Clients’ Preferences for Family Involvement and Subsequent Family Contact Patterns Within OnTrackNY Early Psychosis Services

We report on a prospective CSC cohort using administrative data from the OnTrackNY network, which encompasses 21 EIP programs in New York State (41, 44). Between October 2013 and December 2018, the network served approximately 1,350 individuals with FEP between the ages of 16 and 30 years with a recent-onset (<2 years) nonaffective psychotic disorder. Additional studies describe the development and implementation of the OnTrackNY network in detail (41, 42, 44). As noted above, the OnTrackNY model premises family involvement on the client preferences; at program entry, clients are asked about their preferences for family involvement, aided by several decision-making tools. Client-reported data are collected at admission and quarterly through standardized provider-reported admission, follow-up, and discharge data collection forms to improve care quality. All identifiable information is removed from the data sets, and the deidentified data are approved by the New York State Psychiatric Institute Institutional Review Board. This study included data from 761 participants who enrolled between August 2015 (when data collection for the main outcomes began) and December 2017, which allowed for each participant to be eligible for 1 year of follow-up.

The two target outcomes were client preferences regarding family involvement in their care, with three categorical options: no involvement, involvement with conditions (e.g., a client requests family involvement in some aspects of care but not others), or unconditional involvement. This information was collected at baseline. The second target outcome was family contact with the CSC team, defined as one or more contacts (including e-mail or phone exchanges and attending in person meetings or groups) between one or more family members and CSC team members during the preceding 3 months. This information was collected quarterly during follow-up.

Analyses also included sociodemographic variables, family support system variables (e.g., client’s contact with family, living situation, identified designated support person, and homelessness), functioning and clinical status (e.g., hospitalization, medication adherence, time to first mental health contact after symptom onset, family involvement in the pathway to care, and substance use). A comprehensive list of these variables and their respective categories is shown in Table 1. (Additional information on variable structure is included in an online supplement to this article.)

The standardized clinical measure used was the Mental Illness Research, Education, and Clinical Center Global Assessment of Functioning (MIRECC-GAF) scale, which includes three subscales: symptoms (MIRECC-GAF symptoms), social functioning (MIRECC-GAF SF), and occupational functioning (MIRECC-GAF OF). The MIRECC-GAF OF scale has 10 anchor points, reflecting average level of functioning in the person’s primary role as worker, student, or homemaker during the 30 days before assessment. The MIRECC-GAF SF is similarly structured and considers social interactions, relationship quality and quantity, and interpersonal conflicts in the previous 30 days. The MIRECC-GAF symptoms scale captures patients’ worst level of functioning during the previous 30 days on the basis of suicidality, mood, anxiety, and psychotic symptoms. Subscales are scored from 0 to 100, with higher scores indicating better functioning.

Analyses used data from a censored sample of OnTrackNY clients who were eligible to have received services for 12 months (N=761). To assess the associations between both primary outcomes (client’s family involvement preference at enrollment and family contact with teams over the initial 12 months of follow-up) and baseline variables, descriptive summaries of baseline measures were computed, stratified by baseline involvement preference (no involvement, conditional involvement, or unconditional involvement) and then by family contact pattern (always, mixed, never, or early discharge). The family contact patterns across 1 year of follow-up were characterized in the following way: “always” included those who had family contact in each of the quarterly data collection periods, “mixed” those who had contact in some periods and not in others, “never” those who never had contact, and “early discharge” included participants who disengaged from the program before completing 1 year. The descriptive summaries included proportions for categorical measures, means and SDs for normally distributed continuous measures, and medians and interquartile ranges for skewed continuous measures. Associations between groups and the baseline measures were assessed with chi-square tests, one-way analyses of variance, and Kruskal-Wallis tests, depending on the distribution of the baseline correlate.

