Enhancing Mental Health Care by Scientifically Matching Patients to Providers' Strengths

Patient-Centered Outcomes Research Institute, University at Albany, Psychological and Behavioral Consultants, Outcome Referrals

Research has shown that mental health care (MHC) providers differ significantly in their ability to help patients. In addition, providers demonstrate different patterns of effectiveness across symptom and functioning domains. For example, some providers are reliably effective in treating numerous patients and problem domains, others are reliably effective in some domains (e.g., depression, substance abuse) yet appear to struggle in others (e.g., anxiety, social functioning), and some are reliably ineffective, or even harmful, across patients and domains. Knowledge of these provider differences is based largely on patient-reported outcomes collected in routine MHC settings. Unfortunately, provider performance information is not systematically used to refer or assign a particular patient to a scientifically based best-matched provider. MHC systems continue to rely on random or purely pragmatic case assignment and referral, which significantly "waters down" the odds of a patient being assigned/referred to a high performing provider in the patient's area(s) of need, and increases the risk of being assigned/referred to a provider who may have a track record of ineffectiveness. This research aims to solve the existing non-patient-centered provider-matching problem. Specifically, the investigators aim to demonstrate the comparative effectiveness of a scientifically-based patient-provider match system compared to status quo pragmatic case assignment. The investigators expect in the scientific match group significantly better treatment outcomes (e.g., symptoms, quality of life) and higher patient satisfaction with treatment. The investigators also expect to demonstrate feasibility of implementing a scientific match process in a community MHC system and broad dissemination of the easily replicated scientific match technology in diverse health care settings. The importance of this work for patients cannot be understated. Far too many patients struggle to find the right provider, which unnecessarily prolongs suffering and promotes health care system inefficiency. A scientific match system based on routine outcome data uses patient-generated information to direct this patient to this provider in this setting. In addition, when based on multidimensional assessment, it allows a wide variety of patient-centered outcomes to be represented (e.g., symptom domains, functioning domains, quality of life).

Background and Significance: Mental illness is an extraordinary and highly burdensome public health problem. Unfortunately, even for individuals who access mental health care (MHC), the care is too often substandard. Research has consistently demonstrated that approximately 10-15% of patients will deteriorate or experience harm during treatment. Further, when these rates are combined with no-change rates, only 40% or less of patients meaningfully recover. Importantly, treatment research has illuminated substantial variability in providers' outcomes. Simply put, the MHC provider impacts treatment outcomes, and stakeholders lack systematic access to valid and actionable information to optimize effective patient-provider matches. Without collecting and disseminating performance data, stakeholders lack vital information on which to base health care choices and personalize treatment. Conversely, there potentially is immense advantage to matching patients to providers based on scientific outcome data. Patients, stakeholders, researchers, and clinicians have all endorsed such applied knowledge transfer as a high priority. In response, the investigators have developed and piloted a technology to test this match concept and patient-centered health model. Prominent health care agencies have placed outcome/performance measurement at the center of core initiatives. The Institute of Medicine specifically recommends integrating provider performance data in treatment decision-making. Despite this rhetoric, 2 Cochrane Reviews combined could only identify 4 studies that addressed this question with minimal methodology standards; the results were mixed. Importantly, none involved a targeted dissemination intervention, and none involved MHC. Previous research, including our own, has empirically demonstrated substantial differences in projected treatment effect sizes depending on to which therapist a patient is referred. The key evidence gap is the need for a rigorous test of the effectiveness of a targeted MHC provider-performance dissemination intervention compared to standard/pragmatic referral and case assignment. Relatedly, the Patient-Centered Outcomes Research Institute (PCORI) has called for increased "precision" or "personalized" treatment, with a focus on tailoring. The match algorithm responds directly to this high priority call to customize care in a personal and evidence-based way. Specific Aims: The aim of this comparative effectiveness research (CER) is to test an innovative, scientifically informed patient-therapist referral match algorithm based on MHC provider outcome data. The investigators will employ a randomized controlled trial (RCT) to compare the match algorithm with the commonplace pragmatic referral matching (based on provider availability, convenience, or self-reported specialty). Psychosocial treatment itself will remain naturalistically administered by varied providers (e.g., psychologists, social workers) to patients with complex mental health concerns within a partner clinic network, Psychological and Behavioral Consultants (PsychBC). The investigators hypothesize that the scientific match group will outperform the pragmatic match group in decreasing patient symptoms and treatment dropout, and in promoting patient functional outcomes, outcome expectations, and care satisfaction, as well as patient-therapist alliance quality. Doing so will establish the match algorithm as a mechanism of effective patient-centered MHC. Methods: The investigators will compare the effectiveness of naturalistic MHC either with or without the scientific matching aid with a double blind, individual level RCT. The investigators will first conduct a baseline assessment of PsychBC therapists' (target enrollment N=44) performance (across at least 15 cases) to determine their strengths in treating 12 behavioral health domains measured by the primary outcome tool on which our match algorithm is based -- the Treatment Outcome Package (TOP). The TOP is already administered routinely in our partner network. Based on years of predictive analytic research, this tool classifies therapists as "effective," "neutral," or "ineffective/harmful" for each TOP domain. The blinded therapists will be crossed over conditions. Next, for the trial, new adult outpatients (target enrollment N=281) will be randomly assigned to the Match condition or case assignment as usual (typically based on pragmatic considerations, such as provider availability, convenience, or self-reported specialty). The only patient exclusion criterion will be people who are not the primary decision-maker for their care. Thus, patients will present with a multitude of problems across a spectrum of diagnoses. With therapist assignment as the only manipulation, participating therapists will treat patients fully naturalistically. Treatment outcomes will be assessed regularly through mutual termination or up to 16 weeks. Primary analyses will involve hierarchical linear modeling to examine comparative rates and patterns of change on the outcomes.

