The Following Baby Back Home (FBBH) home visiting program supports families of high-risk low birth weight preterm infants after discharge from a hospital NICU. This study compares the health care use, immunization, and infant mortality rate of low birth weight preterm infants enrolled in FBBH with similar infants not in the program.
From January 2013 to December 2017, 498 children enrolled in FBBH were identified in Arkansas vital statistics records and the Arkansas All-Payer Claims Database. Infants in FBBH were matched with children in a control group on the basis of demographics and medical conditions of the infant. Generalized linear mixed models with double propensity-score adjustment were used to estimate program effects.
In the first year after discharge and compared with a propensity-score matched cohort of control infants, those enrolled in FBBH were significantly more likely to have higher numbers of medical appointments and more compliant immunization history. The odds of dying in the first year of life for control infants was 4.4 times (95% confidence interval: 1.2–20.7) higher than those managed in the program.
A goal of the FBBH home visiting program is to work with parents to educate and support them as they care for their medically fragile infants. We conclude that education and support was instrumental in the infant health care use and outcome differences we observed during the first year of life.
Low birth weight preterm infants with NICU stays are at increased risk for developmental and behavioral morbidities and mortality. Home visiting programs can ensure that families have parenting education, support, and linkages to other public and private services.
In the first year of life, compared with propensity-score matched control group infants, infants in the Following Baby Back Home program had more medical appointments and immunizations and had a lower likelihood of infant mortality.
Low birth weight preterm (LBWPT) infants with NICU stays are at increased risk for developmental and behavioral morbidities and mortality.1 LBWPT infants present challenges to their families because of their medical fragility. The time of transition home from NICU is a time of particular importance during which parents often require support for understanding their infant’s special medical needs2–4 and health care coordination.5,6 Parents of LBWPT infants often face challenges that heighten their need for support, because of increased risks for financial and family stress,7 anxiety, and depression,7,8 which may lead to additional challenges in parenting.9–11
Home visiting can provide families with parenting education, emotional and informational support, and linkages to other resources and services. Home visiting programs targeting families of LBWPT children have been shown to improve some aspects of parenting and parent–child interaction,12–16 as well as children’s cognitive development and growth.14 However, there is inconsistent evidence that home visiting improves health morbidities and infant mortality for LBWPT infants. Some studies have reported significant reductions in hospital and emergency department (ED) use,17–19 whereas others have null or inconsistent impacts as children age.20,21 Programs for LBWPT infants have reported positive impacts for immunization adherence,20,22 but not for mortality.17
Differences across studies limit the ability to draw conclusions about the efficacy of home visiting for LBWPT infants. Variations include heterogeneity in the population (eg, sample sizes, definitions of sample), the study design (eg, experimental, quasi- and nonexperimental), and the interventions (eg, duration, intensity, and staff qualifications). Despite similar heterogeneities found in the general home visiting literature, there is some evidence that interventions provided by nurses and nurse–social worker teams can positively impact child health.23–29
The Following Baby Back Home intervention (FBBH), delivered by registered nurse and licensed social worker teams, supports the families of LBWPT infants after discharge from the NICU. Researchers in one previous study of FBBH employed a quasi-experimental comparison of Medicaid claims and a nonexperimental comparison of immunizations and infant mortality.30 Findings revealed that a greater proportion of FBBH infants received >1 routine and >1 nonroutine medical visit than matched comparisons but no differences in hospitalizations or ED use. A nonexperimental comparison revealed a positive impact of FBBH on immunizations and infant mortality.
The objectives with the current study are to examine whether FBBH infants, compared with matched controls, demonstrate increased health care use and immunizations and lower infant mortality from the time of NICU discharge through their first birthday. We hypothesize that rates of health care use, including well-child, specialist visits, and immunizations, will be higher and that mortality will be lower for infants in FBBH than for controls.
Methods
The FBBH Intervention
The goal of FBBH is to maximize the health and development of LBWPT (<2500 g and <37 weeks’ gestation) infants. FBBH teams assist families in providing treatments at home, track adherence to medical appointments and immunizations, and facilitate medical appointments as needed.30 FBBH monitors infants’ health issues, measures growth, conducts developmental screenings, and completes referrals for evaluation and treatment when health or developmental concerns are identified. FBBH teams conduct family assessments to identify needs and, when identified, provides direct education using the Parents as Teachers curriculum31 and referrals for community services.
