Berg balance scale cutoffs for predicting walking independence in older adults with hip fracture: comparison of a Korean retrospective cohort with a published Japanese cohort
Article information
Abstract
Purpose
This study examined the association between admission Berg balance scale (BBS) score and walking independence at discharge in older adults with hip fracture, derived BBS cutoffs in a Korean cohort, and compared them with published Japanese findings.
Methods
This single-center retrospective cohort study included 263 patients aged 65 years or older with hip fracture and a Functional Ambulation Category (FAC) score of 3 or lower at admission. Ordinal logistic regression modeled discharge walking independence, and cutoffs were identified by receiver operating characteristic analysis using the Youden index. Results were compared with published data from a Japanese cohort of 187 patients.
Results
Age and sex distributions were similar between cohorts, but the Korean cohort had a lower admission BBS score (mean±standard deviation, 15.1±11.7 vs. 25.0±13.9; P<0.001) and a smaller proportion walking independently at discharge (4.2% vs. 31.6%; P<0.001). In the Korean cohort, admission BBS independently predicted greater walking independence at discharge (odds ratio, 1.11; 95% confidence interval, 1.08–1.15), comparable to the published Japanese estimate (1.09). The BBS cutoffs were 25 for independent walking (FAC≥4) and 19 for supervised-or-better walking (FAC≥3), versus 28 and 21 in Japan.
Conclusion
Admission BBS was associated with walking independence at discharge, with an effect estimate similar to that reported in Japan. However, discharge walking outcomes and the derived cutoffs differed between cohorts. The Korean cutoffs remain preliminary and require validation in independent multicenter cohorts before clinical use.
Introduction
Background
Hip fracture is among the most common injuries in older adults, and its incidence in Korea continues to increase with population aging. Korean claims data indicate an age-standardized incidence of 104.06 per 100,000 people aged 50 years or older and a 1-year mortality of 16.55%, approximately 2.85 times that of the age-matched general population [1]. Hip fracture is also common among older adults in neighboring Japan [2].
Only about half of patients regain their pre-fracture walking ability, and many need assistance with daily activities after discharge [3]. Walking independence at discharge is therefore a key prognostic indicator. Balance is a major determinant of walking recovery and is commonly assessed using the Berg balance scale (BBS) [4]. The BBS is reliable and valid in older adults and correlates significantly with walking ability [4,5].
BBS-based prediction of walking outcomes has primarily been studied in stroke [6,7]. However, these criteria are difficult to apply to hip fracture because the mechanisms of balance impairment and the course of recovery differ. In a study of 187 older adults with hip fracture in 3 Japanese convalescent rehabilitation wards, Tamura et al. [8] found that admission BBS significantly predicted walking independence at discharge (odds ratio [OR], 1.09). They proposed cutoffs of 28 points for independent walking and 21 points for supervised walking.
The institutional context of convalescent rehabilitation, however, differs considerably across countries. Japan introduced convalescent rehabilitation wards in 2000 and permits hospitalization for up to 90 days for orthopedic conditions such as femoral and pelvic fractures [9,10]. Under the Korean rehabilitation medical institution scheme, by contrast, the reimbursed inpatient period for the musculoskeletal disease group, which includes hip fracture, is generally 30 days. A revision of the government notice that took effect on March 1, 2025 extended this period to 60 days only for selected patients treated with internal fixation or total hip arthroplasty [11,12]. Because rehabilitation duration can directly affect functional status at discharge, cutoffs derived in one country may not be valid in another with a different system.
Objectives
Using real-world data from a Korean convalescent rehabilitation hospital, this study aimed to (1) estimate the prognostic association between admission BBS and walking independence at discharge and derive cutoffs, (2) compare the results with published summary data from a Japanese cohort [8], and (3) interpret the differences in their clinical and institutional contexts. The study also examined the association between discharge disposition and walking independence, which the previous study did not address.
