What is a falls risk factor? Factors associated with falls per time or per step in a glaucoma cohort

Abstract

BACKGROUND/OBJECTIVES:

Falls jeopardize the health of older adults, making it important to properly identify fall risk factors. Here, we determine if visual field (VF) damage or other factors confer a similar risk to falling when falls are ascertained as a rate over time (falls/year) or over activity (falls/step).

DESIGN:

Prospective, observational cohort study.

SETTING:

Clinic-based recruitment with real-world monitoring of falls and physical activity.

PATIENTS:

Two-hundred thirty subjects with glaucoma or suspected glaucoma.

MEASUREMENTS:

Patients recorded falls using daily calendars with injuries identified via follow-up questionnaire. Annual one-week accelerometer trials were used to estimate steps. VF results from both eyes were merged to determine integrated VF sensitivity, with lower sensitivity indicating greater VF damage. Other potential risk factors for falls (age, gender, race, comorbid illness, and polypharmacy) were determined by questionnaires.

RESULTS:

The cumulative probability of falls was 45.2 and 61.6% at 12 and 24 months, respectively, while the cumulative probability of injurious falls was 23.3 and 40.0% respectively. Greater VF damage was associated with higher rates of fall/steps (IRR=1.40/5 dB decrement in sensitivity; p=0.004), but not with more falls/year (IRR=1.25/5 dB decrement in sensitivity; p=0.07). Several additional variables (older age, female gender, more comorbid disease) were also associated with a higher rate of falls/step (p<0.02 for all), but not with falls/year (p>0.10). Conversely, African-Americans had fewer falls/year as compared to Caucasians (p=0.002) but did not differ with regards to falls/step (p=0.07). Similar results were obtained when injurious falls were analyzed.

CONCLUSIONS:

Risk factors associated with more frequent falls when walking (falls/step) are not properly identified when analyzing falls as a rate over time (falls/year). Given the clinical importance of preventing falls while preserving physical activity, falls assessment integrated with activity measurement is recommended when determining if a risk factor is associated with falls.

INTRODUCTION

Falls are the leading cause of injurious death in older adults¹ and account for most injury-related ER visits and hospitalizations in this age group.² Fall survivors are at risk for numerous secondary consequences including depression, social isolation, and activity restriction.³ Thus, a major effort in aging research has been to identify risk factors conferring a higher risk of falling and to study and validate programs to prevent falls. Clinically, physicians caring for older patients must consider how to prevent falls while also encouraging an active lifestyle to guard against the consequences of sedentary behavior.⁴

Achieving both fall prevention and an active lifestyle is complicated by the inherent connection between falls and activity. Some individuals may not experience more falls over time but will experience more falls per unit of activity (i.e. per step). For example, a completely bed-ridden individual will appear to have a low rate of falling on account of their inactivity. While a few studies suggest accounting for activity when assessing falls,^5,6 most studies of falls risk factors have not done so. Problematically, many conditions predisposing individuals to falls may also produce activity restriction. One example is visual field (VF) damage from glaucoma, which has been associated with higher fall rates^7–9 (particularly for inferior VF damage) and physical activity restriction.^10,11 These data suggest that understanding falls in a given condition (i.e. glaucoma) is best accomplished by evaluating disease severity (i.e. VF damage) as a risk factor for both a higher rate of falls over time (falls/year) and a higher rate of falls per unit of activity (falls/step).

Here, we prospectively assess falls in a cohort of patients with varying degrees of VF damage from glaucoma and measure activity levels via annual one-week accelerometer trials. We hypothesize that risk factors for falls, including VF damage and other patient characteristics, will differ when analyzing falls/step or falls/year as the outcomes, reflecting associations of these same risk factors with physical activity levels. Potential discrepancies would call for a re-evaluation of how falls risk factors are defined in research studies, falls prevention efforts, and clinical practice.

METHODS

Two-hundred thirty patients with glaucoma or suspected glaucoma were recruited from the Johns Hopkins Wilmer Eye Institute between 2013 and 2015. Study inclusion and exclusion criteria are described in detail elsewhere.¹² Briefly, patients were study-eligible if they were ≥57 years and had primary glaucoma (or suspected glaucoma). Patients were excluded if they had other ocular conditions resulting in visual acuity worse than 20/40 in either eye, any ocular or non-ocular surgery in the last 2 months, or any hospitalization in the last month. The study was approved by the Johns Hopkins Institutional Review Board and all participants provided written informed consent.

