Performance of retrospectively and prospectively-reported falls data in predicting future falls: results from the Falls in Glaucoma Study

Abstract

Background

Our objective is to investigate the association between prospectively-reported and retrospectively-reported falls data and the likelihood of subsequent falls in a cohort of patients with and without visual field loss.

Methods

Retrospective falls data (self-reported) was collected at the baseline visit where participants reported if they experienced a fall in the preceding year and their level of fear of falling. Prospective falls data (i.e. falls, near falls, and injurious falls) was collected using fall calendars with participants recording falls daily. Participants reported falls and near-falls monthly over a 2-year period. Logistic regressions assessed whether baseline retrospective falls data was predictive of falls in year 1 and whether prospective falls data in year 1 was associated with falls in year 2. Models were controlled for age, gender, race, degree of visual field (VF) loss, comorbidities, and polypharmacy.

Results

Fallers in the first year of study, as judged by prospective falls data, were 2.4 times (Odds Ratio (OR)=2.41, p=0.003) more likely to report a fall in the second year compared to non-fallers in the first year. Those who prospectively-reported more than one fall (OR=5.74, p-value<0.001), a single injurious fall (months 0 to 12; OR=2.30, p-value=.03), and multiple injurious falls (months 0 to 12; OR= 4.28, p-value=0.025) in year one were significantly more likely to report a fall in the following year. Self-report of a fall in the year preceding the study and fear of falling at baseline did not predict reporting falls in the first study year. Prospectively-reported near-falls in the first year were associated with higher odds of falls in both the first (concurrent) study year (OR=3.49, p<0.001) and the subsequent second year (OR=3.14, p<0.001). Predictive capacity of retrospective and prospective data to identify those at risk of falls in subsequent years did not vary by visual field sensitivity.

Conclusions

Prospectively-reported falls and near-falls are predictive of falls in subsequent years, while self-reported falls based on memory do not predict falls in subsequent years. These findings highlight the importance of prospective collection of falls data to accu­rately identify those at risk of falling who might benefit from fall prevention strategies.

Trial registration

Clinical trial number: not applicable

Supplementary Information

The online version contains supplementary material available at 10.1186/s12877-025-06441-3.

Background

Falls remain a leading cause of mortality and injury for adults over the age of 65, and recurrent fallers suffer lasting physical, financial, and psychosocial consequences, affecting functioning, mobility, perceived autonomy, and mental health [1–4]. Thus, a major focus for falls research has been to identify older adults who have fallen, under the assumption that they are at risk for future falls, which threaten their well-being.

One of the most significant risk factors for falls is a prior history of falls, and risk assessments like the STEADI algorithm developed by the CDC rely on a self-reported history of falls to assess fall risk in elderly adults [5]. For clinicians, identification of fallers also depends on self-reported data, asking patients about their fall history, among other risk factors, to guide decision-making and the initiation of further evaluation [6]. As prior work evaluating fall risk has largely focused on the impact of visual acuity, we focused on a population with a range of visual field damage due to glaucoma through the Falls In Glaucoma Study in order to evaluate the effect of visual field loss on fall-related metrics, particularly given that those with glaucoma are three times more likely to fall [7]. While we have reported extensively on risk factors, circumstances, and characteristics that may help identify future fallers [8–12], less is known about how retrospective data (i.e. a self-reported history of falls) compares to prospective data collection (i.e. using monthly fall calendars or diaries to report falls) in assessing fall risk among at-risk elderly adults.

Prior comparisons of retrospective data and prospective fall diaries in characterizing the prevalence of falls have shown significant limitations. Self-reported data may lead to underreporting due to recall bias, memory, and psychosocial factors, underestimating fall rates and resulting in missed opportunities to implement clinical interventions [13, 14]. These studies are further limited by evaluating prospective and retrospective falls data over the same study period, i.e. by having individuals report on falls at monthly intervals and after the study’s conclusion, then asking participants to recall the total number of falls they experienced over the past year [13, 15]. Comparing the ability of retrospective and prospective falls data over the same period may improve participants’ ability to recall their falls and fails to represent the typical real-world conditions of patients reporting a fall to their provider. Even with the possible improvement in recall in these conditions, imperfect agreement between retrospective and prospective data was still evident, questioning the ability of retrospective data to serve as a meaningful substitute for prospective data. Evaluating retrospective and prospective data over distinct time periods would allow for more clinically relevant comparisons of their respective performance in predicting future fallers.

