A Mixed Methods Evaluation of Early Childhood Abuse Prevention Within Evidence-Based Home Visiting Programs

A Mixed Methods Evaluation of Early Childhood Abuse Prevention Within Evidence-Based Home... Objectives In this large scale, mixed methods evaluation, we determined the impact and context of early childhood home visit- ing on rates of child abuse-related injury. Methods Entropy-balanced and propensity score matched retrospective cohort analy- sis comparing children of Pennsylvania Nurse–Family Partnership (NFP), Parents As Teachers (PAT), and Early Head Start (EHS) enrollees and children of Pennsylvania Medicaid eligible women from 2008 to 2014. Abuse-related injury episodes were identified in medical assistance claims with ICD-9 codes. Weighted frequencies and logistic regression odds of injury within 24 months are presented. In-depth interviews with staff and clients (n = 150) from 11 programs were analyzed using a modified grounded theory approach. Results The odds of a healthcare encounter for early childhood abuse among clients were significantly greater than comparison children (NFP: 1.32, 95% CI [1.08, 1.62]; PAT: 4.11, 95% CI [1.60, 10.55]; EHS: 3.15, 95% CI [1.41, 7.06]). Qualitative data illustrated the circumstances of and program response to client issues related to child maltreatment, highlighting the role of non-client caregivers. All stakeholders described curricular content aimed at prevention (e.g. positive parenting) with little time dedicated to addressing current or past abuse. Clients who reported a lack of abuse-related content supposed their home visitor’s assumption of an absence of risk in their home, but were supportive of the introduction of abuse-related content. Approach, acceptance, and available resources were mediators of successfully addressing abuse. Conclusions for Practice Home visiting aims to prevent child abuse among high-risk families. Adequate home visitor capacity to proactively assess abuse risk, deliver effective preventive curriculum with fidelity to caregivers, and access appropriate resources is necessary. Keywords Home visiting · Maternal and child health · Child maltreatment · Mixed methods research Significance Statement What is already known on this topic? Reducing child mal- treatment is a public health priority. At present, home visit- ing represents the primary available prevention strategy for Electronic supplementary material The online version of this article (https ://doi.org/10.1007/s1099 5-018-2530-1) contains supplementary material, which is available to authorized users. * M. Matone Department of Family Medicine and Community Health, matonem@email.chop.edu University of Pennsylvania, Philadelphia, PA, USA Center for Public Health Initiatives, University PolicyLab, The Children’s Hospital of Philadelphia, of Pennsylvania, Philadelphia, PA, USA Philadelphia, PA, USA Leonard Davis Institute of Health Economics, University Division of General Pediatrics, The Children’s Hospital of Pennsylvania, Philadelphia, PA, USA of Philadelphia, Philadelphia, PA, USA Roberts Center for Pediatric Research, 2716 South Street, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA 19146, USA Philadelphia, PA, USA Mixed Methods Research Lab, University of Pennsylvania, Philadelphia, PA, USA Vol.:(0123456789) 1 3 S80 Maternal and Child Health Journal (2018) 22 (Suppl 1):S79–S91 child maltreatment; however, evaluations of home visiting visitation programs found two programs reduced long-term success on this outcome have been varied. child maltreatment reports or death contrasted by four other What does this study add? Our mixed methods evaluation randomized trials showing no associated effect of home pairs a large administrative dataset that allows us to measure visitation on child maltreatment reports (Nelson et al. 2013; child abuse outcomes reliably across multiple home visiting Rubin et al. 2014). programs with qualitative interviews with key stakehold- As home visiting programs have scaled in the context ers to demonstrate and explore home visiting’s impact on of the federal Maternal, Infant, and Early Childhood Home child maltreatment. We find no evidence of positive pro- Visiting program (MIECHV) (“Patient Protection and gram effects but identify explanatory implementation factors Affordable Care Act” 2010), it is important to refine our potentially limiting program effectiveness. understanding of their role and mechanism, and opportuni- ties for maltreatment prevention. In this large scale, mixed methods evaluation, we aimed to estimate the effectiveness Introduction of three MIECHV-funded home visiting models in Penn- sylvania (PA) on early childhood maltreatment ascertained Child maltreatment is a serious public health problem (Chil- from clinical diagnoses in emergency department and inpa- dren’s Bureau 2016) resulting in innumerable short and tient healthcare encounters. A concurrent qualitative anal- long-term health consequences including trauma, adverse ysis of interviews with program staff and clients explores physical and mental/behavioral health, changes to brain the content of and response to curriculum related to child architecture and development, challenges to educational maltreatment. achievement, and reduced social-emotional functioning and relational attachment (Deutsch et al. 2015; Merrick and Latzman 2014). Unfortunately, prevention is challenging to Data and Methods achieve in public health programming, in part due to the level of support needed to overcome risk factors and bolster This study was performed within the Commonwealth of protective factors for families and communities (Agran et al. Pennsylvania (PA) MIECHV evaluation. The study followed 2003; Frioux et al. 2014; Wood et al. 2012). Early childhood a partially mixed, concurrent, equal status design, in which maltreatment prevention programs (i.e. home visitation) qualitative and quantitative data were analyzed separately intervene at a critical period when children are most at risk and mixed at the stage of interpretation (Leech and Onwueg- (Schatz and Lounds 2007) and promote prevention though buzie 2009). The study was approved by PA’s Department of strengthening protective factors within a family and con- Human Services with human subjects approval by the Chil- necting families with community services. Programs teach dren’s Hospital of Philadelphia’s Institutional Review Board. approaches to child rearing, decreasing parental stress, pro- vide guidance on reducing childhood hazard exposures, and Quantitative Data serve a monitoring function for identifying and responding to maltreatment (Gomby et al. 1999). Analytic Sample Inconclusive and at times conflicting clinical trials and post-implementation studies regarding home visiting’s Data was obtained for clients enrolled in MIECHV funded impact on child maltreatment rates have highlighted the need PA nurse–family partnership (NFP, n = 22), parents as for further attention and evaluation (Institute of Medicine teachers (PAT, n = 9), or early head start (EHS, n = 7) from and National Research Council 2014; Rubin et al. 2014). The 2008 to 2014. Clients were matched to local-area non-client varied success in reducing child maltreatment necessitates women (comparisons) who (1) had similarly aged children further investigation into the conditions under which home identified in birth certificate files and (2) resided in the same visiting programs can achieve prevention, and for whom local implementing agency catchment area (i.e., county or (Howard and Brooks-Gunn 2009). multi-county service area). Inclusion criteria for clients Early results from home visiting evaluations suggested and comparisons were as follows: (1) child affiliated with efficacy in decreasing child maltreatment. Notably, a ran - MIECHV program enrollment was identifiable in PA birth domized controlled trial of the Nurse–Family Partnership certificate files and (2) child affiliated with MIECHV pro- (NFP) resulted in 80% fewer injury and ingestion-related gram enrollment had enrollment in the state medical assis- doctor visits in intervention group (though non-significant tance program (Medicaid) during the outcome observation at 2 years) (Olds et al. 1994). However, following wide- period. scale replication of evidence-based home visiting programs, Clients and potential comparisons were identified in a data are limited in supporting the effectiveness in prevent- multisource administrative data file linked using an itera- ing abuse (Matone et al. 2012). A 2013 review of home tive deterministic approach reliant on unique identifiers 1 3 Maternal and Child Health Journal (2018) 22 (Suppl 1):S79–S91 S81 constructed from social security numbers, names, and dates particular catchment were matched to comparison women of birth that included program enrollment, vital statistics living in the same catchment. Entropy balancing and PSM (birth and death), welfare eligibility, and medical assistance were performed within catchments, and then the samples claim files (Dusetzina et al. 2014). were aggregated. Quasi-experimental Design Description of PSM for EHS and PAT The primary analysis examined if the prevalence and rate of PSM is a matching technique for observational data that child abuse episodes significantly differed between program mimics a randomized control trial by creating pairs (or sets) clients and comparison women for NFP, EHS, and PAT pro- from clients and comparison women with similar values of grams separately. Two