Techniques to Identify Diabetes Treatment Gaps

More than 20,000 articles have been written on how poor medication adherence, which costs the United States between $68 to $146 billion annually, increases adverse clinical outcomes and mortality.

Predictors of poor medication adherence include increased regimen complexity, multiple comorbidities, high medication cost, lack of patient understanding and engagement, and poor relationship with providers.

Researchers have used data mining for decades to determine factors that contribute to nonadherence. A specific type of data mining known as sequence discovery techniques can identify factors precipitating gaps in therapy (eg, delays in prescription refills).

A team of researchers from Humana recently published an article using sequence discovery techniques in The American Journal of Managed Care. These techniques identify sequences of events that precede gaps in therapy.

This study differed from past ones because it used data mining to identify associations between medication adherence and exposures without pre-determined hypotheses. The top exposures identified in this study completely differed from the 18 exposures identified in a 2014 systematic review of 27 studies.

The researchers included patients with at least 6 months of continuous diabetic medication refill history. A gap in therapy was defined as a refill 6 or more days after patients finished their last refilled dose.

Exposure sequences were analyzed 90 days prior to the gap and before the last refill among patients who had no gaps in care.

The 3 main events that preceded gaps in therapy such as prescription refills were as follows:

1. Patients who were prescribed a new medication

• Medication adherence outreach should especially target patients who have multiple out-of-network claims and those who visit a specialty physician after the medication is prescribed.

2. Patients who reversed a prescription claim.

• This is especially relevant if they are then prescribed a new medication or visit a specialty physician.

3. Patients who had multiple out-of-network claims, particularly those who have been hospitalized.

Past reports have shown that a 25% increase in adherence is associated with an average A1C decrease of 0.34%. Eliminating nonadherence would save $13,000 on average per newly diagnosed patient.

This study was limited by the population belonging to one insurer concentrated in the southern United States, a focus on oral antidiabetics only, and documentation errors in the prescription database.

The study authors maintained that their novel approach illuminates previously unknown factors driving medication nonadherence. Health insurers may use a similar method of data mining their own databases to prevent nonadherence and therefore reduce costs.