Development and internal validation of two prediction models of walking dependence and restricted walking speed in individuals with subacute stroke

doi:10.21203/rs.3.rs-3639244/v1

Download PDF

Research Article

Development and internal validation of two prediction models of walking dependence and restricted walking speed in individuals with subacute stroke

https://doi.org/10.21203/rs.3.rs-3639244/v1

This work is licensed under a CC BY 4.0 License

You are reading this latest preprint version

Background

The prognosis of the walking ability of individuals with stroke affects the choice of rehabilitation program. Identifying patients who will need assistance with ambulation at discharge allows clinicians to deliver rehabilitation programs focused on educating caregivers and adjusting the patient’s environment to allow safe transfer within the home. The primary objective of this study was to develop and internally validate a prediction model of walking dependence for patients with stroke admitted to a rehabilitation facility. The secondary objective was to establish a prediction model of restricted walking speed.

Methods

This retrospective cohort study included 476 individuals with subacute stroke who were admitted to a rehabilitation facility. Model 1 was developed to predict the probability of walking dependence. Model 2 was developed to predict restricted walking speed. Walking dependence was defined as a functional independence measure walk score of 5 or less. Restricted walking speed was defined as the ability to walk at 0.93 m/s or less. Potential predictors, including age, leg strength of the affected side, sitting balance, cognitive function, and urinary function, were selected based on the literature and analyzed using logistic regression analysis. Models were internally validated using the bootstrap method. Model performance was assessed by discrimination (area under the receiver operating characteristic curve (AUC)) and calibration (Hosmer–Lemeshow (H–L) goodness-of-fit test and calibration plots).

Results

A total of 236 patients (49.6%) walked dependently at discharge. Of the 240 individuals who achieved walking independence, 121 (50.4%) had restricted walking speed. In model 1, older age, poor leg strength, sitting balance, cognitive function, and urinary incontinence were significantly associated with walking dependence at discharge. The AUCs of models 1 and 2 were 0.93 (95% confidence interval (CI) = .90–.95) and 0.69 (95%CI = .62–.76), respectively. Both models had good calibration confirmed by the H-L test.

Conclusions

The internally validated prediction model of walking dependence had good discrimination and calibration, while the prediction model of restricted walking speed had poor discrimination. The prediction model for walking dependence developed in this study may be useful for planning rehabilitation strategies and setting realistic goals for patients.

stroke

prognosis

walking independence

locomotion

prediction

gait

convalescent rehabilitation

The reported prevalence of stroke is 101 million cases globally, with 33% of cases involving individuals over the age of 70 years and increasing incident cases [1, 2]. Gait disorder is a significant sequela of stroke. Among patients with stroke who cannot walk independently on admission to rehabilitation facilities, approximately 40% remain dependent for walking at discharge [3]. Identifying patients, early after admission, who are likely to be walking-dependent at discharge, allows clinicians to deliver rehabilitation programs that focus on educating caregivers and adjusting the patient’s environment to allow safe transfer within the home [4]. Furthermore, early identification of patients who can achieve walking independence with a restricted walking speed, allows the implementation of rehabilitation plans to increase walking speed and adjust environmental factors to support walking in the community. Even if walking independence in the home is achieved, community ambulation may be limited if the individual walks 0.93 m/s or less [5]. Therefore, prediction models of walking ability at discharge are important for planning rehabilitation strategies and setting realistic goals [6].

Previous longitudinal studies of walking dependence and walking speed in stroke have only reported associations in terms of odds ratios and other measures [7–20]. Since it is clear that the strong associations captured by odds ratios do not translate into predictive ability in treating specific patients, there is a need to use statistical methods focused on predicting outcomes for specific individuals and to demonstrate the discrimination and calibration of prediction models [6, 21–23]. In particular, calibration is considered a primary requirement for the clinical use of prediction models because it indicates the likelihood that an actual event will occur with respect to the individual to whom the prediction model is applied [23]. In addition, the development of prediction models requires a process of internal validation to assess the possibility that models may be overfitted to the development data and not be reproducible [24].

Previous studies reported that older age [7–10], poor leg strength of the affected side [9–14], poor sitting balance [14–18], the presence of cognitive impairment [10, 20], urinary incontinence [19, 20], and lower levels of functional independence [8, 12, 19, 20] were associated with walking dependence and slow walking speed in individuals with stroke. Combining these factors will provide a more definitive ability to predict walking outcomes.

However, to our knowledge, no study has developed and internally validated a walking dependence prediction model and restricted walking speed prediction model with subacute stroke using the statistical methods required for prediction model development. The primary aim of this study was to develop and internally validate a model that predicts walking dependence at discharge in patients with subacute stroke in a rehabilitation facility using age, leg strength, sitting balance, cognitive function, urinary function, and functional independence on admission. In addition, we hypothesized that these variables could predict restricted walking speed among patients with walking independence. Therefore, our secondary objective was to develop and internally validate a restricted walking speed prediction model using the aforementioned predictors.

