Abstract
Trunk function is a key determinant of activities of daily living (ADL) after stroke. While the Trunk Impairment Scale (TIS) has been linked to functional outcomes, its prognostic utility in the recovery phase is less established. This study aimed to develop and validate an ADL
prediction model at discharge using the TIS in recovery-phase stroke patients. This prospective cohort study included 80 first-ever stroke patients admitted to Kiminomori Rehabilitation Hospital in Chiba, Japan. Trunk function (TIS), motor function (Stroke Impairment Assessment Set (SIAS)-M), stroke severity (NIH Stroke Scale/Score (NIHSS)), and ADL ( Functional Independence Measure (FIM)-M) were assessed. Three regression models were compared, and
predictive validity was tested using cross-validation and bootstrap analysis. The TIS model showed the highest
predictive accuracy, outperforming baseline and motor function models. Bootstrap analysis confirmed the independent contribution of one-month TIS to discharge ADL outcomes. The TIS assessed one month after stroke is a reliable
predictor of discharge ADL, supporting its use for individualized rehabilitation planning.
Introduction
Trunk function is a key determinant of activities of daily living (ADL) after stroke, as it provides the foundation for postural control and mobility [1-3]. Trunk function, defined as the integrated capacity for coordinated movement, proprioceptive regulation, and trunk muscular strength, plays a key role in supporting postural stability and efficient functional activity after stroke. Impairment of trunk stability can compromise independence, while targeted trunk training has been shown to improve balance, gait, and quality of life [4-5]. Prognostic prediction after stroke depends on multiple factors, among which trunk function plays a particularly critical role [6-7].
In recent years, shorter hospital stays and the need for seamless transition to recovery-phase rehabilitation have increased the importance of reliable prognostic tools [8-11]. During this stage of heightened neuroplasticity [12-14], accurate prediction of functional recovery supports individualized rehabilitation planning and discharge preparation [6,10].
Trunk control is especially important for basic motor tasks, wheelchair-level ADL, and reducing caregiver burden [15-18]. Several assessment tools have been developed, but many focus only on sitting balance or isolated abilities [19-20]. The Trunk Impairment Scale (TIS) overcomes these limitations by evaluating both trunk balance and functional components [21]. While its association with ADL has been reported, evidence for its predictive utility in the recovery phase remains limited.
Recent findings have also demonstrated that early TIS scores can predict gait independence after acute stroke [22], further supporting its potential as a prognostic indicator.
Therefore, the aim of this study was to develop and validate a prediction model for discharge ADL using the TIS in recovery-phase stroke patients. Such a model may enhance prognostic accuracy and support the design of individualized rehabilitation programs.
Materials & Methods
Participants
This prospective cohort study initially included 116 patients who were transferred from acute care hospitals to recovery-phase rehabilitation hospitals (Kiminomori Rehabilitation Hospital, Chiba, Japan) with a diagnosis of cerebral infarction or cerebral hemorrhage between December 2021 and March 2023.
The inclusion criterion was a first-ever unilateral cerebral infarction. At one month after admission to the recovery-phase rehabilitation ward, participants were required to have a level of consciousness classified as awake without stimulation (Glasgow Coma Scale ≥14).
Exclusion criteria included impaired consciousness (Glasgow Coma Scale ≤14), surgical intervention, stroke deterioration, or death. Stroke deterioration was defined as an increase of ≥4 points in the National Institutes of Health Stroke Scale (NIHSS) score between admission and one month after admission to the recovery-phase ward.
Based on these criteria, the final study population consisted of 80 participants (46 men and 34 women).
All participants received a detailed explanation of the study’s purpose, and written informed consent was obtained. For patients unable to provide a signature, consent was obtained from an authorized representative, such as a family member. The study was approved by the Ethics Committee of Kiminomori Rehabilitation Hospital, Chiba, Japan (approval no. 2021-11), and all procedures were conducted in accordance with the principles of the Declaration of Helsinki.
Methods
Demographic data, including age, sex, and length of stay (LOS), were extracted from electronic medical records. Trunk function was assessed using the 7-item Trunk Impairment Scale (TIS; Fujiwara version) [21]. The scale consists of seven performance-based items, with a total score ranging from 0 to 21 points, where higher scores indicate better trunk function.
Two items are adapted from the Stroke Impairment Assessment Set (SIAS) and assess abdominal muscle strength and postural verticality, while the remaining five items uniquely evaluate trunk verticality perception, rotational trunk strength on both the affected and unaffected sides, and bilateral righting reactions.
The Fujiwara version of the TIS has demonstrated high reliability and validity and is widely used in clinical and research settings to quantify trunk impairment after stroke. Stroke severity was assessed using the National Institutes of Health Stroke Scale (NIHSS) [23]. Motor function on the affected side was evaluated using the motor items of the SIAS (SIAS-M) [24], and ADL were assessed with the motor subscale of the Functional Independence Measure (FIM-M) [25].
Only total scores were analyzed, without reproducing or describing any individual items, in accordance with copyright restrictions.
Although both the motor and cognitive subscales of the FIM were collected, only the motor subscale (FIM-M) was used in the analysis because the study focused on physical ADL performance in relation to trunk function, and the discharge FIM was assessed uniformly on the day before discharge for all participants.
Written confirmations regarding appropriate usage conditions were obtained from Wolters Kluwer (TIS and NIHSS), SAGE Publications (SIAS), and UDSMR/Netsmart (FIM).
To account for variability in physical function, assessments were performed one month after admission and again at discharge.
All evaluations were conducted by the same examiner.
Statistical analysis
The Shapiro-Wilk test was used to examine whether variables followed a normal distribution. To address multicollinearity, Spearman’s rank correlation coefficients were calculated, and variables with |r| ≥ 0.9 were excluded. Variance inflation factors (VIFs) ≥ 10 were also used to identify multicollinearity.
Multiple regression analyses were performed to evaluate the predictive validity of discharge ADL. The baseline model included age, length of stay, and one-month FIM-M as explanatory variables, with discharge FIM-M as the dependent variable. The TIS model additionally included one-month TIS, and the SIAS-M model included one-month SIAS-M. For each model, the coefficient of determination (R²), regression coefficients, and standard errors were calculated to compare predictive accuracy.
To assess generalizability, 10-fold cross-validation was performed, and R² and mean squared error (MSE) were calculated for each fold. In addition, bootstrap analysis (1,000 resamplings) was conducted to further validate model stability and reliability, estimating bias, standard errors, and 95% confidence intervals for regression coefficients and predictive accuracy.
All statistical analyses were conducted using IBM SPSS Statistics for Windows, version 29.0 (released 2022, IBM Corp., Armonk, NY) and R version 4.3.1 (R Foundation for Statistical Computing, Vienna, Austria). Statistical significance was set at p < 0.05.
Results
Participant characteristics
Participant characteristics are summarized in Table 1. The median age of participants was 73 years (interquartile range (IQR): 64-80), and the median length of hospital stay was 139 days (IQR: 97-149). The cohort included 46 males and 34 females, with 41 patients presenting right-sided and 39 left-sided lesions. Stroke type consisted of 50 cerebral infarctions and 30 cerebral hemorrhages.
