兰德-Evaluation-of-an-Intervention-to-Prevent-Falls_66页_797kb

Summary of the Evaluation of an Intervention to Prevent Falls

Core Content

This document presents an evaluation of a fall-prevention intervention designed for older Medicare patients, with a focus on objective measures of health care costs and emergency department (ED) visits for fall-related injuries. The study is based on data from a previously published randomized trial (Cohen et al., 2015) and links it to Medicare claims to provide a more comprehensive and objective analysis.

The intervention, developed by Abt Associates and LifePlans, consists of four components: an in-home fall risk assessment, a customized action plan and toolkit, follow-up coaching phone calls, and quarterly newsletters. The evaluation was conducted by the RAND Corporation at the request of the Assistant Secretary for Planning and Evaluation (ASPE), who sought to assess the cost-effectiveness of the intervention for Medicare beneficiaries.

Main Objectives

• To evaluate the effect of the fall-prevention intervention on Medicare costs and health care utilization.
• To compare the intervention group with both active and administrative control groups.
• To address limitations of the previous study, particularly the reliance on self-reported data and the lack of objective cost measures.

Key Findings

• No statistically significant effect: The study did not find a statistically significant effect of the intervention on health care costs or on the rate of fall-related ED visits.
• Point estimates and confidence intervals:
  • Total health care spending increased by $18 per month (95% CI = -$94 to $130).
  • Injury-related Medicare costs increased by $18 per month (95% CI = -$12 to $48).
  • The risk of falls increased by 4% in the experimental group (95% CI = 26% decrease to 43% increase).
• Cost savings from long-term care: The previous study reported a 33% reduction in long-term care claims, leading to a net cost savings of $1.68 for every $1.00 spent.
• Limitations of the original study:
  • The effect on health care spending could not be measured.
  • Dropout rates may have introduced bias.
  • Self-reported data limited the ability to use the administrative control group as a comparison for falls.

Methodology

• Data Sources:
  • LifePlans data on 7,178 participants, including group assignment, enrollment dates, and health information.
  • Medicare claims data and enrollment files from 2006 to 2014, accessed through the Virtual Research Data Center.
• Analysis Approaches:
  • Intention-to-treat (ITT): Analyzed all participants who were initially enrolled, regardless of whether they completed the intervention.
  • As-treated: Analyzed outcomes for participants who actually received the intervention.
  • Risk-adjusted analyses: Used a difference-in-differences approach to control for baseline differences and seasonal effects.
• Outcome Measures:
  • Self-reported falls: Based on quarterly telephone interviews.
  • ED visits for fall-related injuries: Estimated using a claims-based model (see Appendix A).
  • Medicare costs: Total and injury-related costs per beneficiary per month.
  • Mortality: Evaluated for one- and two-year periods.

Study Design and Population

• The study population included individuals aged 75 and older who had long-term care insurance.
• The trial was randomized into three groups: Experimental Group (EG), Active Control Group (ACG), and Administrative Control Group (ADCG).
• The EG had a higher dropout rate (19%) compared to the ACG (7%).
• Analyses were restricted to those enrolled in traditional FFS Medicare due to data availability.

Limitations and Considerations

• The use of self-reported data introduced recall bias and limited the ability to compare the ADCG with the EG.
• The delayed intervention period and the limited data coverage (up to 2014) affected the as-treated analysis for some participants.
• The study acknowledges that the imprecision of estimates may have masked a clinically meaningful effect.

Conclusion

The fall-prevention intervention did not have a substantial effect on Medicare costs or on the rate of fall-related ED visits. While the original study suggested cost savings from reduced long-term care claims, this evaluation using objective Medicare data did not confirm a significant impact on overall health care costs. The results highlight the importance of using objective measures in evaluating interventions and the potential limitations of relying on self-reported data for fall-related outcomes. The study also underscores the need for more pragmatic trials that can be implemented in real-world settings and provide reliable cost-effectiveness data.

Key Information

• Intervention Components: In-home assessment, customized action plan, follow-up calls, quarterly newsletters.
• Control Groups: ACG (active control) and ADCG (administrative control).
• Data Sources: LifePlans data and Medicare claims (2006–2014).
• Outcome Measures:
  • Self-reported falls.
  • ED visits for fall-related injuries.
  • Medicare costs (total and injury-related).
  • Mortality rates.
• Statistical Methods:
  • Linear regression for cost analysis.
  • Poisson regression for ED visit counts.
  • Risk-adjusted models to control for baseline differences.
• Confounding Considerations:
  • Dropout rates.
  • Seasonality and secular trends.
  • Selection bias in the analytic samples.

This evaluation provides a more rigorous and objective assessment of the fall-prevention intervention, offering important insights for policy and program development.