Relationship between Physical Function at Admission and Walking Ability at Discharge in Older Adults with Vertebral Compression Fractures: An Analysis Using Propensity Score Matching
Article information
Abstract
Background
Vertebral compression fractures (VCFs) are common among older adults, with the highest prevalence observed in Japan. These fractures cause pain, reduce quality of life, and increase the need for physical therapy. This study identified key factors at admission that predict walking ability at discharge in patients with VCFs.
Methods
This retrospective cohort study included 143 patients aged ≥65 years with VCFs. VCFs are fractures in which only the anterior column of the vertebral body collapses. The primary variables assessed upon admission included the revised Hasegawa Dementia Scale (HDS-R) score, grip strength, skeletal muscle mass index (SMI), and phase angle (PhA). Propensity score matching was applied to adjust for background factors, after which a logistic regression analysis using a generalized linear model was conducted to determine whether these variables influenced walking ability at discharge.
Results
Significant associations were observed between walking ability at discharge and HDS-R score at admission (p<0.001, effect size [ES]=0.42), grip strength (p=0.027, ES=0.23), SMI (p=0.025, ES=0.23), and PhA (p<0.001, ES=0.40). Logistic regression analysis indicated that HDS-R score (odds ratio [OR]=1.19, p=0.005) and PhA (OR=3.21, p=0.015) during admission significantly predicted walking ability at discharge.
Conclusion
Walking ability at discharge in patients with VCFs can be predicted based on early assessments. In particular, HDS-R score and PhA at admission may serve as key indicators for prognosis in patients with VCFs.
INTRODUCTION
Vertebral compression fracture (VCFs) is one of the most common fractures among older adults.1,2) Compared with other regions, the incidence of vertebral fractures is particularly high in Asia.3) Among Asian countries, Japan has the highest vertebral fracture prevalence.4) VCFs cause intense pain and decreased quality of life.5) Consequently, in the coming years, the number of patients with VCFs requiring physical therapy is expected to increase.
Several studies have investigated VCFs, consistently reporting a higher prevalence among female.1) Additionally, factors influencing the recovery of activities of daily living (ADL) in patients with VCFs include sex and various pre-existing fractures.6)
Moreover, balance ability and cognitive function at discharge7) have been identified as key factors associated with walking ability at discharge in patients with VCF. In female patients, phase angle (PhA) at admission8) is significantly associated with walking ability at discharge. Thus, the existing research on VCFs has primarily focused on ADL recovery and walking ability.
Despite the clinical relevance of early physical therapy assessment during hospital admission, studies on this aspect are limited. Early evaluation upon admission is crucial for predicting prognosis and designing effective treatment strategies. Early physical therapy intervention for VCFs is widely recognized as beneficial.9,10)
However, pain in patients with VCFs often hinders key assessments, such as knee extensor strength,11) which is closely related to walking ability, and standing balance,7) considering these evaluations may provoke discomfort during measurement. Consequently, conducting early physical therapy assessments upon admission is often challenging. These condition-specific limitations may complicate the early evaluation of patients with VCFs.
Hypothetically, in patients with VCFs who must rest during the acute phase, walking ability at discharge is influenced by body composition parameters such as skeletal muscle mass index (SMI) and PhA—both can be assessed in the supine position. Therefore, this study investigated the relationship between early admission assessment and walking ability at discharge.
Our findings provide insight into the key determinants of walking ability at discharge in patients with VCFs, thereby enhancing the prognostic accuracy of those experiencing severe acute pain. Furthermore, this study informs the development of effective treatment programs. Overall, we sought to highlight the importance of early physical therapy assessment in predicting walking ability at discharge in patients with VCFs.
MATERIALS AND METHODS
Participants
This retrospective study examined patients diagnosed with VCFs based on the Denis classification, defining VCFs as fractures involving the collapse of only the anterior column of the vertebral body.12,13) Therefore, this study included only patients diagnosed with VCFs involving collapse limited to the anterior column of the vertebral body. To ensure diagnostic precision, orthopedic surgeons confirmed VCFs using plain radiography and magnetic resonance imaging. Only patients with fractures strictly limited to the anterior column were included.
