Flaws Revealed in "Non-Adherence" Models

On October 19, 2010, Amy Tenderich of DiabetesMine wrote a post asking whether "'Adherence' Is the New 'Compliance'?". Although the medical profession has worked hard to abandon the attitude of "blaming the patient" (they aren't 100% there - yet - but no one can deny the effort), decades of this attitude has been transferred to such parties as healthcare providers (e.g. insurance companies who pay the bills) and other HIPAA "covered-entities" like pharmacy benefits managers who hope to use this privileged data to make even more money.


Various statistical models are now being marketed by various PBMs and data-mining firms who are selling the sales data back to insurance companies to squeal on so-called "non-adherent" patients, hopefully to help improve medication compliance, which they theorize will save money. In August 2010, for example, CVS/Caremark's Behavioral Change Research Partnership was launched (see HERE) to study how behavioral economics — the science behind the idea of using "nudges" to help consumers improve their decisions – could be applied to health-care behavior, specifically prescription drug choices (meaning whether or not to fill orders). More recently, rival Express Scripts introduced (see HERE) a software system that determines which of its enrollees are most likely to stop using their medications in the middle of a prescribed regimen, with particular focus on diabetes, hypertension and high cholesterol patients. Medco most likely has a similar program, although that company has certainly been quieter about it.

On November 10, 2010, The Wall Street Journal Health Blog featured another post (see HERE) highlighting an interesting hole in this statistical data-mining, and the pharmacies such as CVS and Walgreens who may be playing a role in these flawed assumptions in their models. Low-cost generics programs pioneered by Wal-Mart and Target, but since adopted by most pharmacy retailers, may actually be responsible for polluting the very data they think they know about patients because most pharmacies never bother to submit insurance claims for low-cost generics. The patient/consumer doesn't really care as long as their drugs are inexpensive, but the healthcare plan and the data-mining companies look at this as non-adherence. Good statisticians can factor these data issues things into their models by giving these factors less weight, but right now, it seems the patient annoyance factor may be due to rise based on mis-information from data-mining companies, with no effort to address these issues.

If you are accused by your insurance company of being non-adherent for a medicine that sells as a low-cost generic, you should tell them you got the medicine for $4 and their models probably do not consider the fact that the pharmacies don't submit a claim for those sales, therefore they may be entitled to a refund from the PBM and/or data-mining company that sold them the costly model, but that they can stop harassing you for factually-inaccurate non-adherence.