Eligible subjects for FBBH are LBWPT infants with chronic medical conditions at the time of discharge as identified by a neonatologist. Referrals to FBBH are solicited from all NICUs in Arkansas with most originating from level IIIB and level IV facilities that serve infants from across the state. Recruitment and informational materials are provided to NICUs, which are provided to families on referral.
Services were initiated with 44.5% of referred families. Reasons for not receiving FBBH services included family refusals (21.9%), inability to contact (17.0%), living outside the service areas (8.6%), and full caseload (7.6%). FBBH teams are located in 6 locations across the state, each serving serve up to 45 families (total 270 families). A central administrative team provides model training and monitors fidelity. FBBH provides 2 home visits per month for the first 2 months after enrollment as close as possible to the transition to home, 1 home visit per month until age 1, and alternating home and phone and/or virtual visits every month until age 3. The largest proportion (40%) of families in FBBH remain enrolled at least 1 year, with 23.4%, 22.5%, and 14.1% enrolled <3 months, 3 to 6 months, and 6 to 12 months, respectively.
Data Sources
This study required accessing data included in the Arkansas Transparency Initiative and linking with FBBH participant data by using an anonymous individual-level 44-character, alphanumeric-scrambled identifier constructed by using the last name and date of birth. The following data sources are contained in linked data for this study.
Following Baby Back Home
Participant information and program services (ie, dates of referral, first and last home visit, and discharge) for participants were entered in a Research Electronic Data Capture32 database maintained by FBBH.
Arkansas All-Payer Claims Database
The All-Payer Claims Database (APCD) is a database securely administered by Arkansas Center for Health Improvement and governed by the Arkansas Insurance Department. The APCD includes hospital, ED, office-based, and pharmacy claims for Arkansans with health coverage through nonself-insured private payers, Medicaid, and Medicare. Data for this study were available from January 2013 through December 2018.
Vital Statistics Records
Data fields from birth certificate records were used for propensity-score matching (PSM) FBBH and non-FBBH infants. Death certificate records were used to identify a main outcome of infant mortality in FBBH and non-FBBH matched cohort populations.
FBBH Study Population
APCD and birth certificate records from 2013 to 2017 were used to link with FBBH subjects and to identify a control group. Figure 1 depicts the derivation of the FBBH study population. A further exclusion of controls to match characteristics of FBBH participants required that all infants were discharged from the hospital from the NICU.
Study Outcome Measures
The study outcomes included mortality status, immunization visits, and health use in the first year of life. The number of immunization visits were assessed as a proxy for pediatric immunization series completion. In the first 15 months after birth, guidelines recommend a minimum of 5 encounters (not including influenza vaccination) to complete the immunization series, with the last round of immunizations administered between 12 and 15 months of age.33 We defined immunization coverage as complete if an infant had a minimum of 4 immunization encounters in the first year of life. Health care use measures included the number of hospitalizations post-NICU discharge, ED visits, ED visits for nonurgent reasons (eg, visits coded as evaluation and management), wellness visits, non-wellness outpatient visits, and number of filled prescriptions covered through commercial insurance or Medicaid. Use outcomes were defined by using the Healthcare Effectiveness Data and Information Set Value Set Directory as shown in Supplemental Table 4.34
For health care use outcomes, we further restricted the study population depicted in Fig 1 and potential controls by requiring a continuous enrollment in APCD for 1 year with 1 allowable gap of 30 days to ensure that we captured the entirety of health care uses experienced during the first year of life. The final study population for health care use outcomes consisted of 473 FBBH enrollees.
Main Predictor and Covariates
The main predictor is receipt of the FBBH intervention. Covariates adjusted in analytic models included sex, hospital length of stay after birth, type of health care coverage, mother’s race and ethnicity, and facility level of longest hospital length of stay after birth (using the Arkansas Perinatal Regionalization Levels of Care; I: basic service level to IV: most complex services available), mother’s marital status, rural–urban commuting area code assigned to mother’s zip code at the time of delivery, mother’s smoking status during pregnancy, zip code level median household income quartiles (obtained from Census data from zip code at the time of delivery), weight of the child at birth, Apgar score 5 minutes after birth (range 1 [poor] to 10 [good]), and clinical characteristics observed in the NICU (presence or absence of newborn respiratory disease, history of intraventricular hemorrhage, convulsions, critical congenital heart disease, chromosome abnormality, and use of gastrostomy tube). Codes are provided in Supplemental Table 4.
