Construction of a prediction model for fall risk in older patients with osteoporosis and comorbid with psychiatric disorders

Abstract

BACKGROUND

Older patients with osteoporosis have a high risk of falls, and depression and anxiety can aggravate this risk. At present, there is a lack of an accurate prediction model integrating physical and mental indicators.

AIM

To investigate the effect of comorbid depression/anxiety on fall risk in older patients with osteoporosis and to establish an individualized fall risk prediction model.

METHODS

A total of 130 older patients with osteoporosis admitted between December 2022 and December 2025 were divided into fall (58 cases) and non-fall (72 cases) groups according to whether falls occurred within the past year. Data were collected via questionnaire surveys, medical record reviews, and functional evaluations. Multivariate logistic regression analysis was used to screen for independent influencing factors, and a nomogram prediction model was constructed. Model discrimination and calibration were assessed by receiver operating characteristic curve and Hosmer-Lemeshow test.

RESULTS

The incidence of falls was 44.62%. Univariate and multivariate logistic regression analyses showed that a history of falls [odds ratio (OR) = 4.326, 95% confidence interval (CI): 1.987-9.416], Geriatric Depression Scale score (OR = 1.342, 95%CI: 1.127-1.598), Generalized Anxiety Disorder 7-item score (OR = 1.287, 95%CI: 1.082-1.531), and Timed Up and Go Test duration (OR = 1.218, 95%CI: 1.065-1.393) were independent risk factors for falls (P < 0.05). The Berg Balance Scale score (OR = 0.891, 95%CI: 0.826-0.961) and 25-hydroxyvitamin D levels (OR = 0.912, 95%CI: 0.855-0.973) were independent protective factors (P < 0.05). The prediction model established based on these six indicators had an area under the receiver operating characteristic curve of 0.876 (95%CI: 0.817-0.935), sensitivity of 0.845, and specificity of 0.819. The Hosmer-Lemeshow goodness-of-fit test showed χ² = 7.864, P = 0.447.

CONCLUSION

Comorbid depression/anxiety is an independent risk factor for falls in older patients with osteoporosis. The model integrating psychological and physical function assessment shows favorable predictive performance for identifying high-risk patients and guiding intervention strategies.

Key Words: Osteoporosis; Depression; Anxiety; Risk of falling; Prediction model; Aging

Core Tip: Older patients with osteoporosis are prone to falls, and this study identified depression, anxiety, poor balance, and low vitamin D levels as key contributing factors. Early screening of high-risk individuals was enabled by a predictive model that combined psychological and physical indicators and showed good performance.

INTRODUCTION

Falls are a major global public health problem among older adults and can severely affect them in several ways. They can cause direct consequences such as fractures, soft tissue injuries, and even death. Falls may also lead to secondary consequences, including limited mobility, social isolation, depression, and anxiety. This creates a vicious cycle that severely impairs the quality of life[1,2]. Older patients with osteoporosis are particularly susceptible to fall-related fractures because osteoporosis is characterized by low bone mass and deteriorated bone microarchitecture. With population aging in China, the burden of osteoporosis and other chronic diseases continues to increase[3,4]. In recent years, accumulating evidence has suggested that mental disorders such as depression and anxiety can lead to reduced energy, excessive worry, tension and restlessness, bradykinesia, impaired balance, and decreased physical activity, which are closely associated with fall risk in older patients[5,6]. The specific mechanism underlying the correlation between comorbid psychiatric disorders (depression/anxiety) and fall risk in older patients with osteoporosis has not been fully elucidated. In particular, comprehensive prediction models that integrate psychological assessments and physical function measurements are lacking[7,8]. Accordingly, this study enrolled older patients with osteoporosis admitted between December 2022 and December 2025, analyzed the correlation between comorbid depression/anxiety and fall risk, and constructed an individualized fall-risk prediction model. This study aimed to provide a reference for the early identification of high-risk patients and the formulation of comprehensive intervention strategies in clinical practice.