Study design

This study used a retrospective cohort study design. The study protocol was approved by the ethics committees of the Shinshu University and Toyama Prefectural Rehabilitation Hospital & Support Center for Children with Disabilities (No. 5557 & No. 86). The need for participant consent was waived because of the retrospective design of the study. The study was reported according to the Transparent Reporting of a multivariable prediction model for Individual Prognosis or Diagnosis (TRIPOD) statement [21]. The TRIPOD checklist is included in Appendix 1. We analyzed the medical records of patients admitted to the rehabilitation facility between August 2019 and March 2022. The inclusion criteria for participants were age > 18 years, a diagnosis of cerebral infarction or intracerebral hemorrhage, and a Functional Independence Measure (FIM) walk score of less than 6 points on admission. The exclusion criteria were transfer to an acute care facility or death during hospitalization.

Outcome and associated variables

The primary outcome was walking dependence at discharge from the rehabilitation facility. The secondary outcome was restricted walking speed at discharge from the rehabilitation facility. Walking dependence was defined as a FIM walk score of < 6 [25]. Restricted walking speed was defined as a walking speed of < 0.93 m/s in addition to walking independence [5]. These outcomes were assessed within two weeks before discharge from the rehabilitation facility.

The associated factors were based on published studies. We used age, leg strength, sitting balance, cognitive function, urinary function, and ADL independence [7–20]. The four variables of leg strength, sitting balance, cognitive function, and urinary incontinence were dichotomized, as previously described [26]. We defined leg strength using the knee extension item of the Stroke Impairment Assessment Set (SIAS) [27] and identified a score of < 3 as poor and ≥ 3 as good; a score of 3 points indicated the ability to extend the knee > -10° horizontally in a back-supported sitting position. We defined sitting balance using the sitting items of the Berg Balance Scale (BBS) [28] and identified a score of < 2 as poor and ≥ 2 as good; a score of 2 indicated the ability to hold a sitting position for 30 s unsupported. We defined cognitive function using the Mini-Mental State Examination (MMSE) and identified scores of < 24 as poor and ≥ 24 as good [29]. When the MMSE could not be conducted because of severe impairment of consciousness or aphasia, the participants were included as individuals with poor cognitive function. We defined the presence of urinary incontinence using the FIM score for bladder management and identified a score of < 6 as indicating the presence of urinary incontinence [25]. We defined the levels of independence in ADL with a total FIM score [25]. All potential predictors were assessed within two weeks after admission. The assessments used in this study were intended to be simple and easy to use, based on core clinical assessments recommended for stroke and the most commonly used cognitive assessments worldwide [30, 31].

Sample size

Sample size was determined based on an Event Per Variable (EPV) ≥ 10, which is commonly used in logistic regression analysis for six explanatory variables [32]. Consequently, a minimum of 60 participants in each group was deemed necessary.

Data analysis

Dealing with missing data

Multiple imputation methods were undertaken to estimate missing data over 50 iterations to generate 20 complete datasets for model development [33]. To determine whether imputation led to radically different results, a sensitivity analysis was performed by comparing the outcomes of the imputed data set with complete cases only.

Study population characteristics

The patients were divided into two groups, walking dependent and independent, based on an FIM walk score of < 6 at discharge. The walking independent group was further divided into two groups, restricted walking speed group and unrestricted walking speed group. Univariate analysis was initially conducted to determine whether there was any difference among the groups for each demographic and clinical characteristic using analysis of variance for continuous variables and a χ² test or a Fisher exact test for categorical variables.

Multicollinearity

The correlation coefficient of 0.5 or more was considered multicollinearity for the six variables [34]. When multicollinearity was present, the variable with the higher variance inflation factor (VIF) was excluded from the explanatory variables.

Model development

We used multivariate logistic analysis to develop two models; model 1 identified walking dependence versus independence groups, and model 2 identified restricted walking speed versus unrestricted walking speed groups. We used a backward stepwise selection method (p < .05) to develop a simpler and more accurate model.

Models internal validation and optimism correction

Internal validation of the models was assessed using a bootstrap method followed by a multiple imputation method [35]. Bootstrap samples of the same size as the original sample (B = 1000) were drawn with replacements from the original data, reflecting the drawing samples from the underlying population. Shrinkage factors were estimated to adjust the model coefficients to prevent overfitting the models. For a shrinkage factor less than 0.9, the intercept and coefficients were re-estimated by multiplying by the shrinkage factor [36].

Model performance

The Receiver Operating Characteristic curve (ROC) and Area under the curve (AUC) were used to evaluate the discrimination of the models. Model 1, identified patients who would remain dependent for walking. Model 2 identified patients who would continue to have a restricted walking speed. An AUC of 0.7 to 0.8 is considered acceptable, an AUC of 0.8 to 0.9 is considered excellent, and an AUC of more than 0.9 is considered outstanding [37]. The agreement between the predicted probabilities and observed frequencies was visualized in the calibration plots and confirmed as a good model fit with the Hosmer–Lemeshow test being non-significant. Calibration plots were created by sorting the participants in order of predicted probability, dividing them into 10 groups, and plotting the mean predicted probability for each group versus the observed frequency of the outcome. A calibration plot plots a point on a straight line when the prediction and observation are in perfect agreement. Statistical analyses were performed in R software (version 4.1.1, https://www.r-project.org) using “psfmi” package [38].

Participants

A flow chart of the participants enrolled in this study is shown in Fig. 1. A total of 768 individuals with stroke were assessed for eligibility and 292 were excluded from the study. The main reason for exclusion was having a FIM walk score of 6 or more points (n = 172). Finally, 476 patients were included in the analysis.

