Discriminative ability of functional performance tests on risk of falling among older adults with knee A cross-sectional study

Authors: Lalita Khuna, Theardkhwan Plukwongchuen, Weeranan Yaemrattanakul, Pei-Yun Lee

Abstract

Identifying fall risk among older adults with knee osteoarthritis (OA) is essential for targeted prevention. Although functional tests assessing mobility, strength, and balance are widely used, no consensus exists on the most effective test to identify fall risk in this population. This study aimed to compare functional performance tests and fall risk between older adults with and without knee OA, compare the tests in older adults with and without fall risk in both groups, and determine cutoff scores for these tests to identify fall risk among older adults with knee OA. This cross-sectional study included 106 participants aged ≥ 60 years (53 in each group). The participants completed the Thai falls risk assessment test and performed 5 functional performance tests, including the timed-up and go test (TUGT), functional reach test, alternate-step test (AST), 5 times sit-to-stand test (FTSST), and 10-meter walk test (10MWT). The independent t-test and Mann–Whitney U test were used to compare outcomes, and receiver operating characteristic curves were used to identify the optimal cutoff scores. Older adults with knee OA had a higher fall risk and performed significantly worse on the TUGT, AST, FTSST, and 10MWT than those without knee OA. The TUGT and AST scores effectively distinguished the fall risk between older adults with and without knee OA. cutoff scores of 10.5 and 24.5 seconds for the TUGT and AST, respectively, had acceptable area under curve values, effectively discriminating fall risk among older adults with knee OA. Older adults with knee OA exhibit a higher risk of falls and reduced functional abilities. The TUGT and AST may serve as useful fall-risk screening tools among this population.

1. Introduction

Knee osteoarthritis (OA) is a common disease among older adults, characterized by gradual deterioration of articular cartilage, thickening of bone edges, and reduction in synovial fluid in the knee joint. Such changes often result in pain, reduced mobility, and weakened muscles around the joint,^[1]^ leading to impaired balance and abnormal gait. Consequently, older adults with knee OA are at high risk of falling,^[2]^ which may cause severe injuries and fall-related disabilities.

Regarding the context of physical therapy focused on promotion, prevention, treatment, and rehabilitation, various treatment methods have been developed to manage pain, maintain and inhibit the progression of knee OA, and restore knee function as close to normal as possible. However, in addition to enhancing the treatment and rehabilitation efficacy, identifying simple assessment tools to screen and monitor the functional abilities of older adults with knee OA is equally important for predicting fall risk. Falls often stem from physical impairments such as difficulties with walking, balance, core and leg muscle strength, movement endurance, and other health issues.^[3,4]^ Therefore, assessments indicating fall risk should align with or reflect these physical deficits, which are key contributors to falls.

Generally, healthcare professionals in clinical and community settings use functional performance tests as simple tools for screening and monitoring patients’ abilities.^[5]^ A literature review revealed various functional performance tests, designed to assess specific physical abilities or impairments, capable of indicating fall risk. These include the time-up and go test, which reflects mobility and dynamic balance control, functional reach test (FRT), which measures static balance while standing, alternate-step test, which assesses weight-shifting and stability, 5-time sit-to-stand test, which indicates lower-limb strength and balance, and 10-meter walk test, which evaluates walking speed.^[6–10]^

Given their ability to assess key physical domains such as balance, mobility, lower-limb strength, and gait speed, which are commonly impaired and strongly associated with fall risk, these functional performance tests have been widely utilized across various populations, regardless of the underlying pathology. They were originally validated in other population groups such as community-dwelling older adults, individuals with stroke, Parkinson disease, and spinal cord injury.^[6–10]^ However, their applicability to older adults with knee OA remains limited due to differences in population characteristics and condition-specific impairments. Additionally, a lack of evidence is evident regarding knee OA’s influence or association with fall risk. Therefore, examining these functional performance tests’ abilities to discriminate the risk of falling, particularly among older adults with knee OA, would be beneficial. The identification of the most effective test concerning usability and accuracy in predicting fall risk would assist physical therapists and healthcare professionals in selecting a particular test for screening and monitoring fall risk in older adults with knee OA.

Therefore, this study aimed to compare functional performance and risks of falling between older adults with and without knee OA, compare functional performance between older adults with and without fall risks, in both groups, and estimate the cutoff point, sensitivity, and specificity of functional performance tests for discriminating risks of falling among older adults with knee OA.

