Risk Predictors, Outcomes, and Costs of Falls among Older Adults Living in the Community

TitleRisk Predictors, Outcomes, and Costs of Falls among Older Adults Living in the Community
Publication TypeThesis
Year of Publication2015
AuthorsHoffman, GJ
AdvisorEttner, SL
Number of Pages393
Date Published2015
UniversityUniversity of California, Los Angeles
CityLos Angeles
Thesis TypePh.D.
Accession Number1675056036
KeywordsConsumption and Savings, Health Conditions and Status, Healthcare, Medicare/Medicaid/Health Insurance, Public Policy, Risk Taking

Background: Non-injurious falls (NIFs) and fall-related injuries (FRI) among older adults are common and impose substantial health burdens. Risks include health and health behavior factors, medications, and environmental hazards. Non-injurious falls (NIFs) and fall-related injuries (FRIs) may particularly compromise older adults' ability to engage in social and recreational activities and activities of daily living and can generate chronic stress, resulting in depressive symptoms. Previous studies have used depression to predict falls, but are potentially subject to endogeneity bias by not controlling for other temporal associations between falls and depression. With 2.5 million and .75 million FRIs treated in emergency departments and hospitals, respectively, FRIs also have important Medicare cost implications. However, existing FRI cost estimates are inconsistent, ranging from $2,000 to $26,000, and traditional FRI identification methods likely undercount FRIs. Three studies were conducted to assess relationships between (1) falls and caregiving, (2) falls and depressive symptoms, and (3) FRIs and Medicare costs. Methods: First, 10 survey waves of pooled data from the 2000-2010 Health and Retirement Study (HRS) were used to compare risks of NIFs and FRIs among respondents receiving and not receiving formal and/or informal caregiving assistance in the past two years. Stratified models were used to assess whether cognitive impairment modified the relationship between caregiving receipt and fall risk. Next, using 2006-10 HRS data, "cross-lagged panel" analyses using structural equation modeling were conducted to evaluate temporal associations between falls and depressive symptoms. Structural coefficients of pathways between falls and depressive symptoms (in 2006[arrow right]2008 and 2008[arrow right]2010) were estimated, controlling for baseline (2006) latent factors: poor physical health status (indicated by vision, hearing, cognitive status, activities of daily living, chronic disease, and functional limitations), perceived social support, and social cohesion. Finally, 2007-9 Inpatient, Outpatient, Carrier, Skilled Nursing Facility, Home Health, Durable Medical Equipment, and Hospice Medicare claims and linked 2008 HRS survey data were used to estimate the difference in predicted medical cost change scores (12-month post-index minus 12-month pre-index costs) for an FRI compared to a non-FRI cohort. For this analysis, linear regression models adjusted for sociodemographic characteristics, indicators of health status (sensory impairments, cognitive status, and indices for functional limitations and chronic health conditions), and geographic factors were used. Four different methods--using (1) external cause-of-injury codes (e-codes) and a select set of inpatient ICD-9 diagnostic codes, (2) e-codes, inpatient diagnostic and outpatient diagnostic and procedure codes, (3) e-codes and a broad set of diagnostic codes, and (4) e-codes only--were used to identify FRIs in the claims data. Population: Community-dwelling older adults (ages 65 and older). For the first study, 17,420 unique individuals (172 with cognitive impairment), with a total of 63,542 pooled person-wave observations; for the second study, 13,439 individuals; and for the third study, 5,548 individuals (103 FRI and 5,445 non-FRI) with continuous Medicare Parts A/B coverage and alive during the 24-month study period. Results: For the first study, adjusting for sociodemographic and health characteristics, relative to no care received, older adults receiving low levels of informal care had a 3 percentage-point increase in NIF risk and 2 percentage-point increase in FRI risk while those receiving high levels of informal care had a 4 percentage-point increase in FRI risk. However, those receiving both formal and informal care had a 4 percentage-point (or 20%) decrease in NIF risk. Wald tests showed that those receiving formal care only or formal care and informal care had lower NIF and FRI risks than those receivin informal care only. Among those