The study population comprised individuals who independently ambulated in their daily life before hospitalization, despite using walking aids.
All patients were treated in the convalescent rehabilitation ward of our hospital. The rehabilitation program incorporated physical and occupational therapy, with sessions conducted once daily, 7 days a week. Each session lasted approximately 60 minutes, which is approximately 120 minutes of therapy per day. Physical therapy primarily targeted lower limb muscle strengthening, balance training, and gait rehabilitation. Occupational therapy focused on ADL training, with cognitive interventions incorporated as needed. To ensure patient safety and comfort, all sessions were conducted with careful attention to pain management.
The inclusion criteria were as follows: 208 patients with VCFs, aged ≥65 years, and discharged from the convalescent ward of Kurume Rehabilitation Hospital, Kabutoyama Medical Corporation, between August 2019 and June 2024. The following were the exclusion criteria: patients aged <65 years, previous surgery at an acute care hospital before admission, use of a wheelchair for mobility before hospitalization, and use of pacemakers. Additionally, patients who were transferred to other medical institutions or wards, discharged because of death, or had their prescriptions modified owing to changes in their medical condition were excluded. Patients who developed pneumonia or experienced worsening cardiovascular disease during hospitalization—conditions that hinder participation in routine rehabilitation walking exercises—were also excluded. Furthermore, individuals with moderate-to-severe cognitive impairment, defined as having a revised Hasegawa Dementia Scale (HDS-R) score14) of <13, were not included.
After applying these exclusion criteria, 143 patients were included in the final analysis (Fig. 1). Among them, 63 patients sustained VCFs because of falls, 68 sustained injuries during routine daily activities such as turning over in bed or engaging in hobbies, and 12 had an unknown mechanism of injury. Additionally, 79 of 143 participants (55.2%) had a history of osteoporosis.
Flowchart of the patient selection process. VCF, vertebral compression fracture; HDS-R, revised Hasegawa Dementia Scale.
The ethics committee of Kurume Rehabilitation Hospital approved this study (IRB Approval No. 24-004), and data collection was initiated following this approval. This study was conducted per the ethical principles outlined in the Declaration of Helsinki. Patient information was retrospectively extracted from medical records, with full consideration for the protection of personal data.
Measurements
Baseline data collected included age at admission, sex, number of vertebral fractures at the time of injury, HDS-R score at admission, handgrip strength at admission, SMI at admission, PhA at admission, and length of hospital stay. Additionally, the Functional Independence Measure (FIM) at discharge was recorded.
In our rehabilitation ward, physical and occupational therapies are conducted daily. Physical therapy includes standing exercises, walking training, and ADL training adjusted according to the patient’s pain level. Occupational therapy encompasses ADL and instrumental activities of daily living training, which are also tailored to the patient’s pain tolerance.
Based on the number of vertebral fractures, patients were categorized as having single or multiple vertebral fractures. The HDS-R is a nine-item cognitive screening tool with a maximum score of 30 points, which is widely used in Asia.15-17) Initial cognitive assessments were conducted upon admission.
Grip strength was measured upon admission using a digital handgrip dynamometer (TKK 5401; Takei Scientific Instruments Co., Niigata, Japan).18) Left- and right-hand grip strengths were assessed two times while the patient was seated, and the highest value for each side was recorded.
Body composition was evaluated using a body composition analyzer (InBody S10; InBody Co., Ltd., Seoul, Korea) to determine SMI and PhA via bioelectrical impedance analysis (BIA). This device is suitable for patients who cannot stand or sit. The SMI was measured in the supine position for the upper limbs, lower limbs, and trunk at admission. It was calculated by dividing the sum of the muscle mass of the upper and lower limbs by the square of the patient’s height (m²).19) PhA, reflecting cellular health and nutritional status,20) was calculated using resistance (R) and reactance (Xc) values obtained from the limbs and trunk via BIA at a 50-kHz frequency,21) using the following formula: PhA = arctangent (Xc/R) × (180/π). Per the device measurement protocol, the phase angle of the right side of the body (right arm, right leg, and right trunk) typically represents whole-body assessment. Herein, the right-side measurement was adopted as the representative value.22)
The FIM index assesses ADL performance.23) At discharge, physical therapists, occupational therapists, and nurses evaluated FIM scores and determined walking ability based on the walking subscale. The accuracy of the FIM scores at discharge was subsequently verified by the supervisor of the physical and occupational therapy departments at the rehabilitation center.