Comparisons of the participants’ characteristics and functional outcomes between the groups

Table 1 lists the characteristics of the participants. At discharge from the rehabilitation facility, 236 participants (49.6%) were walking-dependent. Of the participants who could walk independently, 121 (50.4%) had a restricted walking speed. The average days to walking independence were 64.4 (standard deviation 33.0) from stroke onset. Compared to patients in the walking independent group, patients in the walking dependent group were older and had urinary incontinence, poor leg strength, sitting balance, and cognitive function, and lower levels of ADL independence (p < .05). Compared to patients in the unrestricted walking speed group, patients in the group with restricted walking speed were older, had poor leg strength, and urinary incontinence (p < .05). However, there were no significant differences in sitting balance and cognitive function between the groups.

Table 1

Comparisons of participants’ demographic and clinical characteristics according to model groups
	Model 1
Characteristic	Total (n = 476)	Walking dependent (n = 236)	Independent (n = 240)	p-value	Restricted walking speed (n = 121)	Unrestricted walking speed (n = 119)	p-value
Age, y	71.2 ± 13.0	74.6 ± 11.9	67.9 ± 13.2	< 0.001	70.5 ± 12.6	65.2 ± 13.2	0.002
Sex, female	185 (38.9)	93 (39.4)	92 (38.3)	0.810	43 (43.8)	39 (32.8)	0.079
BMI, kg/m²	22.6 ± 4.0	21.7 ± 3.7	23.5 ± 4.0	< 0.001	23.0 ± 3.6	24.0 ± 4.3	0.065
Stroke type, ICH	175 (36.8)	79 (33.5)	96 (40.0)	0.140	46 (38.0)	50 (42.0)	0.527
Previous stroke, presence	122 (25.6)	72 (30.5)	50 (20.8)	0.016	33 (27.3)	17 (14.3)	0.013
No. of days from stroke onset to admission, days	30.3 ± 15.1	33.6 ± 15.9	27.0 ± 13.5	< 0.001	28.1 ± 13.6	25.9 ± 13.3	0.157
Total score of the SIAS-motor, points	9.2 ± 5.2	6.2 ± 5.3	11.6 ± 3.7	< 0.001	10.7 ± 4.2	12.6 ± 3.0	< 0.001
Total score of the BBS, points	22.0 ± 19.0	8.3 ± 11.5	35.3 ± 14.9	< 0.001	30.2 ± 13.8	40.5 ± 14.2	< 0.001
Motor FIM, points	51.3 ± 25.2	34.0 ± 19.0	68.2 ± 17.9	< 0.001	63.2 ± 17.7	73.3 ± 16.8	< 0.001
Cognitive FIM, points	23.1 ± 8.7	18.7 ± 8.2	27.4 ± 6.9	< 0.001	26.7 ± 6.9	28.2 ± 6.9	0.046
Total FIM, points	74.4 ± 32.1	52.6 ± 25.2	95.6 ± 22.5	< 0.001	89.9 ± 21.9	101.4 ± 21.6	< 0.001
Aphasia, no	369 (77.5)	183 (77.5)	186 (77.5)	0.991	100 (82.6)	86 (72.3)	0.054
Leg strength, poor	203 (36.1)	163 (61.8)	40 (15.3)	< 0.001	28 (21.2)	12 (9.3)	0.011
Sitting balance, poor	106 (20.1)	100 (40.4)	6 (0.4)	< 0.001	4 (3.3)	2 (1.7)	0.999
Cognitive function, poor	277 (58.2)	177 (75.0)	100 (41.7)	< 0.001	49 (40.5)	51 (42.9)	0.711
Urinary incontinence, presence	256 (53.4)	201 (85.0)	55 (22.6)	< 0.001	36 (29.8)	19 (15.3)	0.007
Length of hospital stay, days	83.0 ± 31.4	91.5 ± 31.0	74.7 ± 29.5	< 0.001	79.0 ± 31.7	70.4 ± 26.6	0.041
No. of days from stroke onset to discharge, days	113.3 ± 36.4	125.0 ± 35.3	101.7 ± 33.7	< 0.001	107.2 ± 35.4	96.2 ± 31.0	0.014
No. of days from stroke onset to walking independent, days			64.4 ± 33.0		76.2 ± 34.2	52.4 ± 27.1	< 0.001
Values are presented as mean ± standard deviation or number (%). P-values indicate significant differences between the groups at discharge.
BMI, body mass index; ICH, intracerebral hemorrhage; SIAS, Stroke Impairment Assessment Sets; BBS, Berg Balance Scale; FIM, Functional Independence Measure

Multicollinearity

The FIM total score correlated with sitting balance and urinary function at 0.50 and 0.79, respectively. Multivariate logistic regression analysis revealed multicollinearity for the FIM total score with the highest VIF, which was excluded from the analysis.

Model development and internal validation

For Model 1, multiple imputations of missing data were made for 59 of the 476 participants; 42 had missing data for leg strength, 13 for sitting balance, and 4 for urinary function. Table 2 shows the regression coefficients, odds ratios, and regression equations for each model.