primary quasi-experimental methods the propensity score (Stuart 2010). Multivariable logistic were used for causal inference related to program effect on regression models estimate the probability of program par- child abuse: entropy balancing for NFP and propensity score ticipation using available maternal sociodemographic and matching (PSM) for EHS and PAT. Both analytic approaches clinical characteristics—from birth certificate: mother’s are widely used for obtaining covariate balance in observa- age at birth (continuous), race/ethnicity (white/black/His- tional data, but neither approach was suitable for all three panic/other), maternal education (< high school/high school program model analyses for the following reasons. or greater), gestational age (continuous), smoking prior to First, PSM has a disadvantage of dropping subjects una- pregnancy (y/n); from welfare eligibility files: receipt of ble to be matched to counterparts, which creates a biased Temporary Assistance for Needy Families (TANF) or sup- sample. In the case of this study, PSM did not retain a gen- plemental nutrition program prior to or during the first tri- eralizable subset of the clients in the NFP analysis (specifi- mester of pregnancy (y/n); from medical assistance claims: cally, young mothers in rural areas were disproportionately Medicaid eligibility (y/n), maternal diagnosis of substance dropped in attempted PSM). Entropy balancing retained all abuse, depression and/or bipolar disorder in the immediate cases and was the approach used for NFP. While PSM was preconception period or first trimester of pregnancy (y/n). not the optimal approach for NFP, it was chosen for the EHS A separate logistic regression model was performed within and PAT analyses because it allowed for standardized fol- each of the catchment areas for each local implementing low-up time in the outcome observation windows of matched agency. Our matching approach used both caliper and exact sets of clients and comparisons. This is a critical analytic matching on covariates to produce matched sets. Any nearest design feature for programs without standardized enroll- neighbor within a caliper of 0.05 was considered a match ment at a point in time. Unlike NFP, which uniformly enrolls (up to a maximum of four comparison women per client). clients into the program prior to a child’s birth, EHS and Matching was conducted exactly on catchment area, infant PAT programs do not uniformly enroll at a particular age. year of birth, and maternal age (< 18 years of age at birth or Therefore, for each client, PSM allowed for the identification 18 and older). A threshold of 2.5 absolute percentage points of comparisons with similarly aged children at the time of was used to determine balance within each catchment area program enrollment. The analysis then standardized obser- model. Interaction terms were added to the propensity score vation periods for outcome ascertainment within matched model when needed to achieve balance. Analytic weights sets of clients and comparisons using the client’s child’s age were developed within matched sets; each comparison of enrollment and length of time in the program as the refer- woman was given a weight equal to the inverse of the num- ence point (e.g., if client enrolled child at 3 months and was ber of comparison women matched to that client and each observed through month 27, all comparison children for that client was given a weight of 1. These weights were applied client are observed for months 3 through 27). This level of to outcome modeling. modeling flexibility is not possible with entropy balancing, but was also not necessary for NFP given the requirement Description of Entropy Balancing for NFP of prenatal enrollment, which serves as a standardization (i.e., all client and comparison children begin observation Entropy matching is a multivariable weighting technique at birth). that creates a balanced sample by reweighting the control Both entropy balancing and PSM were performed within group (in this case the comparison women) to have the same local implementing agency catchment areas (Matone et al. covariate distribution as the treatment group (i.e., the cli- 2012) to address the possibility that there is confounding ents) using the above described maternal sociodemographic by geography (i.e., the outcomes might vary across sites at and clinical characteristics. In this approach, specifications a community level beyond maternal-level characteristics). for each covariate can be applied as to whether exact bal- Catchments included each implementing agency’s county ance between the two groups should be achieved on the first and contiguous counties. Clients enrolled in a program in a moment (mean), second