Analytic Approach and Model Predictor and Covariates
In this study, we used a cross-sectional prospective quasi-experimental design, with a matched FBBH (treatment) to non-FBBH (control) population. This design is appropriate given the pragmatic limitation of randomizing treatment assignment. In a well-balanced match, where FBBH and non-FBBH are systematically similar (across demographic, socioeconomic, and underlying medical conditions of infants), differences between FBBH and non-FBBH outcomes will determine the effectiveness of the FBBH treatment. An intention-to-treat approach was taken because FBBH infants may have been enrolled in the intervention but parents and guardians may have subsequently declined to receive services at any time after enrollment. This approach will produce conservative analytic results.35
At the first analytic stage, a one-to-one FBBH and non-FBBH infant matching strategy was implemented by using a combination of exact category and greedy nearest neighbor PSM. To be matched, both infant groups had to be identical in the following categories: sex, NICU length-of-stay category, mother’s race and ethnicity, and facility level of longest stay. In addition, by using SAS (SAS Institute, Inc, Cary, NC) proc psmatch,36,37 both groups were matched by propensity score (probability of each group being assigned to the FBBH treatment was equal or nearly equal based on all remaining covariates included in the logistic regression model). The greedy nearest neighbor matching algorithm attempts sequential matches between FBBH and non-FBBH infants where the within-pair propensity-score difference is the smallest. In total, 418 controls were matched among the 473 FBBH study population (88.4%) for health care use outcomes, and 485 controls were matched among the 498 FBBH study population (96.8%) for mortality outcome.
At the second analytic stage, we used a bias correction approach by implementing double propensity-score adjustment38,39 to estimate average treatment effects (ATEs), which captures the difference in the expected value of the outcome by treatment and control groups. All covariates used in the propensity-score estimates were included in the double adjustment. We performed nonparametric bootstrap with 2000 resampling runs to obtain confidence intervals (CIs).40
For health care use outcomes, generalized linear integrated mixed models for each outcome were fitted. An indicator of each FBBH and non-FBBH one-to-one matched dyad was included in the mixed models as a random variable. Count distribution outcome models were fitted by specifying a negative binomial distribution (chosen over a Poisson distribution due to significant overdispersion parameter estimates) with a log-link function. A binomial distribution with a logit-link function was used for binary outcomes. Differences in FBBH and non-FBBH use were determined using ATE, incidence rate ratios (IRRs, for Poisson count), adjusted odds ratios (aOR, for binary outcomes), and P values.
For infant mortality, Firth’s logistic regression model with added covariates method41,42 was applied to compute aOR, as well as the predicted probability used in the calculation of ATE. Firth’s logistic regression is a standard method applied to analyze rare events with small samples and is appropriate to estimate aOR; however, it is known to produce biased predicted probability.43 Puhr and colleagues41 proposed Firth’s logistic regression model with added covariates to correct for this bias by adding the “ghost factor” in the logistic regression and separating the pseudodata from the original data.
This study received data use approval from the University of Arkansas for Medical Sciences Arkansas Biosciences Institute council, as well as the Arkansas Insurance Department Commissioner. This study was also approved under University of Arkansas for Medical Sciences Institutional Review Board (protocol number: 134415). Families consented to participate.
Results
Table 1 presents a comparison of demographic and birth characteristics of those who were excluded and retained in the study based on data availability. With the exception of greater specificity in race and ethnicity (ie, fewer infants assigned to other category) in the study population, no differences were detected.
The characteristics for treatment and control groups are provided in Table 2. For the PSM comparison, SD and variance ratio (VR) statistics are presented. With the exception of median income quartiles, all variables postmatching are balanced across both infant study groups (SD <0.15 and VR within the recommended range of 0.5 to 2.0).37 The PSM resulted in balanced groups for additional family demographics (eg, maternal education, Women, Infants, and Children Special Supplemental Nutrition Program [WIC] receipt) and covariates of preterm birth. There was one difference, favoring the controls, in prenatal risk for preterm birth (see Supplemental Table 5).