Thus, the purpose of this study is to directly compare the efficacy of retrospective and prospective data (obtained via falls diaries) over a longitudinal period to identify elderly adults at risk of future falls. This cohort is from the larger Falls in Glaucoma Study, and included participants with vision loss due to glaucoma as well as normally-sighted controls. By including a range of vision impairment, with many patients having normal vision or mild levels of glaucoma field loss (typical for the normal spectrum of disease), we believe our findings will have greater generalizability to older adults. While previous publications from this study focused on the association of visual impairment with falls, here we utilized retrospective falls data collected at baseline (i.e. self-reported history of a fall in the previous 12 months) and prospectively-reported falls/near-falls data from a multi-year study to predict falls in a distinct future period. We hypothesize that both retrospective and prospective falls data will be associated with higher likelihood of subsequent falls with prospective falls data showing a higher predictive capacity than retrospective methods.

Methods

Study design and population

Our study analyzed a prospective cohort of participants suspected to be at a higher risk of future falls due to a clinical diagnosis of glaucoma. Glaucoma suspects without visual damage were also enrolled as a control group, and both groups were followed with annual visits and prospective falls data collection over a three-year period as part of the Falls In Glaucoma Study (FIGS) [10, 11].

FIGS participants were recruited through the Johns Hopkins Glaucoma Center of Excellence between September 2013 and March 2015 and had a diagnosis of glaucoma or suspected glaucoma based on chart review. Eligibility requirements included age ≥ 60 years by conclusion of the study period, residence within a 60-mile radius of Wilmer Eye Institute, and ability to participate in visual field (VF) testing. Participants were excluded if they had visual acuity below 20/40 in either eye due to comorbid eye disease unrelated to glaucoma, a clinical diagnosis of neovascular or uveitic glaucoma (which would imply other systemic disease that could influence fall risk), recent surgical history (ocular or non-ocular) within the preceding two months, hospitalization within the preceding month, or being confined to a wheelchair or bed.

The study was conducted in accordance with the tenets of the Declaration of Helsinki and approved by the Johns Hopkins University School of Medicine Institutional Review Board. All participants signed written consent.

Identification of falls

Over the course of the study, participants were instructed to record falls and near-falls daily utilizing paper calendars (diaries), which were subsequently collected via mail or e-mail each month (Fig. 1). Falls were defined as “any fall, slip, or trip in which you lose your balance and land on the floor or ground or at a lower level.” [16] Near-falls were described as a trip, slip, or loss of balance where an individual begins to fall but regains balance to prevent a fall to the ground or surface below [17]. Additionally, participants received instructional videos with examples of falls and non-falls to provide further clarification during baseline evaluation [18].

Fig. 1.

Timeline of retrospective and prospective falls data collection

Participants reporting a fall within the previous month received a follow-up call involving a 30-item questionnaire to assess circumstances surrounding the fall, location of the fall, any sustained injuries, as well as required treatment. Individuals were classified as fallers, near-fallers, and non-fallers during each year of study. Individuals who failed to submit their monthly calendars received weekly follow-up calls to facilitate data collection. Missing falls data was collected up to three months following the expected collection date.

Self-reported history of falls & fear of falling

At the baseline study evaluation, just before prospective collection of falls data began, participants retrospectively reported any history of a fall within the previous 12 months. Fear of falling was also assessed at this baseline evaluation and at the one year follow-up visit using the University of Illinois at Chicago Fear of Falling Measure, a validated questionnaire that evaluates participants’ concerns about falling across 18 everyday tasks [19]. To avoid the possibility of ceiling effects, participants were also asked ‘‘If you were to walk on uneven terrain, how worried would you be of falling?’’ and ‘‘If you were to walk outside at night when icy, how worried would you be of falling?’’. Participants indicated their level of concern for falling while performing each task as not worried, a little worried, moderately worried, or very worried. Scores were Rasch-analyzed into person-measure scores along a logs-odds (logit) scale with higher scores indicating greater fear of falling.

Covariates

Participants completed questionnaires to evaluate social and demographic information as well as polypharmacy and comorbidities with associations with fall risk. Demographic information included race, ethnicity, age, education, employment status, and living arrangement (with a friend, alone). Polypharmacy was defined as use of five or more non-ophthalmic medications daily [20]. The number of comorbid conditions was assessed using a standardized list of 15 possible diagnoses [21]. Visual fields were used to assess the severity of glaucoma at baseline utilizing the SITA 24 − 2 test on the Humphrey Field Analyzer-II (Carl Zeiss Meditec, Inc., Dublin, CA, USA). Fields deemed unreliable based on inconsistency with prior examinations or testing artifacts were excluded [8]. Estimation of binocular integrated visual field (IVF) sensitivity was computed through integration of pointwise sensitivity from each eye as previously described to characterize levels of visual field damage [9].

