Effect of Family Navigation on Participation in Part C Early Intervention

Abstract

Objectives:

Part C Early Intervention (EI) services have been shown to reduce autism symptoms and promote healthy development among young children. However, EI participation remains low, particularly among children from structurally marginalized communities. We investigated whether family navigation (FN) improved EI initiation following positive primary care screening for autism compared to conventional care management (CCM).

Methods:

We conducted a randomized clinical trial among 339 families of children (ages 15-27 months) who screened as having increased likelihood for autism at eleven urban primary care sites in three cities. Families were randomized to FN or CCM. Families in the FN arm received community-based outreach from a navigator trained to support families to overcome structural barriers to autism evaluation and services. EI service records were obtained from state or local agencies. The primary outcome of this study, EI service participation, was measured as the number of days from randomization to first EI appointment.

Results:

EI service records were available for 271 children; 156 (57.6%) children were not engaged with EI at study enrollment. Children were followed for 100 days after diagnostic ascertainment or until age three, when Part C EI eligibility ends; 65 (89%, 21 censored) children in the FN arm and 50 (79%, 13 censored) children in the CCM arm were newly engaged in EI. In Cox proportional hazards regression, families receiving FN were approximately 54% more likely to engage EI than those receiving CCM (1.54 (95% CI: 1.09-2.19), p=0.02).

Conclusions:

FN improved the likelihood of EI participation among urban families from marginalized communities.

INTRODUCTION

The American Academy of Pediatrics recommends that children are referred to early intervention (EI) services as soon as they are identified as having an increased likelihood of autism.¹ Timely engagement in EI services, which leverages early brain plasticity,^2,3 has been shown to result in improvements in emotional and social development, academic readiness, and physical abilities^1,4,5 and reductions in autism symptoms.^1,5,6 As a result, the U.S. Department of Health and Human Services included an objective to improve early service use among children under four years old with autism in Healthy People 2030.⁷

In the US, the most common pathway to EI services for children from birth to age three is through federally funded infant and toddler EI Part C programs legislated in the Individuals with Disabilities Education Act (IDEA).⁸ All states offer EI services to children with developmental conditions or delays under Part C. However, eligibility criteria and the scope and comprehensiveness of services vary by state.^9,10 Despite evidence of positive developmental gains among children receiving EI, studies have consistently found that many eligible children, including those with behaviors suggestive of autism, are not enrolled.^11–13 Moreover, studies have shown that marginalized populations, including families who are low-income, from racial/ethnic minority groups, speak a primary language other than English, and/or have lower educational attainment, have limited participation in EI services.^11,13–17 When children from these families receive services, they tend to begin services at later ages¹² and, some research shows, receive less intense services.^18,19

Efforts to improve EI participation, especially among marginalized groups, have focused on improving clinical processes and supports for families of eligible children. We report on the impact of one intervention, family navigation (FN), to support participation in Part C EI among children with an increased likelihood of autism. FN is a care management strategy designed to address structural (e.g., language, transportation) and psychological (e.g., stigma) barriers to care.^20–25 We focus on the period beginning with developmental screening suggesting increased likelihood of autism through engagement in recommended services.

Our previous work has shown that FN improves completion of autism evaluations, such that 82% of children with FN reached diagnostic ascertainment within six months, compared to 68% who received conventional care management (CCM).²⁶ This study expands the investigation of FN’s effect on the autism care continuum, and is, to our knowledge, the first to evaluate the impact of this intervention on EI engagement.