2. Methods

2.1. Participants

This study employed a cross-sectional design and included individuals aged ≥ 60 years, both with and without clinically diagnosed knee OA. Participants with knee OA were recruited from primary care units in southern Thailand. The diagnosis of knee OA was based on the clinical classification criteria of the American College of Rheumatology, which is knee pain accompanied by osteophytes in the knee joint, along with at least one of the following morning stiffness and difficulty with knee movement and/or a crepitus sound during knee movement.^[11]^ Additionally, participants without knee OA were community-dwelling individuals aged ≥ 60 years who had no prior physician diagnosis of knee OA served as a comparison group. The exclusion criteria were selected to minimize confounding effects related to mobility limitations or neurological impairments. Participants were excluded if they a history of hip or knee replacement surgery; a history of below-knee amputation; severe pain in 1 or both knees at rest (scoring > 7 out of 10 on the numerical rating scale); neurological conditions leading to sensory loss, muscle weakness, or balance issues, such as stroke, spinal cord injury, Parkinson disease, dementia, vertigo, or vestibular disorders; and inability to comprehend verbal instructions.

A sample size calculation was performed based on data from Amano et al study,^[12]^ assuming a 21% fall risk among individuals without knee OA and a 79% fall risk among those with knee OA. To attain a 99% confidence level (alpha = 0.01) and 90% statistical power (beta = 0.10), at least 53 participants were required in each group, with a total of 106 participants.

This study was performed following the principles of the Declaration of Helsinki. The study was approved by the Human Research Ethics Committee, Faculty of Medicine, Prince of Songkla University (REC. 65-069-35-2) on April 26, 2022. All participants who met the eligibility criteria provided written informed consent before participating in the study.

2.2. Research protocol

Participants were recruited using purposive sampling to ensure equal numbers in the knee OA and non-OA groups, based on predefined eligibility criteria. Demographic information, including age, sex, body mass index, and underlying diseases (e.g., diabetes, hypertension, and hyperlipidemia), was collected from all participants. All participants were assessed for fall risk and functional performance. In addition, the characteristics of knee OA, including onset time, type of lesion, and severity level using the Oxford Knee Score, were obtained from participants with knee OA. The data used in these analyses were collected from July 2022 to July 2023. The details of the measurements are described below.

2.2.1. Oxford knee score

The Oxford Knee Score has demonstrated strong psychometric properties in its original version, with a Cronbach alpha of 0.93 and good test-retest reliability in patients undergoing total knee arthroplasty.^[13]^ The present study used the 0 to 48 scoring format (12 items scored from 0 to 4), which has been widely adopted in subsequent clinical research. Knee OA severity was classified as severe (0–19 points), moderate (20–29 points), mild (30–39 points), and normal or asymptomatic (40–48 points).^[14]^

2.2.2. Risk of falling assessment

Each group’s participants were further divided into 2 subgroups, those with and without fall risk, based on the results of the Thai falls risk assessment test (Thai-FRAT), which was used to assess their risk of falling. The test demonstrated a sensitivity of 0.92 and a specificity of 0.83, indicating its effectiveness in identifying fall likelihood.^[15]^ Fall risk scores were calculated using the Thai-FRAT based on 6 sex (0: male, female), visual acuity (0: able to read, unable to read more than half of the letters on the 6/12 line of a Snellen chart), balance ability (0: able to maintain a full tandem stance for 10 seconds, unable to do so), use of specific medications (0: not taking-, taking at least one of the sedatives/hypnotics, psychotropic drugs, antihypertensives, diuretics, or 4 or more other medications), history of falls (0: no falls or only 1 fall, 2 or more falls in the past 6 months), and Thai style housing conditions (0: not living-, living in a stilt house 1.5 meters or more above the ground). A score of 0 to 3 points was classified as “no fall risk,” whereas a score of 4 to 11 points was classified as “fall risks.”^[15]^

2.2.3. Functional performance measures

All participants underwent 5 functional performance tests to evaluate their physical function and fall risk. All tests were repeated 3 times, and the average value was calculated for analysis.

The timed-up and go test (TUGT) assesses dynamic balance, basic mobility for daily activities, and fall risk among older adults,^[16]^ demonstrating strong test-retest reliability (intraclass correlation coefficient  = 0.97). The participants were instructed to rise from a standard armrest chair, walk straight for 3 m, turn around a cone, and return to sit on the chair at the fastest and safest pace. The time required to complete the task was measured in seconds using a timer.

The functional reach test (FRT) measures body stability and shows good reliability (ICC = 0.81) in adults and older individuals (21–87 years old).^[17]^ The participants were instructed to stand sideways against a wall with their arms raised at a 90° angle parallel to the floor and subsequently lean forward as far as possible without losing balance or taking a step while maintaining their arms at the same height. The assessor measured the distance (cm) between the starting position and the furthest reach of the middle finger.

