Prevalence and predictors of falls in a health‐seeking older population: An outpatient‐based study

Abstract Background Falls are one of the major causes of disability in older people. A wide range of risk factors for falls are described according to setting – inpatient, nursing homes and community. The aim of this study was to identify the risk factors for falls in an outpatient setting. Methods In this cross‐sectional observational study, 160 consenting subjects were enrolled randomly, from the Geriatric Medicine outpatient department, All India Institute of Medical Sciences, New Delhi, India. Non‐ambulatory, seriously ill subjects were excluded. The subjects underwent brief evaluation including falls and geriatric assessment. They were grouped into fallers and non‐fallers. A multivariable logistic regression analysis was used to identify the factors associated with falls. Results The prevalence of falls was 23.75% (38/160). Women were proportionately higher (26.31%) in the fallers group vis‐à‐vis 19.67% in the non‐fallers group. After multivariate analysis, opioids (odds ratio [OR] 5.24 [95% CI, 2.0 18‐13.611]), vision impairment (OR 2.71 [95% CI, 1.050‐07.011]), fear of falling (OR 3.17 [95% CI, 1.167‐08.629]), instrumental activity of daily living (IADL) impairment (OR 3.41 [95% CI, 1.251‐09.301]), anti‐anginal medications (OR 8.90 [95% CI, 0.997‐79.564]) and self‐employment (OR 5.37 [95% CI, 1.058‐27.329]) were associated with falls. Adequate nutrition (OR 0.82 [95% CI, 0.688‐00.976]) and caregiver support (OR 0.46 [95% CI, 0.275‐00.801]) were protective of falls. Conclusion We identified the multi‐factorial etiology of falls. Patients having any of the above risk factors should undergo detailed fall risk assessment and preventive measures afterwards.

ambulatory subjects and subjects with severe functional impairment, defined as ability to perform only one ADL, 7 were excluded from this study.
A semi-structured interview in Hindi was carried out to identify socioeconomic status, personal history, medical history and medication use. Medications were identified by looking at the health records or by the blister packs. Polypharmacy was defined as taking more than or equal to five medications in the same month. 8 Orthostatic hypotension was defined as a drop in systolic BP >20 mm Hg or diastolic BP >10 mm Hg within three minutes of standing from a lying down position. It was measured by the Omron 7310™ apparatus which uses the oscillometric method. 9 Functional impairment was assessed by Barthel activity-dependent daily living (BADL) and Lawton instrumental activity-dependent daily living (IADL). Impairment in any one domain was considered as dependent. Tinetti performance-oriented mobility assessment (POMA) and timed up and go test (TUG) was used for mobility assessment. Vision was assessed by Snellen chart and E charts for illiterates and hearing impairment by WHO grades. [10][11][12][13] Frailty index was used to identify the frail population, Montreal Cognitive Assessment (MoCA) for identifying cognitive impairment, and the mini nutritional assessment-short form (MNA-SF) was used to identify the nutritional status. For the assessment of depression, the Geriatric Depression Scale (GDS-15 version) was used. 14-17

| Statistical methods
Statistical analysis was carried out using statistical software Stata/ SE version 14.2 (StataCorp LP). Qualitative variables of the study were described as absolute/relative frequency with percentage and quantitative variables by mean (standard deviation)/median (quartile range). To find the association between qualitative independent variables, the chi-square test/Fisher's exact test was used. To assess the association between two quantitative variables, the Pearson/ Spearman correlation coefficient was used. To find the difference in quantitative variables between groups, the t test/Wilcoxon test was used according to the distribution of the data. To find the factors associated with falls, stepwise multivariable logistic regression analysis was used. Variables which were found to be significant under crude association up to a level of 25% and/or clinically relevant were considered for the stepwise procedure. Calibration of predicted probability of the developed model was assessed by Hosmer-Lemeshow (HL) test and specification error by link test. Discrimination ability of the developed model was evaluated using the area under the curve.
Results were presented in the form of odds ratio with corresponding 95% confidence interval (CI). P value <.05 was considered as statistically significant.  Table 1.   Table 2).

| RE SULTS
The discrimination ability of the developed model was found to be satisfactory and the model was able to discriminate a case of fall with probability 0.8376 (Figure 1).

| D ISCUSS I ON
In this observational study, the proportion of fallers was found to be 23.75%. The percentage of women in the fallers group (23.61%) was higher than in the non-fallers group (19.67%), even though the prevalence of falls was a shade less than reported in the English longitudinal aging study (ELSA) (28.4% vs 23.75%). But ELSA also identified that women fall more than males. 18 It was also reported that women tend to report falls more than male peers and seek healthrelated advice in a medicare population. 19 Subjects who cohabitate were less prone to falls in this study and falls were found to be unrelated to socioeconomic status, wealth, personal habits and body mass index. However, while the proportion of fallers was slightly higher in the poorest social strata and people who live alone, these findings may be due to the hospital setting, as these groups come

Variables Non fallers (N = 122) Fallers (N = 38) T/χ 2 values P value
Nutritional assessment to government hospitals. Similar findings were also observed in the

MNA-SF
In the multivariate analysis, antianginals and opioids were significantly associated with falls. These findings were similar to the study by de Jong et al 24 and may be due to sudden hypotension, arrhythmia or a syncopal episode. 25 There was also a significant difference in fear of falls and vision impairment between the groups.
In multivariate analysis also, they were found to be independently associated with falls. The complex interaction of postural control, visual gaze and anxious behavior were well explained by Young et al. 26 Here the people with dementia had fallen more, even though no statistical significance was reached, and depression was also less prevalent in this population reflected by the mean scores.
In this study, frailty was not associated with falls. In a study by Li et al, 27 it was found that for the frail population to experience one fall, the frailty index (FI) should be higher ( In multivariate analysis IADL was also an independent predictor of falls. A study by Nourhashemi et al, 28 found that IADL impairment was associated with a frailer population, fear of falling and associated with falls. Poor nutritional status was also contributing to falls and it was an independent predictor of falls. It was also shown by Chien et al. 29 that poor nutritional status was associated with falls and difficulties in IADL and it can also predict falls independently.
The strengths of this study include an adequate sample size of the representative population, and extensive evaluation of a wide range of fall risk factors described in the literature. It also has certain limitations. It was a cross-sectional study, so causality could not be established. There might be recall bias in recalling the incidence of fall. This was carried out in the outpatient department settings, as an interview.

| CON CLUS ION
Falls are one of the major public health problems. In this study, regardless of socioeconomic and demographic variations, the etiology of falls can be multifactorial. We need a large-scale observational study to evaluate the fall risk factors, and design preventive as well as rehabilitative programs. However, the independent factors identified in this study can be used as a screening tool and older people who are having these factors may be subjected to detailed falls assessment and rehabilitation thereafter.

ACK N OWLED G EM ENTS
The authors acknowledge support from the Department of Geriatric Medicine faculties, residents, physiotherapists and staff for their support in data collection.

CO N FLI C T S O F I NTE R E S T
Nothing to disclose.