pediatrics
May 2017, VOLUME139 /ISSUE 5

Enriched Medical Home Intervention Using Community Health Worker Home Visitation and ED Use

  1. Meghana Anugu, BAa,
  2. Amy Braksmajer, PhDb,
  3. Jiayu Huang, PhDc,
  4. Jie Yang, PhDd,
  5. Kristi L. Ladowski, MPHb, and
  6. Susmita Pati, MD, MPHb,e
  1. bDivision of Primary Care Pediatrics,
  2. cDepartment of Applied Mathematics and Statistics, and
  3. dDepartment of Family, Population and Preventive Medicine,
  4. aSchool of Medicine, Stony Brook University, Stony Brook, New York; and
  5. eStony Brook University Medical Center, Stony Brook, New York
  1. Ms Anugu and Dr Braksmajer drafted the manuscript; Drs Huang and Yang conceptualized and designed the study and conducted acquisition, statistical analysis, and interpretation of data; Ms Ladowski provided administrative, technical, or material support and obtained funding; and Dr Pati conceptualized and designed the study and obtained funding; and all authors critically revised the manuscript for important intellectual content and approved the manuscript as submitted.

Abstract

BACKGROUND AND OBJECTIVES: Community health workers (CHWs) have great potential to extend medical home services and reduce emergent health care use, but evidence in pediatrics is scarce. We evaluated the impact of an existing enriched medical home intervention (EMHI) that directly integrates CHWs into emergency department (ED) visits and hospitalizations for pediatric ambulatory care–sensitive conditions (ACSCs).

METHODS: The EMHI group in this prospective cohort study received home visits from trained CHWs to support adherence to recommended care; the comparison group received usual care (UC). Sociodemographic characteristics were compiled from the EMHI database, and ED and hospitalization information was extracted for study participants from a statewide database. The Wilcoxon signed rank test was used to compare ED data and the Wald test was used to compare hospitalization use for ACSCs between the intervention and UC groups after adjusting for different characteristics between groups by using propensity score matching method.

RESULTS: The study sample included 922 children (225 intervention, 697 UC). After propensity score matching, the analytic sample included 450 children (225 intervention, 225 UC). After propensity score matching, the intervention group was significantly less likely than the UC group to visit the ED for an ACSC (18.2% vs 35.1%; P = .004). We found no differences in ACSC hospitalizations between the 2 groups.

CONCLUSIONS: Our findings suggest that EMHIs using trained CHWs may be a cost-effective model to reduce preventable ED utilization, especially among vulnerable children.

  • Abbreviations:
    ACSC
    ambulatory care–sensitive condition
    CHW
    community health worker
    ED
    emergency department
    EMHI
    enriched medical home intervention
    KFH
    Keeping Families Healthy
    SBC
    Stony Brook Children's
    UC
    Unusual Care
  • What’s Known on This Subject:

    Emergency department use for preventable conditions adds to the cost of health care. There are few evidence-based models for community health worker (CHW) interventions that can serve as sustainable alternatives to reduce emergent health care utilization for pediatric preventable conditions.

    What This Study Adds:

    This enriched medical home intervention using trained CHWs decreased emergency department use for families receiving support in health education, disease management, and between-visit troubleshooting. This model may be a cost-effective blueprint to reduce inappropriate health care utilization and promote healthy outcomes.

    Inappropriate emergency department (ED) utilization is a significant challenge for the US health care system. In 2011, there were 136 million ED visits contributing an estimated 4% to health care spending, one-quarter of those being pediatric visits.1,2 Ten percent to 60% of pediatric ED visits are for nonurgent care matters.35 Underserved and vulnerable groups of children, identified as low-income, public insurance users, and racial/ethnic minorities, display higher rates of ED visits, particularly for nonurgent matters and ambulatory care–sensitive conditions (ACSCs) (ie, conditions that can be managed by appropriate ambulatory care, thereby preventing or reducing the need for ED admission or hospitalization) compared with other children.68 Having a usual source of care significantly reduces ED visitation.6,9 Many vulnerable families do not understand appropriate ways to use primary care, and reliance on EDs leads to fragmented care and missed opportunities for preventive care and education, further contributing to health disparities among underserved children.68,10,11