To judge the representativeness of our recruited sample as compared to the study-eligible patients from the recruitment site, we evaluated, using a short questionnaire, a high proportion of patients who would have been eligible for our study (97%, n=258) over a one-week period. Study-eligible sample completed a short questionnaire allowing comparison to the recruited sample. Recruited participants were of similar race, age, gender and visual filed severity as the study-eligible sample. However, recruited participants were more likely to use an assisted device (35 vs. 13%, p<0.001) or report falling in the past 12 months (42 vs. 23%, p<0.001), suggesting preferential recruitment of subjects at higher risk of falling.

Assessment of falls and fall-related injuries

Falls were defined as unintentionally coming to rest on a ground or a lower level at the baseline visit, and further illustrated using an instructional video.¹³ Participants prospectively kept falls calendars over the 3-year study period, with calendars returned at the end of the month via mail or email. Individuals not returning calendars were contacted by phone and/or email until data were obtained or a period greater than three months passed, at which time data were recorded as missing. Falls indicated in submitted calendars initiated a call to determine if any injuries resulted. Injury was defined as sustaining one or more of the following: pain, bruising, swelling, pulled muscle, sprained ligament, join dislocations or fracture.

Evaluation of physical activity

Patients wore a waist-bound omni-directional accelerometer (Actical, Respironics Inc., Murrysville, PA) for one week each study year, and received a reminder call from a coordinator at least twice during the week to maximize adherence. Accelerometer data were used to project steps over the upcoming year, i.e. until the next accelerometer trial was performed.¹⁴ Participants who did not have at least 4 days of the valid data during any study year were excluded from the analysis (n=13).¹⁴

Evaluation of Vision

Baseline VF testing was performed using the 24–2 test pattern obtained on a Humphrey Field Analyzer II (Carl Zeiss Meditec, Inc., Dublin, CA), with right and left eye pointwise sensitivity data combined to calculate integrated VF (IVF) sensitivity.¹² IVF data were summarized as average overall and inferior field sensitivities, with lower sensitivities representing greater VF damage. These sensitivities were taken as the primary predictor of falls based on their importance with regards to staging glaucoma severity and also because prior research has highlighted the particular importance of VF damage, as opposed to other visual measures, as a risk factor for falls.^7–9 Baseline IVF was chosen for the analysis as previous examination of progression rates for all patients under our care found the rate to be 0.2 dB/year. Thus, we might expect only 0.4 dB of loss over a 2-year period, which would represent a negligible loss of roughly 1% of the visual field (normal sensitivity is 31 dB). Visual acuity (VA) was assessed in each eye using a back-lit ETDRS chart at 4 meters with patient’s presenting correction.

Assessment of other potential falls risk factors

Age, gender, race, and number of non-visual comorbidities¹¹ were gathered using questionnaires. Comorbidities were evaluated as a total number, ranging from 0–5. Glaucoma was not included as one of the comorbidities. The small number of participants with more than 5 comorbidities (n=7) were reclassified to have 5 comorbidities. Non-eye drop medication lists were generated by direct observation of pill containers when possible, or otherwise by patient report, and classified as polypharmacy if 5 or more prescription medications were used.¹⁵ Grip strength was measured by having participants squeeze the Jamar Hand Dynamometer (Sammons Preston Rolyan, Bolingbrook, IL) three times as hard as possible with their dominant hand. Leg strength was measured with participants sitting in a chair by placing a MircoFET2 Dynamometer (Hoggan Scientific LLC, West Jordan, UT) just above the knee. The participant was asked to resist the pressure of the device twice on each leg for 5 seconds without touching the floor with their foot. The maximum value recorded in kilograms was used as a measure of grip or leg strength.

Sample size calculation and Statistical analysis

Sample size was set at 250 patients in order to detect 1.6-fold higher odds of sustaining one or more falls in the first year with each 8 dB decrement in the better-eye VF MD with 80% power and a type I error probability of 0.05, assuming a model R² value of 0.2 when regressing VF sensitivity on other independent variables in the model, and a 10% loss to follow-up rate per year.

Univariate and multivariable negative binomial models were used to identify factors associated with the amount of baseline physical activity (i.e. steps per day). Next, two sets of negative binomial models were used to identify risk factors associated with higher fall rates using the first 24 months of received calendar data. The first set evaluated fall incidence over time (i.e. study years) as the outcome, while the second evaluated fall incidence over steps taken during the time corresponding to the analyzed calendar data. The primary predictor variable in each model was overall IVF sensitivity, with separate models evaluating inferior IVF sensitivities as the primary independent variable. All models controlled for age, race, gender, number of comorbid illnesses, and polypharmacy.

