The provider survey begins with a set of screening questions to identify providers eligible for the survey. To be eligible, providers must be an X-waivered or opioid treatment provider practicing in Arizona, who have prescribed MOUD since receiving the waiver, and have prescribed buprenorphine or methadone between March of 2020 and time of survey. Eligible survey respondents were then asked a series of questions about the types of MOUD treatment offered by their practice, their practice setting, number of MOUD patients, and how their practice responded to the pandemic. Respondents were then asked to identify which of the ten accommodations permitted by federal regulatory agencies were offered to their MOUD patients before March 2020, during Arizona’s “Stay-Home /Shut Down” period March-May 2020, and “currently” (at the time of survey completion). We also asked whether any additional services were provided to MAT patients during these time periods. Respondents were given the option to identify up to two additional accommodations offered to MOUD patients over these three time periods. Figure 1 displays the proportions of surveyed providers who reported implementing each of the 10 new accommodations (or additional services) during the three time periods.
Respondents were then asked whether all MOUD patients were eligible for these services and, if not, to explain eligibility conditions. Respondents were also asked why they did not implement some or any accommodations (as appropriate). This was followed by a series of questions evaluating the respondent’s perception of the importance, efficacy, risks, and difficulties involved in implementing these specific accommodations. Finally, respondents were asked a set of questions measuring stigma associated with MOUD treatment and patients, and a number of questions about the demographics of the respondent.
Development of the MAPI2
The MOUD Access Policy Implementation Index is a designed to be a flexible tool that can be used to measure different dimensions of MOUD accommodation implementation. Here we describe three operationalizations of the MAPI2 that are appropriate for addressing different types of research questions. The first is the unadjusted MAPI2 which is simply the sum of the number of accommodations implemented in each time period by each provider. Figure 2. displays the mean of this unadjusted MAPI2 over our three time periods. On average Arizona MOUD providers who responded to our survey reported providing 1.2 accommodations prior to the pandemic, 3.2 implemented during the COVID shutdown period, and 3.1 at the time of survey completion. These data indicate that the average number of accommodations increased during the COVID shutdown period, relative to pre-COVID, and remained at a similar average level of implementation at the time of survey completion.
Generally, when building a case for a new index, a standard step is to examine the internal validity/consistency of the summary measure (Singleton & Straits, 2018). One way to do this is to examine the degree of correspondence between the responses to the various component questions to make the case that these measures are indeed correlated with one another. Cronbach’s alpha is a widely used measure of inter-item covariance, and the Kuder-Richardson coefficient of reliability is the equivalent of the Cronbach’s alpha for dichotomous variables (Jaison, 2018). The Kuder-Richardson coefficient of reliability for the 10 dichotomous variables, indicating whether each of the 10 new accommodations were implemented or not, produces the following scores: pre-COVID .67, COVID-shutdown .75, time of survey .72. A score in the .7-.8 range is widely considered good/acceptable, while a score between .6 and .7 is more questionable.
However, in this case the purpose of the index is to describe the extent to which this specific set of newly allowed accommodations were implemented. Given this descriptive goal, measures of inter-item covariance are less useful in this context. As opposed to recommended practice in other applications, we would not want to remove infrequently implemented accommodations to improve an internal reliability metric, because it is the implementation of the accommodations themselves that we are measuring with this index, not some underlying characteristic of providers captured by the component items in the index. A low score on the Kuder-Richardson coefficient of reliability would simply mean that there were sets of accommodations that some providers implemented that were not strongly related to their likelihood to implement other types of accommodations. Instead, we find a modestly strong level of inter-item covariance which suggests that providers likely to implement one accommodation were likely to implement others.
The unadjusted MAPI2 provides clear answers to specific questions, such as, how many of the ten possible allowed accommodations were implemented by providers? However, from a patient perspective whether a provider has audio-only telehealth versus audio-video telehealth may be unimportant. A provider could choose not to implement audio-video telehealth because they are focusing on audio-only telehealth which could be viewed as more accessible for their patients. Similarly, a provider may rely on home MOUD delivery, and not implement the use of lock-box delivery or third party MOUD delivery, because they already have a home delivery practice. There are 4 items related to telehealth options, and three related to home delivery of MAT. However, from a patient perspective often what matters is whether a provider has telehealth options or a home delivery as an option or not. Further, it matters to patients whether they are eligible for these services or not.
The adjusted MAPI2 collapses these two sets of overlapping accommodations, weights for patient eligibility, and scales the implemented count of accommodations to the total possible number of accommodations. Specifically, the adjusted MAPI2 is comprised of 7 components, dichotomous indicators of the implementation of: 1) telehealth (audio-only, audio-video or for use by behavioral health support groups); 2) buprenorphine induction via telehealth; 3) increased take-home (multiday) dosing beyond the CFR requirements; 4) license reciprocity to allow for out-of-state dispensing; 5) home delivery of medications (at-home provision of methadone or buprenorphine, use of a lock box for methadone dispensing at home, etc); 6) off-site dispensing (non-pharmacy); 7) any additional services. This count was then weighted by patient eligibility for these accommodations. Providers were asked whether all patients were eligible for the accommodations that they reported implementing. Of the 74 surveyed providers, 26% did not implement any accommodations, 61% reported that all patients were eligible for these accommodations, and 14% reported that not all patients were eligible. If they answered that not all patients were eligible, providers were then asked to explain their eligibility criteria. The most common answers provided were that “stable” patients and existing patients, as opposed to new patients, were eligible for these accommodations. Providers that answered that all patients were eligible for accommodations received an eligibility weight value of 1, while providers that made distinctions between patients (such as stable and unstable or between new and existing patients) received a weight value of .75. Stable vs. unstable rating for a patient is relevant because the regulatory flexibility allowed up to 14 or 28 days of multiday dosing (take home dosing) for unstable or stable patients respectively (SAMHSA, 2020)
Adjusted MAPI 2 = ((telehealth + home delivery + bup induction via telehealth + take-home dosing + license reciprocity + off-site dispensing + additional services)*Eligibility weight)/6
After weighting this score was scaled to a percent out of 6, which is both the total count of possible newly allowed accommodations and the maximum value found in these data. Expressed as a percentage of possible accommodations (roughly weighted by eligibility criteria), this adjusted version of the MAPI2 attempts to operationalize provider implementation behavior with an emphasis on impacts for patients’ access to MOUD treatment. This version of the index is preferable for analyses testing links between provider behavior (in regards to COVID-era accommodations) and patient experiences or community-level impacts.
Finally, we developed a third iteration of this index, the relevant MAPI2. This version of the index is informed by the insight that not all of the newly allowed accommodations are relevant for all providers. Relevance is dependent primarily on the specific type of medicated treatment offered, with the most important distinction being whether or not a provider offers methadone as a treatment. All of the new accommodations are relevant for providers who offer methadone. The relevant MAPI2 score was scaled as a proportion of these 6 total accommodation options for providers offering methadone. For providers not offering methadone their provider-level score is scaled to the 3 accommodations relevant for their patients (telehealth, buprenorphine induction via telehealth, and license reciprocity).
Relevant MAPI 2 for methadone providers = (telehealth + home delivery + bup induction via telehealth + take-home dosing + license reciprocity + off-site dispensing + additional services)/6
Relevant MAPI 2 for non-methadone MOUD providers = (telehealth + bup induction via telehealth + license reciprocity + additional services)/3
This version of the MAPI2 is particularly appropriate for analyses concerned with understanding factors shaping degree of implementation at the provider level.