    Community health worker (CHW) interventions have been suggested as a means of overcoming these obstacles. CHWs are defined broadly as “frontline public health workers who are trusted members of and/or have an unusually close understanding of the community they serve.”12 Current literature suggests that CHW interventions have a positive impact on health outcomes among low-income, minority children with pediatric asthma and improve the quality of life of their caretakers.1315 The wide variation in intervention components, as well as the health outcomes studied, has made it difficult to evaluate the overall effectiveness of existing CHW programs.13 Furthermore, there are few recent studies that examine the effects of home visitation on any-cause ED utilization, particularly among disadvantaged populations and children.16,17 Finally, few home visit programs have been directly integrated with the primary care medical home via an enriched medical home intervention (EMHI) model.16,18 EMHIs are delivered by trained CHWs who work in conjunction with medical homes to assist families with health care navigation, care coordination, and understanding basic health education.6

    The present study examined the effects on ED visits and hospitalizations of an EMHI that included home visits in a diverse cohort of children attending pediatric primary care practices. We hypothesized that compared with usual care, this intervention would lead to decreased ED utilization visits and hospitalization for ACSCs.

    Methods

    Study Population

    In July 2011, Stony Brook University Medical Center–affiliated pediatric primary care practices (Stony Brook Children's [SBC]) launched an EMHI service program entitled Keeping Families Healthy (KFH). SBC operates 5 pediatric primary care offices located across Suffolk County, New York. Patients attending SBC primary care practices are representative of the communities from which they are drawn and closely resemble the racial and socioeconomic distribution of Suffolk County.

    KFH Program

    KFH is a voluntary program offered to all pediatric patients (newborns through those aged <18 years) considered by an SBC clinician to be “at risk” for poor health outcomes. We relied on clinician judgment to identify at-risk children (eg, children with special health care needs) because this approach was the most feasible way to efficiently identify high-risk patients in this real-world service program implementation. KFH uses a direct integration model for providing EMHI services such that the CHW serves as a direct extension of the pediatric office by improving patient–provider communication and reinforcing provider recommendations outside of the office and hospital setting.16,17

    As previously described, KFH CHWs are paraprofessionals with at least a high school education selected from the communities they serve.18 CHWs are matched with families based on the parent/caregiver’s age, experience with the child’s medical conditions, and cultural and linguistic backgrounds. CHWs receive >60 hours of initial training from clinical and program staff that includes home visiting content, program procedures, and preventive health education topics. On average, CHWs visit families monthly for a 45- to 60-minute home visit. After each home visit, CHWs send summary notes to clinicians for review and entry into the child’s electronic medical record. These summary notes help clinicians prepare for the family’s next visit by documenting the support provided and the family and/or CHW concerns regarding the child’s health.

    In addition, CHWs may help families keep track of a child’s medical information, appointments, insurance information, and medications; communicate with primary medical physician and specialists; or troubleshoot issues between office visits (eg, transportation, medication refills). The program is specifically designed to meet the unique needs of each individual family by tailoring the number of home visits and the intensity of the services provided. At the home visits, the CHWs use tablet computers preconfigured with a Research Electronic Database Capture protocol that guides visit content based on the family’s strengths and needs.19 Overall, these visits aim to provide health education and anticipatory guidance, promote healthy decisions and lifestyles, improve disease management, and empower parents and children by engaging them in their own health care and teaching them ways to navigate the health care system. To date, KFH CHWs have provided >9900 home visits for >2100 children aged 0 to 18 years since the program’s launch.

    In the present study, we determined the program’s impact on ED visits and hospitalizations for ACSCs. All research activities were approved by Stony Brook University’s institutional review board. Enrollment occurred in the ambulatory practice setting or via telephone after an ambulatory office visit. The usual care group, a convenience comparison sample (705 children aged <18 years), was actively recruited by trained staff from 1 SBC primary care practice in Setauket, New York, between July 2011 and February 2012. Intervention group participants were defined as having at least 1 home visit by a KFH CHW. The number of home visits among those in the intervention group ranged from 1 to 10 (median, 4; 25th percentile, 2; 75th percentile, 6). The comparison sample was not offered EMHI services, and they continued to receive usual care from their health care provider without any CHW support.