RESULTS

Characteristics of study participants

Average participant age was 70.7 years (SD = 7.6 years) and a wide range of IVF sensitivity was noted (mean=27.3 dB, range=9.8 to 33.6 dB, normal value = 31 dB), Table 1. Subjects averaged 4,046 daily steps at baseline.

Table 1.

Study population characteristics

Demographics	Values (n=230)
Age (years), mean (SD)	70.7 (7.6)
African-American race, n (%)	60 (26)
Female gender, n (%)	110 (48)
Employed, n (%)	81 (35)
Lives alone, n (%)	45 (20)
Education, N (%)
Less than high school	5 (2)
High school	27 (12)
Some college	31 (13)
Bachelor’s degree	57 (25)
More than bachelor’s degree	109 (48)
Health
Comorbid illnesses > 1, n (%)	149 (65)
Polypharmacy, n (%)	75 (33)
Body Mass Index (kg/m^2), mean (SD)	27.0 (4.9)
Grip strength (kg), mean (SD)	31.5 (10.2)
Lower body strength (kg), mean (SD)	17.7 (5.9)
Vision
IVF sensitivity (dB), median (IQR)	28.0 (26.13, 29.73)
MD better-eye (dB), median (IQR)	−2.51 (−5.41, −0.64)
MD worse-eye (dB), median (IQR)	−5.73 (−12.78, −2.70)
Better-eye acuity-logMAR, median (IQR) Snellen equivalent of the acuity-logMAR	0.06 (−0.02, 0.14) 20/23 (20/19, 20/27)

SD=standard deviation, n=number, kg=kilogram, m=meter, IVF=integrated visual field, dB=decibel, IQR=interquartile range, MD=mean deviation, logMAR=logarithm of the minimum angle of resolution.

Patient characteristics associated with physical activity levels

In univariate models, fewer daily steps were observed in patients with worse IVF sensitivity (RR=0.85 per 5dB decrement in IVF sensitivity, p=0.002). Additional factors associated with fewer daily steps included older age, African-American race, more comorbid illness, and polypharmacy (Table 2). Females also took fewer steps than males, though the association was not statistically significant (p=0.09). In multivariable models, less physical activity was noted in patients with worse IVF sensitivity, older age, female gender, African-American race and polypharmacy (p≤0.05 for all). In multivariable models, greater comorbid illness was not associated with fewer daily steps (p=0.41).

Table 2.

Association between daily steps and patient characteristics in univariate and multivariable models

Patient characteristics	Daily steps
	Univariate models RR (95% CI)	Multivariable model RR (95% CI)
IVF sensitivity (5dB worse)	0.85 (0.76 – 0.94)	0.90 (0.81 – 1.00)
Age (5 years older)	0.84 (0.79 – 0.89)	0.84 (0.80 – 0.89)
Male (vs. Female)	1.16 (0.98 – 1.38)	1.21 (1.04 – 1.42)
African Am (vs. Cauca)	0.75 (0.62 – 0.92)	0.74 (0.61 – 0.90)
Number of Comorbidities	0.93 (0.88 – 0.98)	0.98 (0.92 – 1.04)
Polypharmacy (≥5 vs. <5 meds)	0.69 (0.58 – 0.83)	0.81 (0.67 – 0.99)

Bold – significant results, RR – rate ratio, CI – confidence interval, IVF – overall integrated visual field, dB – decibel, Am – American, Cauca – Caucasian, meds – medications

Patient characteristics associated with fall rates

Fall calendar return rates were 94.2%, with a total of 331 falls noted over 5,230 calendar months. The cumulative probability of falling was 45.2 and 61.6% at 12 and 24 months, respectively, while the cumulative probability of an injurious fall was 23.3 and 40.0% at 12 and 24 months (Figure 1).

Figure 1.

Cumulative incidence of falls and injurious falls

Worse overall or inferior IVF sensitivities were similarly associated with higher rates of falls/step (RR=1.40 per 5dB decrement in sensitivity, p=0.004, and RR=1.34 per 5dB decrement, p=0.01, respectively), though neither was associated with more falls/year (RR=1.25 per 5dB decrement, p=0.07, and RR=1.22 per 5dB decrement, p=0.09, respectively) (Figure 2). Older age, female gender, and more comorbid illness were also associated with a higher rate of falls/step (p<0.02 for all), though none were associated with more falls/year (p>0.1 for all). Conversely, African-Americans did not differ from Caucasians in their rate of falls/step (RR=0.67, p=0.07), but had fewer falls/year as compared to Caucasians (RR=0.50, p=0.002).