Methods
Ethics statement
The Institutional Review Board of Immanuel Medical Rehabilitation Hospital approved this study (approval number: 3-70095001-A-2026-012). Written informed consent was waived because the retrospective medical record review posed no additional risk and involved no collection of personally identifiable information. The study complied with the Declaration of Helsinki and the Korean Bioethics and Safety Act. The de-identified dataset analyzed in this study is available as Data Set 1,, together with its codebook (Suppl. 1) and explanatory notes (Suppl. 2).
Study design
This observational study compared single-center retrospective cohort data from Korea with published summary statistics from a Japanese cohort [8] and followed the Strengthening the Reporting of Observational Studies in Epidemiology guidelines [13]. Individual patient data from the Japanese cohort were unavailable. Comparisons were therefore limited to tests based on published means±standard deviations (SDs), and frequencies and were interpreted at the group level without covariate adjustment.
Setting
Electronic medical records were obtained from a single convalescent rehabilitation hospital in Chungcheongbuk-do, Korea. Data were extracted for all 455 admissions in which the reimbursed convalescent rehabilitation period began between January 1, 2021 and June 30, 2025. The index assessment comprised BBS and Functional Ambulation Category (FAC) measurements at the initial evaluation immediately after admission; the outcome assessment comprised BBS and FAC measurements at the final evaluation before discharge.
The 263 patients included in the analysis were admitted between September 30, 2021 and June 25, 2025. Earlier admissions did not qualify because BBS and FAC were not yet routinely recorded: among 24 earlier admissions, an admission BBS was available for only 5 and a discharge BBS for only 1.
Participants and data sources
Korean cohort
Patients admitted with hip-related conditions during the study period were identified using the hospital registry of disease groups eligible for rehabilitation medical institution admission and were enrolled consecutively. Each patient was the unit of analysis; when a patient had multiple eligible admissions, only the first was used.
Patients were eligible if they (1) had a primary diagnosis of hip fracture, (2) were aged 65 years or older, (3) could not walk independently at admission (FAC≤3), and (4) had undergone surgery (internal fixation, hemiarthroplasty, or total hip arthroplasty). Patients were excluded for (1) inability to follow verbal instructions during balance and gait assessment, (2) blindness or severe visual impairment affecting walking, (3) gait impairment caused by other neurological or musculoskeletal conditions, or (4) missing admission or discharge BBS or FAC measurements. Figure 1 shows exclusions at each step. The previous study [8] also required a first fracture caused by a fall; because these criteria could not be verified from our records, a primary diagnosis of hip fracture was used instead.
Comparison cohort
Tamura et al. [8] reported a cohort of 187 older adults with hip fracture from 3 Japanese convalescent rehabilitation wards between April 2018 and June 2020. Their study used the same age and admission FAC criteria (age ≥65 years and FAC≤3), instruments (BBS and FAC), and 3-group classification of walking independence at discharge, allowing direct comparison. However, the Japanese cohort included 5 patients treated conservatively, 1 classified as “other,” and no patients who underwent total hip arthroplasty; its composition by surgical procedure therefore differed from that of the Korean cohort (Table 1).
Variables and measurement
The primary outcome, walking independence at discharge, was classified using discharge FAC as independent (FAC≥4), supervised (FAC=3), or assisted (FAC≤2). The total admission BBS score was the primary exposure, and discharge disposition was the secondary outcome.
Age was prespecified as the sole adjustment variable to replicate the age-adjusted model of Tamura et al. [8], because age consistently predicts walking recovery and the presence of only 11 independent walkers limited the events per variable. Cognitive function, assessed using the Mini-Mental State Examination (MMSE), activities of daily living, assessed using the modified Barthel index (MBI), surgical procedure, and length of stay were prespecified as exploratory covariates. Each was added separately to the primary model to assess the stability of the BBS estimate (Table 2). Data were also collected on sex, discharge disposition, disease group (single-site or multi-site musculoskeletal), and the reimbursed rehabilitation period.
