Trajectories of Fear of Falling and Associated Factors among Stroke Patients

1. INTRODUCTION

Stroke is a neurological disorder caused by interruption of cerebral blood supply due to ischemic or hemorrhagic events, resulting in acute neurological dysfunction [1]. According to the Global Burden of Disease Study 2021, stroke remains a major global health concern and was the second leading cause of death worldwide in 2021, accounting for approximately 7.3 million deaths and more than 160 million disability-adjusted life years. The global burden of stroke has increased substantially since 1990, with a disproportionate impact on low- and middle-income countries [2]. In these settings, stroke often occurs at a younger age, contributing to productivity loss and increased healthcare costs. One survey reported that 25% of previously employed stroke survivors left work or reduced workload after stroke [3]. With advances in acute stroke care and rehabilitation, the number of stroke survivors has increased in recent years.

Stroke often results in weakness on one side of the body, making stroke survivors feel uncoordinated and unsteady. Thus, falling is a common issue among stroke survivors. Based on research studies, stroke survivors have 14% risk of falling in the first month and over 30% risk of falling within ten years [4]. Falls can result in physical injury, emotional distress, activity restriction, and greater dependence [5], which may further contribute to social isolation and poorer quality of life. Thus, many survivors have “fear of falling (FOF)”, defined as concern about falling during daily activities [6]. According to a survey, about half of stroke survivors have FOF, which may further restrict activity participation and social engagement [7].

Most research on FOF in stroke has used cross-sectional designs, and longitudinal evidence remains limited. Previous studies suggest that FOF is dynamic and may vary across stages of recovery [8, 9]. In some individuals, FOF may fluctuate early and then stabilize or improve during rehabilitation [10, 11]. Given substantial heterogeneity in functional recovery, impairment severity, comorbidity burden, and social support, stroke survivors may follow distinct patterns of change in FOF over time. Some individuals may regain mobility and confidence, whereas others may experience persistent or increasing concern about falling. These considerations support the existence of clinically meaningful subgroups with distinct developmental trajectories of FOF. Traditional growth models estimate an average trajectory and assume population homogeneity [12], which may obscure important subgroup differences. Growth mixture modeling (GMM) can identify latent classes characterized by distinct longitudinal trajectories [13]. Therefore, this study applied GMM to identify latent classes of FOF within 1 to 6 months after stroke onset and to examine factors associated with latent class membership.

2. METHODS

3. RESULTS

3.1. Latent Classes and Corresponding Development Trajectories of FOF

In the study, FOF scores at 1, 2, 3, 4, and 6 months post-stroke onset were used as repeated measures. The model was specified as a GMM, and 1- to 4-class models were estimated for the linear growth model, the latent-basis growth mixture model, and the quadratic function model. For the linear growth model, the LMR test was not statistically significant for the 3-class and 4-class models but was statistically significant for the 2-class model. For the latent-basis growth mixture model, the LMR test was not statistically significant for the 3-class and 4-class models but was statistically significant for the 2-class model. For the quadratic function model, the LMR test was not statistically significant for the 2-class, 3-class, or 4-class models. In both the linear growth model and the latent-basis growth mixture model, the AIC, BIC, and aBIC values for the 2-class model were lower than those for the 1-class model. When comparing the 2-class linear growth model with the 2-class latent-basis growth mixture model, the two models had the same entropy value of 0.996, and both the LMR and BLRT tests were statistically significant. The AIC, BIC, and aBIC values for the 2-class latent-basis growth mixture model were lower than those for the 2-class linear growth model; therefore, the 2-class latent-basis growth mixture model was selected as the optimal model. The model fit results for each model are presented in Table 1.

Table 1.

Model	AIC		BIC	aBIC	Entropy	LMR	BLRT
Linear growth model	-		-	-	-	-	-
1-class	5135.824		5168.956	5137.273	-	-	-
2-class	4941.794		4984.866	4943.679	0.996	0.0031	<0.0001
3-class	4891.802		4944.813	4894.121	0.938	0.4722	<0.0001
4-class	4855.448		4918.399	4858.202	0.963	0.0926	<0.0001
Latent-basis growth mixture model	-		-	-	-	-	-
1-class	5103.126		5146.198	5105.010	-	-	-
2-class	4904.826		4957.837	4907.145	0.996	0.0001	<0.0001
3-class	4857.406		4920.357	4860.160	0.935	0.4862	<0.0001
4-class	4821.438		4894.328	4824.627	0.962	0.0740	<0.0001
Quadratic function model	-		-	-	-	-	-
1-class	5074.337		5120.722	5076.366	-	-	-
2-class	4881.616		4941.253	4884.225	0.996	0.0596	<0.0001
3-class	4834.003		4906.893	4837.192	0.934	0.5740	<0.0001
4-class	4767.710		4853.854	4771.479	0.963	0.3769	<0.0001

