Analyzing evidence-based falls prevention data with significant missing information using variable selection after...

Area of Science:

• Gerontology
• Public Health
• Biostatistics

Background:

• Falls are a primary cause of injury and death in older adults.
• Evidence-based fall prevention programs aim to reduce fall-related risks.
• National data on fall prevention programs often contain significant missing values, complicating analysis.

Purpose of the Study:

• To evaluate and compare statistical methods for handling missing data in fall prevention research.
• To assess the performance of multiple imputation-stepwise regression (MI-stepwise) and multiple imputation-least absolute shrinkage and selection operator (MI-LASSO) for variable selection.
• To investigate these methods under various challenging data conditions, including missing not at random (MNAR) mechanisms.

Main Methods:

• Utilized data from 39 Administration for Community Living (ACL) grantees across 22 states (2014-2017).
• Employed multiple imputation techniques to address missing fall efficacy data.
• Conducted simulation studies to compare MI-stepwise and MI-LASSO performance, including scenarios with varying signal-to-noise ratios and missing data patterns.

Main Results:

• Simulation studies provided insights into the performance of MI-stepwise and MI-LASSO under diverse data conditions.
• Evaluated the MI-LASSO method's performance with adjusted tuning parameters to mitigate overselection issues.
• Identified the effectiveness of MI-LASSO in handling complex missing data scenarios, including MNAR.

Conclusions:

• Multiple imputation methods, particularly MI-LASSO, offer robust solutions for variable selection in datasets with missing values common in fall prevention research.
• The findings support the use of MI-LASSO for more reliable analysis of fall prevention program data.
• Addressing missing data is crucial for accurate variable selection and effective program evaluation in gerontology research.