Figure 2.

Association between different rates of falls with degree of visual field damage and other patient characteristics. RR – rate ratio, p – p-value, IVF – integrated visual field, dB – decibels, yr – year, vs. – versus, Af. Amer – African American, Cauc – Caucasian, meds – medications. Associations presented in the figure all come from the same two models (one assessing fall/year as an outcome, the other assessing falls/step), except for the associations with inferior IVF, which come from additional multivariable models which evaluated inferior IVF sensitivity (and not total IVF sensitivity) as the primary independent predictor of the falls/year or falls/step.

Discordant results were again obtained when injurious falls were analyzed (Figure 3), with worse overall IVF sensitivity, older age, female gender and more comorbid illness associated with a higher rate of injurious falls/step (p<0.02 for all) but only more comorbid illness associated with injurious falls/year (p<0.001). African-Americans demonstrated fewer injurious falls/year as compared to Caucasians (p=0.02), but no difference in injurious falls/step (p=0.16).

Figure 3.

Association between different rates of injurious falls with degree of visual field damage and other patient characteristics. RR – rate ratio, p – p-value, IVF – integrated visual field, dB – decibels, yr – year, vs. – versus, Af. Amer – African American, Cauc – Caucasian, meds – medications. Associations presented in the figure all come from the same two models (one assessing injurious fall/year as an outcome, the other assessing injurious falls/step), except for the associations with inferior IVF, which come from additional multivariable models which evaluated inferior IVF sensitivity (and not total IVF sensitivity) as the primary independent predictor of the injurious falls/year or injurious falls/step.

Contrast sensitivity was also examined as a predictor of falls while controlling for age, race, gender, comorbidities and polypharmacy and showed no significant association with falls/injurious falls per year or per step (p>0.13 for both). Also, additional models including IVF sensitivity, falls in the year prior to the baseline visit, and the above covariates were analyzed with a goal of optimally predicting future falls per year and falls per step. In these models, IVF sensitivity remained associated with falls/step (RR=1.32 per 5dB decrement in sensitivity, p=0.03), but not falls/year (RR=1.17 per 5dB decrement in sensitivity, p=0.21). Additionally, prior falls were strongly associated with both falls per year (RR=1.64, p=0.008) and per step (RR=1.59, p=0.012).

DISCUSSION

Individuals with greater VF damage fall more per steps taken, but walk less and do not demonstrate more falls over time. Similarly, older patients, females, and patients with more comorbid illness also fall more per steps taken, but walk less and do not demonstrate more falls over time. The most sensible method for preventing falls is to promote safe physical activity, minimizing falls in the context of an active healthy lifestyle. While falls can also be prevented by restriction of activity, such an approach trades injury avoidance for the negative physical and psychological consequences of inactivity and should be reserved for temporary downturns in health, i.e. hospital stays. As seen in the current study, the factors associated with more falls per step are frequently distinct from those associated with more falls per time, highlighting the limitations of assessing falls in the absence of concurrent activity assessment and suggesting that both falls assessment and activity measurement are required to optimally characterize if a risk factor is associated with falls.

While risk factors associated with more falls/year predispose individuals to the negative consequences of falls, such associations conflate two possibilities: (1) that individuals are unaware of their higher risk of falls/step and do not restrict their activity or (2) that they fall an equal/lesser amount per step but perform more activity (as observed for Caucasians). Thus, relying exclusively on falls/year risks identifying markers of increased activity as falls risk factors. Equally concerning, no association may be found between a putative risk factor and falls/year in the context of clinically-relevant counterbalancing problems, i.e. a higher rate of falls/step and diminished physical activity (as observed for greater VF damage, older age, and greater comorbid illness).

We believe that the discordant effects of a “falls risk factor” on falls/step vs. falls/year is likely to be widespread. Indeed, several factors felt to be important for falls, such as cognitive impairment and Parkinson’s disease, are also associated with restriction of physical activity.^16,17 As such, the importance of these factors with regards to falls may well be underestimated as they increase the risk of falls even as they produce activity restriction, as was noted for VF damage in the current study. Assessing the efficacy of falls interventions without activity assessment also deserves further consideration; many interventions to prevent falls may also increase activity, and no benefit may be noted if increased activity levels mask fewer falls/step. In other words, successful interventions could produce fewer falls with equal or greater physical activity or similar fall rates in the context of greater physical activity. As a specific example, interventions which reduce falls through exercise programs^18,19 are likely to reduce fall rates per step to an even greater degree than fall rates per year, though physical activity was not measured as an outcome in these studies. While some studies have examined physical activity before and after fall prevention efforts²⁰, they have gauged physical activity through self-report, which is poorly correlated with objectively-measured activity^21–23, and also less likely to capture actual activity given its weaker correlations with health measures such as body-mass index, blood pressure, triglyceride levels, and waist circumference²⁴.