All statistical tests were two-sided, with a significance level of 5%. Because of the novelty of the primary outcomes examined in this study and lack of previous research utilizing analogous data points, we conceptualized our analyses as exploratory and hypothesis generating, and we therefore did not employ tests for multiple comparisons. All analyses were run with SAS, version 9.4.

At enrollment, 59% (N=450) of the 761 participants requested unconditional family involvement, 35% (N=266) involvement with conditions, and 6% (N=45) no involvement. Baseline variables examined in relation to these involvement preferences are reported in Table 1. Significant differences were noted in baseline family involvement preferences for family and support system measures, medication adherence, substance use, employment, and participation in secondary-postsecondary education. Compared with clients requesting conditional or unconditional involvement, clients in the no-involvement group were more likely to be older and employed and to have graduated from college; they also were less likely to have a designated support person and more likely to be homeless, live alone, or live with nonfamily members. Family member involvement in pathways to care was significantly less likely in the no-involvement group than in the other groups, but the groups did not differ in the time from symptom onset to first mental health contact. In general, metrics for the conditional-involvement group fell in between those for the unconditional-involvement and no-involvement groups. In addition, family involvement preference was significantly associated with early discharge, with a higher proportion of early discharges in the no-involvement group. Previous hospitalizations and GAF scores were not significantly associated with family involvement preference.

In the study cohort, 67% (N=506) remained enrolled for 12 months. Rates of family contact with CSC teams significantly decreased over time: at 3, 6, 9, and 12 months, 84%, 78%, 74%, and 73%, respectively, of clients’ families had contact with the team (Cochran-Armitage z=4.90, N=761, p<0.001). Longitudinal family contact patterns were as follows: 42% of clients’ families had contact at least once every quarterly period (always, N=323), 3% had no contact at all follow-up points (never, N=22), 21% had a mixed pattern of family contact (mixed, N=161), and 34% disengaged from the program early (early discharge, N=255).

Group differences in baseline characteristics of families with three contact patterns and those with early discharge are reported in Table 2. Racial differences were marked: families of White (non-Hispanic) clients represented a third of those with continuous contact (always) and only 5% of those with no contact (never). Additional cross-tabulations showed that Black (non-Hispanic) race-ethnicity was significantly associated with insurance status (χ²=14.3, df=1, p=0.003) and criminal justice involvement (χ²=3.9, df=1, p=0.047). Age, gender, insurance status, current family contact, living situation, designation of a support person, homelessness, medication adherence, family involvement in pathways to care, and baseline employment all differed significantly across family contact patterns.

Clients with family members who were never in contact with the CSC team were more likely to be older, to be employed at baseline, and to have higher baseline occupational functioning than were clients whose family members had other contact patterns. Clients whose family members had continuous contact (always) were more likely to be younger, male, and living with their parents and to have higher symptom burden and markedly worse occupational functioning. Characteristics of those whose family members had a pattern of mixed contact generally fell in between those in the continuous-contact and no-contact groups, although they were typically much closer to the former.

A statistically significant association was detected between baseline client preferences and family contact patterns (Table 3). Nearly half of those specifying no involvement at baseline were discharged before 12 months. Conversely, families of clients who preferred unconditional involvement were most likely to have continuous contact (always) (48% vs. 37% and 18% in the conditional-involvement and no-involvement groups, respectively). However, for the conditional- and unconditional-involvement groups, the combined percentages of those whose families had continuous team contact and those whose families had mixed team contact were nearly equal (61% and 68%, respectively). Even among families of clients who had preferred unconditional involvement, a small proportion of families (1%) did not make contact at any point during the first year. Additionally, about a third (36%) of families of clients who expressed a preference for no involvement nevertheless had contact with the team during all or some quarterly assessment periods.