Crude and adjusted logistic regression models predicting discharge walking independence in the Korean cohort (n=263)
The BBS contains 14 items scored from 0 to 4 (total, 0–56), with higher scores indicating better balance [4]. The FAC classifies the assistance needed for walking into 6 levels, from 0 (unable to walk) to 5 (independent walking, including stairs), and has established predictive validity and responsiveness in stroke [14,15]. The 3-group classification was selected because FAC 3 denotes supervised walking without physical contact, whereas FAC 4 or higher denotes unsupervised walking on level surfaces [14]. This distinction determines whether a caregiver must be continuously present and is therefore a clinical decision point in discharge planning; the classification also matched that used by Tamura et al. [8]. Admission BBS was analyzed as a continuous variable. The classification was finalized before data review, and no alternative cutoffs were explored.
Trained physical and occupational therapists used standard institutional procedures to administer the BBS and FAC.
In the Japanese cohort, attending therapists also used the original BBS and FAC to assess patients at admission and discharge [8], so the cohorts were considered comparable in their measurement methods. However, the Japanese assessments occurred a mean±SD of 25.6±12.8 and 76.9±25.9 days after injury or surgery, whereas the Korean records did not include injury or surgery dates. Consequently, time since injury or surgery could not be compared, and only length of stay in the convalescent rehabilitation hospital is reported for the Korean cohort.
Bias
Selection bias was addressed through consecutive enrollment and the use of FAC criteria identical to those in the Japanese study. Assessors evaluated patients during routine care without knowing the study hypotheses, making measurement bias unlikely; however, unmeasured confounding could not be excluded.
Study size
All eligible patients from the predefined study period were included; no a priori sample size calculation was performed.
Statistical methods
Analyses were prespecified as primary, secondary, or exploratory. The primary analysis used ordinal (proportional-odds) logistic regression to model 3-level walking independence at discharge as a function of admission BBS and age. The proportional-odds assumption was assessed using the test of parallel lines. Secondary analyses comprised 2 binary logistic regressions (FAC≥3 and FAC≥4) and receiver operating characteristic analyses. Cutoffs were selected by maximizing the Youden index, and discrimination was assessed using the area under the curve. Crude and adjusted estimates are presented for all regressions (Table 2).
Exploratory analyses included (1) models in which surgical procedure, MBI, MMSE, or length of stay was added individually; (2) subgroup analyses by reimbursed period (≤30, 31–59, and ≥60 days) and admission before versus after the notice revision effective March 1, 2025; (3) subgroup analyses by surgical procedure; and (4) analysis of the association between discharge disposition and walking independence. The subgroup analyses were intended to generate hypotheses and were not adjusted for multiple comparisons; interaction tests were not performed because the sample size was insufficient.
Groups were compared using the Kruskal-Wallis test with Bonferroni post hoc testing, the chi-square test, the Wilcoxon signed-rank test, the Mann-Whitney U test, or the Fisher exact test, as appropriate. Normality was assessed using the Shapiro-Wilk test. Between-cohort comparisons were performed using the Welch t-test and the chi-square test.
Complete-case analysis was performed without imputing missing values. Missing data were confined to admission MMSE (n=21, 8.0%) and discharge MMSE (n=29, 11.0%) (Table 3); models that included MMSE therefore used data from 242 patients. There was no post-discharge follow-up and hence no loss to follow-up; Figure 1 identifies the 69 admissions excluded for missing discharge assessments.
Analyses were performed using Python 3 (pandas, SciPy, statsmodels, and scikit-learn), and the results were checked in IBM SPSS Statistics ver. 28.0 (IBM Corp.). The results were identical across platforms; the SPSS results are reported here. Statistical significance was defined as P<0.05.
Results
Participants
Of the 455 admissions identified, 263 patients met the eligibility criteria and were included in the analysis (Figure 1). At discharge, 11 patients (4.2%) could walk independently, 29 (11.0%) could walk with supervision, and 223 (84.8%) needed assistance. Hemiarthroplasty was the most common surgical procedure, followed by internal fixation and total hip arthroplasty. Most patients were in the single-site musculoskeletal group and had an approved reimbursed rehabilitation period of no more than 30 days.