MATERIALS AND METHODS

Study subjects

A total of 130 older patients with osteoporosis who were hospitalized in the Department of Geriatrics and Orthopedics of our hospital between December 2022 and December 2025 were selected as the study subjects. The inclusion criteria were as follows: (1) Age ≥ 60 years; (2) Bone mineral density T-score of the lumbar spine or hip measured by dual-energy X-ray absorptiometry ≤ -2.5, meeting the diagnostic criteria of the Guidelines for the Diagnosis and Treatment of Primary Osteoporosis (2022); (3) Clear consciousness and ability to cooperate with questionnaire surveys and functional assessments; and (4) Voluntary participation in this study and signed informed consent.

The exclusion criteria were as follows: (1) Severe cognitive impairment unable to complete questionnaire assessments; (2) Diagnosis of severe mental illnesses such as schizophrenia and bipolar disorder; (3) Severe cardiac, hepatic, or renal insufficiency, malignant tumors, and other diseases with an expected survival of < 6 months; (4) Unable to complete balance function tests due to physical disability, paralysis, or other reasons; and (5) Incomplete clinical data. This study was approved by the Ethics Committee Nantong Clinical College of Nantong University (The First People’s Hospital of Nantong City) (approval No. NT2025012-16).

Research method

Data collection: Data were collected using a combination of cross-sectional surveys and retrospective reviews of medical records. Questionnaire surveys and functional assessments were completed within 48 hours of admission. All assessments were performed independently by two trained researchers, and the results were double-checked before entry into the database. The following data were collected: (1) Demographic characteristics: A self-designed questionnaire was used to collect data on age, sex, years of education, marital status (married/widowed/divorced), living status (living alone/Living with family members), and body mass index (BMI); (2) Clinical data: Duration of osteoporosis, history of falls, comorbidities including hypertension, diabetes mellitus, coronary heart disease, chronic obstructive pulmonary disease, and osteoarthritis; and medication use, such as benzodiazepines, antidepressants, antipsychotics, calcium supplements, and vitamin D supplements; (3) Bone metabolism-related indicators: Fasting venous blood test results obtained on the first day of admission, including serum calcium, serum phosphorus, alkaline phosphatase (ALP), 25-hydroxyvitamin D, and parathyroid hormone (PTH). All tests were performed uniformly by the hospital’s clinical laboratory in accordance with standard operating procedures; (4) Assessment of physical function: Grip strength: Maximum isometric grip strength of the dominant hand was measured. The test was performed using a Jamar hand dynamometer. Three measurements were obtained, and the highest value recorded. Timed Up and Go Test (TUGT): To assess comprehensive mobility, the time taken for the patient to get up from a chair, walk three meters, turn around, return, and sit down was recorded. Berg Balance Scale (BBS): This consists of 14 items with a total score of 56, with lower scores indicating poorer balance function, and was measured using the Short Physical Performance Battery (SPPB). A total score of 12 was achieved in balance tests, 4-meter walking speed, and chair sit-to-stand tests; (5) Cognitive function was assessed using the Mini-Mental State Examination (MMSE), which assesses five domains: Orientation, memory, attention and calculation, recall ability, and language. The total possible score was 30, with lower scores indicating more severe cognitive impairment; and (6) Psychological assessment: Depressive symptoms were assessed using the 15-item Geriatric Depression Scale (GDS). The total score was 15, and a score ≥ 6 indicated the presence of depressive symptoms. Anxiety symptoms were assessed using the Generalized Anxiety Disorder-7 (GAD-7) scale. The total score was 21; a score ≥ 5 indicated the presence of anxiety symptoms. Fear of falling: Assessed using the Falls Efficacy Scale-International (FES-I), which evaluates the degree of concern about falling during 16 daily activities. The total possible score was 64, with higher scores indicating a greater fear of falling.

Grouping: The patients were divided into fall and non-fall groups based on the occurrence of falls in the past year. A fall was defined as an unexpected involuntary change in body posture that resulted in the body falling to a lower level or the ground. All fall events were confirmed through medical records or by family members, and those caused by syncope, stroke, epilepsy, or other conditions were excluded.

Observation indicators

To compare differences in general demographic data, clinical data, medication use, bone metabolism indicators, physical function, cognitive function, and psychological assessment indicators between the two groups. To screen independent influencing factors for falls in older patients with osteoporosis. To construct and validate a fall risk prediction model.