Table 2

Regression coefficients and odds ratios with 95% CI from the original and internally validated models
	Original Model			Model after internal validation
Variables	Regression coefficient	Odds Ratio (95% CI)	p-value	Regression Coefficient^a
Model 1
Intercept	−6.331		< 0.001
Age (years)	0.048	1.049 (1.023–1.076)	< 0.001
Leg strength (poor)	2.544	12.733 (6.237–25.997)	< 0.001
Sitting balance (poor)	3.189	24.252 (3.129–187.942)	0.002
Cognitive function (poor)	0.929	2.532 (1.374–4.667)	0.003
Urinary incontinence (presence)	2.131	8.420 (4.674–15.168)	< 0.001
Model 2
Intercept	–3.220		< 0.001	–2.784
Age (years)	0.042	1.043 (1.019–1.067)	< 0.001	0.036
Leg strength (poor)	1.387	4.004 (1.721–9.319)	0.001	1.200
Urinary incontinence (presence)	0.783	2.187 (1.127–4.247)	0.021	0.677
To estimate the individual probability of walking dependence or restricted walking speed at discharge: Model 1: P _{(Walking dependence)} = 1/ (1 + e ^{(−(−6.331+0.048Age+2.544Leg strength+3.189Sitting balance+0.929Cognitive function+2.131Urinary incontinence))}) 100% Model 2: P _{(Restricted walking speed)} = 1/ (1 + e ^{(−(−2.784+0.036Age+1.200 Leg strength+0.677Urinary incontinence))}) 100%
CI confidence interval, MMSE Mini-Mental State Examination
^a Regression coefficients after adjustment for overfitting by shrinkage (shrinkage factor model 2 = 0.86); the intercept was re-estimated

Model 1, all variables were selected. Internal validation yielded a shrinkage factor of 0.96. Because of the shrinkage factor ≥ 0.9, the intercept and coefficients were not re-estimated [36]. The regression equations can be used to calculate the probability of walking-dependence.

Model 2, age, leg strength, and urinary incontinence were selected, and sitting balance and cognitive function were excluded. Internal validation yielded a shrinkage factor of 0.87. Because of the shrinkage factor < 0.9, the intercept and coefficients were re-estimated by multiplying by the shrinkage factor [36]. The regression equations can be used to calculate the probability of restricted walking speed.

Model performance

Model 1 had good discrimination with an AUC of 0.928 (0.903–0.947) and the AUC corrected by internal validation was 0.925 [37]. The H–L tests were not significant (p = 0.61).

Model 2 had poor discrimination with an AUC of 0.694 (0.624–0.756) and the AUC corrected by internal validation was 0.672 [37]. The H–L tests were not significant (p = 0.41).

Figure 2 shows that the calibration plot, indicating the agreement between prediction probability and observed frequency, lies almost on the reference line for model 1. For model 2, the points are concentrated around the predicted probability of 0.5, and the larger ones deviate from the reference line by approximately 0.2. Figure 3 shows the ROC curves for both models obtained from the data sets created by the multiple imputation method.

Sensitivity analysis

The sensitivity analyses showed similar results for selected variables and performance for the imputed and the original non-imputed data sets. We concluded that imputation did not result in large differences.

In the present study, we developed and internally validated two prediction models that can be used to predict walking ability at discharge from subacute rehabilitation facilities in patients with stroke. Model 1 predicted the probability of walking dependence with excellent discrimination and accurate prediction and included five predictors: age, leg strength, sitting balance, cognitive function, and urinary function. Model 2 predicted restricted walking speed with good calibration but poor discrimination and included three predictors: age, leg strength, and urinary function.

As walking dependence in patients with stroke is associated with the time spent standing and walking [39], walking-dependent patients can be expected to engage in sedentary behavior. Sedentary behavior is defined as time spent engaged in sitting or lying down activities that require an energy expenditure of 1.0 to 1.5 basal metabolic rates [40]. Sedentary behavior is associated with all-cause mortality, cardiovascular disease, type 2 diabetes, and metabolic syndrome in adults [41]. Previous qualitative studies suggest that stroke survivors could reduce sedentary behavior if they receive appropriate support from service staff and caregivers [42]. A lack of physical opportunities, such as having access to adequate training facilities, can present a barrier to reducing sedentary behavior [43]. Therefore, for patients who are predicted to be walking-dependent at discharge, engaging with them in a way that encourages activity, including measures taken to promote family education and service coordination, will be necessary during hospitalization. Walking speed is a predictor of daily step count. Older adults achieving less than 8000 steps/day were identified for walking speeds less than 0.97 m/s, and stroke patients achieving less than 7500 steps/day were identified for walking speeds less than 0.93 m/s [5, 44]. Exercise and behavior change interventions can increase the number of steps per day [45–47]. Therefore, a more activity-promoting approach is needed for patients predicted to have a restricted walking speed at discharge.

In this study, 236 (49.6%) of the participants remained walking-dependent at discharge from the subacute rehabilitation facility. This result is consistent with results of a previous study reporting that 40% of patients remained walking-dependent at discharge from subacute rehabilitation facilities [3]. Also, 121 (50.4%) of the participants who could walk independently at discharge had a restricted walking speed. However, previous studies reported that individuals with restricted walking speed accounted for approximately 70.0% of community-dwelling people with stroke [5, 48]. The discrepancy in the percentage of participants categorized as having restricted walking speed between studies may be attributed to whether the patient undergoes intensive physical therapy. The participants in this study were admitted to rehabilitation facilities, whereas previous studies have investigated stroke survivors living in the community. Because several types of interventions can improve walking speed of patients with stroke, it is possible that many of the participants in this study achieved unrestricted walking speed because of those interventions [49, 50].