Randomly assigned, by a case-assigning administrator, to naturalistic treatment with a pragmatically matched provider (control group)

Randomly assigned, by a case-assigning administrator, to naturalistic treatment with a scientifically matched provider (experimental group)

We have developed an innovative, personalized Match System based on provider track records determined with a multidimensional outcomes tool - the Treatment Outcome Package (TOP). Specifically, patients are assigned to therapists with previously established strengths (i.e., being historically effective) in treating their primary problems (e.g., depression, anxiety).

Average Z-Scores for the Treatment Outcome Package-Clinical Scales (TOP-CS; Kraus, Seligman, & Jordan, 2005)

The TOP-Clinical Scales consist of 58 items assessing 12 symptom and functional domains (risk-adjusted for case mix variables assessed via 37 items on the companion TOP-Case Mix form, such as divorce, job loss, comorbidity): work functioning, sexual functioning, social conflict, depression, panic/somatic anxiety, psychosis, suicidal ideation, violence, mania, sleep, substance abuse, and quality of life. Global symptom severity was assessed by averaging the z-scores (i.e., standard deviation units relative to the general population mean) across the 12 clinical scales. Higher scores indicate greater impairment. Given that we examined change over the entire treatment period for this outcome (in a longitudinal hierarchical linear model), we provide the average mean and standard deviation for the TOP-CS z-scores across all measurement occasions.

Global psychological distress was assessed with the Symptom Checklist-10 (SCL-10; Rosen, Drescher, Moos, & Gusman, 1999), a 10-item, well validated and widely used self-report inventory that assesses psychological well-being. Total scores can range from 0 to 40, with higher scores indicating greater distress. Given that we examined change over the entire treatment period for this outcome (in a longitudinal hierarchical linear model), we provide the average mean and standard deviation for the SCL-10 total score across all measurement occasions.

Working Alliance Inventory-Short Form, Patient Version (WAI-SF-P; Tracey, & Kokotovic, 1989) Total Score

The WAI is the most widely used alliance measure, assessing patient-therapist agreement on the goals and tasks of treatment, and the quality of their relational bond. This 12-item short form assesses these dimensions from the patient's perspective, with higher scores indicating a more positive relationship (theoretical range = 12 to 84). Given that we examined change over the entire treatment period for this outcome (in a longitudinal hierarchical linear model), we provide the average mean and standard deviation for the WAI total score across all measurement occasions.