Statistical analysis

Statistical analyses were performed using SPSS 26.0 and R software (version 4.2.1). Normality tests were conducted on the measured data. Data with a normal distribution were presented as the mean ± SD, and comparisons between the two groups were performed using the independent-samples t-test. Data with a non-normal distribution were presented as median (interquartile range), and comparisons between groups were performed using the Mann-Whitney U test. Count data were presented as n (%), and comparisons between groups were performed using the χ² test. Variables with P < 0.10 in the univariate analysis were included in the multivariate analysis. Multivariate binary logistic regression analysis (forward stepwise method) was used to screen independent factors influencing falls, and odds ratio (OR) and their 95% confidence interval (CI) were calculated. Based on the independent risk factors identified through multivariate analysis, a nomogram prediction model was constructed using the rms package in R, and internal validation was performed using the bootstrap method (1000 repetitions). The discrimination ability of the model was evaluated using the receiver operating characteristic (ROC) curve, area under the curve (AUC), sensitivity, specificity, positive predictive value, and negative predictive value. Calibration of the model was assessed using the Hosmer-Lemeshow goodness-of-fit test. Statistical significance was set at P < 0.05.

RESULTS

Comparison of general demographic data between the two groups

Among 130 older patients with osteoporosis, 58 experienced falls within the past year, resulting in a fall incidence rate of 44.62%. The fall group exhibited a higher mean age and higher proportions of solitary living and widowed/divorced status than the non-fall group (P < 0.05). No statistically significant differences were observed between the two groups in terms of sex, years of education, or BMI (P > 0.05) (Table 1).

Table 1 Comparison of general demographic data between the two groups, mean ± SD/n (%).

Variable	Fall group (n = 58)	Non-fall group (n = 72)	Statistic	P value
Age (years)	73.58 ± 6.91	69.53 ± 6.28	t = 3.482	0.001
Sex			χ2 = 2.341	0.126
Male	19 (32.76)	32 (44.44)
Female	39 (67.24)	40 (55.56)
Years of education	8.12 ± 3.41	8.73 ± 3.15	t = 1.067	0.288
Marital status			χ2 = 4.523	0.033
Married	31 (53.45)	51 (70.83)
Widowed/divorced	27 (46.55)	21 (29.17)
Living status			χ2 = 5.872	0.015
Living with family	37 (63.79)	59 (81.94)
Living alone	21 (36.21)	13 (18.06)
BMI (kg/m2)	23.28 ± 3.56	23.98 ± 3.31	t = 1.168	0.245

BMI: Body mass index.

Comparison of clinical data and medication use between the two groups

The proportion of patients with a history of falls, prevalence of comorbid chronic diseases, and usage rate of benzodiazepines were higher in the fall group than in the non-fall group (P < 0.05), whereas no statistically significant difference was observed in the usage rate of antidepressants between the two groups (P > 0.05) (Table 2).

Table 2 Comparison of clinical data and medication status between the two groups, mean ± SD/n (%).

Variable	Fall group (n = 58)	Non-fall group (n = 72)	Statistic	P value
Duration of osteoporosis (years)	5.68 ± 3.42	4.97 ± 3.18	t = 1.229	0.221
History of falls	35 (60.34)	18 (25.00)	χ2 = 16.731	< 0.001
Number of comorbidities, median (interquartile range)	4.00 (2.00, 5.00)	2.00 (1.00, 4.00)	Z = -2.937	0.003
Medication
Benzodiazepines	24 (41.38)	13 (18.06)	χ2 = 8.646	0.003
Antidepressants	8 (13.79)	5 (6.94)	χ2 = 1.713	0.191
Antipsychotics	3 (5.17)	2 (2.78)	χ2 = 0.074	0.786
Calcium supplementation	32 (55.17)	48 (66.67)	χ2 = 1.814	0.178
Vitamin D supplementation	28 (48.28)	44 (61.11)	χ2 = 2.142	0.143
Antihypertensive agents	41 (70.69)	43 (59.72)	χ2 = 1.680	0.195
Antidiabetic agents	19 (32.76)	21 (29.17)	χ2 = 0.196	0.658

Comparison of bone metabolism indicators between the two groups

The level of 25-hydroxyvitamin D in the fall group was lower than that in the non-fall group (P < 0.05). There were no significant differences in serum calcium, phosphorus, ALP, or PTH levels between the two groups (P > 0.05) (Table 3).