Model 1 can predict walking dependence with a simple assessment. The calibration plot shows 10 points plotted without extreme bias that lie on a straight line. Therefore, the calculated probabilities can be communicated as frequencies of occurrence of walking dependence when the present model is applied to an individual. However, the use of prediction models should not be used to give up walking independence or to reduce the time of rehabilitation intervention [51]. It is important for patients and caregivers to be able to walk with supervision or with assistance around the home, even if the possibility of daily walking independence is low. In addition, it is more important for caregivers to receive practical information about preventing falls and independent transfers, rather than about the stroke survivor's walking independence [52]. The likelihood of walking dependence calculated by this prediction model will help clinicians, patients, and caregivers to better select rehabilitation strategies and set reasonable goals.

In model 2, discrimination of restricted walking speed was poor [37]. This result supports the idea that factors showing significant associations do not always have prediction ability [5]. In this study, older age, poor leg strength, and the presence of urinary incontinence were selected as predictors. The importance of age as a predictor is consistent with the findings of previous studies showing that older age is a negative factor for stroke gait recovery [6–10]. The use of poor leg strength as a predictor of slower walking speed is supported by studies in subacute and chronic stroke [53, 54]. Urinary function has been shown to be associated with gait speed and activity in several studies in older patients and patients receiving home care [55–57]. However, not all relevant associated factors identified in previous studies could be used as predictors in this study, because of its retrospective design. To improve discrimination, models should include factors associated with walking speed, such as unilateral spatial neglect, executive function, number of steps per day, and improvement in walking ability over a specific period [18, 58–60].

Strengths and limitations

This study has several strengths. First, we applied the statistical methods necessary to develop a prediction model, including reporting calibration, correcting for missing data, sensitivity analysis, and internal validation. To the best of our knowledge, this is the first study to develop a prediction model for walking dependence for patients with subacute stroke, report calibration, and perform internal validation of the model [21]. Second, this study had a large sample size with 476 participants included in the analysis. The sample size is larger than that reported by Kinoshita et al. (n = 374) which is considered to be one of the largest sample sizes used in walking independence prediction research in subacute stroke [17, 51]. Finally, this study used simple prediction variables based on previous evidence. It is well known that the clinical application of prognostic models can be challenging because of model complexity and difficulties with patient assessment, even if the models can predict outcomes with a higher accuracy than clinicians [51]. The present model includes age, leg strength, sitting balance, cognitive function, and the presence of urinary incontinence as variables that are easily assessed during a standard bedside consultation.

Some limitations should be mentioned. The facility in this study is a rehabilitation facility admitting patients with stroke approximately 30 days after stroke onset and providing intensive rehabilitation for approximately three months. Therefore, the results cannot be generalized to patients at different stages of stroke, such as those with acute or chronic stroke, or to patients who do not undergo intensive rehabilitation. Future multicenter studies, including patients with different stages of stroke and rehabilitation status, are needed to increase the generalizability of the findings of this study.

Clinical implications and recommendations for future research

The developed prediction model for walking dependence can be used in clinical practice by performing common and simple assessments of subacute stroke at the bedside during admission. A prediction model to calculate the probability of walking dependence for patients with stroke admitted to a rehabilitation facility can be used to determine how often stroke survivors with the same condition will be walking dependent. This information is particularly helpful in planning rehabilitation strategies and setting realistic goals for clinicians, patients, and caregivers to achieve safe transfers and prevent falls following discharge. Patients with a high probability of walking dependence may be able to walk with assistance and live without fear of falling if education to caregivers and with respect to the patient’s environmental setting is actively provided during hospitalization. However, evidence on the physical and mental impact of planning rehabilitation strategies on patients and caregivers is lacking, and further research is recommended. Furthermore, future studies should confirm the external validity of the prediction model and determine the clinical implications of using prediction models as well as other prediction studies of stroke rehabilitation [61].

The prediction model for restricted walking speed had poor discrimination, so future studies should investigate predictors to improve the model performance.

We have developed and internally validated two prediction models that can be used to predict the probability of walking dependence and restricted walking speed at discharge from subacute stroke rehabilitation facilities. The model for predicting walking dependence showed excellent discrimination and accurate prediction of walking dependence, while the model for predicting restricted walking speed showed poor discrimination. The prediction model calculates the likelihood of walking dependence and has shown promise as a tool for clinicians to help plan rehabilitation strategies and reasonable goal setting.

AUC, area under the receiver operating characteristic curve; BBS, Berg Balance Scale; CI, confidence interval; EPV, event per variable; FIM, functional independence measure; H–L, Hosmer–Lemeshow; MMSE, Mini-Mental State Examination; ROC, receiver operating characteristic curve; SIAS, Stroke Impairment Assessment Set; TRIPOD, Transparent Reporting of a multivariable prediction model for Individual Prognosis or Diagnosis; VIF, variance inflation factor

Ethics approval and consent to participate

This study was conducted according to the guidelines of the Declaration of Helsinki, and approved by the ethics committees of the Shinshu University (registration number 5557, date of approval 10 June 2022) and Toyama Prefectural Rehabilitation Hospital & Support Center for Children with Disabilities (registration number 86, date of approval 15 September 2022). The need for participant consent was waived because of the retrospective design of the study.

