Enhanced Child Psychiatry Access and Engagement via Integrated Care: A Collaborative Practice Model With Pediatrics

We hypothesized that providing highly integrated outpatient child psychiatry consultation to primary care pediatricians, combined with peer-to-peer parent support and outreach, would be associated with improved treatment access and engagement for at-risk youths compared with usual care. We tested this hypothesis by means of a quasi-experimental design in which youths meeting inclusion criteria were nonrandomly selected at the clinic level to receive either the CPM intervention or usual care.

Descriptive differences in sample characteristics and outcomes between treatment groups are shown, before and after propensity-score weighting, in Table 1. Among the 228 youths referred by pediatrics for psychiatric evaluation, 32 were referred from the CPM pilot clinic and 196 were referred from the seven usual-care clinics (Figure 1). Of the 186 youths who accessed a psychiatric evaluation, 29 were from the CPM group and 157 were from the control group. Before applying propensity-score weights, we found no statistically significant difference between groups in evaluation access rates. On average, compared with the control group, the CPM group was significantly older (p<.001), with a mean±SD age of 13.1±4.9 years old versus 8.7±3.6 years old, and was significantly more likely to be Hispanic (63% vs. 47%) or black (25% vs. 19%) and less likely to be white (3% vs. 29%), (p<.01). Compared with the control group, the CPM group was significantly more likely to speak Spanish (41% vs. 10%) and less likely to speak English (56% vs. 71%) (p<.001). Both groups were majority male. After applying propensity-score weights, none of the differences in covariates between the treatment groups remained significantly different.

Among the 99 youths who engaged in treatment, 23 were from the CPM group and 76 were from the control group. Before propensity-score weighting, CPM group youths were significantly more likely to engage in treatment than control group youths (79% vs. 48%; p<.01). As in the sample who accessed treatment, youths receiving the CPM intervention were significantly older (p<.001) and were more likely to be Hispanic or black and less likely to be white (p<.05). They were also more likely than the control group to speak Spanish and less likely to speak English (p<.01). The treatment groups were also significantly different in terms of diagnosis (p<.01). Mood disorders were most common in the CPM group (41%), whereas anxiety disorders were most common in the control group (36%). After propensity-score weights were computed and applied, no covariates remained significantly different between treatment groups (see Table 1.)

All regression model results are shown in Table 2 (treatment access) and Table 3 (treatment engagement) and were estimated with propensity-score weights applied. In the unadjusted model, the odds of achieving treatment access were not significantly different between groups. After the model was adjusted for age, gender, race-ethnicity, and language, however, a significant difference emerged, with members of the CPM group having more than four times greater odds of accessing treatment than the control group (adjusted odds ratio [AOR]=4.16). This translated into significantly different predicted probabilities of treatment access for each group: 92% for youths in CPM versus 75% for youths receiving usual care (p<.05; Figure 2).

For the treatment engagement outcome, results from the unadjusted model indicated marginally significant greater odds of engagement for youths in CPM compared with those in the control group (OR=3.35). After the model was adjusted for age, gender, race-ethnicity, language, and diagnosis, however, the difference increased in magnitude and became highly significant: The odds of engagement for the CPM group were more than seven times higher than those for the control group (AOR=7.54). In terms of predicted probabilities, about 80% of youths in CPM who completed psychiatric evaluation would be expected to engage in further recommended care compared with 49% of youths receiving usual care (p<.01; Figure 2). Therefore, conditioning engagement on access, about 74% of all youths receiving a referral for a psychiatric evaluation via the CPM intervention would be expected to complete the referral and engage in recommended follow-up treatment compared with about 37% of youths receiving a referral under usual care conditions.

Social factors associated with poverty, such as parental illness, transient housing, and legal issues, can present impediments to diagnosis and management of chronic illness among children, including mental illness (24). In this study, the CPM intervention was designed to address these challenges, weaving together elements of integrated child psychiatry consultation with community-based FSS, to create a more patient-centered delivery system for at-risk youths. Findings of significantly higher rates of access and engagement associated with the CPM intervention than with usual care lend preliminary support to the hypothesis that a family-centered, integrated approach can reduce treatment barriers in a population at risk for health disparities (25).