Table 3 Comparison of bone metabolism markers between the two groups, mean ± SD.

Variable	Fall group (n = 58)	Non-fall group (n = 72)	Statistic	P value
Blood calcium (mmol/L)	2.26 ± 0.18	2.29 ± 0.16	1.007	0.316
Serum inorganic phosphorus (mmol/L)	1.13 ± 0.21	1.16 ± 0.19	0.861	0.391
ALP (U/L)	89.47 ± 23.68	83.92 ± 21.45	1.408	0.162
25-hydroxyvitamin D (ng/mL)	18.36 ± 5.72	22.84 ± 6.31	4.196	< 0.001
PTH (pg/mL)	52.38 ± 18.46	47.65 ± 16.92	1.531	0.128

ALP: Alkaline phosphatase; PTH: Parathyroid hormone.

Comparison of physical function and cognitive function between the two groups

Grip strength, BBS, SPPB, and MMSE scores in the fall group were all lower than those in the non-fall group, whereas TUGT time was longer in the fall group (P < 0.05) (Table 4).

Table 4 Comparison of physical function and cognitive function between the two groups, mean ± SD.

Variable	Fall group (n = 58)	Non-fall group (n = 72)	t value	P value
Grip strength (kg)	21.34 ± 5.86	24.78 ± 6.23	3.243	0.002
TUGT (second)	15.68 ± 4.73	11.94 ± 3.86	4.962	< 0.001
BBS score (points)	41.53 ± 7.26	47.82 ± 6.54	5.234	< 0.001
SPPB score (points)	8.16 ± 2.34	9.73 ± 2.18	3.966	< 0.001
MMSE score (points)	25.37 ± 3.14	26.82 ± 2.86	2.773	0.006

TUGT: Timed Up and Go Test; BBS: Berg Balance Scale; SPPB: Short Physical Performance Battery; MMSE: Mini-Mental State Examination.

Comparison of psychological assessment indicators between the two groups

The GDS score, GAD-7 score, FES-I score, and positive rates of depressive and anxiety symptoms in the fall group were all higher than those in the non-fall group, with statistically significant differences (P < 0.05) (Table 5).

Table 5 Comparison of psychiatric and psychological assessment indicators between the two groups, mean ± SD/n (%).

Variable	Fall group (n = 58)	Non-fall group (n = 72)	Statistic	P value
GDS score (points)	7.43 ± 3.26	4.68 ± 2.85	t = 5.138	< 0.001
Positive depressive symptoms	35 (60.34)	21 (29.17)	χ2 = 12.746	< 0.001
GAD-7 score (points)	6.87 ± 3.52	4.21 ± 2.94	t = 4.691	< 0.001
Positive anxiety symptoms	28 (48.28)	16 (22.22)	χ2 = 9.873	0.002
FES-I score (points)	38.46 ± 9.87	27.53 ± 8.64	t = 6.749	< 0.001

GDS: Geriatric Depression Scale; GAD-7: Generalized Anxiety Disorder-7; FES-I: Falls Efficacy Scale-International.

Multivariate logistic regression analysis of factors influencing falls

Variables with a P value of < 0.10 in the univariate analysis were included in the multivariate logistic regression analysis. To assess the multicollinearity between the GDS and GAD-7 scores, the Spearman correlation coefficient and variance inflation factor (VIF) were calculated. The results showed that the correlation coefficient between the two was 0.523 (P < 0.05), VIF = 1.47, indicating a degree of correlation but not reaching the threshold for significant multicollinearity (VIF < 5); therefore, both were included in the multivariate model. The stepwise forward method was used to identify the independent risk factors for falls. After a history of benzodiazepine use was included in the model via stepwise regression, the partial regression coefficient decreased from 1.168 to 0.624 and the P value increased from 0.003 to 0.092 upon the introduction of the GAD-7 score, leading to its exclusion. After the comorbidity burden was included in the model, the P value increased from 0.006 to 0.183 upon the introduction of the TUGT and BBS scores, leading to its exclusion. During stepwise regression, the MMSE score had a P value of 0.089 and was not permitted to enter the model. Ultimately, the results showed that a history of falls (OR = 4.326, 95%CI: 1.987-9.416), GDS score (OR = 1.342, 95%CI: 1.127-1.598), GAD-7 score (OR = 1.287, 95%CI: 1.082-1.531), TUGT time (OR = 1.218, 95%CI: 1.065-1.393) were identified as independent risk factors for falls in older patients with osteoporosis, while BBS score (OR = 0.891, 95%CI: 0.826-0.961), 25-hydroxyvitamin D levels (OR = 0.912, 95%CI: 0.855-0.973) were independent protective factors (Tables 6 and 7).

