The Cost of Adherence Mismeasurement: A Claims-Based Analysis

If you are attending AMCP Nexus in October, I am a co-author on a poster presentation titled “The Cost of Adherence Mismeasurement: A Claims-Based Analysis” co-authored with Felicia Forma, Ethan Scherer, Ainslie Hatch and Darius Lakdawalla.  The presentation abstract is below and here.

BACKGROUND:

Payers often wish to measure how adherence to prescription drugs affects downstream medical costs. Although payers typically rely on claims-based metrics metrics—such as the proportion of days covered (PDC)—these measures overestimate patient adherence when patients do not ingest all doses purchased and underestimate adherence when patients purchase prescriptions out-of-pocket. This study quantifies the effect of these adherence measurement errors on inferences about the benefits of adherence among patients with serious mental illness (SMI)

OBJECTIVE:

Measure the effect of improved adherence to oral atypical antipsychotic medications on inpatient cost after accounting for adherence inaccuracies in claims data.

METHODS:

We derived the statistical bias that occurs in adherence-utilization studies when adherence measurement is inaccurate and conducted a literature review to identify the key bias parameter: the correlation between true and claims-based adherence measures. Using data from Truven MarketScan Commercial and Medicaid databases (2007-2013), we applied our bias-correction methodology to a case study of patients diagnosed with bipolar disorder, major depressive disorder or schizophrenia who initiated atypical antipsychotic therapies. Adherence to oral atypicals was measured using PDC. We calculated the naïve and bias-adjusted effect of adherence on inpatient costs controlling for patient demographics, comorbidities, and prior spending.

RESULTS:

Among the 231,526 SMI patients who initiated atypical therapy, a ten percentage point increase in PDC lowered annual inpatient costs for all patients by $42 (95% CI: -$65 to -$19) per person and for patients with schizophrenia in particular by $86 (95% CI: -$147 to -$25). After adjusting for bias due to mismeasurement, we found that this same increase in PDC decreased inpatient costs by $223 (95%

CI: -$359 to -$106) and $458 (95% CI: -$815 to -$141) per person, respectively. Extrapolating these results to the entire U.S. population of patients with schizophrenia, the effect of a 10% increase in adherence is $0.3 billion using the naïve approach and the $1.5 billion after adjusting for bias.

CONCLUSIONS:

Payers may underestimate the effect of improved adherence on inpatient cost by a factor of 5 or more due to mismeasured adherence in claims data. Improving the accuracy of adherence data—through electronic pillboxes, smart caps, or ingestible sensors—could help provide clearer insight into the full value of improving adherence.

SPONSORSHIP:

Otsuka America Pharmaceutical