METHODS

We conducted a parallel group randomized trial of FN versus CCM among children ages 15-27 months who were identified with increased likelihood of autism during routine primary care visits at one of 11 urban pediatric primary care practices in Boston, MA, New Haven, CT, and Philadelphia, PA. Primary care practices were part of integrated care networks that included developmental and behavioral pediatrics (DBP) specialty clinics that were members of the Health Resources and Services Administration-funded DBP Research Network (DBPnet). Families were recruited from February 12, 2015 through May 17, 2019. Randomization occurred following baseline data collection using randomly permuted blocks of two and four, stratified by primary care site and receipt of prescreening educational materials. Study investigators and staff were blinded to study arm assignment. Parents and/or guardians provided written informed consent that included permission to access their child’s Part C EI service records. Additional details of study methods and consent protocols are available elsewhere.^23,26 This study was approved by the Boston University Medical Campus Institutional Review Board and adheres to the Consolidated Standards of Reporting Trials (CONSORT) guidelines.

Participants

Children were identified as having concern for autism using the Modified Checklist for Autism in Toddlers, Revised with Follow-up (MCHAT-R/F) as part of routine clinical practice at 18- and 24-month well child visits.^1,27,28 Upon study referral, parents repeated the MCHAT-R/F with study staff, using standard administration and scoring procedures, to ensure consistency across sites and that all children were administered the Follow-up interview. Consistent with clinical practice, children who did not demonstrate increased likelihood of autism based on their MCHAT-R/F score could be referred to the study if their primary care clinician had high levels of concern. Children with previous autism diagnoses and in child protective services were excluded from study participation.

Interventions

We have previously described the content and fidelity of FN and CCM.^23,26 Briefly, families assigned to FN were supported by a single community health worker (CHW), called a navigator. Navigators were largely bilingual, bicultural community members, who received training in autism characteristics, motivational interviewing, collaborative problem solving, community resources, and principles of patient navigation. We hired navigators with the goal of maximizing racial, ethnic, and language concordance among families assigned to FN. All Spanish speaking participants had the opportunity to work with a bilingual-bicultural navigator. FN was designed to support families beginning at the time of positive autism screening through engagement in recommended services. In this study, FN began at study entry and continued for 100 days after the completion of the autism diagnostic evaluation, at which point families were expected to be engaged in autism-specific services. FN was manualized. Navigators were guided by a workbook delineating content and tasks for three structured visits that aligned with critical events in the care continuum: at the time of a positive autism screen, immediately after diagnosis, and 100 days following diagnostic ascertainment Additional in-person and remote contacts occurred as needed. Navigators had study-provided cellular telephones to communicate with families via telephone, text, or email and a car service subscription to travel to families’ homes. Communication was initiated by families or navigators. Navigators participated in monthly supervision meetings focused on how to support families’ goal attainment, coordinate care, and address care barriers.²² They also referred parents whose children were not receiving to EI to local agencies, assessed parents’ understanding of EI services, and discussed the complementary features of EI and DBP evaluation. Throughout this process, navigators worked with families to identify challenges and supports to facilitate engagement in both EI services and the diagnostic evaluation.

CCM was an enhanced form of usual care at each site. These enhancements included a protocol for family outreach, including three attempts to reach families by phone to introduce the care manager and remind the family of the autism evaluation date. Mail follow-up was conducted for families who were unreachable by telephone. With the exception of initial outreach by care managers, all contact was initiated by the child’s healthcare team or family. Care managers were existing staff who had use of all available health system resources; care managers had similar race and ethnicity as families at two of three sites.

Outcome

Time to EI participation was operationalized as the number of days from randomization to the first date of a billable EI visit as documented in service records. We used “participation” instead of “enrollment” to reflect ongoing engagement in EI services. The mean number of EI visits per child during their enrollment period was 70.3 (standard deviation (sd)=86.5) and the overwhelming majority of children (75%) had more than five visits. EI service reports were gathered from state agencies in Massachusetts (Department of Public Health) and Connecticut (Office of Early Childhood, Birth to Three) and from individual EI agencies in Pennsylvania. We received information on service date, type (i.e., physical therapy, speech therapy, occupational or developmental therapy, social work, other services), setting (i.e., home, community), program information (i.e., whether services were provided through general or autism-specific programs), and visit duration for each scheduled encounter. We also received information on scheduled visits that were cancelled.

