Implementing the Cultural Formulation Interview in a Community Clinic to Improve Appointment Retention: A Pilot Study

When this project began in 2012, FHMC employed 30 mental health providers. Of 1,500 active patients, approximately 30% identified as non-Latinx White, 30% as Latinx, 32% as non-Latinx Asian, and 8% as non-Latinx Black. In 2014, the first author (N.K.A.) presented the CFI during grand rounds. FHMC’s administrators agreed to allow studies on implementing the CFI in its outpatient psychiatric clinic after agreeing that the CFI fulfilled hospital requirements for staff cultural competence.

The first author conducted separate focus groups with patients, clinicians, and administrators, soliciting feedback on the 16-item CFI, hospital cultural competence policies, and implementation techniques. Two barriers to implementation were identified in all groups: the time needed for the CFI and its lack of instructions on incorporating cultural information into diagnostic and treatment plans. At FHMC, all patients attend a 60-minute intake session with a social worker (session 1), a 45-minute appointment with a psychiatrist (session 2), and a 60-minute psychotherapy session (session 3) within 1 month. Clinicians agreed to implement the CFI after administrators’ assurances that a CFI training session would fulfill continuing education requirements and that all sessions were billable. Patients and clinicians wanted patient preferences from the intake CFI to be recorded for follow-up sessions.

The intake session with the CFI (session 1) and two follow-ups (sessions 2 and 3) were used to introduce a new intervention called the CFI–Engagement Aid (CFI-EA) to improve appointment adherence. The CFI-EA addresses critiques that the CFI collects cultural information without instructions for including it in treatment. The CFI-EA also includes instructions for clinicians on customizing treatments by charting patient preferences for other clinicians to view. With these processes in place, cultural assessment became a method to deliver person-centered care through treatment negotiation. (For a sample outline of the CFI-EA, see the online supplement to this column.)

The CFI-EA addressed themes from patient and clinician focus groups. Patients wanted time to ask questions and negotiate treatment duration, whereas clinicians wanted clarity about treatment goals and adherence. In session 1, clinicians conducted the CFI before an intake history. The clinicians explained the treatment plan and requested feedback. In session 2, clinicians asked, What problems are you concerned about from the treatment I’m recommending? How long would you like treatment? How often would you like to come? What changes will tell you whether the treatment has worked? In session 3, clinicians asked, How has treatment been going so far? What, if anything, has gotten in the way, and how have you adhered to treatment? What has worked well and what more can be done? In each session, clinicians were asked to incorporate patient preferences into treatment plans and discuss differences nonjudgmentally. They communicated treatment preferences through the patient’s chart.

A prepilot open trial in which two social workers and a psychiatrist used the CFI-EA with nine patients was conducted through the NIH’s stage model for developing behavioral (nonpharmacological) interventions (8). A clinician manual specified how the CFI-EA could improve cultural competence, acceptable and unacceptable behaviors with patients, and desired outcomes. Every CFI session was audiotaped to assess fidelity; all 16 CFI questions were asked with six patients, and 15 questions were asked with three patients, with one question omitted because of redundancy.

Nine patients ages 20–67 years old (mean±SD=39.3±19.1) enrolled in the trial. Four were male and five were female. Three patients were non-Latinx Black, three were Latinx, two were non-Latinx White, and one was non-Latinx Asian. English (N=7) and Spanish (N=2) were their primary languages. Their sources of income were employment (N=5), disability insurance (N=2), family assistance (N=1), and pension (N=1). No patient earned more than $53,000 annually. Their primary intake diagnoses by class were depression (N=4), anxiety (N=3), bipolar disorder (N=1), and somatic symptom disorder (N=1). All patients completed sessions in English, the primary language of all participating clinicians, because DSM-5 does not include instructions on CFI use with interpreters. Eight patients completed all three CFI-EA sessions and stayed in treatment for 2 months after CFI-EA session 3. One dropped out because of limited finances. All patients and clinicians agreed in individual debriefing interviews that the CFI-EA could be used without revision in a pilot trial testing the CFI-EA against treatment as usual.

The pilot RCT lasted from April 2017 to September 2018. Patients were randomly assigned to the CFI-EA or control arm through a random number generator. On the basis of NIMHD’s recommendations (7), the trial aimed to enroll three clinicians and 18 new patients in each arm to examine feasibility (the ability to recruit patients and clinicians) and patient appointment retention at 3 months, that is, 2 months after CFI-EA session 3. FHMC’s clinic is referral based, and the intake coordinator presented the study to all patients upon scheduling their first session.

Patients who did not attend sessions were considered nonengaged. Patients who attended sessions and requested referrals because they were resettling or had substance use disorders that the clinic did not treat were considered engaged because they negotiated treatment with clinicians. Any clinician at FHMC who agreed to interview three patients with the CFI-EA or conduct treatment as usual was enrolled. There were no exclusion criteria for clinicians.

Patients who spoke English, underwent intake for any disorder, and were 18–80 years old could enroll. Exclusion criteria included emergencies, such as acute suicidality or homicidality, as well as dementia, mental retardation, and psychosis, which could interfere with responding to CFI-EA questions. All patients received $30 in compensation each session for completing study assessments.