Statistical analysis

For each study year (baseline to 12 months follow-up, 12–24 months follow-up), participants were classified as fallers, near-fallers, or non-fallers if they experienced a fall, near-fall, or neither over this period. Comparisons between groups defined by fall status were assessed with chi-squared analysis (categorical variables) and t-tests (continuous variables). Kaplan-Meier survival analyses were performed to evaluate the probability of falls over specified study periods, with groupwise comparisons in falls rates evaluated through Logrank testing. P-values less than or equal to 0.05 were considered statistically significant.

Logistic regression models were computed to assess whether retrospectively-reported or prospectively-reported events in a prior period of time (falls, near falls, multiple falls, injurious falls), or cross-sectional assessments (i.e. fear of falling), were associated with events (falls) in a subsequent period of time. Fear of falling was assessed as a continuous variable with higher scores indicating greater fear of falling. Visual field loss was assessed as a continuous variable through IVF sensitivity and controlled for in all models. As more severe VF damage is thought to affect fall risk, models were evaluated to assess for interaction between faller/near-faller status and level of field damage to establish if fall risk differed across glaucoma severity. Multivariable models were also adjusted for relevant covariates of visual field loss, gender, age, race, comorbidities, and polypharmacy. Participants in this analysis had at least six months of falls data (median = 24 months, IQR = 24 to 24 months). At the start of the study, for participants who had an initial study month with less than 15 days remaining for falls reporting, that month was excluded from our analysis. Any months with missing data that was not collected within 3 months were dropped from analysis.

Results

244 participants with glaucoma or suspected glaucoma were included in analysis (Table 1). Among all participants, mean age was 71 years (± 7.6); approximately half of participants were female (48.4%) and 28.3% identified as Black. 33% of individuals exhibited polypharmacy and 87.8% had at least 1 comorbidity. Median IVF sensitivity among participants was 28 dB (interquartile range [IQR]: 26-29.7; normal value = 31 dB).

Table 1.

Baseline characteristics by faller status based on the first study year

Variable/ Characteristic	Fallers (n=108)	Near-Fallers Only (n=42)	Non-Fallers* (n=94)	P Value
Age, years, mean (SD)	71.8 (7.3)	68.1 (7.0)	70.4 (7.9)	0.026
Male gender, n (%)	49 (45.4%)	20 (47.6%)	57 (60.6%)	0.081
Black race, n (%)	26 (24.1%)	10 (23.8%)	33 (35.1%)	0.172
Lives alone, n (%)	25 (23.2%)	11 (26.2%)	13 (13.8%)	0.142
Employed, n (%)	34 (31.5%)	16 (38.1%)	36 (38.3%)	0.548
Highest level of education completed, n (%)				0.178
Less than high school	2 (1.85%)	0 (0%)	5 (5.4%)
High school	14 (13.0%)	4 (9.5%)	11 (11.8%)
Some college or 2-year college	14 (13.0%)	10 (23.8%)	9 (9.7%)
4-year college	22 (20.4%)	13 (31.0%)	24 (25.8%)
Master’s or doctorate program	56 (51.9%)	15 (35.7%)	44 (47.3%)
At least high school, n (%)	92 (85.2%)	38 (90.5%)	77 (82.8%)	0.509
At least 1 comorbidity, n (%)	95 (88.0%)	38 (90.5%)	81 (86.2%)	0.774
Polypharmacy, n (%)	40 (37.0%)	12 (28.6%)	29 (30.9%)	0.507
BMI, mean (SD)	27.4 (4.5)	27.5 (5.3)	26.7 (5.7)	0.556
Grip strength maximum (kg) for dominant hand, mean (SD)	29.9 (9.2)	33.9 (10.5)	32.6 (11.2)	0.048
Bilateral max leg strength (kg)	17.4 (5.6)	17.8 (4.8)	17.9 (6.8)	0.841
IVF sensitivity (dB), median (IQR)	27.9 (25.8, 29.4)	28.8 (27, 30.2)	27.7 (25.9, 29.7)	0.137
Best eye visual acuity#, median (IQR)	0.06 (-0.02, 0.14)	0.02 (-0.02, 0.14)	0.08 (0, 0.22)	0.187
OU LogCS score, median (IQR)	1.72 (1.64, 1.76)	1.76 (1.68, 1.8)	1.72 (1.6, 1.76)	0.139
Mean deviation of the better eye, median (IQR)	-2.97 (-5.51, -0.93)	-2.36 (-4.41, -0.33)	-2.55 (-6.38, -0.63)	0.178
Mean deviation of the worse eye, median (IQR)	-5.37 (-11.25, -2.78)	-5.31 (-12.47, -2.43)	-6.74 (-14.2, -3.08)	0.70