Subgroup Analyses

We a priori defined ethnicity and site as theory-based potential effect modifiers. Hispanic ethnicity was examined based on the extensive literature enumerating barriers to care among Hispanic families.^29,30 We examined site as a potential effect modifier given the differences in EI regulations in the three states where the study was conducted.

Sample Size

The original study sample size of 250 was based on the ability to detect a 25% absolute difference in autism diagnostic ascertainment (the main study’s primary outcome) between treatment arms. The sample was expanded to 340 to allow adequately powered sub-analyses of diagnostic ascertainment by gender (a secondary outcome). Power analyses were not conducted for EI participation; as it was not the primary outcome of the original study. EI service records were obtained for 271 children; 68 of the 192 children (42.2%) from Pennsylvania were excluded from analyses because we did not receive their service records. One-hundred fifty-six children who were not participating in EI at study enrollment were included in this analysis.

Analysis Plan

We categorized children based on EI services at study enrollment: (1) no EI services ever; (2) no EI service during the 30 days before study enrollment (3) EI services during the 30 days before enrollment. We defined non-engagement in EI at study entry by combining categories one and two into a single group; children in the third group were considered actively engaged at study enrollment and were not included in this analysis.

We assessed baseline differences among non-engaged children by intervention arm using Pearson’s χ² and two-sample t-tests. Time to EI engagement was descriptively analyzed using Kaplan-Meier curves. Children not engaged in EI services were censored at study exit (100 days post-diagnostic evaluation, n=31), because we believed it was unlikely that FN would affect EI engagement after the study ended, or at age three (n=3) when children were no longer eligible for Part C EI services (children may remain be eligible for EI services after their third birthday through Part B of IDEA). Cox proportional hazards regressions estimated hazard ratios (HR) and 95% confidence intervals (CI) for differences in likelihood of EI engagement. We assessed the main effect of treatment arm, controlling for previous EI enrollment, parent marital and cohabitation status, and public insurance because these factors differed significantly across our intervention groups. We found no differences by site in our models; following Wei et al,³¹ we included site as a clustering variable in all subsequent analyses. Next, we added ethnicity to the model. Finally, we conducted a preliminary investigation of the interaction between ethnicity and treatment group due to our previous findings that the magnitude of effect of FN on diagnostic ascertainment was significantly different for Hispanic children than their non-Hispanic peers.²⁶

Analyses were assessed to ensure that the assumption of proportional hazards were met using scaled Schoenfield residuals. Analyses were conducted in Stata (College Station, TX) with p<0.05 indicating statistical significance.

We multiply imputed the data for the 68 children from Pennsylvania for whom we did not receive service records. Twenty imputations were created in Stata using chained equations.³² Findings based on multiply imputed data were consistent with the unimputed data, which we report on here.

RESULTS

Figure 1 shows the 340 families that were randomized to the intervention (n=1 post-randomization exclusion); 169 were assigned to FN. The only difference between the included (n=93) and excluded (n=68) children from Pennsylvania for whom we did not receive EI service records were that parents of excluded children were less likely to have graduated high school (78% vs. 89%, p=0.05). Overall, 42% (115) of children were receiving EI services at study enrollment.

Figure 1.

Participant Flow Diagram

^aChild was enrolled from a non-participating primary care site.

Of the 156 children were not receiving EI at enrollment, 87 (56%) received FN and 69 (44%) received CCM (Table 1). Children receiving FN were more likely to have public health insurance (FN: 90%; CCM: 75%, p=0.02) and less likely to have had married or cohabiting parents (FN: 46%; CCM: 62%, p=0.04). Ninety-five (61%) children screened with “high risk” scores (≥8) on the MCHAT-R/F and 58 (37%) were identified as “medium risk” (3-7); only three (2%) had low risk scores (0-2).²⁸ There were no statistically significant differences in MCHAT-R/F score by arm (p=0.18). Twenty-two (14%) children received EI in the past but not during the 30 days prior to study enrollment ($\overline{\mathrm{X}}=249.4$ days (sd=198.9)). There were no differences in previous EI receipt (p=0.21) or number of days to prior EI services (p=0.96) by treatment arm.