    Statewide data for ED visits and hospitalizations were abstracted from the New York State Statewide Planning and Research Cooperative System database for January 1, 2011, through December 31, 2013.20 Only those children who had at least 12 months of postenrollment follow-up by December 31, 2013, were included in the analyses. This approach resulted in an analytic sample of 922 participants (697 usual care; 225 intervention).

    Measures

    Primary Outcome

    Our primary outcome measures were as follows: (1) number of preventable ED visits; and (2) number of preventable hospitalizations during the 12-month follow-up period for each participant. From the literature, 21 ACSCs were identified for use in this analysis of preventable ED visits and hospitalizations (Supplemental Table 5).2127 The number of preventable ED visits was further classified into 3 ordinal categories (none, 1, and ≥2) for subsequent analyses because these data had a heteroscedastic distribution. Similarly, the number of preventable hospitalizations was classified into 2 groups (none versus any) because these data also had a heteroscedastic distribution.

    Primary Predictor

    Study group was the primary predictor (intervention versus usual care).

    Covariates

    Covariates known to influence ED and hospital utilization were collected by using a validated family psychosocial risk questionnaire.21,24,28,29 These variables included characteristics of the child, family, home and/or child care environment, and neighborhood. These data were collected in the field by KFH staff (ie, by CHWs for the intervention group and trained research assistants for the usual care group) at the time of enrollment via tablet computer and automatically uploaded to the Research Electronic Database Capture system by using electronic case reporting forms. Each form was reviewed by the KFH team member collecting the data and signed electronically before submission in real-time.

    Statistical Analyses

    Differences in participant demographic characteristics between the usual care and intervention groups were examined by using standardized differences in the original sample. Conventionally, a standardized difference >0.2 indicates a significant discrepancy between 2 groups. We estimated the adjusted population-average difference in the number of preventable ED visits and hospitalizations by using propensity score matching methods to adjust for significant differences in the original sample between the 2 groups in a number of participant characteristics. Propensity scores were first calculated based on 26 variables, including child, family, and neighborhood characteristics, using a logistic regression model. The matched sample was then created by using propensity score matching based on nearest-neighbor 1-to-1 matching with replacement. In the matched sample, usual care and intervention groups have the same number of participants. We then examined the difference in demographic characteristics between the 2 groups in the matched sample using standardized differences. Crude analyses were performed on the difference in the number of preventable ED visits and preventable hospitalizations between the 2 groups, using a Wilcoxon ranked sum test with normal approximation to its test statistic and the Wald test based on the logistic regression model, respectively. We then performed analyses on the difference in number of preventable ED visits and preventable hospitalizations, using the matched samples and applied the bootstrapped Wilcoxon signed rank test with normal approximation to its test statistic and the Wald test for logistic regression.

    Missing data included ∼13% missing in income; no data were missing for child sex, child race/ethnicity, child age, and insurance. All other variables had <5% missing. To alleviate the effect of missing data, multiple imputation according to the chained equations method was performed before matching. The final results were based on analyses conducted on 10 imputed data sets.30,31 The variance of the test statistics combined both within-imputation variances and cross-imputation variance. All analyses were performed by using SAS version 9.3 (SAS Institute, Inc, Cary, NC).

    Results

    The primary analytical sample before propensity score matching included 922 participants, 225 of whom were in the intervention group and 697 were in the usual care group. Table 1 summarizes the characteristics of all 922 participants. After propensity score matching, the analytical sample included 450 participants: 225 participants in the usual care group and 225 participants in the intervention group. Table 2 summarizes the characteristics of the matched participants. The majority of children were aged <1.5 years at enrollment; sex distribution was nearly equal. After propensity score matching, more than one-half of children were Hispanic (59%), 80% had Medicaid/self-pay, and 15% had special health care needs. After matching, intervention participants remained less likely than the control group to have mothers rate their health rate as excellent/very good/good and read to their children daily.

