We conducted a retrospective cohort study of patients with stroke admitted to a convalescent hospital. The inclusion criterion was patients with subacute stroke admitted to Kyoto Rehabilitation Hospital between January 2020 and October 2023. The exclusion criteria were patients with infratentorial lesions, subarachnoid haemorrhage, multiple lesions, a history of stroke, needing assistance in ADL before the stroke, an onset-to-admission interval of more than 120 days, hospitalization for 6 days or less, sudden deterioration of symptoms, no paralysis, FIM-walking score ≥6 on admission, and incomplete data. Among the 544 patients with hemiplegia admitted during the same period, 206 were included in the analysis after 338 who met the exclusion criteria were excluded (Fig. 1). All patients underwent physical and occupational therapy and speech therapy, if needed, for 2.5–3 h daily, 7 days a week, from admission to discharge, in accordance with the Full-time Integrated Treatment (FIT) programme for standardizing training content and intensity (17). This programme aimed to facilitate a standardized approach among therapists by setting exercise difficulty levels on the basis of motor learning principles, focusing on ADL and walking training. Additionally, the FIT programme helped minimize variability in therapists’ levels of expertise or treatment approaches through a team-based structure, utilizing a dual therapist system with designated primary and secondary therapist roles (17). In addition, the rehabilitation plan, goal setting, and length of stay during hospitalization were determined at regular conferences held by the attending physician, nurses, physical therapists, occupational therapists, speech therapists, medical social workers, and national registered dietitians.

This study was approved by the Ethics Committee of Fujita Health University (approval number: HM22-504; August 10, 2021) and registered in the University Hospital Medical Information Network-Clinical Trials Registry (UMIN-CTR UMIN 000053209). Since this was a retrospective observational analysis using only existing data, no written or verbal consent was required or obtained from the patients. Study details were disclosed to the patients so that they could decline their data’s use for research (opt-out).

Upon admission, data on patient age, sex, stroke type, paralysed side, number of days from onset to hospital admission, number of days from onset to hospital discharge, total lower-extremity motor function score of the Stroke Impairment Assessment Set (SIAS LE) (18), total FIM-motor item score (FIM-M) (10), FIM-walking score (a subitem of FIM-M) (10), total FIM-cognitive item score (FIM-C) (10), and FAC score (7) were recorded. Additionally, the primary endpoint variable, FIM-walking, was assessed every 2 weeks from admission to discharge. In our study, we defined “real-life walking” as the performance of practical walking in a structured hospital ward. Specifically, patients were evaluated when they rose from their hospital beds, opened doors to their rooms, navigated around other patients who might be walking in the hallway, and proceeded to their designated destinations, such as the restroom or dining area. Additionally, this assessment included conditions during daytime and nighttime and immediately upon awakening. The FIM (10) is an ADL evaluation method consisting of 18 evaluation items, 13 of which address motor and 5 of which address cognitive issues. Each item is rated on a 7-point Likert scale from 1 to 7, depending on the degree of assistance needed. The FIM-walking subitem is evaluated under conditions of actual walking performance in real-life situations. Independence in this study was defined as an FIM-walking score of 6 or higher (6 = modified independence, 7 = complete independence). Complete independence (score of 7) means that the individual can safely walk at least 150 ft (50 m) without assistive devices. Modified independence (score of 6) means that the individual can walk at least 150 ft (50 m) without supervision but with the support of an orthosis, cane, or walkerette. Specifically, a FIM-walking score of 6 or higher indicates that the patient can independently ambulate in the actual ward setting without the need for assistance or supervision. FIM scores of 1–5 are determined by the level of physical assistance required for walking and the presence of supervision. The reliability and validity of this scale have been confirmed in patients with stroke (19). In this study, FIM assessments and recordings were conducted by physical and occupational therapists who received training and certification in FIM use.

Patients scoring 6 or higher at the point of discharge were allocated into the Independent group, whereas those scoring 5 or lower at the point of discharge were included in the Non-Independent group. The number of patients for analysis was set at 102 in the Independent group and 104 in the Non-Independent group (Fig. 1). Patient information on admission was compared between the Independent and Non-Independent groups using the Mann–Whitney U test for ordinal variables and the Pearson χ² test for nominal variables. We then performed a survival-time analysis using improvement to a FIM-walking score of 6 or higher as the event during hospitalization, referencing previous studies (6), Kaplan–Meier plots, and log-rank tests. Survival-time curves were compared and stratified by age, FIM-M score, FIM-C score, and FAC score on admission. The number of weeks from admission until the FIM-walking score improved to 6 or higher was assigned as the time variable. Patients discharged without a FIM-walking score of 6 or higher during hospitalization were censored at the number of weeks from admission to discharge.

Additionally, the probability of achieving a FIM-walking score of 6 or higher during hospitalization was calculated using a multivariate Cox proportional hazards model. Age (3, 4), FIM-M score (20), FIM-C score (21), which have previously been reported to influence walking ability, and FAC score on admission were included as explanatory variables to estimate hazard ratios (HRs). The forward selection (likelihood ratio) method was used for variable extraction. To account for bias in the distribution of scores and because high or low scores for each variable may have reciprocal effects on the results, the variables were stratified as follows on the basis of previous reports. Age was stratified into 64 years or younger (low age), 65–80 years (middle age), and 81 years or older (high age) (6); FIM-M score was stratified into 13–38 points (low ADL), 39–64 points (middle ADL), and 65–91 points (high ADL) (22); and FIM-C score was stratified into 5–14 points (low cognition), 15–24 points (medium cognition), and 25–35 points (high cognition) (22). FAC scores were classified as 0, 1, 2, or 3 or more points. These classifications were used in the model.