Consent for publication

Not applicable.

Availability of data and materials

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Competing interests

The authors declare that there are no conflicting interests.

Funding

Not applicable.

Authors' contributions

HF and KM were involved in the design of the study conception. HF, HM, TS were involved in the acquisitions of subjects, and data collection. HF and YK conducted data analysis. HF wrote drafts of the manuscript. KO had substantively revised draft. HF, KO, YK, and KM were involved in the interpretation of the results. All authors read, critically reviewed and approved the final manuscript.

Acknowledgements

The authors certify that no other persons have made substantial contributions to this

manuscript.

Feigin VL, Brainin M, Norrving B, Martins S, Sacco RL, Hacke W, et al. World stroke organization (WSO): global stroke fact sheet 2022. Int J Stroke. 2022;17:18–29.
GBD. Stroke collaborators. Global, regional, and national burden of stroke and its risk factors, 1990-2019: a systematic analysis for the Global Burden of Disease Study 2019. Lancet Neurol 2021. 2019;20:795–820.
Preston E, Ada L, Dean CM, Stanton R, Waddington G. What is the probability of patients who are nonambulatory after stroke regaining independent walking? A systematic review. Int J Stroke. 2011;6:531–40.
Kwah LK, Herbert RD. Prediction of walking and arm recovery after stroke: a critical review. Brain Sci. 2016;6:53. https://doi.org/10.3390/brainsci6040053.
Fulk GD, He Y, Boyne P, Dunning K. Predicting home and community walking activity poststroke. Stroke. 2017;48:406–11.
French MA, Roemmich RT, Daley K, Beier M, Penttinen S, Raghavan P, et al. Precision rehabilitation: optimizing function, adding value to health care. Arch Phys Med Rehabil. 2022;103:1233–9.
Ishiwatari M, Tani M, Isayama R, Honaga K, Hayakawa M, Takakura T, et al. Prediction of gait independence using the Trunk Impairment Scale in patients with acute stroke. Ther Adv Neurol Disord. 2022;15:17562864221140180.
Paolucci S, Bragoni M, Coiro P, De Angelis D, Fusco FR, Morelli D, et al. Quantification of the probability of reaching mobility independence at discharge from a rehabilitation hospital in nonwalking early ischemic stroke patients: a multivariate study. Cerebrovasc Dis. 2008;26:16-22. https://doi.org/10.1159/000135648.
Hirano Y, Hayashi T, Nitta O, Takahashi H, Nishio D, Minakawa T, et al. Prediction of independent walking ability for severely hemiplegic stroke patients at discharge from a rehabilitation hospital. J Stroke Cerebrovasc Dis. 2016;25:1878-81. https://doi.org/10.1016/j.jstrokecerebrovasdis.2015.12.020.
Friedman PJ. Gait recovery after hemiplegic stroke. Int Disabil Stud. 1990;12:119–22.
Hsu AL, Tang PF, Jan MH. Analysis of impairments influencing gait velocity and asymmetry of hemiplegic patients after mild to moderate stroke. Arch Phys Med Rehabil. 2003;84:1185–93.
Viosca E, Lafuente R, Martínez JL, Almagro PL, Gracia A, González C. Walking recovery after an acute stroke: assessment with a new functional classification and the Barthel index. Arch Phys Med Rehabil. 2005;86:1239-44. https://doi.org/10.1016/j.apmr.2004.11.015.
Jørgensen HS, Nakayama H, Raaschou HO, Olsen TS. Recovery of walking function in stroke patients: the Copenhagen Stroke Study. Arch Phys Med Rehabil. 1995;76:27-32. https://doi.org/10.1016/s0003-9993(95)80038-7.
Veerbeek JM, Van Wegen EEH, Harmeling-Van der Wel BC, Kwakkel G, EPOS Investigators. Is accurate prediction of gait in nonambulatory stroke patients possible within 72 hours poststroke? The EPOS study. Neurorehabil Neural Repair. 2011;25:268-74. https://doi.org/10.1177/1545968310384271.
Smith MC, Barber PA, Stinear CM. The TWIST algorithm predicts time to walking independently after stroke. Neurorehabil Neural Repair. 2017;31:955–64.
Makizako H, Kabe N, Takano A, Isobe K. Use of the Berg Balance Scale to predict independent gait after stroke: a study of an inpatient population in Japan. PM R. 2015;7:392-9. https://doi.org/10.1016/j.pmrj.2015.01.009.
Kinoshita S, Abo M, Okamoto T, Tanaka N. Utility of the revised version of the ability for basic movement scale in predicting ambulation during rehabilitation in poststroke patients. J Stroke Cerebrovasc Dis. 2017;26:1663-9. https://doi.org/10.1016/j.jstrokecerebrovasdis.2017.02.021.
Louie DR, Eng JJ. Berg Balance Scale score at admission can predict walking suitable for community ambulation at discharge from inpatient stroke rehabilitation. J Rehabil Med. 2018;50:37–44.
Singh R, Hunter J, Philip A, Todd I. Predicting those who will walk after rehabilitation in a specialist stroke unit. Clin Rehabil. 2006;20:149-52. https://doi.org/10.1191/0269215506cr887oa.