The AST, adapted from the stool-stepping task of the Berg balance scale, evaluates the ability to transfer weight and maintain body stability [9] and has good test-retest reliability in older adults (ICC = 0.78).^[18]^ Participants were instructed to place their left and right feet (without shoes) on a set of stairs measuring 18 × 40 × 60 cm (height × width × length) 8 times with each leg alternately as quickly and safely as possible. The time (sec) required to complete 8 steps for each foot was measured using a timer.^[19]^

The 5-time sit-to-stand test (FTSST) assesses leg muscle strength and has excellent rater and test-retest reliability (ICC = 0.84–1.00) in older adults with knee OA.^[20]^ The participants were instructed to stand up from a standard-height chair with the hips and knees fully extended and subsequently sit down 5 times as quickly and safely as possible. The time (second) required to complete the task was measured using a timer.

The 10-meter walk test (10MWT) evaluates walking speed and demonstrates good test-retest reliability (ICC = 0.85) in individuals with knee OA.^[21]^ The participants were instructed to walk at a comfortable pace along a 10-meter walkway, with or without a walking aid. The assessor recorded the time taken at 4-meter intervals. Walking speed was calculated using the following walking speed = distance/time (m/second).^[22]^

Prior to the main data collection, the inter-rater reliability of the 5 functional performance tests was assessed using a separate sample of 20 older adults. Three trained assessors independently administered each test. All assessors involved in the main study were licensed physical therapists with prior experience in administering these assessments and followed standardized testing protocols to ensure consistency in measurement. The ICCs were calculated using the 2-way mixed effects model (consistency). The results demonstrated excellent inter-rater reliability for all tests as TUGT = 0.987 (95% CI: 0.973–0.994), FRT = 0.933 (0.858–0.971), AST = 0.985 (0.969–0.994), FTSST = 0.970 (0.937–0.987), and 10MWT = 0.973 (0.943–0.989).

2.3. Statistical analysis

Descriptive statistics were used to analyze the baseline demographics, knee OA characteristics, and the study’s findings. The Shapiro–Wilk test was used to assess data distribution normality. To compare the findings between participants with and without knee OA, as well as with and without fall risk, categorical data were analyzed using a chi-square test, whereas continuous data were compared using either an independent t-test or the Mann–Whitney U test depending on the data distribution. Receiver operating characteristic (ROC) curves were used to estimate the optimal cutoff scores to discriminate the risk of falling among participants with knee OA. The area under the receiver characteristic curve (AUC) of ≥0.50 is regarded as an acceptable level of discriminative ability.^[23]^ All statistical analyses were conducted using SPSS version 25 (IBM Corp., New York). Statistical significance was set at *P *< .05 was considered statistically significant.

3. Results

A total of sample comprised 106 older adults (average 70.1 ± 5.6 years), of which 53 participants had knee OA and 53 participants did not. Most participants were female (75.5%), had underlying diseases (73.6%), and their body mass index classified them as overweight. Most participants with knee OA were diagnosed with the condition for more than a year and experienced mild-to-moderate disease severity. No significant differences in age, sex, or body mass index were found between participants with and without knee OA. The participants with knee OA had significantly more underlying diseases (*P *= .002) and a higher risk of falling (*P *= .049) than those without knee OA (Table 1).

The functional performance results indicated that participants without knee OA performed significantly better in several tests than those with knee OA. Particularly, these participants required less time to complete the TUGT, AST, and FTSST and exhibited faster walking speeds in the 10MWT (*P *< .001, Table 1) than those with knee OA. These findings emphasize the notable differences in mobility, strength, and gait speed between the 2 groups. However, the FRT outcomes did not differ significantly between participants with and those without knee OA (*P *= .109, Table 1).

When comparing the characteristic data and functional performance results between the subgroups with and without fall risk, distinct differences emerged. Among the participants with knee OA, the severity of the condition significantly differed between those with and without fall risk (P < .022). Conversely, significant sex differences were observed among participants without knee OA (*P *< .021, Table 2). Moreover, participants without fall risk performed significantly better on the TUGT and AST than those with fall risk, regardless of knee OA status (*P *< .05, Figs. 1A and C). Furthermore, only participants without knee OA and fall risk showed significantly better performance on the FTSST than those with fall risk (*P *= .015), whereas no difference was observed in those with knee OA (P = .052, Fig. 1D). No significant differences in FRT and 10MWT were found between participants with and those without fall risk in either group (*P *> .05, Figs. 1B and E).