SPSS software (version 16.0 for Mac; IBM Japan Ltd, Tokyo, Japan) was used for statistical analyses. The significance level was set at 5%.

None of the patients declined participation or passed away during hospitalization. Table I shows patient information on admission. The Independent group was significantly younger (mean [standard deviation (SD)]: Independent group, 71.12 [13.67] years; Non-Independent group, 77.64 [9.00] years; p < 0.001), had a shorter time from onset to admission (mean [SD]: Independent group, 24.72 [13.73] days; Non-Independent group, 29.87 [15.07] days; p = 0.003), had a shorter time from onset to discharge (mean [SD]: Independent group, 105.66 [39.35] days; Non-Independent group, 140.95 [48.80] days; p < 0.001), and had higher SIAS LE scores (median [interquartile range (IQR)]: Independent group 12 [9–14]; Non-Independent group, 6 [1–12]; p < 0.001) than did the Non-Independent group on admission. Compared with the Non-Independent group, the Independent group had greater FIM-M (median [IQR]: Independent group, 50.5 [38–60]; Non-Independent group, 19 [13–30]; p < 0.001), FIM-walking (median [IQR]: Independent group, 1 [1–4]; Non-Independent group, 1 [1–1], p < 0.001), FIM-C (median [IQR]: Independent group, 27 [21–31]; Non-Independent group, 16.5 [10–21]; p < 0.001), and FAC (median [IQR]: Independent group, 2 [1–3]; Non-Independent group, 1 [0–2]; p < 0.001).

The median time from admission to achieving an FIM-walking score of 6 or higher was 4 weeks (IQR: 2–8 weeks). From admission to discharge, 102 patients (49.5%) could walk independently in real-life situations. Fig. 2 shows the relationships between age, FIM-M, FIM-C, and FAC on admission and the time to achieving a FIM-walking score of 6 or higher using Kaplan–Meier curves without adjustment for covariates. The log-rank test was significant (P < 0.05) in each analysis, indicating that the time to achieving an FIM-walking score of 6 or higher and the percentage of patients who achieved this goal differed based on age, ADL, cognitive function, and “test-setting walking” ability on admission (Fig. 2).

Multivariate Cox proportional hazards analysis identified independent factors positively associated with achieving a FIM-walking score of 6 or higher early during hospitalization: age ≤64 years (HR 1.92, 95% confidence interval [CI] 1.21–3.06, p < 0.01), FIM-C ≥25 (HR 2.42, 95% CI 1.52–3.86, p < 0.01), and FAC ≥3 (HR 1.98, 95% CI 1.22–3.21, p < 0.01) on admission (Table II). On the other hand, FIM-M ≤ 38 (HR 0.23, 95% CI 0.13–0.40, p < 0.01) and FAC = 0 (HR 0.18, 95% CI 0.06–0.62, p < 0.01) on admission were identified as factors that inhibited early achievement of walking independence.

A limitation of this study is that although the FIM score was used to measure “real-life walking” performance during hospitalization, how accurately this measurement could reflect patients’ walking ability in more complex and unpredictable everyday situations outside the hospital environment remains unclear. Community ambulation poststroke has been reported to be affected by more complex factors than walking in a structured environment such as a hospital ward (35). Additionally, this study included patients with haemorrhagic stroke as well as severe stroke, who are expected to take longer to achieve independence (6), but the study was not validated after censoring at the point of discharge. Future research should include situations outside ward life and follow-up assessments after discharge. Furthermore, the content and intensity of rehabilitation may affect the outcomes of walking performance (36–38). However, owing to the difficulty in collecting information on the specific content and intensity of rehabilitation during hospitalization, this remains a challenge for future studies. Different countries and cultures may have different rehabilitation outcomes (39). The study highlights that, compared with high-resource settings, clinical practice guidelines for stroke rehabilitation are often underutilized in low- and middle-income countries owing to barriers such as limited resources, which could affect recovery outcomes (40). Our study represents a step towards understanding the specific patient population in Japan, but further validation is needed in diverse rehabilitation settings. Finally, the possibility of selection bias cannot be ruled out because most severely affected patients might not have participated in the study due to excluding patients with a history of stroke or multiple lesions. Owing to the retrospective design of this study, selective outcome reporting bias cannot be eliminated, and measurement accuracy may be limited. Future research should aim to validate these results through prospective studies.

Among patients with subacute stroke who were admitted to a convalescent rehabilitation ward, those who were younger, had greater cognitive function, and exhibited greater “test-setting walking” ability on admission were likely to experience earlier improvements in “real-life walking” performance. Additionally, patients with lower ADL independence and reduced “test-setting walking” ability on admission were less likely to show improvement. This information could facilitate the development of more efficient rehabilitation plans.