A major contribution of this study is the reporting of data concerning client preferences for family involvement in one of the largest implementations of CSC in the United States. At baseline, a substantial number of clients requested family involvement—full involvement or with some conditions. Of these clients, the percentages whose families had either continuous contact or mixed contact with the CSC team in the first year were very similar (61% and 68%, respectively), suggesting that centering client preferences does not adversely affect family contact with the CSC team and may in fact increase it. Ultimately, assessing the comparative effectiveness of different strategies to involve families in services will be of considerable clinical importance to the field. Although the OnTrackNY statistics appear promising, limitations of the data used in both this study and other reports, as well as differences in both timescale and constructs measured, preclude our making meaningful comparisons among outcomes. Future research efforts should better delineate the kind, type, and circumstances of family contact; the conditions requested by clients and their rationales; and the impact of clients’ preferences on clients’ and family members’ interactions with EIP and CSC services.

Racial-ethnic differences in family contact with services have been noted in the literature (38, 45–47). We found significant but attenuated differences in client preferences among racial-ethnic groups and more robust differences in actual family contact. Black participants were significantly more likely than those from other racial-ethnic groups to disengage from the program early, and their families were more likely to fall into the no-contact and mixed-contact groups, mirroring recent analyses of data from the RAISE NAVIGATE trial (38) Although these findings may be mediated or moderated by socioeconomic differences (48–50), they foreground the importance of meeting the needs of clients and families from minority groups, who are at elevated risk for disengagement and premature discharge.

It is unsurprising that CSC clients who displayed signs of greater independence at baseline, as indicated by older age, higher likelihood of being employed, greater educational attainment, and greater likelihood of living independently, were more likely to prefer no or conditional family involvement. Similarly, we would expect less need for family involvement among clients with fewer needs—that is, those with higher occupational and social functioning, higher rate of employment, greater educational attainment, and less severe symptoms. More surprisingly, however, preferences for less family involvement were also predicted by public insurance, potentially a proxy for lower socioeconomic status, and by race (being Black vs. White). These findings suggest hidden subgroups in this cohort—one potentially reflecting greater client independence and lower needs and another reflecting a preference for less family involvement because of the additional barriers that socioeconomic disadvantage (and its intersections with race and cultural stigma) may present (48–51). Future research may better identify these potential subgroups, because lack of family contact with services may reflect a logical outcome for low-needs clients, but it may also reflect evidence of disparities driven by socioeconomics or race for others.

The fact that nearly half of those who preferred no family involvement disengaged prior to 1 year may suggest that a preference for no family involvement is a red flag for potential disengagement; however, given the patterns described above, such a preference may also be driven by lower needs and higher functioning associated with “positive” early discharge resulting from a desire for a lower level of care (e.g., medication management only). A more nuanced understanding of variable lengths of engagement with EIP and CSC programs is important and will depend on identifying diverse latent groups and latent trajectories among clients.

A small subgroup of clients’ families had no contact with the CSC team; however, compared with family involvement in mainstream community mental health services (43, 52), family involvement across OnTrackNY was exceptionally high. Other CSC programs could learn from and replicate these practices. Although the data set for this study did not allow for investigation of preferences for family involvement in and family contact with group interventions, including the multifamily groups that are commonly implemented in EIP and CSC programs, such research could help explain how to best meet the needs of family members.

The observational nature of our data does not allow causal inferences. Furthermore, the measures studied likely interacted in multiple and complex ways; however, small bivariate category sizes in our sample precluded more complex modeling of these relationships. Potentially important metrics were not available in the data set, including more comprehensive indicators of family socioeconomic status and acculturation, as well as perceived family conflict (21, 23). The only available family contact metric in the data set (any contact in the previous quarter) was imprecise and did not allow us to examine diverging contact patterns or to distinguish between deep and sustained versus more superficial involvement.

The first-person experience of involvement, both from client and family member perspectives, was also absent from this study, and future research would benefit from strategic use of mixed methods to deepen our understanding of client and family involvement decisions and their associated impact. Finally, preferences are dynamic rather than static, and measurement of changes in preferences over time is important. This study focused on preferences expressed during enrollment, which is both a limitation and a subject worthy of additional investigation in future studies.