The ROC curve analysis identified cutoff scores of 10.5 seconds and 24.5 seconds for the TUGT and AST, respectively, both yielding acceptable and significant AUC values for determining fall risk among participants with knee OA (sensitivity = 77.3% for both tests; specificity = 54.8% and 61.3% for the TUGT and AST, respectively; AUC = 0.67 and 0.68, respectively, P < .05). Although the AUC value for the FTSST cutoff score exceeded the acceptable level, the difference was not statistically significant. The AUC for the FRT and 10MWT cutoff scores were below the acceptable threshold for identifying the risk of falling among participants with knee OA (Table 3).

4. Discussion

This study examined functional performance tests’ (TUGT, FRT, AST, FTSST, and 10MWT) ability to differentiate falling risk among older adults with knee OA by comparing test outcomes among individuals with and without knee OA, within subgroups with and without a risk of falling, and using ROC curve analysis. The findings indicated that functional performance tests, except the FRT, could significantly differentiate between older adults with and without knee OA. Among those with knee OA, performance on the TUGT and AST significantly differed between individuals classified as being at risk of falling and those not at risk, according to the Thai-FRAT tool.

The significant differences in the functional performance tests, including the TUGT, AST, FTSST, and 10MWT, between older adults with and without knee OA (Table 1) suggested that these tests effectively differentiated performance levels between these groups. This study’s findings align with those of previous studies, which showed that individuals with knee OA performed significantly worse than those without knee OA on the TUGT, stair-climbing test, chair-stand test, and walking speed test. Additionally, a previous study showed that these physical function performance tests are highly effective in distinguishing individuals with knee OA from healthy controls.^[24]^ Conversely, the FRT showed no significant difference between older adults with and without knee OA (Table 1), consistent with Osaki et al,^[25]^ who found similar FRT outcomes in women with and without radiographic knee OA. This may be owing to the FRT primarily challenging static balance by relying on ankle and hip strategies,^[26]^ making it insufficient for distinguishing performance in individuals with and without knee OA.

A greater proportion of older participants with knee OA were classified as at risk of falling, aligning with prior study of Kelsey et al,^[27]^ who found that approximately 30% of older adults with knee OA experience falls. Another study reported a fall frequency of 63.2% among older adults with severe knee OA,^[28]^ which is consistent with this study’s finding that knee OA severity significantly differed between those with and without fall risk (Table 2). Greater OA severity is likely to increase pain, stiffness, and dysfunction, leading to instability, muscle weakness, and a higher fall likelihood.

When comparing the subgroups with and without fall risk, the TUGT and AST findings suggested that these tests effectively distinguished the risk of falling in participants with and without knee OA (Figs. 1A and C). These findings are coherent with that of Alencar et al,^[29]^ which found a statistically significant difference in TUGT performance for functional mobility among older women with knee OA, with median times of 10.08 seconds for non-fallers and 11.59 seconds for fallers (*P *= .038). A systematic review and meta-analysis reported that older adults who were fallers had significantly longer mean TUGT times than non-fallers.^[30]^ Additionally, a study by Tiedemann et al^[9]^ on community-dwelling older adults found that multiple fallers took significantly longer to complete the AST than non-multiple fallers (*P *= .007). Therefore, the TUGT and AST effectively differentiated the risk of falling among older adults with and without knee OA.

The TUGT, in which a clinician observes whether an individual exhibits difficulty or unsteadiness when standing up from a chair, walking 3 m, turning, returning, and sitting, is recommended as a simple fall risk screening tool.^[16]^ The established TUGT cutoff score in this study (Table 3) corresponds with thresholds reported in previous literature and highlights the potential utility of this test in identifying individuals who may benefit from targeted fall-prevention strategies. A previous systematic review found that cutoff times separating non-fallers from fallers ranged between 10 seconds and 32.6 seconds, with all retrospective studies showing a significant positive association between TUGT time and fall history, particularly with high odds ratios in older adults.^[31]^ To our knowledge, limited evidence supports the TUGT as an indicator of fall risk among older adults with knee OA. However, a study on older adults with hip OA found that a cutoff score of 10 to 11 seconds for the TUGT could help predict frequent near falls, suggesting its potential for assessing future fall risk among older adults with hip OA.^[32]^

Additionally, the AST is a modified version of the stool-stepping task, which was originally part of the comprehensive Berg balance scale, designed to assess clinical balance performance, and has been shown to predict fall risk among older adults.^[33]^ To our knowledge, no evidence supports the AST as an indicator of fall risk in older adults with knee OA. However, a study on community-dwelling older adults demonstrated that a cutoff score of 10 seconds for the AST has reasonable sensitivity and specificity for identifying multiple fallers.^[9]^ In contrast to the findings of the present study, this can be explained by variations in the study population and fall data collection methods.