Walking ability was categorized per previous research.24,25) Patients were classified into two groups according to their FIM walking subscale score at discharge: those scoring ≥6 points were assigned to the independent walking group, whereas those scoring ≤5 points were placed in the assisted walking group.
Statistics Analysis
For statistical analysis, group comparisons for sex and the number of vertebral fractures were conducted using the χ² test, whereas age at admission, HDS-R scores, grip strength, SMI, and PhA were compared using t-tests. Effect sizes (ES) were calculated using Cohen’s d and φ coefficients to measure the magnitude of differences. ES were interpreted as small (0.1–<0.3), medium (0.3–0.5), and large (>0.5).26) To control for background factors, including age at admission, sex, and the number of vertebral fractures, allowing for the selection of the independent and assisted walking groups for analysis, propensity score matching was employed.
A binary logistic regression analysis was then performed, with walking ability at discharge as the dependent variable and HDS-R score at admission, handgrip strength, SMI, PhA, and length of hospital stay as independent variables. Finally, the cutoff value indicating walking independence was determined using the receiver operating characteristic (ROC) curve.
Statistical analysis was performed using the Statistical Package for the Social Sciences (version 28.0; IBM, Armonk, NY, USA), with a significance level set at 5%.
RESULTS
Participant Characteristics before Propensity Score Analysis
Among the injury sites, 56 cases (39.2%) involved thoracic fractures, 74 cases (51.7%) involved lumbar fractures, and 13 cases (9.1%) involved thoracolumbar fractures. Before propensity score matching, 143 patients (28 men and 115 women; mean age, 85.4±6.4 years) were included in the analysis. At discharge, 95 patients (66.4%) could walk independently, while 48 patients (33.6%) required assistance.
A comparison between the independent and assisted walking groups at discharge revealed that the independent walking group had significantly higher HDS-R score at admission (p<0.001, ES=0.38), handgrip strength (p<0.001, ES=0.29), SMI (p<0.001, ES=0.23), and PhA (p<0.001, ES=0.40) than the assisted walking group. In particular, the differences in the HDS-R score and PhA at admission were the most pronounced (Table 1).
Characteristics of physical function before propensity score matching
Participant Characteristics after Propensity Score Analysis
After adjusting for background factors using propensity score matching, 94 patients (19 men and 75 women; mean age, 86.1±6.5 years) were included in the analysis. At discharge, 48 patients (51.1%) could walk independently, while 46 patients (48.9%) required assistance.
A comparison between the independent and assisted walking groups revealed that HDS-R score at admission (p<0.001, ES=0.42), handgrip strength (p=0.001, ES=0.23), SMI (p=0.001, ES=0.23), and PhA (p<0.001, ES=0.40) were significantly higher in the independent walking group. Even after propensity score analysis, the differences in the HDS-R score and PhA at admission remained particularly pronounced (Table 2).
Characteristics of physical function after propensity score matching
Results of Logistic Regression Analysis after Propensity Score Matching
To control the potential confounding variables, propensity score matching was performed before conducting logistic regression analysis. In Model 1, the HDS-R score, handgrip strength, SMI, and PhA at admission were included as independent variables, with walking ability at discharge as the dependent variable. Reportedly, walking ability at discharge was significantly associated with the HDS-R score (odds ratio [OR]=1.20; 95% confidence interval [CI], 1.08–1.33; p=0.001) and PhA (OR=2.74; 95% CI, 1.16–6.48; p<0.001). Model 1 demonstrated a statistically significant model fit (χ² test, p<0.001) and adequate discriminative accuracy (Hosmer–Lemeshow test, p=0.832; classification accuracy=74.4%).