Table 1.

Baseline Characteristics of Participants Enrolled in Project EARLY who were not Engaged in EI at Study Enrollment

	Family Navigation (n=87)		Care Management (n=69)
	% or $\overline{\mathrm{X}}$	(n or sd)	% or $\overline{\mathrm{X}}$	(n or sd)	p
Intervention Characteristics
Site					0.92
Boston, MA	40	(35)	44	(30)
New Haven, CT	24	(21)	23	(16)
Philadelphia, PA	36	(31)	33	(23)
Positive MCHAT-R/F Screening Eligibilitya	98	(85)	100	(69)	0.21
MCHAT-R/F Score ($\overline{\mathrm{X}}$ (sd))	8	(3.3)	8	(2.8)	0.18
Child Characteristics
Sex: Male	75	(65)	77	(53)	0.76
Born in US	64	(56)	55	(38)	0.24
Public Health Insurance	90	(78)	75	(52)	0.02
Gestational Age <37 weeks	9	(8)	10	(7)	0.84
Diagnosisa					0.21
Autism Spectrum Disorder	70	(57)	64	(32)
Language or Other Developmental
Disorder	16	(13)	28	(14)
Other	14	(11)	8	(4)
Engaged in EI >29 Days before Study
Enrollment	17	(15)	13	(9)	0.47
Parent & Family Characteristics
Parent Race/Ethnicity					0.36
Hispanic	29	(25)	41	(28)
Non-Hispanic	71	(62)	60	(41)
Black, non-Hispanic	58	(50)	44	(30)
White, non-Hispanic	7	(6)	7	(5)
Other, non-Hispanic	7	(6)	9	(6)
Parent <25 years old	18	(16)	16	(11)	0.75
Preferred Language					0.05
English	87	(76)	78	(54)
Spanish	8	(7)	20	(14)
Other	5	(4)	2	(1)
Currently Working (outside home)	55	(48)	58	(40)	0.73
High School Graduate	82	(71)	78	(54)	0.60
Married or living with partner	46	(40)	62	(43)	0.04
Parental Worry about Child’s Development	7	(2.8)	7	(3.3)	0.45
Only Child	41	(36)	32	(22)	0.22
Family Receiving
WIC	69	(60)	65	(45)	0.62
Cash Assistance	22	(19)	17	(12)	0.49
SNAP	70	(61)	58	(40)	0.12

MCHAT-R/F = Modified Checklist for Autism in Toddlers, Revised with Follow-up; WIC = Special Supplemental Nutrition Program for Women, Infants, and Children; SNAP = Special Supplemental Nutrition Program for Women, Infants, and Children Supplemental Nutrition Assistance Program

^aDiagnostic information was not available for 5 children in the CM intervention arm.

Protocol Adherence

As described in our previous work,²⁶ protocol adherence was acceptable for FN content and motivational intervention fidelity in the original study. Thirty-four of 87 (39%) families had at least two face-to-face contacts with their FN prior to EI engagement. However, the average number of contacts from study enrollment through EI engagement among the current sample was 17.1 (sd=13.5); the majority occurred via phone ($\overline{\mathrm{X}}=8.4$ (sd=6.6)) and text ($\overline{\mathrm{X}}=6.9$ (sd=7.3)). Six of the 87 families (7%) who were not engaged in EI at study enrollment disengaged, defined as at least ten weeks during which there was no contact with the navigator. Navigators contacted EI agencies/staff on behalf of 49 (58%) families.

Among parents assigned to CCM, 6 (9%) parents initiated contact with the care manager between study enrollment and EI engagement. The average number of total contacts between care managers and parents was approximately one; the majority were phone contacts.