Based on the optimal model, two latent classes were identified. In Class 1, the estimated intercept mean was 59.97 (p < 0.001), and the slope mean was -2.61 (p < 0.001), indicating a higher initial FOF level followed by a gradual decline over time. This class was labeled the “higher-FOF declining group” and accounted for 60.0% of participants. In Class 2, the estimated intercept mean was 48.95 (p < 0.001), and the slope mean was 1.74 (p < 0.001), indicating a relatively low initial FOF with a gradual increase over time. This class was labeled the “lower-FOF increasing group” and comprised 40.0% of the sample. Previous research has estimated the minimal clinically important difference (MCID) for the FES-I to be approximately 5.5 points using distribution-based methods [18]. In the present study, the baseline difference between the two classes was approximately 11 points, nearly twice the reported MCID. This difference is likely to be clinically meaningful rather than a minor statistical variation. The variances of the intercept and slope factors were constrained to be equal across classes to ensure model stability and parsimony. The developmental trajectories of FOF are presented in Fig. (1), and parameter estimates are reported in Table 2.

Table 2.

Parameter	Estimate	SE	t	p	95%CI Lower	95%CI Upper
Latent Class 1	-	-	-	-	-	-
Mean	-	-	-	-	-	-
Intercept	59.966	0.229	261.399	<0.0001	59.516	60.416
Slope	-2.607	0.265	-9.852	<0.0001	-3.126	-2.088
Variance	-	-	-	-	-	-
Intercept	2.896	0.411	7.055	<0.0001	2.092	3.701
Slope	1.540	0.363	4.242	<0.0001	0.828	2.251
Latent Class 2	-	-	-	-	-	-
Mean	-	-	-	-	-	-
Intercept	48.949	0.243	201.718	<0.0001	48.473	49.424
Slope	1.742	0.232	7.520	<0.0001	1.288	2.196
Variance	-	-	-	-	-	-
Intercept	2.896	0.411	7.055	<0.0001	2.092	3.701
Slope	1.540	0.363	4.242	<0.0001	0.828	2.251

Note: In the model, the variances of the intercept and slope were constrained to be equal across latent classes. Therefore, identical variance estimates across classes are reported.

4. DISCUSSION

The dynamic nature of FOF has been discussed, describing it as a process that may fluctuate across different stages of recovery [20]. The present study supports this view by identifying two distinct developmental trajectories: the “higher-FOF declining group” and the “lower-FOF increasing group.” The “higher-FOF declining group” showed a gradual reduction in FOF over time. One study reported that the proportion of patients with FOF was lower at six months post-stroke (42%) than during the acute phase (51%) [21]. Improvements in physical function, mobility, and balance during rehabilitation may increase confidence in daily activities and may be associated with reduced FOF [22]. Higher FOF in the acute phase may also relate to uncertainty about recovery [23]. Clinically, this pattern may reflect heightened perceived threat early after stroke, followed by attenuation as recovery progresses. Management for this subgroup may emphasize early risk control and support, including safety education, graded balance and gait training, confidence-building practice for transfers and walking, and brief cognitive-behavioral components to address catastrophic beliefs and reduce avoidance.

In contrast, the “lower-FOF increasing group” exhibited a progressive rise in FOF. Prior studies have linked FOF to reduced social interaction and physical activity [15, 24, 25]. Although short-term activity restriction can reduce fall exposure, persistent avoidance may contribute to deconditioning [26], which could be associated with higher fall risk and greater FOF. FOF is associated with psychological stress, anxiety, and depression [27], which may reduce engagement in rehabilitation [28]. Clinically, this trajectory may indicate an emerging subgroup in which FOF increases over time despite a relatively low initial level of FOF and may reflect challenges encountered during later rehabilitation or after discharge. For this subgroup, post-discharge screening should be used to detect early worsening of FOF, and an upward trajectory should prompt more frequent reassessment. When escalation is observed, interventions can focus on psychoeducation, graded exposure in a controlled setting, home-environment risk assessment and modification, and structured caregiver guidance. The aim is to address rising FOF early, before it is accompanied by sustained avoidance, reduced participation, and functional decline [29].