SD standard deviation, IVF integrated vision field (in decibels), IQR interquartile range

*Non-fallers represent participants without a prospectively-recorded near-fall or fall

^#Normal visual acuity in LogMAR units is 0 (corresponding to 20/20 vision). Values for all groups were less than 0.1, which would correspond to 20/25 vision, below the threshold for visual impairment

^##Mean deviation values range from +2 to -30 dB, with lower values worse. Values between -6 and -12 dB are considered moderate disease, while values below -12 dB are considered severe disease

In Kaplan-Meier analyses, prospective report of a fall in year one of the study (baseline to month 12) was associated with a higher probability of a fall in the second study year (month 12 to 24; log-rank p-value < 0.001) (Fig. 2). Conversely, prospective report of a fall in year one of the study (baseline to month 12) was associated with a higher probability of a fall in the second study year (month 12 to 24; log-rank p-value < 0.001) (Fig. 3).

Fig. 2.

Cumulative probability of not falling in year 2 by prospectively-reported fall status in year 1

Fig. 3.

Cumulative probability of not falling in year 1 by self-reported fall status prior to baseline

In logistic regression analyses, a higher likelihood of falls in the second study year (months 12 to 24) was associated with prospective report of one or more falls in year one (months 0 to 12, OR = 2.41, p-value = 0.003), two or more falls in year one (OR = 5.74, p-value < 0.001), and report of an injurious falls in year one (OR = 2.68, p-value = 0.004) (Table 2). A higher likelihood of falls in the second study year (months 12 to 24) was also associated with reporting of a single injurious fall during year one (months 0 to 12; OR = 2.30, p-value = 0.03) and multiple injurious falls during year one (months 0 to 12; OR = 4.28, p-value = 0.025). Near-falls in study year one (months 0 to 12; OR = 3.14, p-value < 0.001) and year two (months 12 to 24; OR = 3.20, p-value < 0.001) both showed a significant association with report of a fall in study year two (months 12 to 24). A higher likelihood of falls in the first study year (months 0 to 12) was associated with near-falls over the same period (months 0 to 12; OR = 3.49, p-value < 0.001) (Table 2).

Table 2.

Associations between prior falls/near falls and fear of falling with future falls

Predictor Variable	Prospectively defined Fall in year 1 (OR, 95% CI, p-value)	Prospectively defined Fall in year 2 (OR, 95% CI, p-value)
Self-reported fall in year before study (Yes vs. No)	1.17, 0.69-2.00, p = 0.56
Prospectively reported near fall in year 1 (Yes vs. No)	3.49, 2.01–6.05, p < 0.001	3.14, 1.73–5.73, p < 0.001
Fear of falling at baseline (continuous)	1.03, 0.92–1.17, p = 0.54
Prospectively reported fall in year 1 of study (Yes vs. No)		2.41, 1.34–4.34, p = 0.003
Prospective report of multiple falls in year 1 of study
One fall		1.41, 0.71–2.80, p = 0.33
Two or more falls		5.74, 2.51–13.11, p < 0.001
Prospective report of injurious fall in year 1 (Yes vs. No)		2.68, 1.36–5.27, p = 0.004
Prospective report of multiple injurious falls in year 1 of study
One injurious fall		2.30, 1.08–4.89, p = 0.03
Two or more injurious falls		4.28, 1.20-15.34, p = 0.025
Total number of injurious falls in year 1 (risk per each additional fall)		2.09, 1.27–3.43, p = 0.004
Prospectively reported near fall in year 2 (Yes vs. No)		3.20, 1.72–5.93, p < 0.001
Fear of falling at year 1 (continuous)		1.26, 1.09–1.47, p = 0.002

OR odds ratio, CI confidence interval

In contrast, no association was found between falls in in the first study year (months 0 to 12) and a retrospective report of falls in the year prior to the start of the study (months − 12 to 0; Odds Ratio (OR) = 1.17, p-value = 0.56). Likewise, fear of falling at baseline showed no association with risks of falls during the first year of study (OR = 1.03, p = 0.54) while fear of falling at month 12 was associated with higher risk of falls in year two (OR = 1.26, p = 0.002). No association was found between reporting of a single fall during year one (months 0 to 12) and falls during the second study year (months 12 to 24; OR = 1.41, p-value = 0.33). Predictive capacity of retrospective and prospective data to identify those participants likely to fall in subsequent years did not vary by visual field damage (all p values > 0.12, OR range from 0.11 to 1.80). The distribution of fallers/near-fallers/non-fallers in year 1 or year 2 of study by faller status in the preceding year is presented in Additional file 1.

