Skip Navigation
AHRQ--Agency for Healthcare Research and Quality: Advancing Excellence in Health Care
  • Home
  • Search for Research Summaries, Reviews, and Reports

Abstract - Final – Feb. 11, 2010

Developing Methods to Assess Health Outcomes in the Presence of Competing Risks for Effectiveness Research


Background: Caring for older adults is an exercise in balancing competing risks. Clinicians, necessarily, use diagnostic tests, interventions, surgical procedures, and medications in the face of known and unknown risks. Complicating the care of older adults is the almost universal presence of diseases besides those that are being targeted with the interventions. Although drugs, devices, and procedures may have been demonstrated to be efficacious and safe in the narrow populations in pre-marketing trials, this cannot be assumed to be true for other populations such as among much older people. Additionally, the effectiveness of drugs, devices, and procedures when used in a usual care setting may differ importantly from their demonstrated efficacy in controlled settings.

Objective/Research Questions: The aim of this project is to develop methods to transparently balance risks against benefits, which include the competing risks for adverse events and death faced by older adults.

Study design: Cohort study using observational data for the development of statistical models.

Methods: Data from the Women’s Health Initiative Observational Cohort (WHIOS) and FIT/FLEX clinical trial data will be used. A cohort will be created of women who have osteoporosis, defined by baseline bone densitometry (a T-score below —2.5) when available (measured in 6442 women at 3 sites), self-report of osteoporosis, or use of a bisphosphonate or other therapy for osteoporosis (e.g. calcitonin), excluding hormonal therapy. There will be no exclusion based on age but women will be stratified so that one stratum is similar to the women enrolled in the clinical trials and the other stratum is older. The primary clinical outcomes to be studied will be fractures (hip, spine, and other) and death. Outcomes will be modeled by exposure status while controlling for predictors of fractures (fall and fracture history, medications, body weight, race, age, bone mineral density, smoking, postural instability, self-rated health, neuromuscular disease). Prediction models will be used that are based on a direct regression of the cumulative incidence function for an event. The direct regression methods will be extended to incorporate adjustment for imbalance in pre-treatment variables and unmeasured confounders. Differences among groups will be controlled for with multivariate analysis.

Expected Outputs: Methods reports

Expected date of project completion in month, year format: Fall 2011.