The study used a quasi-experimental design, with the intervention implemented in a community mental health clinic and compared prospectively with routine care in the remaining community mental health clinics within a large urban public behavioral health system. Criteria for admission to behavioral health care clinics include a diagnosis of SMI; concomitant substance use disorder, co-morbid medical problem, or cognitive impairment; frequent use of acute psychiatric care services and / or involvement of the legal system; and circumstances that justify the need for intensive case management (eg, homelessness, aggressive behavior, adhesion difficulties). The intervention clinic was chosen on the basis of strong pre-existing relationships with the research team, an investment in improving primary care by clinical leadership, and a high percentage of patients with complex psychosocial and medical needs. The intervention clinic provides outpatient care, case management and support services to approximately 600 adults insured with SMI each year.
The routine care group continued to receive regular care in their mental health clinics and could also receive care from primary care providers in other clinics, without explicit integration of services. While providers of different specialties had access to common electronic health records, patients in routine care seeking non-psychiatric medical care had to go to separate clinics and laboratories for screening without additional support.
Participants were adults (aged 18 years or older) diagnosed with SMI and continuously enrolled in the public behavioral health system from January 2014 to December 2015. The exclusion criteria included being in prison and / or in a short-term care facility closed during the study period. This study received Institutional Review Board (IRB) approval from the University of California, San Francisco. The IRB approved a consent waiver for the clinic’s patients because the recruitment procedures involved routine review of medical records, did not adversely affect participants’ rights and well-being, and posed minimal risk to subjects and their privacy. .
The development of CRANIO has been previously described . In short, the study researchers used the Behavioral Change Wheel model, a scientific implementation framework that identifies goals for behavioral intervention and has been shown to be effective in SMI patients. . Feedback from focus groups of psychiatrists and patients informed the design and implementation of the intervention, [25, 27] which consisted of the following four components:
Patient-centered and team-based care
CRANIUM used pre-existing resources in specialized mental health clinics (the psychiatrist and case manager) and added a primary care counselor to 0.1 FTE. Rather than co-locating primary care providers in a federally qualified health center as has been attempted previously , the primary care provider has been integrated as an electronic consultant (eConsultant), available to answer questions (e.g., taking medication, connecting to primary care services). A peer navigator was also integrated into the team and prepared the laboratory cards, accompanied the selected patients to the laboratory facilities and entered the results into the electronic medical record (EHR).
Management of the panel with patient records
The CRANIUM registry included test results from three separate EHRs operating throughout the health system: EHR for Mental Health (AVATAR), EHR for Primary Care (Invision), and Laboratory Results (LabCorp Beacon). Each month, research staff extracted data from these electronic health records on patients who had treatment plans due and compiled the information into a single separate electronic database. The information was distributed to psychiatrists and case managers via a personalized spreadsheet. Lab sheets were pre-filled for all identified patients and provided to the psychiatrist. The staff and research team met quarterly to conduct panel management and troubleshooting for those in need of screening or treatment.
Training courses and protocols for psychiatrists on metabolic screening and HIV testing
Psychiatrists were trained to order an annual screening for hypertension, A1c, total cholesterol, high-density lipoprotein (HDL), and low-density lipoprotein (LDL). Since people with SMI are also at an increased risk of HIV but have low test rates, [12, 28] we have included the annual HIV test. Lab sheets were pre-filled for all identified patients and provided to the psychiatrist. To promote consistent screening, psychiatrists received a personalized monthly registry-based list of patients with no labs or vital signs. For those patients who lacked laboratories, based on individual needs and preferences, a peer navigator was available to accompany patients with missing laboratories to the appropriate laboratory facilities, and an on-site nurse was available to design the laboratories.
Training and protocols for the treatment of diabetes, hypertension and dyslipidemia
To mitigate concerns previously reported by psychiatrists in prescribing non-psychotropic drugs,  The primary care consultant provided all psychiatrists with a one-time group training course on drug management recommended by the guidelines for common metabolic abnormalities. This training is now available through SMI Adviser . In addition, easy-to-use, evidence-based drug algorithms were available in all treatment rooms and online. Patients with cardiovascular risk factors were highlighted to facilitate treatment discussions during panel management meetings.
The CRANIUM surgery was delivered in 2015 over twelve months at a modest cost: $ 74 per patient per year. 
Results of the study
This report focuses only on the screening results. The feasibility results have been published elsewhere . The main outcomes of interest were individual screening for diabetes (A1c), fasting lipids, and HIV testing at least once in the year before and after surgery. We did not include annual blood pressure measurement as a result due to a lack of reliable data (between arms, 72% missing or unmeasured outcomes in 2015). Inconsistent information about participants’ previous diagnoses of diabetes, hypercholesterolemia, and HIV precluded our ability to exclude people with these pre-existing conditions from analysis. Demographic data (age, sex, race / ethnicity) were also collected.
Chi-squared tests and t-tests were used to compare baseline demographics between study arms. Since abandonment information was only available from the intervention site and not from usual care sites, we conducted an intention-to-treat analysis and included all participants before and after the intervention. However, we have provided information on abandonment in the Supplement. Mixed-effect logistic models were used with nested random effects by clinic and participant and including the interaction term for the intervention condition and time to estimate the differences between arms in change from pre-post-intervention to A1c and lipid screening. and HIV test rates, first without adjustment, then adjusted for age, sex, race and ethnicity. A double face P. value <0.05 was considered statistically significant. The planned sample of 5,000 is estimated to provide 80% power in 2-tailed tests to detect differences of 5-6 percentage points between arms in variation in screening and test rates. The estimates were obtained by approximating the covariance matrix of the coefficients estimated by the random effects logistic model used for the analysis and implemented in R (R Foundation for Statistical Computing, Vienna, Austria, 2020). Analyzes were performed using Stata MP version 15.1 (College Station, Texas).