Skip to main content
Conference Coverage

Predicting Diabetes Medication Adherence With Pharmacy Claims, Public Data

Edan Stanley

Researchers at AMCP Nexus 2022 created metrics to quantify social determinants of health (SDoH) data to better predict medication adherence.

“SDoH are environmental and socially based factors that influence individual and group health outcomes,” explained the authors of the poster. “While these factors play a significant role in health outcomes, metrics quantifying these factors are not readily available in pharmacy administrative claims data.”

Combining publicly available data sets from the North American Industry Classification System (NAICS), the Census, and merging with patience claims and provide data from a pharmacy benefit manager, researchers created zip code-based metrics to quantify SDoH, as well as assess their predictive validity specifically for adherence to diabetes medications.

A total sample of 37,789 patients aged ≥18 years, who met PQA inclusion criteria for diabetes, were included.

Economic stability was measured using several factors including the following:

  • Count of residential zip codes to capture housing stability,
  • ratio of housing units to people,
  • proportion of the population employed within zip code,
  • and mean population salary within zip code.

Health care access was measured using the following:

  • distance traveled by patient to prescriber,
  • distance traveled to pharmacy,
  • flag that denotes whether a patient had a primary care physician (PCP),
  • and the ratio of population to PCPs within zip code.

Researchers calculated the proportion of days covered (PDC) for diabetes medications using the PQA guidelines. Regression models were used to assess the relationship between PDC and SDoH.

According to the results of the analysis, patients with 2 zip codes had lower PDC values compared with those with only 1 (3 + (B = -0.04, P < .01 or B = -0.03, P < .01). Average PDC increases were observed as the ratio of housing units to people increased.

In terms of distance traveled, patients traveling 26+ miles to a prescriber had lower PDC values than those who only traveled 5 miles (B = -0.02, P < .01). Additionally, patients traveling between 1 to 15 miles or greater than 16 miles, had lower PDC values than patients traveling less than 1 mile.

Both having a primary care provider (PCP) and living in a zip code with 150 or fewer residents per PCP was associated with higher PDC values, compared with those without PCPs or in areas with ≥501 residents per PCP.  

Researchers found employment and mean population salary were not predictors of PDC.

Overall, researchers concluded these metrics quantifying SDoH can be used to predict medication adherence.

Reference:
Hunter B, Brown-Gentry K, Prasla K, Santilli M. Social determinants of health: combining publicly available data with pharmacy claims to create metrics that predict medication adherence. J Manag Care Spec Pharm. 2022;28(10-a suppl):S1-S137. doi:10.18553/jmcp.2022.28.10-a.s1