Primary Outcome: EI Service Engagement among Families not Engaged at Study Enrollment

The Kaplan-Meier plot (Figure 2) shows time to EI participation for the 115 children (65 FN, 50 CCM) newly engaged with EI prior to the end of the study. The difference between treatment arms emerges by day 50, when 29 (33%) children in FN and 18 (26%) children in CCM have initiated EI services and peaks approximately 100 days after study enrollment (FN: 48 (55%); CCM: 27 (39%)). Among children newly engaged with EI, the overwhelming majority (109 (95%)) were actively participating in EI ($\overline{\mathrm{X}}=65.7$ service dates (sd=57.4), median=59). Children receiving FN ($\overline{\mathrm{X}}=75.1$ (sd=56.1)) had more EI visits than their peers receiving CCM ($\overline{\mathrm{X}}=53.4$ (sd=57.3), p=0.04).

Figure 2.

Time to Part C EI Services, by Treatment Group

EI= Early Intervention; FN=Family Navigation; CCM=Conventional Care Management. Hazard ratio from Cox proportional hazards model controlling for whether the child was engaged with EI services >30 days before study enrollment, clustered on site. Time 0 represent study enrollment. These are the estimated probabilities based on the Kaplan-Meier curve. Censoring was set at the end of the study (n=31) or age three (n=3), whichever came first.

The Cox proportional hazard model, controlling for participation in EI >30 days before enrollment, parental marital or cohabiting status, and insurance type, clustered on site, showed that families receiving FN were approximately 54% more likely to participate in EI services than those receiving CCM (1.54 (95% CI: 1.09-2.19), p=0.02). All subsequent analyses control for the variables listed within this paragraph and are clustered on site. Sensitivity analyses using a 60-day cut-off for EI services prior to study enrollment were consistent with the 30-day analysis.

The Effect of Ethnicity

The HR for ethnicity was non-significant when it was added to the model described above (HR=0.73 (95% CI: 0.50-1.09), p=0.13). The HR for treatment group remained significant (HR=1.49 (95% CI: 1.10-2.03), p=0.010). Based on our a priori hypothesis that ethnicity would modify the effect of treatment on EI participation, we assessed the intervention x ethnicity interaction (HR=0.95 (95% CI: 0.79-1.15), p=0.60). Despite finding non-significance, we conducted exploratory analyses based on our previous work showing differential effects of intervention by ethnicity. We created a four-category variable representing this interaction (FN, non-Hispanic; FN, Hispanic; CCM, non-Hispanic; CCM, Hispanic) to assess within arm differences. We found no significant differences by ethnicity within the FN or CCM arms.

DISCUSSION

Clinical processes to improve Part C EI participation are necessary to ensure that eligible children^{10–12,16,17} benefit from the positive developmental outcomes related to timely services.^1–6 Study findings suggest that FN improves EI participation, by up to 54%, compared to CCM. Examining the pattern of initiation of services depicted on the Kaplan-Meier plot suggests that FN may be particularly helpful for families for whom initial EI referrals do not result in timely enrollment. EI participation was similar from study enrollment to approximately 50 days postenrollment by arm, likely representing referral to EI in response to primary care autism screening. The benefit of FN appears to begin approximately 50 days after study enrollment, as demonstrated by the separation of the FN and CCM curves. FN may support increased EI participation through two mechanisms. Frequent contact with families addresses structural barriers as they emerge and ensures successful connection EI programs. In addition, FN is designed to provide culturally relevant education related to child development and address stigma that may interfere with EI engagement.

We hypothesized that the effect of FN would be modified by ethnicity based on our previous analysis of diagnostic ascertainment.²⁶ Our exploratory analyses did not support this hypothesis. We found that FN improved EI participation equally well across ethnic groups, supporting FN’s potential to improve equitable access to services. In comparison with our previous work, the magnitude of the effect of FN was similar among Hispanic and non-Hispanic families, raising questions regarding whether FN may have differential impacts for Hispanic families depending on the clinical process.

