This chapter summarizes the discussions that took place at the end of the day, with the intent of synthesizing key themes and identifying the most promising approaches that meet SAMHSA’s data collection goals of measuring recovery from substance use and mental disorders. Wilson Compton (National Institutes of Health) began by saying that at the beginning of the day he saw recovery as having, at the minimum, two components: one related to self-identity and the other related to behaviors or other external characteristics. The workshop discussions expanded his view and highlighted the need for multidimensional measures and measures that can capture the process of recovery.
Compton said that he is struck by the fact that measuring recovery seems more applicable to a treatment context than to a cross-sectional population survey context. In particular, he commented that he is not sure how measuring recovery could fit in the framework of the National Survey on Drug Use and Health (NSDUH). A survey such as the NSDUH seems to be the only type of vehicle that is suitable for producing national estimates, but it is not clear how data of defensible quality could be produced without imposing a major burden on respondents. Perhaps asking a small number of questions and providing a disclaimer that this is the best that one can do in a cross-sectional population survey is the only option.
In addition, if there is interest in measuring the process of recovery, Compton argued, there would be tremendous advantages associated with a longitudinal design. He added that some longitudinal studies exist and
some were discussed throughout the day, but those are not large-scale, national surveys that would be suitable to address SAMHSA’s current goals.
Compton said that perhaps the most realistic approach may be simple measures, such as whether the person once had a problem but no longer does or just asking people whether they consider themselves to be in recovery, and, if applicable, how long the person has been abstinent. The discussion of the quality-of-life measures was interesting, he noted, and measures of that type could also be asked in a cross-sectional population survey. Compton said that he was intrigued by the positive mental health concept proposed by Corey Keyes (see Chapter 5) and will want to learn more about the measures and whether they are ready to be used in a large-scale population survey. He noted that it is clear that the concept of flourishing is somewhat distinct from the symptom and problem-based approach and could enrich the data.
Overall, Compton said that he was concerned that a prevalence estimate of recovery could be highly dependent on how the questions are asked: consequently, perhaps the best strategy would be to ask it in a number of different ways and then analyze the data. Some of the questions that have been used in the past to generate prevalence estimates could be included to provide additional context.
Compton said that one of the outstanding questions is whether recovery needs to be measured annually. He argued that the methods discussed seem to imply that collecting data on this topic every few years would be sufficient. It is also not likely that the prevalence of recovery would change substantially from year to year. If so, Compton said, it might be possible to combine measuring recovery with a design that has several alternating modules on topics that do not change rapidly.
Sherry Glied (New York University) noted that she started off with the same general idea as Compton that there are two main aspects of recovery: the identity aspect and the symptomatology or remission aspect. The workshop discussions made her think that there are several key categories of concepts that are relevant. One of the relevant categories is the service provider role, and whether service providers have characteristics that facilitate recovery. This is an important question and therefore important to measure in some way. She argued that provider-oriented measures could be aligned with patient-centered care measures developed for other conditions.
In terms of the person-oriented measures, Glied said that it is obvious that identity is important, and one might want to just ask people whether they identify as being in recovery. Another concept that seemed to surface as key is remission. Remission and recovery are not the same thing,
although there are overlaps in the two concepts. It appears, Glied noted, that work is needed to improve methods for asking about remission.
Another category of concepts includes such elements as housing, work, education, quality of life, impairments associated with the condition, and the extent to which a person is able to surmount those impairments. From a policy perspective, she argued, learning that there are many people who report that they are in recovery but have a terrible housing situation would be very useful information.
A final category of concepts relevant to recovery based on Glied’s grouping consists of items such as self-efficacy, flourishing, and engagement. She argued that these are the types of questions that would be useful to ask of everyone, not only of people in recovery from substance use or mental disorder. A case could even be made for wanting to collect these data in a way that is not explicitly in the context of recovery.
Glied commented that it is clear that a large volume of research is being carried out on recovery in small, focused studies. The question, she said, is what are the advantages of a national population survey on this topic. One advantage is that a representative survey can produce a prevalence estimate, which can be used in a variety of different ways, including for advocacy purposes. Furthermore, the data can be used to conduct subgroup analyses, such as by cohort, race, ethnicity, income, and education, which may not be possible with smaller, more targeted studies. Finally, using national data may make it possible to have consistent comparisons across conditions, even across different mental health conditions, across different substance use conditions, or across mental health and substance use conditions, as well as conditions in other areas.
Glied argued that increased consistency in the measures used across national surveys would make them more robust. Consequently, it would be worthwhile to review the metrics that have been used in various national surveys and map them onto the concepts of interest to SAMHSA. It would also be useful to think about which surveys ask key relevant questions that could perhaps enable SAMHSA to add a one-time question about whether someone considers herself or himself to be in recovery and then continue to monitor changes in relevant areas without necessarily having to ask the question again.
As others have noted, Glied said, she also believes that it would be useful to do some longitudinal studies of highly targeted samples. However, a longitudinal study would be much more expensive than a cross-sectional design. She said that she does not like the idea of special studies, such as a follow-up study, because these types of efforts are often only funded once and then disappear. If a question is added to an existing survey, it is likely to continue to remain on the survey.
