Prehospital and Posthospital Fall Injuries in Older US Adults

Key Points Question What are older adults’ risks of fall injuries in the periods surrounding hospitalization? Findings This retrospective cohort study using national survey and linked Medicare data observed spikes in older adult fall injury risk in the periods just before and after hospitalization. Risk increases were particularly pronounced for those who experienced an inpatient fall injury. Meaning These findings suggest that efforts to improve coordination of fall injury risk during care transitions into and out of the hospital are needed.

To identify cognitive impairment, we followed existing practice by using the Telephone Interview of Cognitive Status (TICS), an alternative to the Mini-Mental State Examination. 46,47 TICS scores range from 0-35, with higher scores indicating better functioning. Respondents with scores of ≤8 were considered impaired. 28 Because not all respondents were able to complete the TICS, we followed precedent by using information reported from proxy respondents regarding respondent's cognitive status, which were assessments of the respondent's memory (0-4); limitations in instrumental activities of daily living (0-5); and the respondent's difficulty in completing the interview because of cognitive impairment (0-2). 28 This resulted in a scale with a range of 0-11, with higher scores representing worse cognition; individuals whose proxies reported scores of ≥6 were additionally considered cognitively impaired.

Race/Ethnicity
The HRS offers seven response categories for race plus an "other" category; the authors collapsed categories with small numbers (American Indian, Alaska Native, Asian, Native Hawaiian, Pacific Islander, other) into a single "Other" category. These were collapsed, due to small cell sizes, into four categories: African-American, Hispanic, non-Hispanic White, and Other.

Frailty
The frailty measure is based on an indicator developed by Cigolle et al.
29 using variables from the Health and Retirement Study and includes 4 domains: physical, nutritive, cognitive, and sensory. Individuals with difficulties in two or more domains were considered frail. Problems with physical functioning were indicated when dizziness was a persistent problem, if the respondent had two or more falls in the prior 2 years, or difficulty lifting 10 pounds. Problems with nutritive functioning were indicated if the respondent had a 10% or greater weight loss in the prior 2 years or a BMI of <18.5 kg/m 2 . Problems with cognitive functioning were indicated using self and proxy reports, as indicated above. Problems with sensory functioning were indicated by fair or poor vision or hearing despite the use of corrective lenses or hearing aides. The time variables for periods 1 and 3 are interpreted as the average weekly change in the odds of a fall injury within each of those respective periods. For instance, an OR of 1.10 in period 1 indicates that the odds of a fall injury in week 2 of period 1 (or 3) are 110% those of the odds in week of period 1 (or 3). However, the ORs in periods 2 and 4 indicate the difference in the odds of a fall injury compared to those in periods 1 and 3, respectively. For instance, if the OR for periods 1 and 2 are both >1, that indicates that the slope of period 2 is greater than the slope of period 1, i.e., that the average odds of a fall injury for each week spent in periods 1 and 2 increase, but the odds of injury increase more in period 2 than in period 1.
It is possible, however, to use the slopes of periods 1 and 2, and then of periods 3 and 4, to compute ORs for period 2 and 4 that do not reflect the relative change in odds of an injury compared to periods 1 and 3, respectively; instead, for periods 2 and 4, we can compute the average change in the odds of a fall injury for those periods. In other words, we can produce ORs that are interpreted in the same ways for periods 2 and 4 that we interpret them for periods 1 and 3. 1 The odds ratios and marginal effects for piecewise logistic regression models estimated using each of the two fall injury algorithms (conservative and aggressive approaches) are shown below. 'Baseline' refers to 6 to 1 month before hospitalization; 'Before hospitalization' refers to less than 1 month before hospitalization; 'After hospitalization' refers to from discharge to 1 month following discharge; 'Follow-up' refers to 1 to 6 months following discharge. Estimates were obtained from a piecewise logistic regression model that included dummy variables for each of the 4 time periods of interest (e.g., baseline, before hospitalization periods), using 2008-2014 data for Medicare beneficiaries with linked Health and Retirement Study data.  (0.00, 0.01) * p < 0.05. + pp = percentage point a Only included observations for individuals for whom the time between the date of their HRS survey and start of the study period (6 months prior to the anchor hospitalization) was ≤365 days. b Dropped first hospitalizations for individuals with more than one hospitalization. c Fall injury diagnoses were included in a single fall injury episode if they occurred within 180 days of one another (i.e., if an ankle fracture was treated in the emergency department on April 2, 2013 and a foot injury was treated in a physician's office on September 12, 2013, both diagnoses would be considered to be part of a single fall injury episode that occurred on April 2, 2013.