Petrilli S, Durufle A, Nicolas B, Pinel JF, Kerdoncuff V, Gallien P. Prognostic factors in the recovery of the ability to walk after stroke. J Stroke Cerebrovasc Dis. 2002;11:330-5. https://doi.org/10.1053/jscd.2002.130124.
Moons KGM, Altman DG, Reitsma JB, Ioannidis JP, Macaskill P, Steyerberg EW, et al. Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD): explanation and elaboration. Ann Intern Med. 2015;162:W1-73. https://doi.org/10.7326/M14-0698.
Varga TV, Niss K, Estampador AC, Collin CB, Moseley PL. Association is not prediction: A landscape of confused reporting in diabetes - A systematic review. Diabetes Res Clin Pract. 2020;170:108497.
Van Calster B, Nieboer D, Vergouwe Y, De Cock B, Pencina MJ, Steyerberg EW. A calibration hierarchy for risk models was defined: from utopia to empirical data. J Clin Epidemiol. 2016;74:167–76.
Binuya MAE, Engelhardt EG, Schats W, Schmidt MK, Steyerberg EW. Methodological guidance for the evaluation and updating of clinical prediction models: a systematic review. BMC Med Res Methodol. 2022;22:316.
Fiedler RC, Granger CV. The functional independence measure: A measurement of disability and medical rehabilitation. In: Functional Evaluation of Stroke Patients. Chino N, Melvin JL, editors. Tokyo: Springer Japan; 1996. p. 75–92.
Preston E, Ada L, Stanton R, Mahendran N, Dean CM. Prediction of independent walking in people who are nonambulatory early after stroke: a systematic review. Stroke. 2021;52:3217-24. https://doi.org/10.1161/STROKEAHA.120.032345.
Liu M, Chino N, Tuji T, Masakado Y, Hase K, Kimura A. Psychometric properties of the stroke impairment assessment set (SIAS). Neurorehabil Neural Repair. 2002;16:339-51. https://doi.org/10.1177/0888439002239279.
Berg KO, Maki BE, Williams JI, Holliday PJ, Wood-Dauphinee SL. Clinical and laboratory measures of postural balance in an elderly population. Arch Phys Med Rehabil. 1992;73:1073–80.
Tsoi KKF, Chan JYC, Hirai HW, Wong SY, Kwok TC. Cognitive tests to detect dementia: a systematic review and meta-analysis. JAMA Intern Med. 2015;175:1450-8. https://doi.org/10.1001/jamainternmed.2015.2152.
Pohl J, Held JPO, Verheyden G, Alt Murphy M, Engelter S, Flöel A, et al. Consensus-based core set of outcome measures for clinical motor rehabilitation after stroke-a Delphi study. Front Neurol. 2020;11:875.
Saa JP, Tse T, Baum CM, Cumming T, Josman N, Rose M, et al. Cognitive recovery after stroke: a meta-analysis and metaregression of intervention and cohort studies. Neurorehabil Neural Repair. 2021;35:585–600.
Iba K, Shinozaki T, Maruo K, Noma H. Re-evaluation of the comparative effectiveness of bootstrap-based optimism correction methods in the development of multivariable clinical prediction models. BMC Med Res Methodol. 2021;21:9.
Steif J, Brant R, Sreepada RS, West N, Murthy S, Görges M. Prediction model performance with different imputation strategies: A simulation study using a North American ICU registry. Pediatr Crit Care Med. 2022;23:e29–44.
Vatcheva KP, Lee M, McCormick JB, Rahbar MH. Multicollinearity in regression analyses conducted in epidemiologic studies. Epidemiology (Sunnyvale). 2016;6. doi:10.4172/2161-1165.1000227.
Wahl S, Boulesteix AL, Zierer A, Thorand B, van de Wiel MA. Assessment of predictive performance in incomplete data by combining internal validation and multiple imputation. BMC Med Res Methodol. 2016;16:144.
Chen L. Overview of clinical prediction models. Ann Transl Med. 2020;8:71.
Mandrekar JN. Receiver operating characteristic curve in diagnostic test assessment. J Thorac Oncol. 2010;5:1315–6.
Heymans M. psfmi: prediction Model Pooling, Selection and Performance Evaluation across Multiply Imputed Datasets. https://cran.r-project.org/web/packages/psfmi/index.html. Accessed 26 October 2023.
Skarin M, Sjöholm A, Nilsson ÅL, Nilsson M, Bernhardt J, Lindén T. A mapping study on physical activity in stroke rehabilitation: establishing the baseline. J Rehabil Med. 2013;45: 997–1003.
Pate RR, O’Neill JR, Lobelo F. The evolving definition of “sedentary.” Exerc Sport Sci Rev. 2008;36:173–8.
de Rezende LFM, Rodrigues Lopes M, Rey-López JP, Matsudo VKR, Luiz Odo C. Sedentary behavior and health outcomes: an overview of systematic reviews. PLOS ONE. 2014;9:e105620.
Hall J, Morton S, Fitzsimons CF, Hall JF, Corepal R, English C, et al. Factors influencing sedentary behaviours after stroke: findings from qualitative observations and interviews with stroke survivors and their caregivers. BMC Public Health. 2020;20:967.
Reicherzer L, Wirz M, Wieber F, Graf ES. Facilitators and barriers to health enhancing physical activity in individuals with severe functional limitations after stroke: A qualitative study. Front Psychol. 2022;13:982302.
Middleton A, Fulk GD, Beets MW, Herter TM, Fritz SL. Self-selected walking speed is predictive of daily ambulatory activity in older adults. J Aging Phys Act. 2016;24:214–22.