Compared with participants with elementary school education or less, those with a middle school education were less likely to belong to the increasing group. One potential explanation is that lower educational attainment is often linked to poorer health literacy, which can limit the ability to obtain, understand, and use rehabilitation information. This, in turn, may reduce adherence to recommended mobility practice and coping strategies [30]. Among stroke survivors, higher health literacy has been related to greater self-efficacy, suggesting that limited health literacy may undermine confidence in managing post-stroke challenges and weaken self-management behaviors [31]. In fall-prevention research, inadequate health literacy has been noted to hinder understanding of educational materials and participation in interventions, which may contribute to a gradual increase in concerns about falling [32]. Consistently, adequate health literacy has been associated with a lower prevalence of falls in community-dwelling older adults, which may result in lower fall exposure and less escalation of FOF [33]. Because only the middle school category reached statistical significance, it should be interpreted cautiously and replicated in larger samples. One possible explanation for this non-linearity is that, although the odds ratio for junior college and above remained in the protective direction, it did not reach statistical significance. This estimate may have been underpowered, as only a small number of participants in this category were observed in the lower-FOF increasing group, potentially resulting in unstable estimates and reduced statistical power. Therefore, the lack of statistical significance should not be interpreted as evidence of no association. In addition, broad educational categories may not fully reflect differences in health literacy, engagement in rehabilitation, or coping capacity.

Participants without a fall history had lower odds of belonging to the increasing group than those with a fall history. This pattern is consistent with evidence in stroke populations, including a meta-analysis identifying fall history as a significant correlate of higher FOF after stroke [34]. Another meta-analysis also reported a significant association between falls and FOF in both acute and chronic stroke cohorts, suggesting that individuals who have fallen tend to report greater FOF [35]. Falls can be important events that make people judge their fall risk as higher and feel less confident about their balance, which may lead to avoiding activities. Reduced activity and less walking practice can then cause loss of strength and balance, which increases the chance of another fall and strengthens the fear. This cycle has been described in stroke-related models of FOF and fall risk [36]. Longitudinal studies also show that early fall experiences are linked to later FOF, highlighting the need for early screening and timely interventions to prevent worsening over time [37].

Participants with hemorrhagic stroke had higher odds of belonging to the increasing group than those with ischemic stroke. This finding is consistent with reports that hemorrhagic stroke is often accompanied by greater initial neurological impairment and a slower or delayed recovery, which may prolong feelings of instability during mobility and increase concerns about falling over time [38, 39]. Differences across studies in case severity, rehabilitation intensity, and the timing of assessment may partly explain variation in the strength of this association. Similar associations between stroke type and FOF have also been reported [40]. In addition, participants with three or more chronic illnesses had higher odds of belonging to the increasing group than those with two or fewer chronic illnesses. This aligns with evidence linking multimorbidity to poorer physical reserve and a higher risk of falls and FOF [41, 42]. A higher comorbidity burden may increase functional limitations, fatigue, and medication use, which can reduce gait stability and balance confidence and thereby contribute to worsening FOF [41].

Clinically, patients with a fall history, hemorrhagic stroke, or three or more chronic illnesses may warrant proactive monitoring and tailored support. Targeted strategies such as accessible health education, fall-prevention and confidence-building interventions, and coordinated chronic disease management may help mitigate the progression of FOF.

CONCLUSION

The FES-I can be administered at 1, 2, 3, 4, and 6 months after stroke onset to monitor FOF. An upward trend, defined as two consecutive increases, may indicate worsening FOF over time. Routine assessments typically focus on physical function, such as mobility and balance. Monitoring of FES-I adds fall-related psychological and behavioral information. These findings highlight the importance of early recognition of patients who may be more likely to follow an increasing FOF trajectory, particularly those with an elementary school education or below, prior falls, hemorrhagic stroke, and three or more chronic illnesses.

ETHICS APPROVAL

This study was approved by the Ethics Committee of Chengde Central Hospital, Hebei Province, China (Approval No. CDCHLL2024-403). All procedures were conducted in accordance with the Declaration of Helsinki. Before enrollment, physicians assessed each participant’s capacity to provide informed consent, and participants were enrolled based on their own agreement. Each participant was assigned a unique study ID. Names and phone numbers were stored separately from the research data. Baseline and follow-up data were recorded using study IDs. All files were password-protected and accessible only to the research team.

LIMITATIONS

This study has several limitations. Convenience sampling from a single region limits generalizability. Requiring verbal communication excluded participants with aphasia or cognitive impairment, potentially affecting the identified trajectories. The first assessment was conducted one month after stroke onset, so an acute-phase baseline was unavailable, and early rapid changes were missed. FOF was self-reported and collected through repeated face-to-face interviews, which can be influenced by misunderstanding, social desirability, and interviewer effects. Inter-rater reliability and standardization checks were not reported. The two-stage variable selection strategy may introduce instability. Although some confounders were retained a priori, the selection of the remaining variables still depended on univariate significance, which may have missed confounders that were not significant in the univariate analysis but modified the relationship. Key psychological factors such as anxiety, catastrophizing, and activity avoidance were not assessed, limiting mechanistic interpretation. Rehabilitation exposure, including therapy type and intensity, was not measured, so its impact on FOF trajectories could not be evaluated. Finally, GMM involves iterative estimation and is susceptible to local maxima and model instability, which can affect class solutions and the reliability of trajectory classification.

REFERENCES