THE ADHERENCE ILLUSION: WHY ADHERENT-ONLY RESULTS CANNOT GENERALLY INFORM REAL WORLD DECISIONS
Author(s)
Gabriel Gbidye, MS, RN1, Brian Rittenhouse, PhD2;
1MCPHS University, Boston, Boston, MA, USA, 2MCPHS University, Boston, MA, USA
1MCPHS University, Boston, Boston, MA, USA, 2MCPHS University, Boston, MA, USA
OBJECTIVES: Adherent only cost-effectiveness analyses appear to provide actionable evidence but generally cannot be implemented because adherence status is not known at treatment initiation. We develop a decision analytic framework that uses adherent only results but shows that non adherent outcomes and assumptions about adherence diagnostic accuracy are also required.
METHODS: A decision maker may either ignore adherence and treat all patients according to intention to treat (No Adherence Test) results or may “Test” for adherence and assign appropriate treatments conditional on the label (perhaps falsely) “adherent” or “non adherent” in the Test. As in any diagnostic problem, Test results depend on the prevalence of adherence, test sensitivity and specificity, and the costs and QALYs associated with each adherence group. Using adherence prevalence and diabetes risk differences estimated from the DPP literature, we constructed adherent and non adherent cost and QALY profiles and combined them using Bayes’ rule to evaluate the expected outcomes of a Test strategy. We then compare Test and No Test strategies.
RESULTS: In a base case with 54 percent adherence prevalence from the DPP and 90 percent conjectured sensitivity and specificity, we determined optimal treatment conditional on the Test result and calculated expected costs and QALYs under Test versus No Test. The Test strategy produced lower expected costs (7,340 dollars vs 7,498 dollars) but fewer QALYs (2.0435 vs 2.099). The incremental cost effectiveness ratio for No Test compared with Test was approximately 3,000 dollars per QALY, indicating that No Test is preferred at conventional willingness-to-pay thresholds. Results varied substantially with adherence prevalence and test accuracy, demonstrating that adherent only results alone are insufficient to determine optimal policy.
CONCLUSIONS: Without explicit consideration of non-adherent patient outcomes and uncertainty in identifying who will adhere, with attendant false positives and negatives, adherent only analyses offer only an illusion of actionable evidence.
METHODS: A decision maker may either ignore adherence and treat all patients according to intention to treat (No Adherence Test) results or may “Test” for adherence and assign appropriate treatments conditional on the label (perhaps falsely) “adherent” or “non adherent” in the Test. As in any diagnostic problem, Test results depend on the prevalence of adherence, test sensitivity and specificity, and the costs and QALYs associated with each adherence group. Using adherence prevalence and diabetes risk differences estimated from the DPP literature, we constructed adherent and non adherent cost and QALY profiles and combined them using Bayes’ rule to evaluate the expected outcomes of a Test strategy. We then compare Test and No Test strategies.
RESULTS: In a base case with 54 percent adherence prevalence from the DPP and 90 percent conjectured sensitivity and specificity, we determined optimal treatment conditional on the Test result and calculated expected costs and QALYs under Test versus No Test. The Test strategy produced lower expected costs (7,340 dollars vs 7,498 dollars) but fewer QALYs (2.0435 vs 2.099). The incremental cost effectiveness ratio for No Test compared with Test was approximately 3,000 dollars per QALY, indicating that No Test is preferred at conventional willingness-to-pay thresholds. Results varied substantially with adherence prevalence and test accuracy, demonstrating that adherent only results alone are insufficient to determine optimal policy.
CONCLUSIONS: Without explicit consideration of non-adherent patient outcomes and uncertainty in identifying who will adhere, with attendant false positives and negatives, adherent only analyses offer only an illusion of actionable evidence.
Conference/Value in Health Info
2026-05, ISPOR 2026, Philadelphia, PA, USA
Value in Health, Volume 29, Issue S6
Code
EE222
Topic
Economic Evaluation
Disease
SDC: Diabetes/Endocrine/Metabolic Disorders (including obesity)