Wang SCY, Kassavou A. Digital health behavioural interventions to support physical activity and sedentary behaviour in adults after stroke: A systematic literature review with meta-analysis of controlled trials. Behav Sci (Basel). 2023;13. doi:10.3390/bs13010062.
Oliveira SG, Ribeiro JAM, Silva ÉSM, Uliam NR, Silveira AF, Araújo PN, et al. Interventions to change movement behaviors after stroke: a systematic review and meta-analysis. Arch Phys Med Rehabil. 2023. doi:10.1016/j.apmr.2023.07.011.
Espernberger K, Fini NA, Peiris CL. Identity, social engagement and community participation impact physical activity levels of stroke survivors: A mixed-methods study. Clin Rehabil. 2023;37:836–50.
Perry J, Garrett M, Gronley JK, Mulroy SJ. Classification of walking handicap in the stroke population. Stroke. 1995;26:982–9.
Mehrholz J, Thomas S, Elsner B. Treadmill training and body weight support for walking after stroke. Cochrane Database Syst Rev. 2017;8:CD002840.
Gamble K, Chiu A, Peiris C. Core stability exercises in addition to usual care physiotherapy improve stability and balance after stroke: A systematic review and meta-analysis. Arch Phys Med Rehabil. 2021;102:762–75.
Stinear CM, Smith MC, Byblow WD. Prediction tools for stroke rehabilitation. Stroke. 2019;50:3314–22.
Krishnan S, Pappadis MR, Weller SC, Fisher SR, Hay CC, Reistetter TA. Patient-centered mobility outcome preferences according to individuals with stroke and caregivers: a qualitative analysis. Disabil Rehabil. 2018;40:1401–9.
Aguiar LT, Camargo LBA, Estarlino LD, Teixeira-Salmela LF, Faria CDCM. Strength of the lower limb and trunk muscles is associated with gait speed in individuals with sub-acute stroke: a cross-sectional study. Braz J Phys Ther. 2018;22:459–66.
Gomes Costa RR, Ribeiro Neto F, Gonçalves CW, Carregaro RL. Accuracy and cut-off points of different models of knee extension strength analysis to identify walking performance in individuals with chronic stroke. Braz J Phys Ther. 2021;25:610–6.
Suhr R, Lahmann NA. Urinary incontinence in home care: a representative multicenter study on prevalence, severity, impact on quality of life, and risk factors. Aging Clin Exp Res. 2018;30:589–94.
Spaltenstein J, Bula C, Santos-Eggimann B, Krief H, Seematter-Bagnoud L. Factors associated with going outdoors frequently: a cross-sectional study among Swiss community-dwelling older adults. BMJ Open. 2020;10:e034248.
Simone de Souza Vasconcelos K, de Carvalho Bastone A, Alvarenga Vieira R, Cristina de Souza Andrade A, Correa Dias R, Domingues Dias JM, et al. Examining the factors associated with functional capacity of community-dwelling older adults using the ICF framework: a cross-sectional study from the Frailty in Brazilian Older Adults Study (FIBRA). Physiother Theor Pract. 2023;39:2454–69.
Handlery R, Regan EW, Stewart JC, Pellegrini C, Monroe C, Hainline G, et al. Predictors of daily steps at 1-year poststroke: A secondary analysis of a randomized controlled trial. Stroke. 2021;52:1768–77.
Tilson JK, Sullivan KJ, Cen SY, Rose DK, Koradia CH, Azen SP, et al. Meaningful gait speed improvement during the first 60 days poststroke: minimal clinically important difference. Phys Ther. 2010;90:196–208.
Sakai K, Hosoi Y, Harada Y. Walking ability associated with executive dysfunction in patients with stroke: A cross-sectional study. Brain Sci. 2023;13. doi:10.3390/brainsci13040627.
Stinear CM, Byblow WD, Ackerley SJ, Barber PA, Smith M-C. Predicting recovery potential for individual stroke patients increases rehabilitation efficiency. Stroke. 2017;48: 1011–1019.

No competing interests reported.

Appendix1.pdf

Download PDF

Editorial decision: Revision requested
21 Jan, 2024
Reviews received at journal
12 Dec, 2023
Reviewers agreed at journal
05 Dec, 2023
Reviewers invited by journal
29 Nov, 2023
Editor assigned by journal
29 Nov, 2023
Editor invited by journal
20 Nov, 2023
Submission checks completed at journal
20 Nov, 2023
First submitted to journal
20 Nov, 2023

You are reading this latest preprint version

Development and internal validation of two prediction models of walking dependence and restricted walking speed in individuals with subacute stroke

Status:

Version 1

Abstract

Background

Methods

Results

Conclusions

Figures

Background

Methods

Study design

Outcome and associated variables

Sample size

Data analysis

Dealing with missing data

Study population characteristics

Multicollinearity

Model development

Models internal validation and optimism correction

Model performance

Results

Participants

Comparisons of the participants’ characteristics and functional outcomes between the groups

Multicollinearity

Model development and internal validation

Model performance

Sensitivity analysis

Discussion

Conclusion

Abbreviations

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1