Sharpen the focus

Our discussions revealed the need to clarify and focus our efforts to achieve consensus. The following summarizes the questions raised, decisions made and the supporting rationale.

Which aspects of measurement-based care (MBC)?

Many people view MBC as entailing the regular use of simple clinical ratings and ordinary laboratory tests to accomplish a variety of clinical tasks including selecting a treatment, personalizing treatment dosing, minimizing side effects, and making timely changes in treatment in the face of inadequate response [63]. Typically, MBC methods are applied at the clinician-patient interface, but other stakeholders (e.g., administrators) may also use the same measures to manage program resource allocation. Researchers and regulators also want measures to assess the safety and efficacy of treatments and quality of care. For purposes of brief MBC in primary care or specialty care, in this workgroup we focused on measures to enhance dosage optimization and timely revision in treatments when outcome is suboptimal.

We found that we needed to clarify the conceptual distinctions between MBC approaches typically defined by and designed for various clinical tasks as noted above: treatment sequences—sometimes called algorithms—and disease management protocols.

Treatment algorithms define the principles, rationale, and evidence for which treatment to initiate and which to follow depending on therapeutic and adverse effects achieved with the prior treatment steps [64]. The algorithm may recommend one or a range of reasonable treatment options at each step. The combination of measurement of outcomes at each step and a systematic organization of the steps is associated with better outcomes and less expensive care in depression [64, 65] and other conditions such as panic disorder, generalized anxiety disorder, bipolar disorder, etc. [11, 14].

Disease management protocols are population-based efforts that have a broad set of components designed to improve chronic disease outcomes. They often recommend procedures to screen for, establish the history, and define a differential diagnosis. These protocols also typically recommend treatment options, ways to personalize treatment titration or delivery, recommend treatment algorithms, specify multi-dimensional approaches to enhancing symptom reduction, relapse prevention, functional restoration, and minimization of side-effect burden, as well as addressing issues of adherence and lifestyle changes. Disease management protocols often involve multiple treatment team members.

Our discussions led us to provide a context for MBC as shown in Fig. 1. However, MBC is only one ingredient within a complex array of factors that affect outcomes.

Fig. 1

(Copyright © 2018) [66]. American Psychiatric Association. All Rights Reserved

Factors affecting treatment outcomes. Adapted with permission from the American Journal of Psychiatry, Volume 175, Issue 12, “Improving Depression Outcome by Patient-Centered Medical Management,” A. John Rush and Michael E. Thase, p. 1188.

In addition to the host of parameters that affect treatment outcomes, Fig. 2 shows the range of important treatment outcomes of interest. Consensus formed around the idea of measuring those outcomes that were specific to OUD and that were clinically most relevant to assessing the effectiveness of the treatment and its rapid and comfortable dosing obtained at the patient-clinician interface in primary care or specialty care. We reached this consensus for two reasons. First, the effectiveness of MBC has been most robustly demonstrated when it has been used to adjust medication doses and to assess symptomatic treatment outcomes, for example in mental health arenas such as depression [11,12,13]. Second, other measures are likely needed to accomplish different clinical tasks or to assess whether other outcomes were achieved, such as predicting relapse.

Fig. 2

Core OUD signs and symptoms. Permission to adapt this figure [67] was obtained under the terms of the Creative Commons CC BY license

Phases of treatment

There was a consensus to recognize that roughly speaking the phases of OUD treatment approximate those outlined by Rush and Thase [66] for the patient-centered delivery of medications for depression (see Table 2). The measure being considered for medication dose personalization would typically be used early in treatment, whereas the measure to assess overall treatment outcome would be used somewhat later.

Table 2 Patient-Centered Treatment PhasesWhich disorders: OUD or SUD?

While we initially aimed at SUD, we found that an initial focus on medication treatment for OUD would likely be more productive because the types of treatment being dose adjusted will vary across different SUDs and because of greater availability of FDA-approved medications for OUD.

Core symptoms or associated symptoms?

We discussed whether to focus on DSM-5 core symptoms of OUD or to also include symptoms of commonly associated conditions, such as pain, sleep, anxiety, depression, etc. The focus on core OUD symptoms was chosen because the other conditions are likely more useful in tailoring other aspects of OUD care (e.g., counseling) and are often already measured in well-validated brief measures in primary care and specialty settings.