Average daily steps were taken as the primary measure of physical activity in the current study as walking is a primary means of activity in older adults, and guidelines for physical activity frequently focus on the number of steps which should be taken daily.²⁵ Additionally, the amount of walking serves as a reasonable measure of the extent to which an individual subjects themselves to the risk of a fall. Of course, many falls may occur outside of the context of walking, i.e. during postural changes, though such postural changes may often occur just before or immediately after a bout of walking. Further research is needed to determine if a measure of activity other than steps can better capture the exposure to falling, and what exactly that measure of activity would be.

Our study findings regarding the relationship between glaucoma damage and fall rates were somewhat out of line with prior studies. Specifically, we did not find an association between the severity of glaucoma damage and the rate of falls over time, which lies in contrast to prior studies which either found a higher rate of falls in a glaucoma group as compared to a group of controls, or a higher rate of falling in patients with greater levels of VF damage.^8,26–29 Of note, most of these studies evaluated falls retrospectively, such that the associations may have been biased towards a positive finding, particularly if those with glaucoma were more likely to remember or report their falls. Also of note, one prior study examining falls prospectively in a glaucoma cohort did find an association between VF damage and the rate of falls over time, though it is unknown if an even greater association between falls/step would have been observed as physical activity was not assessed. While prior studies have found an association between contrast sensitivity and fall rates per year in older adults³⁰, we did not observe such an association in our study despite the effects of glaucoma on contrast sensitivity. These data suggest that VF damage is particularly important in capturing the relationship of vision to falls, at least in the context of glaucoma, though difficulties measuring VF damage makes it less practical to integrate into visual screening efforts.

While our primary models were run to determine the association of poor vision (i.e. VF damage) to fall rates, we also were interested in predictive models designed to determine which patients with glaucoma would most benefit from a falls intervention. In these predictive models, previous self-reported falls and VF damage both predicted falls/step, suggesting that it is important to consider both when identifying patients who are most likely to fall in the future.

Our findings also confirm prior studies linking VF damage to physical activity restriction.^10,11 These findings have been observed in both clinic populations as well as in population-based samples. Additionally, we observed that older age, female gender, African-American race, more comorbid illnesses and polypharmacy were all associated with less physical activity (fewer steps). Indeed, many of these associations (i.e. age, female gender, and comorbid illness) have also been identified in other cohorts of older adults.³¹

Our study has several limitations. First, recruited participants were demographically similar to the study-eligible sample, but were more likely to self-report a fall in the past year and have mobility challenges (i.e. use an assistive device), suggesting preferential recruitment of high-risk subjects and potentially those who are frailer. Moreover, our study population may not be representative of the glaucoma population in the United States. Physical activity over each year was inferred from a one-week accelerometer trial, though prior studies suggest that even 4 days of data can be used to infer long-term activity patterns.¹⁴ Also, both medications and comorbidities were evaluated based on a total number. While this approach allows for the idea that medical conditions and medications may interact with each other to produce falls, it does not consider that some conditions/medications may place individuals at the greater risk of falling than others. Further, it is possible that some of the patients developed other ocular conditions that would have placed them at even higher risk for falls, though this is unlikely to have occurred preferentially in those with more severe VF damage. Strength of this study include prospective falls data collection, objective measurement of the physical activity, and dual characterization of the fall rates over time and over physical activity.

Our findings call for the routine use of both falls and activity assessment when defining falls risk factors, judging falls interventions, and caring for older adults clinically. As illustrated here, failure to do so risks missing factors associated with more falls/step masked by diminished activity, or misidentifying risk factors based on their associations with greater physical activity.

IMPACT STATEMENT.

We certify that this work is novel. The ideal method for reducing falls is to minimize falls per unit of activity (i.e. per step), thus allowing for fewer falls in the context of an active lifestyle. Very few studies have examined risk factors of falls per step and none, to our knowledge, have compared the risk factors associated with falls per time and falls per step. Here, we demonstrate that the risk factors associated with falls per year are distinct from those associated with falls per step, suggesting a re-evaluation of how researchers and clinicians think about what constitutes a falls risk factor.