In Model 2, the length of hospital stay was included as an additional covariate. After this adjustment, significant associations persisted among walking ability at discharge, the HDS-R score (OR=1.19; 95% CI, 1.05–1.34; p=0.005), and PhA (OR=3.21; 95% CI, 1.25–8.23; p=0.015). Moreover, Model 2 demonstrated good statistical fit (χ² test, p<0.001), with a Hosmer–Lemeshow test result of p=0.078 and an improved classification accuracy of 76.4%. No independent variables exhibited multicollinearity (variance inflation factor <5) (Table 3).
Results of the logistic regression analysis
ROC analysis of the HDS-R score at admission yielded an area under the curve (AUC) of 0.74, with an optimal cutoff of 22.5 points (sensitivity=72.9%, specificity=31.1%) (Table 4, Fig. 2). Similarly, the ROC analysis of PhA at admission showed an AUC of 0.77, with a cutoff of 3.75° (sensitivity=72.9%, specificity=21.7%) (Table 4, Fig. 3).
Results of ROC analysis
Receiver operating characteristic curve for estimating the cutoff value of the revised Hasegawa Dementia Scale that determines walking ability at discharge. The phase angle cutoff was 22.5 point (sensitivity=72.9%, specificity=31.1%, area under the curve=0.74, 95% confidence interval 0.64–0.85).
Receiver operating characteristic curve for estimating the cutoff value of the phase angle that determines walking ability at discharge. The phase angle cutoff was 3.75° (sensitivity=72.9%, specificity=21.7%, area under the curve=0.77, 95% confidence interval 0.67–0.87).
DISCUSSION
Patients with acute-phase VCFs often experience severe pain, leading to prolonged bed rest and limited early physical therapy assessment. This study identified early predictors of walking ability at discharge. Notably, the HDS-R score, handgrip strength, SMI, and PhA at admission were significantly higher in the independent walking group than in the assisted walking group. These associations remained significant even after adjusting for age, sex, and the number of vertebral fractures during injury. Notably, the HDS-R score and PhA were the strongest predictors of walking ability at discharge.
Cognitive function and balance have been identified as key factors influencing walking independence in patients with VCFs.27) Herein, the findings align with previous research, further reinforcing these associations. Notably, the OR for the HDS-R score at admission was 1.19, indicating that each one-point increase in HDS-R score corresponded to a 1.19-fold greater likelihood of achieving independent walking at discharge. This finding underscores the critical role of cognitive function in improving walking ability.
Furthermore, the calculated cutoff value for the HDS-R score in this study was 22.5 points. Umehara et al.28) reported that an HDS-R score of <23 points at admission served as a reliable predictor of ADL decline 1 year after hip fracture surgery. Given the close similarity between these cutoff values, our study findings gain further credibility.
Additionally, the ROC analysis of the HDS-R score at admission yielded an AUC of 0.74. Although not optimal, this value indicates sufficient diagnostic accuracy.29,30) With a relatively high sensitivity of 72.9%, the HDS-R score during admission is a practical and clinically meaningful predictor of walking ability at discharge in patients with VCFs. Nevertheless, the AUC for the HDS-R score at admission had a 95% CI of 0.64–0.85, with the lower bound falling below 0.7. This may reflect the influence of a limited sample size in the present study; thus, caution is warranted while generalizing these findings.
From a clinical perspective, patients with cognitive impairment often struggle to understand and follow movement instructions, which can significantly hinder their ability to regain independent walking. These findings highlight the need to address cognitive and physical factors in rehabilitation. Following VCFs, patients frequently experience severe pain, leading to prolonged immobility and decreased activity levels and eventually contributing to cognitive decline and delirium. To enhance daily activity levels and support functional recovery, it is essential to implement effective pain management strategies, ensure proper positioning, facilitate early mobilization using wheelchairs or reclining chairs, and incorporate gross motor exercises and cognitive tasks during optimal alertness.