By focusing on core DSM-5 symptoms of OUD, we avoid the inadvertent measurement of associated symptoms that occur in the context of concurrent SUDs. We recognize that polysubstance use that involves misuse of other substances is common in those with OUD, which can adversely affect patient outcomes and undermine the sustained impact of OUD medications, even if opioid use is initially reduced by properly dosed medications for OUD. Given that broad assessment of these other factors is not compatible with a brief MBC measure, we urge providers to inquire about factors that may be undermining outcomes in situations where a medication dose that was working well is suddenly not preventing opioid use.

In support of using DSM-5 criteria for OUD to assess treatment outcomes, recovery subgroups from buprenorphine treatment at long-term follow-up assessment have been aligned with DSM-5 OUD severity criteria (past 3 months) and OUD outcomes [68]. Also, a systematic review and meta-analysis suggests the use of a measure of craving, one of the diagnostic criteria of DSM-5, to estimate risk of drug use or relapse for assessment and treatment [69]. Further investigation of DSM-5 core symptoms for assessment and treatment is still needed.

Which interventions?

We chose to focus initially on medication treatment because it can be provided in primary care and specialty care settings. Further, because pharmacological interventions need to be initiated and adjusted in ways that are distinct from psychosocial interventions, rather than attempting to develop scales that would be suitable for both types of interventions, a focus on psychopharmacological intervention seemed a feasible first step. Nevertheless, a measure of core symptoms, if agreed to, should be as useful in assessing the effects of psychosocial, medication and even device-based interventions on OUD.

Which clinical settings?

The intention was to identify measures that would help clinicians rapidly and accurately personalize the doses of OUD medications when initiating medications and assess treatment outcome suitable for use in both primary and specialty care settings by clinicians, patients, and families. In OTP settings where treatment adjustments are protocolized MBC may allow for better tailoring of dose adjustments to larger increments in dose at different frequencies depending on the individual patient.

We agreed that if more extensive/longer measures of outcomes were needed, say in specialized care settings, they could be added to the core outcome measure. A single measure facilitates patient management and communication among clinicians and across settings (e.g., programs that have expert primary care clinicians initiate medications and then hand them off to primary care providers when the patient is stable, or for “hub and spoke” OUD treatment systems [70]), as well as the compilation of programmatic outcomes by administrators. Such a universally agreed to measure could be part of the core outcome package for research studies of OUD treatments, facilitating the transfer of knowledge between researchers and clinicians.

Who would be using these measures?

We chose to focus primarily on measures for routine clinical care, as opposed to the research users because researchers could and often need to include longer or additional measures not easily administered in routine care. Patient self-report measures of substance use symptoms and brief screens for SUD have demonstrated validity in primary care even when results are documented in the EHR [23, 29, 30, 71]. Our concern was optimizing doses and obtaining accurate, practical, and timely assessment of treatment outcome, rather than achieving a complete understanding of all the obstacles to and opportunities for patient improvement.

Two measures or one?

We initially hoped to have a single measure that could be used for personalization of medication induction and dosing and as a reasonably precise measurement of overall OUD treatment effectiveness. During our discussions we realized that measures to accomplish these two clinical tasks (personalized medication dosing; assessment of treatment effectiveness) might both overlap in terms of some signs and symptoms and not overlap in terms of others.

Dose adjustments (which occur within 1–3 weeks of initiating medication) precede the assessment of overall treatment effectiveness. Treatment effectiveness may not be fully known for weeks to months. That is, these two tasks may be completed at different times. Secondly, parameters that best inform dose adjustment, may or may not be elements that are most useful in assessing treatment effectiveness. That is, both when to measure and what to measure to bring precision to each clinical task are likely to be distinct. This conclusion initially supported the notion of developing two metrics, one for each task.

What are the desirable features of a treatment outcome measure?

There was a consensus that having both self-report (for use in the clinic) and clinician-completed versions (for research and regulatory approval) of any measure would be very useful and that crosswalks should be created between such instruments. Self-reports can promote patient and family participation in treatment and can be used outside the treatment site (e.g., by smart phone). Clinician ratings may bring greater between-patient reliability. In addition, clinician ratings can be compared with self-reports, especially if the items are identical. We recognize that the accuracy of both self-reports and clinician ratings (both of which typically rely on patient reporting) is highly influenced by patient recollection and forthrightness.

There was a consensus that both dose adjustment and outcome measures must be brief—preferably 5–6 and no more than 9 items to facilitate ease of use and portability to smart phones. A global (i.e., brief, complete) rating such as the widely used pain scale or the Clinical Global Impression of Severity measure [72] may suffice in many instances.

Is measurement-based care for OUD more effective than treatment as usual?