Discussion

Results and initial conclusions

In this longitudinal cohort study evaluating falls over several years, retrospectively-reported falls (i.e. a positive response to a question asking about a fall within the previous 12 months) could not predict future falls, while falls reported as part of a prospective period of follow-up were highly predictive of future falls. These results cast doubt on the importance of retrospectively-collected falls data, particularly with regards to their ability to predict future fall events. Self-reported fear of falling was similarly inconsistent in predicting with future falls, limiting the relevance of such data for providers assessing fall risk in clinic. Stronger predictors of future falls included prospective report of multiple falls or injurious falls. These data suggest that simple methods for screening individuals at high risk of future falls (i.e. retrospective report of a fall, or high fear of falling), may not be effective at identifying future fallers. Instead, prospective falls data may be most beneficial in identifying those at risk of falls who require targeted fall prevention strategies. Our findings also support the notion that prospective reporting of falls should be taken as the gold standard for research purposes.

Limitations of retrospective data

Our finding that retrospective reporting of a past fall has limited association with future falls extends upon past work which has raised questions about the significance and accuracy of retrospective falls data. Specifically, prior work has identified significant differences in prospectively and retrospectively recorded fall rates [22]. Hannan et al. found that participants often failed to accurately recall falls over a three-month period despite concurrently completing monthly falls diaries as part of the study, demonstrating that self-reported data underestimates fall rates even as participants closely track their falls [15]. Similar results were noted for a one-year study in which subjects were asked to monitor their falls prospectively for 12 months, and then asked about falls over the prior year (at the end of the 12 months of prospective falls collection) [13]. In this study, retrospective assessment frequently led to under-reporting of falls, though 6% of subjects not prospectively reporting a fall also retrospectively reported a fall at the end of the year [13]. Similarly, in that work, researchers noted modest agreement between the falls assessment methods for both falls and recurrent falls (Kappa values just under 0.6). Our study design differed from these past studies. Since retrospective assessment of falls over the same period of prospective falls reporting may be unexpectedly high due to greater attention regarding falls events, we examined the association of both a retrospective report of falls and prospective falls data reporting with prospectively-collected falls data in the following year. We found that only prospectively-collected falls data, but not retrospectively-collected falls data, could predict falls in an independent future period, suggesting that prospective data collection better captures the “falls phenotype”.

Several reasons could account for our findings and those in prior studies. Retrospective assessment may be limited by memory and recall [23, 24]. Additionally, social desirability bias may lead patients to downplay falls due to the negative stigma attached to aging, disability, and perceived frailty. For example, when self-reporting other measures related to falls, older adults have been shown to consistently overestimate their physical activity levels and their ability to complete mobility tasks compared to objective assessment [25, 26]. With retrospectively-reported falls data, individuals may employ past notions of what constituted being a fall, as opposed to definitions given to them before a prospective period of follow-up, leading to inconsistencies across individuals, and across time for the same person [23]. Our findings that retrospectively-reported falls data could not predict persons experiencing a fall the following year is perhaps even more concerning than prior studies showing differences in/under-reporting of falls using the two methods. While under-reporting implies miss events or inaccurate rates, the inability to predict future events highlights that retrospective data may not accurately capture an individual’s true fall risk, preventing the implementation of timely interventions to prevent future falls.

Better predictive capacity of prospective data

In contrast, prospective of falls, particularly recurrent, injurious, and near falls were strongly associated with an increased risk of future falls (as judged by prospectively-collected falls data from a subsequent 12 months of follow-up). Among these falls outcomes, recurrent fallers (particularly those with recurrent injurious falls) exhibited the highest risk of a future fall, consistent with prior studies highlighted how recurrent fallers may struggle to return to regain function after their falls and/or have strong pre-existing risk factors such as gait abnormalities, poor balance, or decreased strength predisposing them to falls [27–29]. Near-fallers have been shown to exhibit similarities to fallers with regards to slower walking speeds and performance on balance and mobility assessments, which may explain their ability to predict future falls when evaluated prospectively [30]. As near-falls can be assessed prior to any injuries or harm that may stem from falls, they may represent a useful tool for preventing falls early - before the psychosocial and functional consequences of falls take room.