This study also identified a strong association between PhAs at admission, as measured by body composition analysis, and walking ability at discharge. Specifically, for each 1° increase in PhA, the likelihood of achieving independent walking at discharge increased by 3.21 times. This underscores the role of PhA as a critical marker of muscle quality and cellular health.31) Research has previously linked improvements in PhA to enhanced ADL among patients with osteoporotic fractures.32) Additionally, PhA indicates nutritional status in older patients with hip fractures,33) and its improvement has shown a significant correlation with FIM scores.34) These findings align with previous studies, suggesting that enhancing PhA—and consequently nutritional status—can contribute to better walking outcomes in patients with VCFs.
This study determined a PhA cutoff value of 3.75°. Comparatively, Kubo et al.33) reported a cutoff value of 3.96° for identifying malnutrition in elderly patients with hip fractures. The similarity between these values reinforces the validity of the present findings, suggesting that nutritional status plays a pivotal role in walking recovery, even among patients with VCFs. Furthermore, the ROC analysis of PhA upon admission yielded an AUC of 0.77, indicating moderate but sufficient diagnostic accuracy.29,30) Given its relatively high sensitivity (72.9%), PhA at admission serves as a meaningful clinical predictor of walking ability at discharge for patients with VCFs. Nevertheless, the AUC for the PhA at admission had a 95% CI of 0.67–0.87, with the lower bound falling below 0.7. This may reflect the influence of a limited sample size in the present study; thus, caution is warranted while generalizing these findings. These findings emphasize the importance of improving muscle quality and nutritional status in walking recovery and highlight the need for targeted rehabilitation and nutritional interventions to enhance functional outcomes.
Conducting comprehensive early physical therapy assessments in patients with VCFs during the acute phase is challenging because of severe pain. However, a key strength of this study is that even with limited assessment parameters, clinically significant insights can be obtained by analyzing the relationship between PhA and walking ability at discharge. Early assessments should focus on evaluating cognitive function and body composition parameters, such as PhA, as these factors reflect muscle quality and nutritional status, both of which influence rehabilitation outcomes. This study underscores the need of integrating cognitive function and nutritional status evaluations in early physical therapy assessments for patients with VCFs to optimize rehabilitation and recovery.
This study had several limitations. First, its single-center design restricted the generalizability of the findings. Future research should validate these results in a multicenter setting. Second, walking independence at discharge may have been influenced by variations in rehabilitation programs. In this study, participants received physical and occupational therapies, including standing exercises, gait training, and ADL exercises; however, the intensity and complexity of the programs were not standardized. Future studies should explore the impact of different rehabilitation protocols. Conducting a thorough initial assessment in patients with VCFs is challenging owing to the intense pain experienced during the acute phase.
In conclusion, this study identified the HDS-R score and PhAs at admission as key independent predictors of walking ability at discharge in patients with acute-phase VCFs, even in the presence of pain. Notably, these associations remained significant after adjusting for confounding factors such as age, sex, number of vertebral fractures, and length of hospital stay, thereby reinforcing their clinical relevance in rehabilitation planning. Moreover, this study established cutoff values for the HDS-R score (22.5 points) and PhA (3.75°), which provide a practical framework for assessing prognosis and guiding early rehabilitation strategies. These thresholds serve as valuable reference points for setting individualized rehabilitation goals and optimizing physical therapy interventions to enhance patient outcomes.
The findings emphasize the critical role of early assessment in predicting functional recovery in patients with acute-phase VCFs. Utilizing HDS-R and PhA as assessment tools enables clinicians to evaluate walking potential without imposing excessive strain on patients experiencing severe pain. This proactive approach facilitates early intervention, improves prognostic accuracy, and supports the development of targeted rehabilitation plans that maximize mobility and independence.
Notes
We would like to acknowledge all patients who agreed to participate in this study.
CONFLICT OF INTEREST
The researchers claim no conflicts of interest.
FUNDING
None.
AUTHOR CONTRIBUTIONS
Data curation, KH, EN, JI; Formal analysis, KH, HO; Investigation, KH, HO; Methodology, KH HO; Project management, KH, HO, JT, HS; Supervision, HO; Writing_original draft, KH; Writing_review & editing, KH, HO.