Logically, more precise measurement to personalize medication doses or to establish treatment effectiveness, might bring greater precision to clinical decision-making in the acute management of patients with OUD and greater consistency in their care amongst practices and clinicians. This in turn could improve outcomes (e.g., greater engagement and retention in treatment, lower amounts, and less frequent opioid use, better daily function and QoL, less frequent severe or prolonged relapses, etc.). However, there is yet no randomized controlled trial evidence that MBC approaches when applied to OUD medication management are more effective than treatment as usual. There is evidence from a meta-analysis that higher buprenorphine dose is associated with improved retention in buprenorphine maintenance treatment [73]. There was consensus that if these measures could be agreed to, the impact of MBC in OUD patients must be assessed to determine whether there is any advantage to this approach in patients with OUD given opioid agonists or other treatments.

Domains for dose adjustment and assessment of effectiveness

This section details our discussion about the elements/parameters or domains that seem to be essential to accomplishing each task: (1) OUD medication dose adjustment for each patient (personalized dosing) and (2) assessment of acute treatment effectiveness on the core signs and symptoms of OUD.

Medication dose adjustment

MBC when applied to dose personalization requires that both side-effect burden and the therapeutic effect be assessed to identify the optimal cost benefit balance for each patient. Personalized dosing may be required because some medications need to be given at higher doses to some patients and at lower doses to others due to pharmacokinetic and pharmacodynamic variability across patients and differing tolerance because of previous drug usage.

The following five parameters were considered sufficient measures for medication dose adjustment: withdrawal symptoms, opioid use, magnitude (severity and duration) of the subjective effects when opioids are used, craving, and side effects. The consensus was for a brief rating of the 4 OUD-specific parameters (either by a few separate items or a global overall assessment) to adjust dosing, and a global assessment or verbal question for side effects. In addition, adherence to the prescribed medication must be checked at each patient-clinician interaction, potentially informing switching to a different medication or formulation (e.g., methadone, long-acting buprenorphine, etc.). It is expected that the psychosocial consequences such as failed role obligations, opioid use despite social problems, and the discontinuation of important activities may NOT change as substantially/meaningfully during this initiation period of 1–3 weeks [74, 75] as during the ensuing weeks to months.

Assessment of treatment effectiveness

Given the prior work done by Marsden and colleagues [16] (“Marsden measure”) and Bradley and colleagues [47], and the usual approaches to assessing treatment outcomes in other psychiatric and general medical conditions, we focused our initial discussions on the criterion signs and symptoms that define OUD based on DSM-5-TR [26]. Table 3 synopsizes the Marsden measure [16] which recommended 5 conceptually separate domains that were adapted from the 11 DSM-5-TR criteria for OUD [26].

Table 3 Domains adapted from DSM-5 Substance Use Disorder Criteria

The first four domains (A–D) of the Marsden measure [16] are specific to OUD because opioids are specified in the individual question items, although should be generalizable to similar addicting drugs: the frequency type and amount of use of an opioid (A); the physiological manifestations of tolerance or dose reduction of an opioid (B); the subjective experiences in terms of desire, control, and urges/craving for these kinds of substances (C); and the use of opioids in dangerous circumstances (D). We recognize that domain D, while specific to opioids, will be variably applicable to patients depending on their type of use and social circumstance.

Domain E of the Marsden measure [16] about negative social consequences, broadly includes daily function and QoL and is not specific to OUD. Negative social consequences are applicable to any SUD and indeed, most other neuropsychiatric and many general medical conditions. Further, negative social consequences can be caused by other concomitant SUDs and general medical and psychiatric conditions (e.g., chronic depression, pain, sleep wake disorders, etc.). An additional reason for separating the measurement of core OUD symptoms (Domains A–C) and function/QoL (Domain E) is that the latter is very likely to improve after the signs and the symptoms (A–C) of OUD are brought under some degree of control for a meaningful period, as suggested by some studies [74, 75].

Separating the assessment of function and QoL (negative social consequences), as in the Marsden measure, would allow for the use of any of a host of brief, psychometrically sound scales that are widely accepted assessments of function and quality of life, though less often applied to SUDs or OUD to date. The workgroup consensus was that measurement of function and quality of life (which would include negative social consequences) separately from the core OUD signs and symptoms per se would have some major advantages.