Descriptive and outcome data
Table 3 presents baseline characteristics and missing data by walking-independence group at discharge. Age and admission BBS, MMSE, and MBI differed significantly among the 3 groups, but sex and length of stay did not. Post hoc testing showed lower admission BBS in the assisted group than in either of the other 2 groups. BBS improved significantly between admission and discharge (Wilcoxon signed-rank test, P<0.001), with improvement observed in 201 of the 263 patients.
Main results
In the primary adjusted ordinal model, higher admission BBS independently predicted greater walking independence at discharge (OR, 1.11; 95% confidence interval [CI], 1.08–1.15), whereas older age predicted less independence (Table 2). The proportional-odds assumption was met. The crude estimates and results of the 2 secondary binary models were consistent with the primary finding.
Cutoffs and comparison with the Japanese cohort
The BBS cutoffs that maximized the Youden index were 25 points for independent walking and 19 points for supervised-or-better walking, compared with 28 and 21 points, respectively, in the Japanese cohort (Table 4, Figure 2). Discrimination for independent walking was numerically higher in the Korean cohort, but discrimination for supervised-or-better walking was similar between cohorts.
Receiver operating characteristic curves for admission Berg balance scale (BBS) predicting independent walking (Functional Ambulation Category [FAC]≥4) and supervised-or-better walking (FAC≥3) at discharge in the Korean cohort. Marked points indicate the cutoff that maximized the Youden index for each outcome. The area under the curve (AUC) was 0.86 (95% confidence interval [CI], 0.77–0.96) for independent walking and 0.83 (95% CI, 0.77–0.89) for supervised-or-better walking.
The cohorts did not differ significantly in age or sex distribution. However, the Korean cohort had lower BBS at admission (Cohen d=−0.78) and discharge and a markedly less favorable distribution of walking independence at discharge (Table 1, Figure 3). The distribution of surgical procedures also differed. Nearly half of the Korean patients were discharged within 30 days, and almost all were discharged within 90 days.
Distribution of walking independence at discharge in the Korean and Japanese cohorts. Although age and sex distributions were comparable, the proportion achieving independent walking (FAC≥4) differed markedly between cohorts (4.2% vs. 31.6%, P<0.001). Japanese cohort data were obtained from Tamura et al. [8]. FAC, Functional Ambulation Category.
Other analyses
In the analysis by reimbursed rehabilitation period, patients in the ≥60-day group stayed approximately twice as long as those in the ≤30-day group. The groups did not differ in admission or discharge BBS, change in BBS or FAC, or the proportion achieving supervised-or-better walking, and no patient in the 31–59-day or ≥60-day group achieved independent walking (Table 5). Patients admitted after the March 2025 revision stayed longer and had higher discharge BBS, but admission BBS and functional change did not differ (Table 6). Both subgroup analyses had small samples and insufficient statistical power.
Functional change and discharge outcomes according to the reimbursed convalescent rehabilitation period (n=263)
Functional change and discharge outcomes before and after the March 1, 2025 revision of the government notice (n=263)
Outcomes did not differ significantly across surgical procedures, and including procedure in the primary model did not change the admission BBS estimate (Table 7). This estimate also remained stable and significant when MBI, MMSE, or length of stay was added separately (Table 2). In the model that included MMSE, MMSE significantly predicted the outcome, whereas age was no longer significant.
The proportion discharged to a non-hospital setting, such as home or a nursing home, increased progressively with walking independence and reached 100% in the independent group (Table 8). Each increase of 1 level in walking independence was associated with greater odds of discharge to a non-hospital setting (OR, 3.64; 95% CI, 1.38–9.62).
