• No results found

Total Potential Office Based Bupernorphine Treatment Capacity

3.4 Model Testing .1 Calibration .1 Calibration

3.4.2 Face Validation

I reached out to all expert panel members via email for face validation interviews, received four responses, and conducted interviews with two panel

members: Dr. Alane O’Connor, and Dr. Andrew Saxon. Interviews were two hours long, and covered boundary, basis, and representation assessment (see Section 2.4.1), as well as critical model assumptions, policies, and future conditions that should be considered in policy analysis.

At the outset of the interview, I introduced Drs. O’Connor and Saxon to the research questions, to the basic structure of the spatial ABS capacity model, and explained that the purpose of the interview was to check assumptions about model scope (boundary assessment), assumptions that informed the use of data (basis

assessment), and model logic that drives key dynamics (representation assessment), and to share early findings prior to final external validation.

In general, the experts were comfortable with the boundary of the model, agreeing that a model that is narrowly investigating questions of capacity might not need rich details on patient experience, for example. They were both interested in the spatial dimension of the model since they both had experience working with rural communities. As to basis assessment, they seemed comfortable with populating a real

map with people with OUD and providers placed plausibly according to population density and the NSDUH survey. However, this may have stemmed more from a lack of familiarity with ABS methods and a reluctance to “red flag” something unfamiliar. They agreed that there was a large variance in the patient loads of providers, which was reported by Stein, et al (2016) and suggested several reasons that some providers choose to serve different numbers of patients. Dr. O’Connor’s practice experience did not conform to the data used for patient retention, which was several years old. In her experience, by the time patients are admitted to her program, they are heavily invested in treatment and are better retained, and suggested that retention in the model may be overly conservative.

Regarding key model logic, both agreed that diversion was likely primarily not for recreational purposes but for self-treatment, or to maintain active opioid use when preferred drugs might not be available. This supports model assumptions about reasons for diversion, which has implications for diversion outcomes of policies that expand access. They also agreed that many people self-treat with buprenorphine while waiting to start formal treatment, sometimes for years. On the other hand, both also

questioned the validity of the assumption that people with OUD select providers on the SAMSHA searchable list. Dr. O’Connor asserted that word of mouth was a much more likely source of information on potential providers, especially in rural areas.

Interviews concluded with open questions about the impact of policies

implemented since modeling began in 2014, including the patient limit expansion policy

and NP/PA prescribing policy, and what unforeseen circumstances may have had outsized impact on their practices and patients. Dr. O’Connor had already obtained her DATA 2000 waiver by the time the interview took place. The model was originally intended to test very modest penetration of NP and PA prescribing, at the 3-5% level, mirroring the uptake among primary care physicians. Dr. O’Connor expects much higher uptake by NPs, and suggested testing the impact of uptake among 50-70% of primary care NPs in states with high NP autonomy. She also expects that initially patients will be shifted from overloaded specialty providers to less distant NPs resulting in a reallocation of treatment without apparent expansion until those specialist spots are backfilled and word of mouth brings new patients to NP prescribers.

Both providers were not surprised that early model experiments with the patient limit policy didn’t produce large gains in the number of people who receive

buprenorphine in a given year, though Dr. Saxon was dismayed by it. In Dr. O’Connor’s experience, while some providers, such as those who run intensive outpatient recovery programs, may have expanded their practices dramatically after increasing their patient limit, many may have elected to expand to 275 patients as a safety valve just in case they approach the 100 patient limit. Dr. Saxon also explained that even addiction medicine specialists may be reluctant to build their practices around buprenorphine prescribing, preferring a more diverse patient mix in part because practice barriers such as high paperwork burden and low reimbursement remain.

I closed by asking what surprised them in 2014-2017 and what to look out for when modeling future policy. The emergence of synthetic fentanyl as a street drug was completely unanticipated and resulted in increased opioid deaths. State policies

squeezing down on prescription opioid diversion and the resultant uptick in heroin use wasn’t totally unanticipated, but it did drive up overdoses and would not have been captured in a policy model that doesn’t differentiate between opioids, including this one. I was made aware of a major change in insurance that could have a large impact on access to treatment. A person with an extremely high deductible insurance plan may be insured on paper, but that insurance wouldn’t pay for treatment, leaving that person effectively uninsured and paying out of pocket, making treatment unaffordable.