Translational relevance

Despite the challenges involved in collecting prospective data, its significant relative strength (as opposed to retrospectively-collected data) allows providers to properly target individuals who would benefit from evidence-based interventions. Specifically, recommendations for fall prevention by the USTPF include exercise and physical therapy, and prior clinical trials have shown that individuals who begin and continue exercising after a fall have decreased risk of future falls [31, 32]. Maintaining and improving a patient’s baseline after a fall may allow a patient to fully recover and prevent negative consequences to their quality of life and functioning longitudinally. Recurrent fallers have also been shown to benefit from exercise and mobility training and thus, providers should encourage the benefits of exercise interventions to all patients at risk of falls [33].

Prospective data collection through diaries may facilitate improved reporting of falls to providers, promoting awareness among patients and improving their understanding of the circumstances and factors that led to their falls. Addressing modifiable risk factors through home modifications, improved lighting conditions, medication review, and physical therapy may significantly improve the likelihood of falls among those identified as high risk [34–37]. While impugned here, retrospective assessment of falls requires far fewer resources and may still have relevance in some settings. Wapp et al. found that asking patients to identify the number of falls within the past year allowed for prediction of future falls [38]with those reporting more than four falls at risk for future falls.

Limitations

Limitations of our study include recruitment of patients who receive care at a single tertiary health center, which may not be generalizable to the overall population of elderly adults. Patients reported a history of prior falls within the past year, however, we did not evaluate if patients accessed interventions after past falls, such as strength/mobility/balance training, redesigning of their home, changes in medications, or other adaptations that may alter their subsequent fall risk. Patients were asked to complete a daily fall calendar to record falls, requiring either some literacy and cognition of patients themselves or high levels of attendance and awareness of caretakers to record falls on behalf of patients, which may not be representative of the full elderly population. Further study should examine methods to include such vulnerable patients with visual impairment to record falls and near-falls to assess risk and prevention. Finally, the outcome predicted (future falls) was gathered prospectively, such that it might be better predicted by previous prospective data due to methodological reasons alone. If more objective methods of falls assessment are validated (i.e. wearable methods for detecting falls), it would be worthwhile repeating the current study design with this objective fall assessment as the predicted outcome. While fear of falling at year 1 was associated with falls during year two, this outcome is likely related to increased awareness of fall risk among participants and limits its applicability to the general population at risk for falls.

Conclusions

Our study highlights a significant advantage of prospective recording of falls and near-falls, as opposed to retrospective collection of falls data, in prediction of future falls. While we could not determine that prospectively-collected data were “more correct”, they did predict future falls, while retrospectively-collected falls did not, suggesting that prospectively-identified falls best capture real events (i.e. falls, or the timing of falls) relevant to subsequent fall risk. On the other hand, caution should be used in relying on a self-reported history of falls to assess a patient’s risk for subsequent falls. Particularly high risk of future falling is present in persons who prospectively report recurrent and injurious falls, and these individuals should be considered for proven fall-prevention interventions. Our results suggest that investigators should emphasize the use of prospective methods to evaluate falls in research studies to best capture events reflective of the individual’s fall risk. Future studies should look to improve the ability of retrospective questionnaires to align with prospective methods given their relative cost effectiveness, widespread use, ease of implementation, and to develop simple objective (i.e. wearable) methods for falls assessment.

Abbreviations

FIGS

Falls in Glaucoma Study

VF

Visual Field

IVF

Integrated Visual Field

IQR

Interquartile Range

OR

Odds Ratio

Authors’ contributions

Study concept and design: PYR, GX, TK, LA, AM; Data Acquisition: GX, TK, LA, ZY, SB, XC, AM, PYR, MD; Analysis and interpretation: MD, GX, LA, SB, XC, AM, PYR; Drafting of manuscript: MD, LA, AM, SB, XC, ZY, PYR; Critical revision: PYR, MD, XC, LA, AM; All authors read and approved the final manuscript.

Funding

This research was supported by the National Institutes of Health (Grant Number: EY022976). The sponsor had no role in any aspect of the analyses or preparation of the manuscript.

Data availability

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

The study was conducted in accordance with the tenets of the Declaration of Helsinki, approved by the Johns Hopkins University School of Medicine Institutional Review Board, and written informed consent was obtained from all participants.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.