Table 3 presents differences across FBBH treatment and non-FBBH control matched population for main outcomes under study using generalized linear integrated mixed models.
Health Care Use
Compared with control infants, FBBH infants had higher use across all health care outcomes. FBBH infants were more likely to have been hospitalized (aOR = 2.31; 95% CI: 1.6–3.30) with a higher rate of hospitalization (IRR = 2.68; 95% CI: 1.82–3.95). The average length of stay for FBBH infants was 9.9 (SD = 21.95; range 1–257) compared with 10.6 (SD = 21.05; range 1–218) for control infants. FBBH infants also experienced more hospitalizations (mean = 0.11, SD = 0.02) in the month after NICU discharge than controls (mean = 0.04, SD = 0.01).
When examining ED visits, FBBH infants were more likely to have used the ED at least once (aOR = 2.19; 95% CI: 1.66–2.89) with a higher rate of visits (IRR = 1.81; 95% CI: 1.46–2.23). Infants in FBBH also experienced more ED visits (mean = 0.25; SD = 0.03) within the month after discharge from NICU than controls (mean = 0.14; SD = 0.02). Examining ED visits coded as evaluation and management, FBBH infants were more likely to have sought nonurgent emergency care at least once (aOR = 2.21; 95% CI: 1.66–2.94) with a higher rate of visits (IRR = 1.92, 95% CI: 1.51–2.43).
Infants in FBBH also had a higher rate of wellness and non-wellness office-based visits (IRR = 1.59; 95% CI: 1.37–1.85 and IRR = 1.81; 95% CI:1.23–2.65), and a higher rate of prescription medications (IRR = 1.24; 95% 95% CI: 1.01–1.43) than infants not receiving the FBBH treatment.
Immunizations
FBBH infants had a higher rate of immunizations (IRR = 1.49; 95% CI: 1.49–1.65) than infants not receiving the FBBH treatment, with a higher proportion of FBBH infants having had 4 or more immunization visits in the first year of life (aOR = 2.37; 95% CI: 1.80–3.14).
Infant Mortality
Compared with FBBH infants, a higher proportion of control infants did not survive their first year of life. Specifically, the odds of death for the control infants were 4.44 times higher (95% CI: 1.22–20.67) than those served in FBBH, with 7 (1.44%) deaths in the control group and 1 (0.21%) death in the FBBH intervention group. Reasons for death for non-FBBH infants included sudden unexpected infant death (n = 3), other specified cerebrovascular disease (n = 1), congenital hydrocephalus (n = 1), pachygyria (n = 1), and multiple congenital malformations (n = 1). For the infant served by FBBH, the reason for death was adrenocortical insufficiency.
Discussion
Home visiting provides a mechanism to support families with LBWPT infants as they transition from hospital to home. Our findings reveal that FBBH had the effect of significant increasing all types of health care use and immunizations. The program also lowered the risk of infant mortality in the first year of life where 7 infants in the non-FBBH matched control group died compared with 1 infant enrolled in FBBH.
The current study extends the literature of home visiting for LBWPT infants. FBBH home visiting teams work with parents to educate and support them as they care for their LBWPT infants, to assure that they receive the medical care necessary to maintain the health of their infants, and to facilitate compliance with their medical regimens. Our findings contribute novel information, being among the first to document increases in wellness and non-wellness clinical visits for infants supported by home visiting. Unlike some of the literature examining home visiting for LBWPT infants, which reports reductions in hospitalizations and ED use for home-visited infants,17–19 our findings document increased hospitalizations and use of emergency care for infants in FBBH. Increased ED use in a medically complex group is expected, and in many cases, warranted.44 Our data also suggest that ED visits of FBBH infants were more likely to be for evaluation and management, which may be representative of enhanced, condition-specific education that families receive. Similar findings have been reported by another home visiting program for LBWPT infants, where hospital and ED use is higher when infants are younger but decreases as children age.45 A nurse–social worker home visiting program for low-income families has also resulted in increased ED use for more minor reasons.29
FBBH positively impacted the health of medically fragile LBWPT infants when measured by immunizations and mortality. There is evidence from the LBWPT home visiting20,22 and nurse home visiting literature that programs positively impact immunization adherence.46 In addition to in-home monitoring, FBBH infants also had more contact with primary health care providers, increasing opportunities for immunizations. Our study is the first to report positive impacts on infant mortality as a result of home visiting for LBWPT infants.17 We speculate that the FBBH’s careful monitoring of the health of these infants and increased use of health care likely contributed to the lower infant mortality rate.
