Optimal adherence by claims was lower among ATSM respondents reporting more missed days for blood pressure and cholesterol but not diabetes pills

Description #

Convergent validity was partial: claims-based optimal adherence (CMG ≤ 20%) declined as ATSM-reported missed days increased, with statistically significant gradients for blood-pressure pills (p=0.02) and cholesterol pills (p<0.01) but not for diabetes pills (p=0.19). So the language-concordant ATSM tracked the objective measure for two of three medication classes.

"the prevalence rates of optimal adherence were lower for ATSM respondents who reported more missed days, with significant differences across response categories for blood pressure (p = 0.02) and cholesterol (p < 0.01) but not for diabetes (p = 0.19)." (Ratanawongsa, 2018, p. 5)

Methods Context #

What? #ⓘ

The observable: prevalence of optimal adherence (CMG ≤ 20% from pharmacy claims) within each ATSM self-reported missed-day category (0, 1, 2–7 days), separately for diabetes, blood-pressure, and cholesterol pills.

"The prevalence of optimal adherence among respondents varied by indication: 83% for diabetic pills, 91% for blood pressure pills, and 85–87% for cholesterol pills." (Ratanawongsa, 2018, p. 4)

How? #ⓘ

Fisher's exact tests comparing optimal vs suboptimal CMG across the three ATSM missed-day categories, computed per therapeutic indication, among participants with a calculable CMG. Insulin was excluded because claims data do not provide a fixed days-supply.

"we used Fisher's exact tests to compare the prevalence of optimal adherence (CMG ≤ 20%) vs. suboptimal adherence (CMG > 20%) across the 3 categories of participant responses." (Ratanawongsa, 2018, p. 4)

Who? #ⓘ

The subset of the 210 SMARTSteps participants with a calculable CMG for each indication (e.g., diabetes pill n=226 dispensing-eligible; ATSM diabetes respondents n=170), all SFHP low-income multilingual diabetes patients.

"Among ATSM respondents, 75 (33.2%) for diabetes, 19 (19.3%) for blood pressure, and 55 (36.4%) for cholesterol reported at least 1 missed pill day." (Ratanawongsa, 2018, p. 4)

Other Notes #

The non-significant diabetes-pill gradient may reflect the short 7-day self-report recall versus the 1-year CMG capture window (see caveat).

Caveats #

• Short 7-day self-report recall against a 1-year claims window plus selection bias limit the ATSM validity estimates The self-report recall window (7 days) differed from the claims-based CMG capture period (1 year), which can attenuate convergent validity (and may explain the non-significant diabetes-pill gradient). CMG was not calculable for some participants, the statistical comparison is conservative, prescribing data could bias CMG, and the sample is subject to selection bias with limited generalizability; the study was underpowered for subgroup analyses.