Bradley and colleagues’ preliminary work for the More Individualized Care: Assessment and Recovery Through Engagement (MI-CARE) trial [76] illustrates the application of these principles in their adaptation of the PROMIS Item Bank v1.0—Short form 7a substance use measure [44, 47] and a psychometrically validated substance use symptom checklist [30]. The OUM was adapted from the substance use PROMIS measure by substituting “opioids” for “drugs.” The baseline timeframe was 3 months but the time frame for follow-up was the past 2 weeks to support measurement-based care [47]. Seven items of the OUM are each rated on a 0–4-point scale using the PROMIS response options (“never” to “almost always”; see Table 4) yielding a total score of 0–28. This scale assesses a range of outcomes. Six of the seven items assess the signs and symptoms of OUD, while only one assesses the effect of the OUD on interpersonal relations. Some items (e.g., “I have an opioid problem”) may need to be revised or potentially deleted if not responsive to change. This measure is expected to reflect the overall severity of the OUD [44] and initial results in a sample of 49 primary care patients made it appear feasible with variability in responses [47].

Table 4 PROMIS-based Opioid Use Monitor (OUM)Summary and next stepsMedication dose personalization

A consensus was developed indicating that a brief rating—either a global rating or a few separate questions that assess the severity of withdrawal symptoms, craving, amounts of opioids used, and subjective response to use of opioids during treatment and a single item global rating of side effects are sufficient to adjust the dose of agonists to each patient.

Whether the use of such a 4-element tool or a global rating to personalize dose adjustment of agonists in the treatment of OUD produces better retention, good patient or clinician satisfaction, and/or different “final” doses deserves study as compared to treatment as usual without any measures.

While no specific brief measure incorporating all four of these elements was identified or recommended, as an example the PROMIS-based 7-item OUM used in the MI-CARE trial described above to assess symptom severity and support MBC includes questions pertaining to each of the recommended dose-personalization elements except for withdrawal [47]. Further, the OUM includes 3 craving questions, which are used as a separate craving score (0–12) in the registry of the trial. The usefulness of the OUM craving questions suggests that an off-the-shelf craving tool could also be considered for the dose adjustment portion of MBC but would require research to determine if it is sufficient. The possibility of adapting an existing tool or developing a new tool with the recommended elements for medication dose personalization should be considered.

Assessment of therapeutic effects

There was consensus that core signs and symptoms of OUD based on some of the 5 DSM-5 domains in the Marsden measure [16], should be the basis for assessing treatment outcome. A select list (e.g., 5–7 items) of these core symptoms and signs rated over the prior 1–2 weeks and used after a clinically reasonable period on a treatment (e.g., 1–2 months) would determine the effect of the treatment on the condition (OUD), such that a clinical decision could be informed by the outcomes obtained (e.g., whether to change the treatment, augment it, etc.).

Domains A and B (see Table 3) of the five DSM-5 domains from the Marsden measure [16] also pertain to dose adjustment. Further, the OUM, discussed above for dosing adjustment, is also appropriate for assessing treatment outcomes, so it is possible that adaptations of a single measure could serve both functions, although maybe not optimally, or that there be two separate optimal measures.

Other considerations and next steps

There was consensus that measurement of function and quality of life (including negative social consequences) should proceed independently with brief validated measures that are widely used in mental health and general medical conditions. The performance of these measures in patients with OUD deserves study.

There was agreement that if more than one outcome measure was recommended or developed, these measures should be assessed and when possible, crosswalks established between them in terms of total severity. Each measure should be assessed in different populations and age groups, but we should avoid tailoring different instruments to different sociodemographic or clinical subgroups. Nevertheless, care should be taken in the measure development and validation processes to consider and assess potential disparities in measure performance in different patient subgroups such as race, ethnicity, age, chronicity, culture and lived experience.

Next steps could be to select and adapt or develop de novo brief measures for OUD treatment medication dose personalization and treatment effectiveness assessment along the lines previously discussed that include the consensus recommended domains and the signs and symptoms of OUD. Whether the use of these measures in clinical care of patients with OUD produces better outcomes (e.g., retention, less drug use, better symptom control, better QoL and daily function or better prognosis) as compared to treatment as usual without measurements deserves investigation. None of the existing measures could be fully recommended to accomplish these aims for OUD because of lack of clinical evidence, absence of certain essential (i.e., recommended) domains, and/or excessive length and complexity. However, the PROMIS-based measure currently in use for OUD (i.e., the OUM) [47] could be considered for adaptation. Alternatively, a brief measure for dose personalization could be developed de novo, or the evaluation of the clinical impact by a single-question global scale that encompasses the 4 suggested parameters could determine whether MBC has clinical utility in arriving at the personalized dose in an expeditious fashion. Once the measures to personalize dosing and assess treatment effectiveness are developed or agreed to by experts, patient acceptability and psychometric studies are also needed.