Silverstein M, Mack C, Reavis N, Koepsell TD, Gross GS, Grossman DC. Effect of a Clinic-Based Referral System to Head StartA Randomized Controlled Trial. JAMA. 2004;292(8):968-971. doi:10.1001/jama.292.8.968
Author Affiliations: Department of Pediatrics, Boston Medical Center, Boston University School of Medicine, Boston, Mass (Dr Silverstein); Robert Wood Johnson Clinical Scholars Program (Drs Silverstein, Koepsell, and Grossman), Harborview Medical Center (Mr Mack and Dr Grossman), and Departments of Pediatrics (Ms Reavis and Dr Grossman) and Epidemiology (Dr Koepsell), University of Washington, Seattle; Dr Gross is an independent consultant in Jacksonville, Fla.
Context Early childhood development programs such as Head Start have proven
benefits for impoverished children. However, few physicians assist families
Objective To test if a primary care–based intervention is efficacious in
increasing Head Start attendance.
Design, Setting, and Participants Randomized controlled trial of 246 Head Start–eligible children
aged 0 through 4 years recruited in spring 2003 from 4 health clinics in Seattle,
Interventions List of Head Start telephone contacts provided to families of all children
and, for those in the intervention group, a computer-generated packet containing
a physician referral letter (and a physical examination form and immunization
record, if available) mailed directly to Head Start by study personnel.
Main Outcome Measure Head Start attendance by January 2004.
Results The 123 children analyzed in each study group were similar at baseline.
Overall, 72 children (29%) were successfully connected with Head Start (ie,
actively attending or on a waiting list) by January 2004. Among the intervention
group, 50 children (41%) were successfully connected with Head Start, contrasted
with 22 (18%) in the control group (adjusted difference, 17%; 95% confidence
interval [CI], 8%-27%). Among the intervention group, 31 children (25%) were
actively attending Head Start, contrasted with 14 (11%) in the control group
(adjusted difference, 12%; 95% CI, 3%-21%). Only 2 clinics contributed children
to Head Start waiting lists. Among children from these clinics, 19 of 87 (22%)
in the intervention group got onto a Head Start waiting list, vs 8 of 94 (9%)
in the control group (adjusted difference, 13%; 95% CI, 5%-21%). To get 1
child either into Head Start or onto a waiting list, we needed to refer 4
Conclusion Facilitating an initial connection to Head Start on families' behalf
substantially increased Head Start attendance.
The integration of community resources with health care delivery is
an important component of quality medical care.1,2 Although
much has been written about referral patterns between primary care physicians
and specialists,3,4 little is
known about how primary care clinicians integrate their services with those
of other community-based organizations. For children, one important evidence-based
community resource is high-quality preschool. Early childhood development
programs produce sustained cognitive, social, and educational benefits for
low-income children.5- 10 In
the United States, the largest of these programs is Head Start. Any family
at or below the federal poverty level is eligible to enroll its 3- to 4-year-old
children in Head Start, and its 0- to 3-year-old children in Early Head Start.
Social and educational benefits have been observed among Head Start graduates11; early results of a randomized controlled trial of
Early Head Start supports its effectiveness across a range of outcomes.12
In 2002, the Centers for Disease Control and Prevention recommended
publicly funded development programs for impoverished preschool children and
suggested the promotion of such programs as part of well-child care.13 A subsequent study, however, showed that few pediatricians
assist families with Head Start enrollment,14 a
finding that prompted experts in the field to call for better, more systematic
connections between clinicians and providers of early childhood services.15 We therefore undertook a randomized controlled trial
of a clinic-based referral system to Head Start.
Clinicians, office staff, and research assistants at 4 clinics in Seattle,
Wash, recruited a convenience sample of children aged 0 through 4 years to
participate in the study. Patients' siblings and children present in clinic
for reasons other than medical care (eg, dental care, social work consultation)
were also eligible. Children were excluded if they were in obvious distress,
previously enrolled in Head Start, unaccompanied by a primary caregiver, or
if their families were unable to provide any contact information. Only children
eligible for Head Start were included; eligibility was determined by a computerized
screening instrument (available from the authors on request) that considered
the child's age, family income and receipt of Temporary Assistance to Needy
Families, and whether the child was in foster care. Children were enrolled
between March 6 and May 13, 2003. The Seattle Children's Hospital and Regional
Medical Center institutional review board approved this study. We obtained
written informed consent (and, after April 15, 2003, a written Health Insurance
Portability and Accountability Act release) from every family.
The objective of the intervention was to facilitate initial contact
between families and Head Start, and to transfer the medical documentation
required for Head Start enrollment. Families of all children in the control
and intervention groups were given a language-appropriate telephone contact
list of all Head Start agencies in the metropolitan Seattle area. For intervention
children, a referral packet was also generated by computer and mailed directly
to Head Start by study personnel; the packet contained a physician referral
letter, including information for Head Start to contact the family; a physical
examination form; and the child's immunization record. The second and third
items were included only if available. Every Head Start agency in the target
area participated in the project. None altered its established enrollment
criteria to prioritize children from the study, and all signed a memorandum
of understanding prior to study participation.
Families reported their primary language, whether the child was the
family's first or had any special health care needs, and whether the family
had previous experience with Head Start enrollment.
Our primary outcome was Head Start attendance by January 2004. To obtain
this information, a designated employee at each Head Start agency indicated
by standardized checklist whether each child in both study groups was attending
Head Start, on a waiting list, or neither. To test whether the intervention
affected other steps leading to Head Start enrollment, we conducted a telephone
survey in June 2003. We asked families whether they had been in contact with
anyone from Head Start and, if so, whether the family or Head Start had been
responsible for making this contact.
