Patient Outcomes of Flexible Assertive Community Treatment Compared With Assertive Community Treatment

We conducted a quasi-experimental controlled study comparing functioning and satisfaction and the working alliance between patients treated by ACT teams that had been reconfigured to FACT teams and patients who remained in ACT teams. The study was embedded in a broader evaluation of FACT in the Capital Region of Denmark (10, 11). We did not assign patients to treatment. Instead, patients were allocated to treatment on the basis of administrative considerations, making the study a natural experiment (15). Patients from nine urban ACT teams in the Capital Region of Denmark were included. The FACT and ACT groups were similar in socioeconomic distribution and catchment area. Six of the nine ACT teams were reconfigured to FACT teams in 2018 and 2019 by integrating ACT teams and CMHTs. This reconfiguration affected all patients treated by the ACT teams and CMHTs, so patients were not individually selected for transfer to FACT. The patients of the remaining three ACT teams constituted the control group. The baseline assessments took place in 2018 and 2019 before the six ACT teams were reconfigured into FACT teams, and follow-up assessments took place approximately 1 year after. The ACT teams followed the nationally recognized ACT model (7), and the FACT teams worked according to the Dutch FACT manual (8).

We retrieved a list from the Danish National Patient Register of a sample of patients who received treatment in the nine ACT teams at baseline (before FACT was implemented) (16). Using this list, we asked the case managers in these teams to inform the patients about the study and invite them to participate. We tried to avoid bias by recruiting only the patients on the list rather than recruiting volunteers or patients chosen by the case managers. However, if a case manager did not find it suitable to invite a patient, that patient was excluded. Reasons for exclusion by case managers included concerns about the patient’s capacity to provide valid informed consent or meaningful answers, concerns that the patient was too unwell to participate, or lack of contact of the ACT team with the patient.

If the patient agreed to participate, we scheduled an assessment without the case manager. However, in some circumstances, the case manager was present at the patient’s request or for personal safety (e.g., if the patient was court ordered to receive treatment). Most participants preferred to be interviewed at home. Other participants were interviewed at the team office. Because of COVID-19 restrictions, some follow-up interviews were conducted via telephone (N=15).

The primary outcome was the difference in personal and social functioning between FACT- and ACT-assigned patients as measured with the Personal and Social Performance (PSP) scale. The PSP is observer rated and a valid and reliable scale for measuring social functioning of patients with schizophrenia (17, 18). The PSP scale measures functioning in the month preceding the interview in four domains: socially useful activities, personal and social relationships, self-care, and disturbing and aggressive behaviors (19). Possible scores range from 1 to 100, with higher scores indicating better functioning; ratings are based on the patients’ verbal reports of their daily living and clinical observations (17). The assessments were conducted by researchers who were trained in using the PSP scale.

We also measured the working alliance with the Working Alliance Inventory (WAI). The WAI is a self-report questionnaire that measures different aspects of the professional relationship between patient and health care provider (20). The WAI incorporates questions about communication, trust, and agreement on goals and tasks between the patient and the health care provider. In this study, we used the WAI short form with 12 items rated by the patients (WAI-C) (21); possible scores range from 12 to 84, with higher scores indicating a better working alliance. Finally, client satisfaction was measured with the self-report Client Satisfaction Questionnaire–8 (CSQ-8); possible scores range from 8 to 32, with higher scores indicating higher client satisfaction (22, 23).

The sample size calculation was based on the PSP scale to estimate treatment effects, a priori defined to be the between-group difference in scores after 1 year. A Cohen’s d of 0.7 was considered a clinically relevant difference on this scale, assuming a difference of seven points between the groups and a pooled standard deviation of 10 points (17). With a two-sided alpha of 0.05 and statistical power of 80%, 33 participants in each group were predicted to be required to detect a clinically relevant Cohen’s d effect size. However, in a nonrandomized study, the sample size must be increased to account for covariance adjustment (24). We therefore chose a more conservative effect size for the sample size calculation, that is, a Cohen’s d of 0.5, assuming a difference of five points between the groups and a pooled standard deviation of 10 points. With a two-sided alpha of 0.05 and statistical power of 80%, we anticipated that 64 participants in each group were required to detect a clinically relevant Cohen’s d effect size.

We used an intention-to-treat approach in which patients were analyzed in the treatment group to which they were allocated through the recruitment procedure, and all patients discharged from the teams were retained in the analysis. We did not use imputation methods to deal with missing data because a large proportion of participants was lost to follow-up (25). Differences in baseline characteristics between the two treatment groups were assessed with chi-square tests for categorical variables, t tests for normally distributed continuous variables, and two-sample Wilcoxon rank sum tests for nonnormally distributed continuous variables. Differences between the FACT and ACT groups in mean overall functioning, client satisfaction, and working alliance at follow-up were analyzed by using analysis of covariance adjusted for baseline differences between the two groups. Analyses were done in R, version 3.6.1.