Table 6 Variable assignment table for multivariate logistic regression analysis.

Variable	Assignment	Variable	Assignment
Fall (dependent variable)	0 = no fall, 1 = fall occurred	Grip strength	Continuous variable
Age	Continuous variable	TUGT	Continuous variable
Marital status	0 = married, 1 = widowed/divorced	BBS score	Continuous variable
Living status	0 = living with family, 1 = living alone	SPPB score	Continuous variable
History of falls	0 = no, 1 = yes	MMSE score	Continuous variable
Number of comorbidities	Continuous variable	GDS score	Continuous variable
Benzodiazepine use	0 = no use, 1 = use	GAD-7 score	Continuous variable
25-hydroxyvitamin D	Continuous variable	FES-I score	Continuous variable

TUGT: Timed Up and Go Test; BBS: Berg Balance Scale; SPPB: Short Physical Performance Battery; MMSE: Mini-Mental State Examination; GDS: Geriatric Depression Scale; GAD-7: Generalized Anxiety Disorder-7; FES-I: Falls Efficacy Scale-International.

Table 7 Multivariate logistic regression analysis of fall risk factors in older patients with osteoporosis.

Variable	β	SE	Wald χ2	OR	95%CI	P value
History of falls	1.464	0.398	13.527	4.326	1.987-9.416	0.001
GDS score	0.294	0.089	10.912	1.342	1.127-1.598	0.002
GAD-7 score	0.252	0.088	8.214	1.287	1.082-1.531	0.005
TUGT	0.197	0.068	8.394	1.218	1.065-1.393	0.004
BBS score	-0.115	0.038	9.136	0.891	0.826-0.961	0.003
25-hydroxyvitamin D	-0.092	0.033	7.776	0.912	0.855-0.973	0.006
Constant	-2.867	1.724	2.767	0.057		0.096

OR: Odds ratio; CI: Confidence interval; GDS: Geriatric Depression Scale; GAD-7: Generalized Anxiety Disorder-7; TUGT: Timed Up and Go Test; BBS: Berg Balance Scale.

Construction and validation of fall risk prediction model

Using the Bootstrap self-validation method to calculate the calibration curve and AUC after optimistic correction yielded a value of 0.862 with an optimism index of 0.014, suggesting that the model exhibited a low degree of overfitting and possessed good generalization ability. Based on the six independent influencing factors identified through multivariate logistic regression analysis, the predictive model equation was constructed as Logit (P) = ln [P/(1-P)] = -2.867 + 1.464 × X1 + 0.294 × X2 + 0.252 × X3 + 0.197 × X4 - 0.115 × X5 - 0.092 × X6; the fall risk probability prediction formula was P = 1/[1 + e^{-Logit (P)}]. Here, X1: Previous fall history (0 = no fall history, 1 = fall history within the past year), X2: GDS score (continuous variable, 0-15 points), X3: GAD-7 score (continuous variable, 0-21 points), X4: TUGT time (continuous variable, second), X5: BBS score (continuous variable, 0-56 points), and X6: 25-hydroxyvitamin D (continuous variable, ng/mL). Model discrimination evaluation: ROC curve analysis showed that the AUC of the predictive model was 0.876 (95%CI: 0.817-0.935). The sensitivity at the optimal cutoff value determined by the maximum Jodden index was 0.845, with a specificity of 0.819, positive predictive value of 0.788, and negative predictive value of 0.867, indicating the good discriminative capacity of the model. Model calibration evaluation: Hosmer-Lemeshow goodness-of-fit test yielded χ² = 7.864 and P = 0.447. Furthermore, the sample size for this study was 130 cases; six predictor variables were included in the final model; and there were 58 cases of the event (falls). The events-to-variable ratio was approximately 9.67:1, which is consistent with the rule of thumb, indicating that the sample size met the stability requirements for model construction.