    TABLE 1

    Original Sample Characteristics Before Propensity Score Matching

    TABLE 2

    Sample Characteristics After Propensity Score Matching

    Comparison results based on the unmatched and matched sample are displayed in Tables 3 and 4. The unadjusted results in Table 3 show that similar proportions of participants in both the control and intervention groups visited the ED ≥1 time (18.8% vs 18.2%, respectively). In the matched sample, however, 18.2% of participants in the intervention group had ED visits compared with 35.1% of control participants. In fact, participants in the intervention group had significantly fewer preventable ED visits than the control group in the matched sample (z = −2.66; P = .004). The most common ACSCs for ED visits were severe ear, nose, and throat infections (54.8%), asthma (16.5%), and seizure disorders (5.5%).

    TABLE 3

    Number of Preventable ED Visits in 12-Month Study Follow-Up Period, Unadjusted and Matched Samples, for EMH Intervention and Usual Care Groups

    TABLE 4

    Preventable Hospitalizations, Unadjusted and Matched Samples, for EMH Intervention and Usual Care Groups

    No differences were found between the control and intervention samples for preventable hospitalizations in either the unmatched or matched samples. Similar proportions of participants in both the control and intervention samples were hospitalized. In the unmatched sample, 17 usual care participants (2.4%) and 7 intervention participants (3.1%) were hospitalized. In the matched sample, 7 participants (3.1%) in each group were hospitalized. The top 2 ACSCs for hospitalizations were the same as those for ED visits (ie, severe ear, nose, and throat infections [22.6%] and asthma [23.2%]) followed by dental conditions (13.8%) and bacterial pneumonia (9.0%).

    Discussion

    Our findings illustrate the positive impact of this EMHI using trained CHWs on decreasing preventable ED visits. Inappropriate ED utilization remains a formidable and costly challenge to the US health care system. Although the CHW intervention has been proposed as a means to improve health outcomes among disadvantaged children, to the best of our knowledge few studies have examined the effects of such interventions on ED visits. Our findings show that this EMHI using trained CHWs directly integrated into the primary care medical home is one way to reduce ED utilization for preventable conditions. These findings are particularly important as the US health care system moves increasingly toward a value-based purchasing environment.

    The long-term and financial implications of our EMHI’s impact on reducing ED visits in the matched results have the potential to be substantial. Although the present study focused on health care utilization, our EMHI includes a broad array of support services that are likely to have enduring and wide-ranging influence on health behaviors and outcomes. Our team’s qualitative analyses found that the tasks of the CHWs most consistently focused on the following areas: medical appointment logistics, medication maintenance, health education, facilitating families’ understanding of the physician’s recommendations, and connecting families with local resources.33 With this approach, KFH aims to have a far-reaching impact on health and well-being and support family self-sufficiency in the long term by giving families tools to improve their navigation of our complex, often-confusing health care system. Rigorous evaluation of the cost-effectiveness of this EMHI would be a useful extension of our study.

    Although extensive literature exists to support the positive impact of CHW interventions, interventions vary substantially.34,35 CHWs assume a variety of roles and job titles across settings, such as lay health worker or patient navigator.16,17 Coker et al16 recently evaluated the impact of a new model for pediatric well-child visits on ED utilization by using a multicomponent intervention that included a practice-based master’s level patient educator to deliver anticipatory guidance. In this model, children in the intervention group were less likely than the control group to have ≥2 ED visits (intervention, 10.4%; usual care, 21.6%; effect size, 0.47), but there was no difference between groups in well-child visit or sick/urgent care utilization. In addition, this analysis could not separately quantify the impact of the patient educator versus the other intervention components on ED visits.16 In another model, ED-based patient navigators approached adults during “primary care related–ED visits” in-person or via telephone shortly thereafter to provide basic health education, connect patients with needed local resources, and make 1 telephone call within 3 to 10 days to follow up and provide additional support.17 In this adult sample, the intervention group had significantly fewer ED visits 12 months’ postintervention (0.9 fewer ED visit preintervention vs postintervention; P < .05) but not at 24 months’ postintervention.17 Our study contributes to this literature by reporting a 50% reduction in ED use among a population with risk factors (eg, 80% Medicaid/uninsured) known to be associated with increased ED use and other adverse health outcomes.9,21