Discussion
Key results
Among older adults with hip fracture in this cohort, admission BBS was independently associated with walking independence at discharge, with an association similar in magnitude to that reported in Japan. Discrimination for supervised-or-better walking was also nearly identical between cohorts. The estimate remained stable across crude, adjusted, and exploratory models, suggesting that the association does not depend on a particular model specification. Nevertheless, the cutoffs derived from the Korean cohort were lower than the Japanese values, and the cohorts differed substantially in their distribution of walking outcomes at discharge.
Length-of-stay policies and functional status at discharge in Korea and Japan
Although the cohorts were similar in age and sex, the proportion achieving independent walking differed by approximately 8-fold. Differences in the permitted duration of convalescent rehabilitation offer one possible institutional explanation.
Japan introduced convalescent rehabilitation wards in 2000 and permits hospitalization for up to 90 days for femoral, pelvic, spinal, hip, or knee fractures and up to 150 or 180 days for cerebrovascular conditions [9,10,16]. In Korea, the reimbursed inpatient period for the musculoskeletal disease group, which includes hip fracture, is 30 days for single-site conditions and 60 days for multi-site conditions. Effective March 1, 2025, a revision extended this period to 60 days only for single-site hip, pelvic, or femoral fractures treated with internal fixation or total hip arthroplasty; hemiarthroplasty remained excluded from the extension [11,12].
Our cohort directly reflects these constraints: most patients were subject to the 30-day limit, and their actual stays clustered around it. Very few stayed beyond the 90-day period permitted in Japan. In addition, nearly half underwent hemiarthroplasty, for which the 30-day limit remained unchanged after the revision, and most were admitted before the revision took effect.
However, our subgroup analyses did not demonstrate that longer stays directly translated into functional gains. Patients approved for 60 days stayed more than twice as long as those approved for 30 days or less, but their functional improvement was no greater, and none achieved independent walking. There are 2 possible interpretations. First, eligibility for 60 days likely selected patients with poorer function or delayed recovery, creating confounding by indication; their lower admission BBS is consistent with this explanation. Second, a subgroup of 13 patients lacked sufficient statistical power to detect a clinically meaningful difference. Thus, this study considers length-of-stay policy only as a possible context for the between-cohort differences and does not directly evaluate the effect of longer rehabilitation.
Previous studies have reported associations between rehabilitation duration and functional recovery. An evaluation of Japanese convalescent rehabilitation ward policy found that increased rehabilitation provision had a small but significant effect on functional gains at discharge; 69.1% of patients were discharged home [9]. In Korea, insufficient continuity of post-acute rehabilitation and early discharge have repeatedly been identified as barriers to functional recovery and reintegration into the community [11].
Timing of transfer from acute to convalescent care
Length of stay cannot fully explain the differences because Korean patients already had substantially poorer balance when convalescent rehabilitation began. This baseline difference may largely reflect when patients transferred from acute care.
First, acute-care length of stay differs between the countries. A comparison of claims data from Japan, Korea, and Taiwan found that patients with hip fracture stayed longer in the hospital where surgery was performed in Japan than in Korea; Japanese patients also tended to remain in the same hospital from surgery through convalescent rehabilitation [17]. Second, the regulatory requirements differ. In Korea, patients with single-site musculoskeletal conditions must enter a rehabilitation medical institution within 30 days of onset or surgery (60 days for multi-site conditions), whereas Japan abolished the onset-to-admission time limit for convalescent wards in 2020 [11,16]. Third, admission BBS in the Japanese cohort was assessed a mean of 25.6±12.8 days after injury or surgery [8]. A multicenter study of 882 Japanese patients also found that one-third could already walk independently at discharge from acute care [18]. However, because the Korean records did not contain injury or surgery dates, assessment timing could not be compared statistically.
These findings suggest that Japanese patients transfer to convalescent wards after substantial initial recovery in acute care, whereas Korean patients begin convalescent rehabilitation earlier and with poorer function. Differences in the distribution of surgical procedures further indicate that the cohorts were not equivalent. The comparisons should therefore be interpreted as observational evidence that admission BBS was associated with different distributions of walking outcomes across the 2 rehabilitation settings; they do not establish that the delivery systems caused these differences.
