Since my initial blog entry on the subject, Part 1 - The Opiate Epidemic: Applying Better Matching and Analysis, I have received feedback that my position came across as a bit heavy-handed on enforcement and less focused on identifying abusers for treatment. I intended Part 1 to share the improved application of real-time hyper-accurate matching in PDMPs, and yes, it focused on enforcement while not applying enough to identifying those who need recovery support earlier and getting them into treatment sooner, where it has a higher probability of success.
Getting back to the matching part being a lynchpin in the strategy, even in getting abusers into recovery programs earlier, let's look at three different vectors enabling opiate abuse and see the role matching plays:
- Identify Abusers Earlier in Addiction Process to Drive More SuccessfulIntervention: This Vector is not an enforcement issue, and it improves the well-being of addicts by setting them on a path to recovery. Well-meaning practitioners need a PDMP solution that a) reduces or eliminates screen-facing time and increases patient-facing time, b) provides real-time information them from their current state accurately, but also see inter-state prescribing history as the patients will move naturally, and they may shop across state lines to avoid detection. Right now, states operate their Prescription Drug Monitoring Programs (PDMPs) with home-grown software or an off-the-shelf platform called Appriss. Depending upon the practitioner's approach to prescribing controlled substances, they are either checking their PDMP system manually before writing the prescription on a pad locally, or they are implementing ePresciption of Controlled Substances (EPCS) and may have either an integrated or non-integrated PDMP check. Every one of those state PDMPs use, at best, Probabilistic Matching. During the development of Referential Matching, it was easy to show that probabilistic matching has a false negative rate of 25-40% (the system fails to match). Almost all of the PDMP solution present a list of potential patient identity overlaps for which they have to take the time and identify the positive matches. The practitioner becomes the manual matcher overcoming probabilistic matching limitations! Further, the practitioner is only being presented with those identities that are probabilistically "close" - it doesn't present those that are more distant (e.g., marriage with a last name change and move, hispanic maternal-paternal last name errors, and simple keying errors) - a large part of the false negatives.
- Identify Individuals That Are Not Abusers, but Obtain Legally and Become a Distributor: For Vector 2, I need to return to the diagram in the Part 1 blog entry, "Sources of Prescription Pain Killers for Non-Medical Use" (I have re-inserted it above too). The source "Prescribed by ≥1 physicians", takes advantage of the right interventional care for addicts described in Vector 1. Two of the remaining three largest sources: "Given by a friend or relative for free" and "Bought from a friend or relative" are situations where the physician prescribes the opiate, but there is no contact with the real abuser. This occurs when patients who obtains the opiate, subsequently chooses to provide some portion of opiates to another individual. This could seemingly be a one-time event to help an acute pain, or this could be repeatedly handing over the medication. To address Vector 2 we would leverage a process of education and cultural change, by changing the behaviors of those prescribed such medication (keeping them from giving it to others). Additionally, practitioners and pharmacists are already improving the use of non-opioids or slowly reducing opioid dose when the pain should be abating. For more info, the initiative CVS recently announced reducing the strength of opiates and increasing the frequency of pharmacy visits.
- Identify Practitioners Who Participate in Distribution With or Without Knowledge: The last of the top four sources on the diagram, "Bought from a drug dealer or other stranger", is Vector 3. While not always the source of opiates for the dealer, practitioners may be prescribing to a person claiming false ID attributes to escape PDMP detection (e.g. dead persons, foreigners who have returned home, aged individuals who have moved from outpatient chronic care to inpatient care, etc.). The practitioner may be a willing or unwilling participant in the ruse, and we do not need to make doctors experts in identity fraud. In this case, Referential Matching wins again as it can identify a person using identity attributes from a dead or non-existent person.
To help drive this change, what motivations influence making use of the PDMP mandatory and what technical capabilities increase the efficacy of PDMPs?
- Practitioners need a PDMP that reduces, or eliminates, the time needed to check for potential abuse and increases the time directly caring for the patient.
- Use real-time Referential Matching to reduce false negatives and increase confidence in the true positive matches so practitioners don't have to do that work.
- Drive increased adoption of EPCS; while cumbersome to enroll, it allows for automated verification against PDMP and eventually improves the practitioner's time meeting chronic care prescription needs.
- Abusers will be more easily identified and targeted for recovery efforts earlier in their addition if we can cross state boundaries and see patient prescription histories nationally...its a basic health information sharing challenge.
- State laws govern the implementation of PDMPs, and rightfully, states are concerned about the privacy of their citizens. Again, improved matching helps eliminate the possibility of false positives to assure legislators that sensitive health information is not being associated with the wrong individuals.
- The PDMP process needs to help identify those individuals who are using identity theft and false claims of pain to build a supply for dealers.
- Referential matching provides the ability to identify potential false claims of identity by associating identity attributes with persons who could not possibly be at the practitioners office (because they are dead, left the country, or in chronic care elsewhere).
- Practitioners should be made aware of individuals on disability leave as part of their disability monitoring includes their continued need for medication that make it impossible to perform their jobs. Consequently, they report continued pain to remain on disability.
While I can be accused of having a tool (radically improved, more timely, patient identity matching) looking for problems to fix, the patient matching challenge is not a new concept. Since there has never been a better mechanism, the problems of probabilistic matching continue to hamper the healthcare industry. Our national legislators have become more aware of this challenge, and they know that new and better solutions exist. This has driven a clear demand for better matching recommendations from the Government Accountability Office (GAO) (see Senator Warren's letter).
Again, I remain interested in your thoughts and opinions.