CFI-EA clinicians were trained according to the DSM-5 CFI field trial protocol, which consisted of reading the core CFI, watching its video simulation, and practicing it through behavioral simulations (3). Clinicians read the CFI-EA manual for sessions 2 and 3. CFI-EA clinicians received written feedback after their first case to improve fidelity, defined as asking each CFI-EA question. Training lasted 3 hours. All appointments were in English.

Two social workers and one psychiatrist enrolled in each arm (N=6). Clinicians in the CFI-EA and control arms were different in order to avoid crossover effects, although prepilot CFI-EA clinicians could be pilot CFI-EA clinicians. Five clinicians identified as female. The ages of five clinicians ranged from 38 to 67 years (mean=51.4±10.9), and one did not give her age. Four were born abroad, and five named a language other than English as their primary language. They identified as non-Latinx White (N=2), Latinx (N=2), and South Asian (N=2). All six clinicians stayed in the trial until completion.

Eighteen patients enrolled in each arm (N=36) and signed informed consent forms in session 1. In the CFI-EA arm, 12 patients identified as female and six as male; 10 patients in the control arm identified as female and eight as male. Ages ranged from 18 to 56 years (CFI-EA, mean=29.6±11.5; control, mean=34.1±12.5). Four CFI-EA patients were born abroad compared with five patients in the control arm, one of whom listed Puerto Rico as her country of origin. Thirteen patients across both arms named English as their primary language. Patients identified as non-Latinx White (CFI-EA, N=1; control, N=1), Latinx (CFI-EA, N=9; control, N=7), non-Latinx Black (CFI-EA, N=6; control, N=6), non-Latinx Asian (CFI-EA, N=2; control, N=2), and multiracial (CFI-EA, N=0; control, N=2). Their sources of income were employment (CFI-EA, N=7; control, N=8), family assistance (CFI-EA, N=4; control, N=5), public assistance (CFI-EA, N=3; control, N=2), and disability insurance (CFI-EA, N=1; control, N=2); three patients in the CFI-EA arm and one in the control arm did not answer. Most patients did not report annual income; one patient in the CFI-EA arm earned $120,000, and one in the control arm earned $84,000, with all others earning ≤$60,000.

Patients’ primary intake diagnoses by class were depression (CFI-EA, N=7; control, N=8), anxiety (CFI-EA, N=6; control, N=4), and adjustment (CFI-EA, N=2; control, N=1). Also, two patients in the CFI-EA arm had bipolar disorder, and one had a psychotic disorder. In the control arm, two patients had attention-deficit hyperactivity disorder, one had a substance use disorder, and one had borderline personality disorder.

All patients attended session 1. Session 2 was attended by 89% (N=16) of individuals in the CFI-EA arm and 83% (N=15) of those in the control arm. This difference was statistically nonsignificant. Two patients in the CFI-EA arm and three in the control arm discontinued treatment and did not return for session 2 despite meeting inclusion criteria (Table 1).

At session 3, two patients in each arm became ineligible to participate after leaving New York City. CFI-EA clinicians referred two patients for substance abuse services on the basis of their treatment preferences. The eligible sample for session 3 was 25 patients (CFI-EA, N=12; control, N=13). One hundred percent (N=12) of eligible patients in the CFI-EA arm and 92% (N=12) in the control arm completed session 3, with one patient in the control arm discontinuing; this difference was statistically nonsignificant. Twelve patients in each arm attended their 2-month follow-up appointments.

This is the first study exploring CFI use in a community setting through an RCT. By discussing time and financial constraints with administrators, clinicians recommended reforms in hospital procedures to facilitate CFI implementation. Administrators responded by providing continuing education hours for CFI training and billing for CFI sessions. These steps helped ensure 100% clinician retention in both trials. Administrators should enlist clinicians before implementing the CFI. Clinics differ in specific procedures, but specifying common objectives across stakeholders is a reproducible goal.

In the pilot trial, two patients in the CFI-EA arm versus four in the control arm discontinued care by session 3, despite meeting inclusion criteria. This finding supports the feasibility of CFI-EA implementation.

Our work had several limitations. First, limited sample size prevented analyses of how patient and clinician characteristics affected appointment retention. Second, all patients and clinicians spoke English, although some patients had other primary languages, which could have affected their communication and satisfaction with care. Third, our study recruited at one site, potentially limiting generalizability, although our process of specifying consensus goals through focus groups helped our study fit real-world conditions. Fourth, clinicians self-selected into the CFI-EA arm, which could have introduced bias compared with randomization. Fifth, social workers conducted psychotherapy, and psychiatrists handled medication management to optimize reimbursements in split treatment. Studies are needed with variety in clinician disciplines and settings. Sixth, financial compensation may have incentivized attendance among patients in both arms, reducing the CFI-EA’s effect.

Nonetheless, our studies illuminate how to build stakeholder consensus to implement the CFI within time and financial constraints. The CFI-EA’s promising results among lower- and middle-income, racially and ethnically diverse patients offer a model for cultural competence without racial or ethnic patient-clinician matching. Future work can clarify whether improved appointment retention and treatment adherence through the CFI-EA decreases symptoms and improves quality of life to alleviate disparities.