Our study has several strengths. The quasi-experimental PSM design is rigorous and accounts for potential selection bias. Our analysis also employed an intention-to-treat design, the most conservative method for demonstrating program impacts. Our outcomes were based on Medicaid and private insurance medical claims, increasing generalizability over previous studies. In addition to strengths, there are also limitations. We were restricted in our ability to incorporate analysis on all infants who received FBBH. Population exclusions resulted from infants being younger than age one, dates of birth falling outside the age range of APCD data, and multiples of the same sex not being uniquely identifiable within the APCD. An additional limitation inherent in the use of administrative data are missing data, for which we made equivalent matches in the propensity scoring algorithm. Finally, APCD billing codes limited analysis to the number of visits in which immunizations occurred, rather than the recommended immunization series.
Conclusions
Results suggest that nurse–social worker home visiting teams following carefully designed intervention protocols can positively affect the health of LBWPT infants. Much effort is spent to keep infants alive in the NICU. The support provided by FBBH appears to increase families’ use of health care and prevent the deaths of medically fragile LBWPT infants.
Acknowledgments
Faculty and staff in the Department of Pediatrics in the College of Medicine at the University of Arkansas for Medical Sciences designed and implemented the FBBH intervention program. The authors are grateful for their cooperation in completing this evaluation. The Arkansas Center for Health Improvement is a health policy organization that maintains and administers the Health Data Initiative and the Arkansas Transparency Initiative data warehouses. Contained within the Arkansas Transparency Initiative data warehouses are the Arkansas APCD and Arkansas Department of Health Vital Records. Access to the APCD for this study was provided by support from the Arkansas Biosciences Institute/Arkansas Insurance Department/Arkansas Center for Health Improvement Collaboration. The authors thank J. Mick Tillford, PhD, Joseph W. Thompson, MD, MPH, and Ping Hu, MS, for their analytic support and review of the manuscript in development. Funding for FBBH services is provided through the Arkansas Department of Human Services, Division of Medical Services and the Health Resources and Services Administration – Maternal, Infant, and Early Childhood Home Visiting Program.
Dr McKelvey conceptualized and designed the study, drafted the initial manuscript, and reviewed and revised the manuscript; Dr Lewis designed the methodology, conducted the analyses, drafted the initial manuscript, and reviewed and revised the manuscript; Drs Beavers and Casey supervised the intervention, conceptualized the construction of outcomes, and reviewed and revised the manuscript; Ms Irby supervised the intervention and the collection of data about intervention group participants, and reviewed manuscript; Dr Goudie designed the methodology, drafted the initial manuscript, and reviewed and revised the manuscript; and all authors approved the final manuscript as submitted and agree to be accountable for all aspects of the work.
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
FUNDING: Funding for this study was made in part possible by the Health Resources and Services Administration of the US Department of Health and Human Services under grant number X10MC29461 of the Affordable Care Act – Maternal, Infant, and Early Childhood Home Visiting Program, awarded to the Arkansas Department of Health. Support has also been provided in part by Arkansas Biosciences Institute, the major research component of the Arkansas Tobacco Settlement Proceeds Act of 2000. The funder did not participate in the work. The information, content, or conclusions expressed in this material are those of the author(s) and should not be construed as the official position or policy of, nor should any endorsements be inferred by, Health Resources and Services Administration, US Department of Health and Human Services, or the US Government.
POTENTIAL CONFLICT OF INTEREST: The authors have no conflicts of interest relevant to this article to disclose.
COMPANION PAPER: A companion to this article can be found online at www.pediatrics.org/cgi/doi/10.1542/peds.2020-049747.
- aOR
adjusted odds ratio
- APCD
All-Payer Claims Database
- ATE
average treatment effect
- CI
confidence interval
- ED
emergency department
- FBBH
Following Baby Back Home
- IRR
incidence rate ratio
- LBWPT
low birth weight preterm
- PSM
propensity-score matching
- VR
variance ratio
- WIC
Women, Infants, and Children Special Supplemental Nutrition Program