DISCUSSION

Recent epidemiological data[9,10] indicate that approximately 30% of community-dwelling older adults experience at least one fall per year, and that this incidence rises to 50% among those aged 80 and over. Previous studies have shown that falls result from the interaction of multiple risk factors, with reduced muscle strength, balance impairment, abnormal gait, household safety hazards, inadequate lighting, and the use of sedatives, hypnotics, or antihypertensive drugs being the primary factors[11,12]. Most previous studies have focused on physical function and environmental factors, while relatively little attention has been paid to psychological factors.

Older adults with osteoporosis are at a high risk of falling, and studies have indicated that those with depression and anxiety are more likely to develop osteoporosis. Furthermore, previous studies have shown that the chronic pain, limited mobility and postural changes caused by osteoporosis may contribute to depression and anxiety in older adults[13]. Similarly, mental disorders can negatively influence the progression of osteoporosis by affecting treatment adherence, lifestyle, and bone metabolism, thereby forming a complex, bidirectional relationship[14,15]. In the present study, falls were classified as binary outcomes (fall/non-fall) based on their occurrence within the previous year[16]. Although this approach is commonly adopted in epidemiological studies, it may overlook important information regarding fall frequency, fall recurrence, and injury severity. Previous studies have suggested that recurrent falls and severe fall-related injuries may have different risk profiles compared with single falls. Future prospective studies should collect detailed information on fall frequency and severity to further refine the risk stratification. Our study revealed that the 1-year incidence of falls among older patients with osteoporosis was 44.62%, which was higher than that in the general community-dwelling older adults. This finding suggests that osteoporosis increases the risk of falls. This finding is consistent with the results of Demeco et al[17] and Moon et al[18], who observed that the fall group had significantly poorer balance, muscle strength, and gait parameters than the non-fall group. Therefore, older patients with osteoporosis require close clinical attention, and fall prevention should be regarded as a core component of clinical management. Combined univariate and multivariate logistic regression analyses revealed that a history of falls, GDS score, GAD-7 score, and TUGT time were independent risk factors for falls (P < 0.05), whereas BBS score and 25-hydroxyvitamin D level were independent protective factors (P < 0.05). An in-depth analysis in the context of clinical practice suggests the following potential mechanisms: The GDS and GAD-7 scales are used to quantify the levels of depression and anxiety. Severe depression and anxiety may increase the risk of falls through various mechanisms. On the one hand, depression is often accompanied by a loss of pleasure and fatigue, leading to reduced physical activity, which in turn results in decreased muscle strength and impaired balance[19,20]; On the other hand, anxiety can cause autonomic nervous system dysfunction, leading to palpitations, dizziness and unsteady gait, thereby increasing the likelihood of a fall[21,22]. Furthermore, depression and anxiety may impair attention and judgment, delay reactions to environmental hazards (such as slippery surfaces or obstacles), and further increase the risk of falls. Previous studies have shown that depression increases the risk of falls through multiple pathways, while falls and fractures may, in turn, exacerbate depressive symptoms[23]; falls also increase the risk of anxiety and depression in older adults, creating a vicious cycle[24]. The ability of physical function measures to predict fall risk is well established. The present study further supports this view by demonstrating that a prolonged time on the TUGT and a lower BBS score serve as independent risk and protective factors, respectively. The TUGT is widely used for evaluating mobility, balance, and fall risk. A prolonged TUGT time indicates diminished muscle strength in the lower extremities, impaired balance control, and abnormal gait. These factors contribute directly to falls[25,26]. As patients take longer to perform activities such as standing up, walking, turning around, and sitting down, it becomes more difficult for them to maintain stability, which naturally increases their risk of falling. The BBS is a comprehensive scale for evaluating balance function that reflects the ability to maintain balance in various postures and during different activities. A higher BBS score indicates better balance and a stronger capacity to handle sudden events such as uneven ground or minor collisions during daily activities, thereby reducing fall risk[27,28]. Vitamin D deficiency is highly prevalent in older adults. It affects not only calcium-phosphorus metabolism and bone health, but also muscle function, balance, and neuromuscular coordination. A study by Saedon and Lok[29] confirmed that older women with higher serum 25-hydroxyvitamin D levels exhibit better standing balance, gait performance and chair-stand ability, and have a reduced risk of falling. This may be because vitamin D promotes intestinal calcium absorption and maintains the calcium concentration in muscle cells, which supports normal muscle contraction and relaxation. This reduces the risk of sarcopenia and consequently improves muscle strength and physical balance. Although lower serum 25-hydroxyvitamin D levels were independently associated with an increased risk of falls, several factors influencing the vitamin D status were not assessed, including seasonal variation, sunlight exposure, outdoor activity, and adherence to vitamin D supplementation. These unmeasured factors may have introduced residual confounding and should be incorporated into future prospective studies.