Outcome Expectation (OE) Subscale of the Credibility/Expectancy Scale (CEQ; Devilly, & Borkovec, 2000)

The OE subscale of the CEQ is the most widely used and psychometrically sound measure of patients' expectations for the personal efficacy of treatment. The three OE items range from 1-9 or 0-100% (in 10 percentage point increments), with higher ratings indicating greater expectation for improvement. Given that the OE CEQ items are assessed on different scales, we re-scaled the items to the same metric before creating a total score (theoretical range = 3 to 27). Given that we examined change over the entire treatment period for this outcome (in a longitudinal hierarchical linear model), we provide the average mean and standard deviation for the OE subscale across all measurement occasions.

Domain-Specific Impairment on the Most Elevated Domain of the Treatment Outcome Package-Clinical Scales (TOP-CS)

The TOP-CS consists of 58 items assessing 12 symptom and functional domains (risk-adjusted for case mix variables assessed via 37 items on the companion TOP-Case Mix form, such as divorce, job loss, comorbidity): work functioning, sexual functioning, social conflict, depression, panic/somatic anxiety, psychosis, suicidal ideation, violence, mania, sleep, substance abuse, and quality of life. Domain-specific impairment reflects each patient's scores on their most elevated problem domain (i.e., the domain most elevated at baseline). These scores were standardized z-scores (i.e., standard deviation units relative to the general population mean), with higher scores indicating greater impairment. Given that we examined change over the treatment period for this outcome (hierarchical linear model), we provide the average mean and standard deviation for the most elevated TOP domain across all measurement occasions. Note that this measure was positively skewed so we log-transformed it.

Early treatment discontinuation was operationalized as a patient discontinuing treatment after 2 or fewer sessions, whereas early continuation was operationalized as attending 3 or more treatment sessions. For analyses, early treatment discontinuation was coded 1 and early continuation was coded 0.

The Overall Provider Quality subscale of the TOP Satisfaction Scale assesses the extent to which patients are satisfied with their mental health care provider. This subscale reflects the average of 4 items, with higher scores indicating greater satisfaction (theoretical range = 1 to 6).

Assessed after 16 weeks of treatment or at the point of naturalistic treatment termination, whichever comes sooner

Inclusion: Being 18-70 years of age Being the primary, informed decision-maker for one's care Willingness to be randomized to condition and to complete a few study-specific measures Exclusion: Being under 18 or over 70 years of age Not being responsible for one's own treatment decisions Unwillingness to be randomized to condition and/or to complete a few study-specific measures

Upon request on a case-by-case basis, we will share all IPD that underlie results in a publication (e.g., data extraction for meta-analyses). As appropriate, based on specific scientific journals' standards, we will provide analytic code used to generate published results.

We will begin sharing IPD upon request 6 months following the publication of our peer-reviewed final research report for PCORI, and we will maintain our de-identified data set indefinitely given possible future requests (e.g., to extract data for meta-analyses).

We will share fully de-identified IPD, a corresponding data dictionary, and the above supporting information to professional scientific colleagues (with established reputations for publishing original psychotherapy research in accordance with ethical guidelines) for the purposes of replication efforts or meta-analyses. For other usages, such as requests to conduct novel secondary analyses, we will consider such collaborations on a case-by-case basis and clearly determine authorship arrangements in writing. Review requests, including a detailed plan for how the data will be used, should be directed via email to the PI, Dr. Michael Constantino (mconstantino@psych.umass.edu). Dr. Constantino reserves the right to deny individual share requests if the usage details are questionable or unclear.

Boswell JF, Constantino MJ, Coyne AE, Kraus DR. For whom does a match matter most? Patient-level moderators of evidence-based patient-therapist matching. J Consult Clin Psychol. 2022 Jan;90(1):61-74. doi: 10.1037/ccp0000644. Epub 2021 Jun 10.

Constantino MJ, Boswell JF, Coyne AE, Swales TP, Kraus DR. Effect of Matching Therapists to Patients vs Assignment as Usual on Adult Psychotherapy Outcomes: A Randomized Clinical Trial. JAMA Psychiatry. 2021 Sep 1;78(9):960-969. doi: 10.1001/jamapsychiatry.2021.1221.