with cognitive impairment, older adults receiving informal and formal care generally had reduced NIF and FRI risks compared to those receiving no care. For the second study, both the NIF (χ 2 (dfs = 266) = 2,646.39, p <0.001, Comparative Fit Index (CFI) = 0.944, Root Mean Square Error of Approximation (RMSEA) = 0.026) and FRI (χ2 (dfs = 267) = 2,511.52, p <0.001, CFI = 0.945, RMSEA = 0.025) models had acceptable fit. Depressive symptoms predicted future NIFs (unstandardized structural probit coefficients = 0.85 for 2006[arrow right]2008 and 0.83 for 2008[arrow right]2010, p <0.001), a relationship potentially mediated by psychotropic drug use; however, NIFs did not predict future depressive symptoms. Depressive symptoms predicted future FRIs (unstandardized structural probit coefficients = 1.00, p <0.001 for 2006[arrow right]2008 and 0.95, p <0.001 for 2008[arrow right]2010). For the third study, average, unadjusted medical costs during the 12-month follow-up period were $23,429 and $9,818 for FRI and non-FRI cohort individuals, respectively. The estimated change score difference between the FRI and non-FRI cohort was $14,879 (p<0.001) using the first FRI identification method using e-codes and a select set of inpatient ICD-9 diagnostic codes. Change score differences were $6,587 for inpatient, $2,476 for outpatient, $4,595 for SNF, and $1,141 for HH. Overall change score differences were $11,678 (p<0.001), $8,963 (p<0.001), and $12,757 (p<0.001) using the second, third, and fourth FRI identification methods (using varying sets of e-codes and inpatient and outpatient diagnostic and procedural codes), respectively. An FRI increased the risk of persistently high costs using all identification methods. Findings were robust to the inclusion of pre-index costs as a predictor and respondents who died during the study. Conclusion: For the first study, caregiving receipt significantly predicted falls, but was not uniformly negatively associated with NIF and FRI risk. The 20% risk reduction among those receiving formal and informal care was comparable to common falls prevention interventions. While informal care may encourage more activity and risk-taking, formal caregivers may prevent risky behaviors associated with falls and cognitively impaired individuals may particularly benefit from caregiving in preventing falls. For the second study, compared to latent indicators of poor physical health status and social factors, depressive symptoms are strong predictors of older adult NIFs and FRIs. Controlling for these confounders, a one standard deviation increase in depressive symptom levels was associated with a 22% and 28% increase in the risk of future NIFs and FRIs, respectively. However, NIFs and FRIs were not associated with greater odds of future increased depressive symptoms. For the third study, FRIs were associated with substantial and sustained increases in Medicare spending. Substantial increases in medical costs associated with FRIs were driven by inpatient (44%), SNF (31%), outpatient/carrier (17%), and home health (8%) costs and FRIs increased the probability of having high cost in the four quarters following an index FRI from 11.6% to 19.5%, reflecting a 67% increase. The findings suggest a wide range of per-FRI as well as total FRI-related Medicare costs, depending upon the method used to identify FRIs in the claims data; per-FRI costs ranged from approximately $9,000 to $15,000 while total FRI-related Medicare costs ranged from $5 to $49 billion due to differing proportions of beneficiaries experiencing FRIs depending on the FRI identification method used. Implications: These findings suggest a potential opportunity for population health management of falls. To scale up falls prevention for falls costing Medicare billions of dollars per year, policymakers and providers should consider offering assistance to informal caregivers through educational outreach. Falls assessments should also include caregiving as an essential component. Prevention effo ts should include depressive symptomatology as a main falls risk factor and geriatric evaluations should emphasize the depressive symptoms--falls relationship. Current approaches to claims-based FRI identification and current FRI surveillance techniques may be inadequate. Given substantial, sustained costs of FRIs and known effectiveness of falls prevention efforts, Medicare should consider reimbursement for geriatric FRI risk prevention.

Endnote Keywords

Public health policy

Short TitleRisk Predictors, Outcomes, and Costs of Falls among Older Adults Living in the Community
Citation Key6339