We estimated a sample size of 100 in each study group to show a 20%
difference in Head Start attendance with 95% certainty and 80% power, assuming
a statistical worst-case scenario that 50% of children would attend Head Start.
Children were randomly assigned to study groups within each clinic using
a computerized random number generator (Figure
1). Telephone survey administrators, investigators, and Head Start
personnel reporting enrollment data were blinded to study allocation.
We assessed intervention effect by intention-to-treat analysis, estimating
relative risk and risk differences with log-binomial or binomial regression,16 adjusting for clinic as a fixed effect and correcting
for family clustering using robust standard error estimates.17 Children
were considered siblings if they had the same guardian and lived at the same
address. Because only children from clinics 2 and 4 got onto Head Start waiting
lists, only children from these clinics were included in waiting list–specific
We assessed effect modification by clinic by adding clinic × study
group interaction terms to the base regression models. To check for residual
confounding, we estimated intervention effect by multivariable logistic regression,
adjusting for child's age and sex, household size, primary language, parents'
previous experience with Head Start enrollment, receipt of Temporary Assistance
to Needy Families, and presence of special health care needs. We used logistic
regression for this purpose because convergence could be achieved across a
wider range of covariate combinations than with binomial or log-binomial regression.18 Statistical analyses were performed using Intercooled
Stata 7.0 (Stata Corp, College Station, Tex).
Research assistants screened 366 children for Head Start eligibility.
Of these, 115 were ineligible. Three additional children were excluded prior
to randomization: 2 for having incomplete contact information and 1 at the
parent's request. Of the 248 children randomized, 124 were allocated to each
study group. One child was withdrawn from each group because both proved to
be duplicates of previously randomized children. The analysis included 123
children in the intervention group and 123 in the control group. Among these,
there were 4 sets of siblings, comprising 9 children in total.
Within each clinic, the proportion of children randomly assigned to
the intervention group ranged from 46% to 57% (Table 1). There were no clinically meaningful differences between
groups with regard to age, sex, household size, English being the family's
primary language, or previous parental experience with Head Start enrollment.
The survey response rate was 75% (78% of intervention and 72% of control
families). Fifty-seven percent of intervention families reported being in
contact with Head Start, contrasted with 36% of control families (adjusted
difference, 21%; 95% confidence interval [CI], 7%-35%) (Table 2). Of those families reporting contact with Head Start, 85%
of intervention families reported that Head Start had initiated the contact,
contrasted with 32% of control families (adjusted difference, 54%; 95% CI,
Overall, 72 children in the study (29%) were either actively attending
Head Start or on a waiting list by January 2004. Although 46 children enrolled
in Head Start, 1 child in the control group dropped out prior to data collection,
leaving 45 (18%) actively attending and 27 (11%) on a waiting list. In the
intervention group, 50 children (41%) were either actively attending Head
Start or on a waiting list, contrasted with 22 (18%) in the control group
(adjusted difference, 17%; 95% CI, 8%-27%) (Table 2).
Thirty-one children in the intervention group (25%) were actively attending
Head Start, contrasted with 14 children in the control group (11%) (adjusted
difference, 12%; 95% CI, 3%-21%) (Table
2). Two Head Start attendees from the control group had siblings
in the intervention group, and therefore possibly benefited from the intervention.
Only children from clinics 2 and 4 got onto Head Start waiting lists. Among
the children at these 2 clinics, 19 of 87 (22%) in the intervention group
were on a waiting list at the time of data collection vs 8 of 94 (9%) in the
control group (adjusted difference, 13%; 95% CI, 5%-21%).
Sample size limitations precluded reliable analysis of effect modification
by clinic. Multivariable regression models controlling for slight imbalances
in baseline characteristics demonstrated no change in the effect of the intervention.
A simple, computerized referral system can be effective in the primary
care setting in promoting Head Start attendance. By facilitating an initial
contact between families and Head Start, we were able to increase Head Start
attendance substantially, compared with providing families with a list of
telephone contacts. To get 1 child into Head Start, we needed to refer 7 children;
to get 1 child either into Head Start or onto a waiting list, we needed to
Studies have shown that when health professionals contact medical specialists
on patients' behalf, follow-up improves.3,4 Few
studies, however, have examined completion of referrals to community-based
organizations. In a case series, Needlman et al19 reported
poor follow-up among mothers with depression referred to community resources,
and Rushton et al20 reported suboptimal follow-up
among children with psychosocial problems referred to mental health services.
Our study adds to this literature by offering a strategy to refer children
to Head Start from the primary care setting.
Our study has several limitations. In randomizing by child, we inevitably
introduced intrafamily contamination when siblings were assigned to different
study groups. Families of control children, by being screened for Head Start
eligibility and getting a list of local Head Start resources, received a potentially
helpful service; and our follow-up telephone survey possibly acted as a reminder
intervention to both study groups. Such limitations, however, likely only
attenuated the effect of the intervention relative to that of the controls.
Our study included a small number of practices in a single geographic
area, its population was nonrandomly selected from among those present at
community clinics, and it was designed as a trial of efficacy, not effectiveness.
Additionally, the centerpiece of the intervention was a free-standing computer
program that required a clinic computer for its operation. On these counts,
the generalizability of the study may be questioned. Furthermore, in locales
having relatively fewer Head Start slots than Seattle, our intervention might
preferentially place children on waiting lists as opposed to into programs.
Although this might lead to program expansion in such areas, it would be less
helpful to the actual families referred.
Considering these limitations, it appears that using a mailed referral
packet to facilitate initial contact between families and Head Start may be
an effective strategy for promoting Head Start attendance from the physician's
office. Although the results of this study are not necessarily generalizable
beyond the interface between primary care and Head Start, they do raise questions
concerning how primary care clinicians might refer low-income patients to
other community resources outside the medical system.