Evaluation of a Cross-sectional Coordinated, Severity Stepped, Evidence-based Care Model for Mental Disorders

Participants will be recruited from regular care settings (in-patient and out-patient services, general practitioners, psychotherapists) and via the RECOVER homepage. Upon fulfillment of the inclusion criteria and after having obtained informed consent, participants will undergo a complex baseline psychiatric diagnostic assessment, including among other measures the global assessment of functioning (GAF), severity of illness (CGI-S) and specification of diagnosis (SCID I and SCID II). Based on these three instruments, participants will be divided into four degrees severity: Grade 1 = mild, Grade 2 = medium, Grade 3 = medium to high, Grade 4 = severe mental disorder. Accordingly, there will be four strata of similar sample sizes (planned n=268, including 30% drop out rate; see power analysis below). Each participant will then be randomized either to the RECOVER treatment condition or the treatment as usual (TAU) condition. Randomization will be conducted by an independent person not involved in the RCT or diagnostic assessment, who will receive the pseudonymised participant ID and stratum number. A previously with the STATA 14 program defined list will be used to achieve randomization. When the planned sample size per stratum is achieved, no further patients will be included in this stratum.

For a complete list of measures, please refer to outcome measures section. In addition, secondary data will be obtained from external data owners, essential for the achievement of RECOVER's research objective - the determination of health costs. These include personal patient-related data such as data from health insurances companies (e.g. data from inpatient and outpatient medical services, rehabilitation services, transport, medical aid, drug/ medication data, home assistance and care services, work disability and sickness benefit/ allowance data). These will be inquired from the health insurance for the years 2016 to 2021 at the 12-month point in time and continuously assessed within the quality assurance study. Subsequently, a monetary valuation is carried out using standardized monetary valuation rates.

Serious adverse events (SAE) will be documented from the first contact (i.e. registry) between patient and study staff and will be included in the participant's record. A copy of each SAE protocol will be directly forwarded to the study's central management. Each SAE is assessed in terms of its causal relationship between the event and the study procedures (""serious undesirable, study-related event"). For each SAE, the SAE protocol must be filled in (with pseudonymised data only) and will be immediately discussed with the study management, who will check the protocol for plausibility and completeness and evaluate the event. Events that have not yet been formally completed are tracked. If the undesired event has been resolved, a final documentation is provided on the SAE protocol. The executive committee is informed by the study management and asked for an assessment. If all assessments have shown that the suspected case is an SAE, all study staff will be informed by the study management immediately, but at the latest within 7 days of notification.

The sample size is based on a power calculation to detect a statistically significant difference between the intervention (RECOVER) vs. standard care (TAU) in order to yield a small to medium standardized mean difference between the groups (Cohen's f of 0.175) after 12 months. At a power of at least 80% with a type 1 error rate of 5%, 2-sided testing and 10% of declared variance due to baseline, at least 233 study participants are required. For testing hypotheses involving the interaction of intervention and stratification subgroup, sample size increases to n=383. Diagnostics and therapy are carried out in approx. 50 clusters with approx. 21 participants each and an intraclass correlation (ICC) of .05 is assumed for the primary, continuous outcome. This results in a design effect of approximately 2.0 and a total sample size of 890 patients (including 30% dropout). With an estimated participation rate of 50%, 2140 patients must be addressed for participation.

Statistical analysis will be supervised by the local Department of Medical Biometry and Epidemiology. The analysis of primary and secondary outcome data is performed according to the intention-to-treat (ITT) principle. Missing values for the follow-up time points will be accounted for within mixed linear and logistic regression models. For sensitivity analyses, missing values will be replaced by multiple imputation and per-protocol (PP) analyses. The analysis of therapy utilization and productivity losses is performed descriptively as well as with difference-in-difference regressions (to analyze the measured costs over time). In addition, multivariate analyses will be carried out. Remission and response data are analyzed using logistic regression, while remission (non-clinical, i.e., performance under critical cut-off) and response (i.e. reduction in symptom severity of 50%) are specifically determined for each psychiatric condition during follow-up as compared to the baseline measurement. Changes in symptoms, level of functioning and health associated quality of life are analyzed with difference-in-difference-regression. To estimate cost efficiency, incremental cost-efficiency-relation (ICER) is calculated. To estimate the ICER's uncertainty, cost-efficiency-acceptance-curves based on net-benefit-regressions are calculated. All secondary outcome parameters are descriptively investigated as well as using multivariate analyses (e.g. logistic regression, difference-in-difference-regression). To assess inhibiting and promoting factors for the RECOVER model, qualitative measures are applied.