Positively Frightened: Spillover Effects of False-Positive Mammography on Medication Adherence

This study estimates the potential spillover effects of mammogram false-positives on anti-hypertensive and dyslipidemia medication adherence.  The issue of false-positive test results, particularly for mammography, has gained attention in health services research, however few studies have examined how these results affect future screening.  Additionally, none have examined the potential spillover effects onto other types of preventive health care utilization, such as medication adherence.  Due to data limitations, all analyses focus on the spillover effects on medication adherence rather than the direct effects on future mammography.  This study has important implications for optimal screening behavior as well as the cost-effectiveness of screening.

I first create a quasi-Bayesian theoretical model of preventive care that explicitly allows for imperfect test results, including false-positives.  The model includes both secondary (mammogram) and tertiary (medication adherence) preventive care decisions.  The model also allows for both a negative effect of a false-positive on adherence due to reduced beliefs in the efficacy of future treatments as well as a positive effect where the false-positive may act as a wake-up call to improve preventive health behaviors.   

The claims-based analyses use both commercial and Medicaid Marketscan claims data, with all models estimated separately by insurance status.   Models are also estimated separately for anti-hypertensive and dyslipidemia medication adherence.  The study estimates difference-in-difference models comparing medication adherence before and after screening between those who experience a false-positive to those who experience a true-negative.  Using diagnostic codes from the medical literature, the analysis is restricted to all women between the ages of 40-64 who receive a screening mammogram (no mammogram within the prior 12 months), do not ever have any claims for breast cancer, and have at least one claim for an anti-hypertensive or dyslipidemia prescription.  The false-positive group is defined as those receiving a screening mammogram then a second mammogram, a biopsy, a breast ultrasound, or other radiological follow-up test within 3 months of the screening mammogram Similarly, the true-negative group is defined as women who receive a screening mammogram but no follow-up screening.  I further test whether the effects are due to increased exposure to medical attention by dividing the false-positive group into those with a single follow-up visit compared to those with multiple.  Using several definitions, the adherence dependent variable is defined as days without medication, an indicator for having medication at least 80 percent of days, stopping medication completely, and any changes in medication.  Additional analyses separate the false-positive group by length between screening  and resolution of false-positive and by invasiveness of follow-up testing (i.e. biopsy vs. non-biopsy).  Further sensitivity analyses also vary the diagnostic codes used for each procedure, vary the length of the pre-and post-periods, vary the length of the follow-up test period, separate by age group, and separate by race/ethnicity.

Preliminary results provide support for the “wake-up call” hypothesis; I find evidence that a false-positive diagnosis of breast cancer is associated with an increase in medication adherence for dyslipidemia drugs and I find weak evidence of medication adherence for hypertensive drugs.