Although child psychiatry evaluations involving multidisciplinary teams were also available to the control group population, one explanation for the higher rates of access and engagement in the CPM group could be that the initial locus of care for evaluation was in the familiar pediatric care setting. Families reported great trust in their PCPs; this relationship could have contributed to sharing of key information with the CPM team and the building of a connection that facilitated treatment initiation. The FSS role was crucial; family history and current stressors identified by the FSS would not otherwise have been known to the team. These details aided the development of individualized treatment recommendations that were more likely to be endorsed by the family. The jointly created treatment plan, developed with the PCP, the parent, and the child, was then written down and provided to the parent, along with contact information. The fact that treatment engagement was sustained even after care was transitioned to other clinicians indicates the need for further research to confirm whether there is a CPM effect that persists after the team’s evaluation.

Clarifying the right diagnosis for youths with complex health care needs is a crucial step in efficient care delivery. Active involvement of child psychiatrists in the initial process of triaging consultation requests may be one reason why access and engagement rates were higher for the CPM model. Because the pilot population was known to face access barriers, the CPM team recognized that they might get only one opportunity to assess the child’s psychiatric diagnosis and treatment needs and share them with the family. They thus attempted to compress the objectives of multiple usual care visits into one CPM visit. Trainees who worked in both intervention and usual-care clinics observed that the CPM evaluation process was more intensive. They saw this as the result of active partnership among child psychiatry, FSSs, families, and pediatricians and noted that the shared approach to evaluation contributed to continuity of intent with regard to treatment planning.

Additional mechanisms for increased treatment engagement may include the following: achievement of a high level of diagnostic clarity sooner, which helps children get appropriate treatment; brief interventions onsite at the primary care clinic, with follow-up advice for pediatricians and families, strengthens recommendations; and individualized hand-offs to treatment resources reduce the need for families to start over. Postevaluation contact may have served to further facilitate initiation of follow-up treatment. The integrated CPM team also created a culture within primary care of shared practice management of youths with complex needs, including those with comorbid general medical conditions.

Because this was a nonrandomized, quasi-experimental study, this research is subject to confounding factors that could have influenced the results. First, because sample selection and assignment to treatment were not randomized, compositional and contextual differences between treatment groups could account for some of the differences in outcomes that we observed. We were able to partially address this risk by propensity-score weighting the control group to achieve statistical balance with the intervention group across a limited but relevant set of pretreatment covariates. We simultaneously controlled for these covariates in multivariate regression models, yielding doubly robust estimates to the extent that at least one of the techniques was correctly specified (26,27). However, our set of covariates was relatively small, and additional unobserved confounders that we did not account for could remain.

Another potential limitation stems from the fact that treatment assignment took place at the site (i.e., pediatric clinic) rather than at the individual level. Although this approach was a strength in that it limited the potential for spillover effects, we cannot rule out the possibility that the pilot clinic implementing the CPM intervention was characteristically different from the remaining control group clinics in ways that gave youths treated in the pilot clinic a better chance of achieving treatment access and engagement regardless of intervention type.

In addition, our intervention group sample size was relatively small, limiting the precision of our estimates as well as our ability to test hypotheses about the degree to which the CPM may have been associated with different results for different subpopulations of youths needing psychiatric care. Nonetheless, our propensity-score weighted multivariate models revealed between-group differences that were statistically significant and large in magnitude; after holding other factors constant, the CPM was associated with youths being twice as likely to achieve both access to treatment and treatment itself.

These findings serve as preliminary evidence suggesting that the CPM for integrated pediatric care should be further studied to isolate potential causal effects on the odds that children and families will receive needed mental health treatment. Additional areas for future investigation include heterogeneity of effect among subpopulations and across delivery systems; effectiveness of substituting other child mental health clinicians into the child psychiatrist role (given workforce capacity) or coordinating with other resources, such as telephone-based child psychiatry consults; and the unique effects of the FSS role on CPM outcomes.