (57.3%) patients had dysphagia.

Table 2

Baseline characteristics of the patients.

Background factors	Median (IQR)
Age (years)	76.5 (70, 79.7)
Number of days of hospitalization (days)	80 (49.2, 136.2)
Laboratory data
Albumin (g/dL)	3.6 (3.2, 3.9)
Total lymphocyte count (/μL)	1,581 (1,299.4, 1953.6)
Total cholesterol (mg/dL)	161 (144, 179.7)
BMI (kg/m2)	19.6 (17.3, 22.7)
Weight loss rate (%)	0.7 (-5.3, 3.3)
Grip strength (kg)	17.9 (12.9, 26.8)
FIM (score)
Total	79.5 (34, 100.7)
Motor	49.5 (15.7, 68)
Cognitive	29 (14.2, 35)
Gain	6 (0, 13.7)
Background factors	N. (%)
Male sex	53 (55.2)
Route of hospitalization
Acute care hospital	59 (61.5)
Home	34 (35.4)
Others	3 (3.1)
CONUT score
0–1	27 (28.1)
2–4	46 (47.9)
5–8	23 (24.0)
9–12	0 (0)
PD severity
HY stage 1–3	43 (44.8)
HY stage 4–5	53 (55.2)
Dysphagia
No	41 (42.7)
Yes	55 (57.3)
Adjusting anti-Parkinsonian drugs
No	65 (67.7)
Yes	31 (32.3)

IQR, interquartile range; BMI, body mass index; FIM, Functional Independence Measure.

CONUT, Controlling Nutritional Status; PD, Parkinson’s disease; HY stage, Hoehn and Yahr stage.

Identification of confounding factors

Variables that were significantly associated with the FIM gain, at the 10% significance level, included PD severity (p<0.001), dysphagia (p = 0.005), cognitive FIM (p = 0.089), and grip strength (p = 0.080). Furthermore, variables that were significantly associated with the CONUT score, at a 10% level of significance, included PD severity (p = 0.011), dysphagia (p = 0.001), cognitive FIM (p = 0.001), and grip strength (p<0.001). Thus, PD severity, dysphagia, cognitive FIM, and grip strength were identified as confounding factors. Fig 2 shows the relationship of the four confounding factors, and these factors were strongly related to each other. We selected PD severity as a representative confounder.

Fig 2

Relationships of the four confounding factors.

The values in the figure represent the p-value between the two variables. PD, Parkinson’s disease; FIM, Functional Independence Measure.

Relationship between the CONUT score and FIM gain

The piecewise linear model with knot = 3 was selected as the best fitted model. The estimated parameters of the best fitted model are summarized in Table 3. Details of the model formula and process of estimating parameters in the formula are presented in the (S1 File). From the table, the relationship between the FIM gain and the CONUT score, after adjusting for age, sex, and PD severity, is given as follows:

If the CONUT score is ≤3:

FIM gain = 5.864 + 1.465 × CONUT score + 0.067 × age + 1.383 × sex − 8.948 × PD severity,

and if the CONUT score is >3:

FIM gain = 16.421− 2.054 × CONUT score + 0.067 × age + 1.383 × sex − 8.948 × PD severity,

where sex = $$ \{\begin{array}{c}0:\mathrm{m}\mathrm{a}\mathrm{l}\mathrm{e}\\ 1:\mathrm{f}\mathrm{e}\mathrm{m}\mathrm{a}\mathrm{l}\mathrm{e}\end{array}$$, PD severity = $$ \{\begin{array}{c}0:\mathrm{n}\mathrm{o}\mathrm{t}\mathrm{s}\mathrm{e}\mathrm{v}\mathrm{e}\mathrm{r}\mathrm{e}\\ 1:\mathrm{s}\mathrm{e}\mathrm{v}\mathrm{e}\mathrm{r}\mathrm{e}\end{array}$$.

Table 3

Estimated parameters and knot in the best fitted model.

Parameter	Coefficient	p-value	95%CI
Intercept	5.864	0.602	-16.528	28.257
CONUT score ≤3	1.465	0.183	-0.729	3.659
CONUT score >3	-2.054	0.026	-3.847	-0.260
Age	0.067	0.665	-0.248	0.383
Sex	1.383	0.549	-3.277	6.044
PD severity	-8.948	<0.001	-13.360	-4.535
Knot of piecewise linear model	3
$$ (between subject variance)	32.898
$$ (within subject variance)	61.760

CI, confidence interval; CONUT, Controlling Nutritional Status; PD, Parkinson’s disease.

Assigning mean values respectively to age, sex (0: male, 1: female) and PD severity (0: not severe, 1: severe) in the best fitted model the estimated relationship of the CONUT score and FIM gain may be illustrated in Fig 3. The fixed effect was 1.465 for a CONUT score ≤3 and -2.054 for a CONUT score >3, with a random effect of 1. Fig 3 and Table 3 reveal that while FIM gain shows no significant decrease when the CONUT scores are ≤3 (slope coefficient, 1.465; p = 0.183), it decreases significantly when the CONUT scores are >3 (slope coefficient, -2.054; p = 0.026).

Fig 3

Estimated relationship of the CONUT score and FIM gain.

Mean values are assigned respectively to age, sex (0: male, 1: female), and PD severity (0: not severe, 1: severe) in the best fitted model. FIM, Functional Independence Measure; CONUT, Controlling Nutritional Status.