By the end of the study or age three (whichever came first), 89% of children in FN had initiated EI services. This finding represents a ten percentage point increase in EI participation for families receiving FN compared to families receiving CCM and is substantially higher than the percent of eligible children enrolled in the nation (45.7%) and our study sites (Connecticut: 57.1%, Massachusetts: 70.6%, Pennsylvania: 73.3%).²⁴ Our results may understate the potential impact of FN, given that our CCM protocol likely provides more assistance than is typically available in real world practice and that this study was conducted in states that rank among the highest in the US regarding enrollment of eligible children. As other studies have shown, FN plays a role in reducing structural (e.g., spoken/written language, transportation) and psychological (e.g., social and emotional support) barriers to care that underlie challenges to EI participation.^20–25 Importantly, we found that FN worked equally well across primary care sites in three states with unique Part C EI systems, lending support to its potential for replication.

Since we began this study seven years ago, evidence of the role that CHWs (navigators are a type of CHW) play in promoting health equity among structurally marginalized groups has accumulated significantly.^33,34 However, few studies of CHWs have focused on children and even fewer on equitable access to developmental services. This study begins to fill an important gap in our understanding of how CHWs support positive health outcomes among individuals across the lifespan. Despite the promise of FN to improve participation in services, debates about the sustainability of CHW services and broader FN programs are ongoing. Currently, funding for FN relies on a patchwork of funding mechanisms.^35–37 However, as of 2021, 21 states have developed sustainable pathways that could support CHWs and the implementation of FN in pediatric primary care settings, either through Medicaid fee-for-service reimbursement for CHWs services or through Medicaid demonstration projects (Medicaid 1115 waivers) that provide higher per person per month rates for practices that utilize CHWs.³⁶ While such funding mechanisms only affect Medicaid recipients, over 40% of US children who experience the greatest inequities are covered by Medicaid.³⁸

Limitations

Our study has several limitations. First, we excluded 68 children from Pennsylvania because we could not obtain their EI records, raising concerns about selection bias. However, few differences in baseline characteristics between children with and without EI records were found and multiply imputed analyses were consistent with the findings reported here. Second, the 30-day cut-off used to categorize children who were actively engaged in EI at study enrollment is not a standard measure. In the absence of an accepted standard, this cut-off is clinically reasonable and findings were stable when we conducted sensitivity analyses using a 60-day cut-off. Third, the analysis was powered on the original study’s primary outcome (time to diagnostic ascertainment) and was, therefore, underpowered for the outcome explored in this analysis. As a result, we may not have detected significant differences that would be present in an adequately powered study. Post hoc sample size calculations determined that a sample size of 655 children would have been required to detect 25% difference in EI engagement, using p=0.05 and 80% power. Additionally, our findings may not be applicable to states with different Part C EI service structures, rural primary care sites, and more advantaged children who are covered by private insurance. Finally, we assessed EI participation based on state agency, not parent reports, although we collected information on EI from both sources. We view this as a strength of our paper and plan to assess parent reports of EI engagement in future manuscripts.

CONCLUSION

Our findings suggest that FN is a promising strategy to address the persistent and well-documented under-enrollment of families from marginalized groups in Part C EI. FN increased participation in Part C EI services in a sample of children from racial, ethnic, or linguistic minority and/or low income families. Integration of FN into primary care practices, especially those whose patient populations largely represent marginalized groups, may support improved developmental outcomes given the association of EI with developmental gains for children with autism. Future analyses should examine whether FN has differential effects among racial and ethnic subgroups as part of efforts to advance equitable access to EI services.

What’s New.

Part C Early Intervention is associated with improved developmental outcomes and reduced autism symptoms. However, participation remains low, especially among marginalized populations. This study suggests that family navigation is a promising strategy to improve engagement among children from marginalized communities.

Data Sharing Statement:

Deidentified participant data and a data dictionary are available through the National Database for Autism Research (NDAR), beginning 7/1/2021. Data can be accessed according to NDAR guidance.