    In contrast to these models, KFH uses an EMHI model that directly integrates trained lay-person CHWs with pediatric medical homes to facilitate adherence to recommended clinical care. This study specifically determined the positive impact of home visits by CHWs on reducing ED utilization. Further research is warranted to consider the impact of other intervention components, such as number of home visits, length and content of home visits, and remote support, on ED visits and hospitalizations. Models combining CHWs with other professionals (eg, nurses, social workers) will likely be needed to have a substantial impact on a range of health outcomes at the population level. Creating a seamless patient experience that leverages the strengths of proven models will be challenging in our existing competitive, multipayer health care environment. To that end, the Center for Medicare & Medicaid Innovation is rigorously evaluating Accountable Health Communities models designed to address certain health-related social needs on a large scale.36

    The present study has several limitations. Although a randomized controlled trial design is the most robust approach to control for confounding factors, our study evaluated an existing EMHI in which clinicians referred participants they deemed to be “high risk.” To address this issue, we used propensity score matching to correct for significant discrepancies in participant characteristics between the intervention and the usual care groups. Analysis of program operational data revealed selection bias (program participation rates are ∼62% among those referred) that is comparable to other intervention studies, including randomized controlled trials, among vulnerable populations.18,37 Thus, our study mirrors what is expected from real-world implementation of this program for high-risk populations. Based on anecdotal data from referring clinicians and KFH staff, we know that patients who decline enrollment typically are from unstable housing situations, have complex mental illnesses, and often perceive they do not need help. Notably, these same groups are also likely to benefit from the KFH program, and further investigation of ways to promote participation among these groups is needed.

    Another limitation of the present study is its generalizability because our study population was derived from communities solely in Suffolk County, New York. Furthermore, our hospitalization analysis was limited by sample size. Replication in larger sample populations and in different geographic regions would advance the evidence base about the impact of this model. Finally, we did not have data about access to outpatient care or continuity of care and were unable to determine the precise mechanisms by which decreased ED visitation occurred. Although we believe that our outcomes can be attributed to improved communication between families and their primary care physicians, enhanced coordination of care, and improved adherence to preventive recommendations, a more rigorous analysis of the components of the program (eg, number, length and content of visits, remote support) is warranted.

    Conclusions

    In the post–Patient Protection and Affordable Care Act era, national efforts to provide comprehensive, high-quality care for populations via robust CHW interventions have yet to be adequately evaluated. Our findings contribute to a growing evidence base regarding the impact of CHW interventions on improving health outcomes. Specifically, KFH as a structured EMHI delivered by CHWs is an effective model to reduce the likelihood of ED visitation for ACSCs among children at high risk. Application of this model nationally has the potential to lead to significant cost savings by reducing ED utilization. In extension, our EMHI model has the potential to serve as a blueprint for other populations as US health care evolves to an increasingly value-based purchasing environment.

    Acknowledgments

    We thank our CHWs and the participating families for their contributions to operating the KFH program. We also thank Hua Wang and Minqin Chen for their administrative and technical contributions to the project. The CHWs received financial compensation for their efforts via employment contracts administered by the Federation of Organizations.

    Footnotes

      • Accepted February 3, 2017.
    • Address correspondence to Susmita Pati, MD, MPH, Division of Primary Care Pediatrics, State University of New York at Stony Brook, Health Sciences Center T11-020, Stony Brook, NY 11794-8111. E-mail: susmita.pati{at}stonybrookmedicine.edu
    • FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.

    • FUNDING: The program was initially funded by a New York State Department of Health Health Care Efficiency and Affordability Law Phase 6 (HEAL-NY) Primary Care Infrastructure grant. Currently, the program is supported by surplus funds from Stony Brook’s Hospital–Medical Home Demonstration Initiative. Research activities were funded by the Stony Brook School of Medicine Dean’s Targeted Research Opportunity award and the Department of Pediatrics Research Award (Principal Investigator: Dr Pati).

    • POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.

    References