Interpretation of cutoffs
Because cutoffs depend on the sample’s functional distribution, lower admission BBS in a population is statistically expected to shift the cutoffs downward. Clinically, applying thresholds from one country to another with different systems and patient composition risks setting targets too high for Korean patients and underestimating which patients can reach them. However, these cutoffs were derived and evaluated in the same dataset using the Youden index, and only 11 patients achieved independent walking. They should therefore be treated as candidate thresholds that have not undergone internal or external validation.
Age did not differ among outcome groups in the Japanese cohort [8], but it independently predicted the outcome in our cohort. Its loss of significance after adjustment for MMSE suggests that its effect may be mediated by, or collinear with, cognitive function; this possibility requires further investigation.
New evidence on discharge disposition
Our analysis of discharge disposition, an outcome not addressed in the previous study, showed progressively more discharges to non-hospital settings as walking independence increased. Early prediction of walking prognosis from admission BBS may therefore help set rehabilitation goals and support earlier discharge planning, caregiver counseling, and preparation for community linkage. A previous study at the same institution similarly found that discharge disposition was more closely associated with recovery of walking, balance, and activities of daily living than with surgical procedure [12]. This finding is consistent with the absence of procedure-related differences in our cohort.
Limitations
This study has several limitations. First, the Japanese comparison relied on published summary statistics, preventing covariate adjustment and limiting interpretation to group-level comparisons. Second, the influence of length-of-stay policies was inferred from differences in cohort characteristics; small subgroup samples and confounding by indication prevented adequate evaluation of extended stays. Third, the Korean data were collected retrospectively at a single center, and the small number of independent walkers reduced the precision of that group’s cutoff. Fourth, missing injury and surgery dates prevented calculation, adjustment, or between-cohort comparison of the interval from injury or surgery to assessment. Fifth, walking aid use was not recorded, precluding replication of the previous study’s aid-specific cutoff analysis. Sixth, a primary diagnosis of hip fracture was used because a first fracture caused by a fall could not be confirmed. Seventh, excluding 69 admissions without discharge BBS or FAC may have introduced selection bias. Eighth, pre-fracture walking ability, comorbidities, and postoperative pain could not be adequately adjusted for, and the sparse outcome restricted the number of variables in multivariable models. Ninth, participants may partially overlap with those in a previous study at the same institution [12].
Generalizability
These findings apply to older adults with hip fracture who underwent convalescent rehabilitation under the Korean rehabilitation medical institution scheme. The single-center data may not generalize directly to institutions with different patient populations or rehabilitation provision. In particular, the derived cutoffs depend on the sample’s functional distribution and may not apply to other cohorts. However, the association between balance at admission and walking independence at discharge was similar in magnitude in 2 countries with different systems and populations, suggesting that it may apply across a relatively broad range of settings.
Conclusion
Among Korean older adults with hip fracture, admission BBS was significantly associated with walking independence at discharge, with an association similar in magnitude to that reported in Japan. The BBS cutoffs were 25 points for independent walking and 19 points for supervised-or-better walking, compared with 28 and 21 points, respectively, in Japan; the distribution of walking independence at discharge also differed substantially. Clinical and institutional factors, including functional status at admission, rehabilitation duration, and rehabilitation delivery systems, may explain these differences. However, the data cannot separate the effects of individual factors or establish causality. BBS cutoffs should therefore be interpreted in the context of the patient population and rehabilitation setting, and the 25- and 19-point thresholds require validation in independent multicenter cohorts before clinical use.
Notes
Conflict of interest
No potential conflict of interest relevant to this article was reported.
Funding
None.
Data availability
A data file is available from https://doi.org/10.5124/jkma.26.0100.
Data Set 1. De-identified electronic health data for the 263 Korean patients, with an accompanying codebook and note.
Supplementary materials
Supplementary files are available from https://doi.org/10.5124/jkma.26.0100.
Codebook for Data Set 1.
Note for Data Set 1.