Kim Mueser (Boston University) followed up on Glied’s comment
that some aspects associated with recovery, such as well-being and positive psychological health, could be useful to measure outside the context of recovery. He agreed with Glied that this would be useful, but he emphasized that some subjective aspects of recovery are really defined with respect to how one perceives a particular illness that they have experienced. For example, items such as empowerment, mental health recovery, and proactive coping, may really only be appropriate for people with an identified mental health condition because it has to do with their current position relative to that disorder. Glied asked whether, if that is the case, a national population survey is the right vehicle for detailed, specific questions. Mueser replied that targeted follow-ups of people in treatment may be a more appropriate context for those types of questions.
Mueser added that if the goal is to measure recovery in a general population survey, it is important to note that in the context of substance use, the concept of recovery resonates with a broader population than just those who are receiving treatment for substance use disorder. However, this is not necessarily the case in the context of mental illness, where the term recovery is not typically used outside of the treatment system.
Christine Grella (University of California, Los Angeles) noted that an argument for doing a nationally representative survey, in addition to the small-scale targeted studies that are more frequent in the field of recovery research, is that most smaller studies suffer from a selection bias. For example, study participants are typically recruited through recovery organizations or from clinical settings. She argued that collecting recovery data from a general population sample could mean tapping into populations that might surprise researchers. In addition, national data on a much broader range of people than before could inform policies and efforts to promote overall population health and counteract the negative images that are so prevalent about mental illness and substance use.
Hortensia Amaro (University of Southern California) said that national data could help lift the stigma associated with substance use and mental disorders and could raise support for funding dedicated to these areas. Corey Keyes (Emory University) commented that state-level data would be useful—particularly the rates of those who have recovered and those who were in recovery and relapsed—in the context of the state prevalence rates of mental illness, serious mental illness, and substance abuse. The availability of this information could help with holding states accountable and also with showcasing states that are leaders in supporting recovery.
Alexandre Laudet (Center for the Study of Addictions and Recovery, National Development and Research Institutes, director emeritus) agreed with Compton that it is not necessary to collect data annually, but she argued that it would be useful to do it at least every 5 years. She said that
it appears that there are two broad reasons for needing a prevalence estimate. First, SAMHSA provides block grants for the treatment of substance use and mental disorders, and the agency needs to know how many people may need recovery support services. Second, it is important to collect national data on people in recovery from substance use or mental illness because these populations and processes are poorly understood. For example, many of the current addiction-related laws stand in the way of people with substance use disorders getting their lives back together. It would be very useful to know if advocacy efforts and a potential decrease in stigma would prompt more people to pursue recovery. Successful efforts to promote parity could be another development that could lead to more people accessing care, which in turn would be important to monitor.
Graham Kalton (Westat) proposed a modification to the suggestion that it may not be necessary to collect the data every year. He noted that it is possible that 1 year’s data would not provide large enough sample sizes to be able to do subgroup analyses by different types of substance use or mental illness diagnoses. To avoid this potential problem, SAMHSA could plan on collecting data on recovery for 2 or 3 consecutive years to accumulate enough cases for such subgroup analyses and then pause for a few years.
Dean Kilpatrick (Medical University of South Carolina) reminded the group that SAMHSA’s overall goal is to expand the collection of behavioral health data in several areas that include, in addition to recovery, specific mental illness diagnoses with functional impairment, serious emotional disturbance in children, and trauma. He wondered whether measuring these areas in isolation would be more difficult than designing a study that integrates all of these topics and measures them adequately. Based on the workshop discussions, he said, it seems that integrating a broad range of behavioral health measures into one survey could have major advantages and produce some very rich data. Such a new survey could perhaps be fielded every few years and maybe fielding the NSDUH could be paused for those years. Kilpatrick acknowledged that this suggestion may not be acceptable for NSDUH stakeholders.
Michael Dennis (Chestnut Health Systems) commented that taking a major national survey offline for several years would be extremely costly due to the expenses associated with shutdown and startup. The costs of the gaps could exceed the costs of fielding the survey every year. However, a feasible alternative might be to collect data from half of the sample in 1 year and another half of the sample in another year. If the data are only needed every 5 years, they could be collected from 20 percent of the sample each year, instead of collecting data from everyone every 5 years. Dennis also agreed with Glied’s point that taking a survey offline could mean that it would stay offline, due to loss of funding in the interim.
Neil Russell (SAMHSA) said that he appreciated the input that collecting data every year may not be necessary and that the potential strategy of conducting one-off follow-up studies might be associated with a higher risk of a study being discontinued. He commented that it was good to learn that measuring recovery from substance use and mental disorder is a task that can be accomplished in some form. It is also useful to understand that most of the data that now exist on recovery are from specific populations and subgroups, and there are no nationally representative data from the general population that address the goals that SAMHSA has set forth for this effort. He also appreciated the perspectives on the importance of prevalence data on recovery from substance use and mental disorders.