Independent researchers conducted all assessments, and participants were informed that individual scores and comments were anonymous and would not be shared with their clinicians. All participants gave written informed consent. The Danish Data Protection Agency approved the project through the Capital Region of Denmark (approval no. RHP-2017-006). The Danish Patient Safety Authority deemed our protocol to be exempt from formal approval because it was regarded as a quality assurance project (no. 3-3013-2207/1).

The baseline characteristics of FACT and ACT patients were similar in most areas (Table 2). The two groups significantly differed in functioning (FACT group mean±SD PSP scale score=36.8±7.6; ACT group PSP scale score=43.3±13.1, p=0.001) and substance use (23% for the FACT group vs. 40% for the ACT group, p=0.051) at baseline. We adjusted the results at follow-up for baseline PSP ratings and substance use to account for these potential confounding covariates. The median time from ACT at baseline to FACT reconfiguration was 163.5 days (range 30–517 days). In cases in which the reconfiguration was delayed, this delay was reflected in the time to the follow-up assessment, ensuring time for any changes to occur; thus, the median time from reconfiguration to follow-up for the FACT group was 297 days (range 207–643 days).

Of the 131 participants assessed at baseline, 71 consented to participate in the 1-year follow-up assessment. We found no significant differences between those who participated at baseline and those who participated at follow-up with regard to the measured baseline characteristics (age; diagnosis; sex; substance use; employment; PSP, CSQ, and WAI scores; number of bed-days; and number of outpatient contacts in the year before baseline).

At the 1-year follow-up, 33 of the 71 patients who agreed to participate had been discharged from the teams (19 FACT patients and 14 ACT patients). PSP ratings indicated a higher level of functioning in the ACT group (Table 3). After adjusting for PSP ratings and substance use at baseline, we found that ACT patients still had on average a higher level of functioning than those in the FACT group (adjusted difference in means=−2.9, 95% CI=−5.8 to −0.1). However, no statistically significant differences in client satisfaction or the working alliance were found between the two groups.

We found a statistically significant difference in patient functioning between the FACT and ACT groups. However, the literature indicates that a difference of 2.9 points on a scale that runs to 100 may be too low to be considered clinically meaningful (typically defined as a difference of seven points between groups) (17, 18).

Our study makes an important contribution to the limited research on the effectiveness of FACT compared with ACT. Research on the effect of FACT has provided mixed results and has mainly focused on mental health care use (10, 11). A U.K. pre-post study reported that patients who transferred from ACT to FACT had fewer admissions, inpatient days, and outpatient contacts (11). These results could not be replicated in a Danish controlled study (10), which found no differences in inpatient days between FACT and ACT. The Danish study also demonstrated that FACT was likely to be a more intensive service in terms of increased outpatient contacts compared with ACT. When we consider the results of this study together with the few previous studies on mental health care use, whether FACT is a beneficial alternative to ACT remains uncertain. Our findings emphasize the importance of promoting interventions for the group of patients previously targeted by ACT teams.

This study has several implications for future research. First and foremost, future studies should aim to evaluate the FACT model through randomized controlled trials. Moreover, the literature on FACT has not assessed whether this model effectively sustains contact with patients. Extensive studies have shown that ACT is better than standard care at maintaining patient contact (3–6). We strongly recommend that future studies include measures of engagement with services in the evaluation of FACT compared with ACT. Finally, future studies on functional outcomes should also include a comparison of FACT with CMHTs.

To our knowledge, this is the first study using a control group to compare FACT with ACT on patient functioning, the working alliance, and client satisfaction. We thereby overcame the main limitation of previous studies that lacked control group comparisons. The PSP assessments were conducted by independent researchers who were not involved in the treatment of patients. This study feature is a strength because the researchers could not influence the amount of support given to the study participants.

Our study also had several limitations. First, randomization to FACT or ACT was not feasible. Unknown or unmeasured confounders could have therefore biased the results. Second, the interviews were carried out by researchers who were not blind to treatment condition. This lack of blinding may have influenced the researchers’ reporting of the results. Third, before the recruitment of participants, we calculated the sample size but did not recruit enough participants to the control group. This underrecruitment may have resulted in a lack of statistical power and a greater probability of a type II error. Fourth, a substantial drop-out rate reduced the sample size available for analyses at the 1-year follow-up, which may have caused a biased estimate of the parameters and weakened the generalizability of the findings. No differential attrition was found between the two groups, but we cannot exclude the possibility that participants and nonparticipants may have differed on unobserved data. However, we could compare the two groups on register-based outcomes, which were available for all participants regardless of attrition status. Fifth, the two groups differed in PSP scores at baseline. However, we adjusted for this difference in the analyses. Finally, we did not conduct fidelity assessments, so we could not assess the extent to which the teams in our study were practicing according to their defined models of care.

The FACT teams in this study were in the start-up phase, whereas the ACT teams were well established. This differential between the two groups could have resulted in two types of bias. It could have led to improved effectiveness of FACT because of the pioneer effect, meaning that a newly established team may have more engaged and motivated staff, or it could have reduced effectiveness because the FACT teams had limited experience with the FACT model compared with ACT. Moreover, some patients changed case managers during the transfer to FACT, whereas ACT case managers stayed the same throughout and could have benefited from previous alliances with patients.