This study constructed a predictive model based on six independent risk factors, integrating multidimensional information, including a history of falls, psychological status (GDS, GAD-7), physical function (TUGT, BBS), and bone metabolism markers (25-hydroxyvitamin D). This reflects the multifactorial nature of the risk of falls in older patients with osteoporosis and is consistent with the core principles of comprehensive geriatric assessment[30]. The model demonstrated good discriminatory power (AUC = 0.876) and calibration (Hosmer-Lemeshow test, P = 0.447), indicating that it can effectively distinguish between high- and low-risk individuals and that there is good consistency between the predicted probability and the actual observed frequency. Furthermore, all six indicators in the model are readily available in clinical practice and offer satisfactory clinical applicability. Healthcare professionals can rapidly calculate an individual’s risk probability based on the patient’s specific circumstances, thereby identifying high-risk patients and formulating appropriate intervention strategies targeting modifiable factors. These strategies include psychological support, exercise training, nutritional guidance, medication management, and environmental adjustments, enabling the precise assessment of older patients with osteoporosis and effective control of fall risk. This study had several limitations. First, this was a single-center retrospective study with a relatively small sample size (130 participants), which may have increased the risk of selection bias and restricted the generalizability of the findings. Although the events-per-variable ratio approached the commonly recommended threshold, the possibility of model overfitting could not be completely excluded because of the number of candidate predictors included in the regression analysis. Second, falls occurring within the previous year were identified through self-reporting and family confirmation, introducing potential recall bias, particularly among older adults with mild cognitive impairment. Third, fall outcomes were classified only as “fall” or “non-fall”, without detailed information regarding fall frequency, recurrent falls, injury severity, or fall-related hospitalization. Consequently, the heterogeneity of the fall events may not have been fully captured. Fourth, several clinically relevant variables including benzodiazepine use, comorbidity burden, and cognitive function were not included in the final model. Although these factors may share variance with the stronger predictors, their clinical importance should not be overlooked. Fifth, the potential collinearity between depressive symptoms (GDS) and anxiety symptoms (GAD-7) was not formally assessed through correlation analysis or VIFs. Sixth, data on factors affecting serum vitamin D levels, such as seasonal variation, sunlight exposure, dietary intake, outdoor activity, and adherence to supplementation were unavailable and may have introduced residual confounding factors. Finally, the model was internally evaluated only by ROC and calibration analyses and lacked bootstrap and external validation cohorts. Therefore, its predictive performance should be interpreted cautiously before its broader clinical application.

CONCLUSION

In conclusion, comorbid depression and anxiety are independent risk factors for falls in older patients with osteoporosis. The prediction model integrating psychological assessment and physical function measurement exhibited favorable predictive performance, which can be used for the early identification of high-risk patients and to guide the formulation of comprehensive intervention strategies. This study also has inherent limitations, such as recall bias due to its cross-sectional and retrospective design as well as the relatively small sample size. In the future, larger-sample, multicenter prospective cohort studies are warranted to further investigate the relationship between various influencing factors and falls as well as the long-term application efficacy of the prediction model.