Identifying observable medication use time in administrative databases: a tutorial using nursing home residents.

Area of Science:

• Pharmacoepidemiology
• Health Services Research
• Gerontology

Background:

• Nursing home residents often have multimorbidity and polypharmacy, making them a key population for pharmacoepidemiologic research.
• Medicare claims are frequently used but have limitations in observing medication use during hospitalizations or post-acute care due to payment structures.

Purpose of the Study:

• To develop and validate algorithms for identifying observable medication use time in nursing home residents using Medicare claims.
• To enhance the accuracy and scope of pharmacoepidemiologic studies in the nursing home population.

Main Methods:

• Linked Medicare fee-for-service (FFS) claims with Minimum Data Set (MDS) clinical assessments for US nursing home residents (2013-2020).
• Defined "observable medication use time" as days when residents were enrolled in Medicare Parts A, B, and D, not receiving post-acute care, and not hospitalized.
• Classified 2.7 billion nursing home days across 12.3 million residents.

Main Results:

• Identified 1.1 billion observable medication use days (72.4% of Medicare-enrolled days).
• Observable medication use increased from 27.3% within 100 days of admission to 89.4% after 100 days.
• Algorithms identified 68% more person-time and 51% more residents compared to standard 100-day definitions.

Conclusions:

• The developed algorithms significantly expand the observable medication exposure time in nursing home residents.
• This methodology improves the validity of pharmacoepidemiologic research by providing a more complete picture of medication use.
• Enhanced data capture enables more robust studies on medication effects and safety in this vulnerable population.