Therefore, the TUGT, which assesses mobility and dynamic balance, and the AST, which evaluates weight-shifting and stability, likely capture the motor control and functional deficits associated with fall risk in older adults, regardless of knee OA status. These tests can detect subtle deficits contributing to fall risk, such as reduced balance and slower response times, making them valuable tools for screening fall risk across populations.

In contrast, the FTSST did not significantly differentiate fall risk among individuals with knee OA, even though its cutoff score showed an acceptable level of classification. These findings suggest that this test may not be as sensitive in capturing fall risk among this population. A previous study combined the 1-leg stand test and FTSST (cutoff scores of 5.3 seconds and 7.9 seconds, respectively) to distinguish fallers from non-fallers among community-dwelling older adults with knee OA, demonstrating high accuracy as a screening tool.^[14]^ Therefore, the use of the FTSST alone may not effectively differentiate fall risk among older adults with knee OA, emphasizing the need for fall-risk assessments that account for the specific biomechanical limitations associated with knee OA. Similarly, the 10MWT distinguished individuals with and without knee OA but failed to differentiate fall risk within the groups. This lack of distinction in fall risk may be owing to the 10MWT predominantly measuring walking speed rather than directly assessing stability or balance-related factors that are more closely related to fall risk. Consequently, the AUC values falling below the acceptable threshold suggest that in the context of knee OA, the 10MWT may have limitations as a standalone screening tool for fall risk and may be more effective when used alongside other balance-focused assessments.^[34]^

This study had some limitations that should be considered. Firstly, this study used the Thai-FRAT, a locally developed and validated screening tool with high sensitivity and specificity in older Thai older adults. While its multidimensional approach – covering balance, medication use, fall history, and environmental risk factors – is suitable for community settings, the tool was designed specifically for Thai populations. This may limit its generalizability and comparability with internationally recognized fall risk assessments. Secondly, the sample size was calculated for group comparisons rather than ROC analysis, which typically requires a larger sample to ensure stable AUC estimation. Although acceptable and statistically significant AUCs were found for the TUG and AST, these results should be interpreted with caution. Thirdly, although some functional performance tests demonstrated AUC values below the conventional threshold of 0.70 for clinical decision-making, their modest discriminative ability may still offer practical value in preliminary screening contexts. Fourthly, this study used a cross-sectional design which captured data at a single point in time, making it impossible to confirm the causality between the variables. Finally, this study lacked information on other factors related to knee OA status and fall risk, such as pain levels and muscle strength around the knee joint. Therefore, prospective studies should examine fall events over time, include a larger number of individuals with knee OA and collect and analyze relevant factors to further clarify and improve the effectiveness of functional performance tests as fall risk screening tools for older adults with knee OA.

5. Conclusion

The findings revealed that older adults with knee OA were at a higher risk of falling and performed significantly worse on mobility and balance tests (TUGT, AST, FTSST, and 10MWT), except the FRT, than those without knee OA. When comparing the subgroups with and without fall risk, the TUGT and AST results effectively distinguished fall risk in both groups. Regarding older adults with knee OA, the cutoff scores from the ROC analysis – 10.5 seconds and 24.5 seconds for the TUGT and AST, respectively – demonstrated acceptable sensitivity and specificity for identifying this with and without a risk of falling. Thus, the TUGT and AST may be useful screening tools for determining fall risk among older adults with knee OA.

Acknowledgments

The authors thank all patients for their participation in this study. We are grateful to Tanya Saetang, Pronchita Cherdchoo, and Wassana Suwattanakul for their assistance with data collection.

Author contributions

Conceptualization: Lalita Khuna.

**Data ** Lalita Khuna, Theardkhwan Plukwongchuen.

**Formal ** Lalita Khuna.

**Funding ** Lalita Khuna.

Investigation: Lalita Khuna.

Methodology: Lalita Khuna.

**Project ** Lalita Khuna.

Resources: Theardkhwan Plukwongchuen, Weeranan Yaemrattanakul.

Software: Lalita Khuna.

Supervision: Weeranan Yaemrattanakul, Pei-Yun Lee.

Validation: Weeranan Yaemrattanakul, Pei-Yun Lee.

Visualization: Pei-Yun Lee.

**Writing – original ** Lalita Khuna.

**Writing – review & ** Lalita Khuna, Theardkhwan Plukwongchuen, Weeranan Yaemrattanakul, Pei-Yun Lee.