moment (standard deviation), or 1 3 S82 Maternal and Child Health Journal (2018) 22 (Suppl 1):S79–S91 higher moments (Hainmueller and Xu 2013). As is intended Sensitivity Analyses with this methodology, there is automatic balance created between the samples after conducting entropy balancing, so Two sensitivity analyses were conducted to test if the rela- no additional balance checks or model adjustment to create tionship found between program participation and abuse balance was necessary. Covariates used in entropy balancing outcome was robust to potential confounders that could not were the same as included in PSM. be included as covariates in the PSM or entropy balancing model. We tested for confounding between program par- Abuse and Injury Episode Creation ticipation and abuse by maternal psychosocial risk factors by separately including each risk factor in the primary out- The primary outcome for this study was the presence of an come model and examining if the estimated odds ratio effect abuse episode or high risk injury episode (composite meas- for the program participation, adjusting for the risk factor, ure) with a secondary outcome that identified the presence changed by 10% or more. The risk factors ascertained from of any injury episode. Outcome measures were derived from the literature to be confounders are (1) maternal previous child Medicaid claims. Episodes were created to conserva- involvement with child protective services (CPS) before tively count unique instances of abuse and injury recogniz- pregnancy and (2) intimate partner violence (IPV) measured ing that multiple claims/encounters may exist for a single after conception (Berlin et al. 2011; Eckenrode et al. 2000). event. The methodology of collapsing claim encounters to To identify clients involved in CPS, NFP clients and create episodes is described in Matone et al. 2012 and fur- comparisons residing in Philadelphia were linked to county ther in Online Appendix A. child welfare records via first name, last name, date of birth, Abuse episodes were those in which an ICD-9 code indi- and gender. Child welfare systems are administered at the cated child abuse (995.50-5, 995.59), as well as high risk county-level in PA; Philadelphia represents the largest injuries (HRI), specific types of severe injuries considered county in the state and produced a sample large enough for highly suspicious for abuse without the presence of a medi- sensitivity analysis. For any clients and comparisons suc- cal diagnosis of abuse in the medical record. These episodes cessfully linked to child welfare records, dates of protective feature injuries that include fractures of the femur, radius, service were provided. We identified mothers with childhood ulna, tibia, fibula, humerus, ribs, or traumatic brain injuries CPS involvement (prior to pregnancy). within the first 24 months of life (without the presence of Regarding the second sensitivity analysis, IPV was iden- ICD-9 codes indicating an injury due to a motor vehicle tified in maternal medical encounters as an ICD-9 code of crash) (Wood 2010) (Online Appendix B). 995.8x during the observation window of child’s date of Injury outcomes included: superficial injuries, a compos- conception through the first month of life. Even though we ite of dislocation, fracture, and crush injuries, poisonings, identified IPV in some clients after program enrollment and burns identified through ICD-9 codes. occurred or after the program services could have started, we The observation window for episodes were claims during deemed this time window as most meaningful for measuring 0 to 24 months of life for NFP cohort and, for EHS and PAT IPV for two reasons: (1) to reflect a baseline risk proximal to cohorts, 24 months post-enrollment, up to 6 years of life. program enrollment and (2) to increase likelihood of ascer- Right-censoring of episodes occurred for children whose tainment in medical assistance files given increased health observation periods exceeded the study end period of 2014. seeking during pregnancy and increased risk period for IPV. While rates of IPV during pregnancy vary depending on Outcome Models the samples studied and measures used, the prevalence of IPV during pregnancy is elevated compared to women of The primary exposure was NFP, PAT, or EHS program par- non-reproductive age and may be increased compared to ticipation. For EHS and PAT analyses, a weighted condi- non-pregnant patients (Hellmuth et al. 2013; Jasinski 2004). tional logistic regression model was used to examine the Less biased screening (i.e. more universal screening) may unadjusted association between program participation and occur during the prenatal period due to recommendations by the primary outcome. For the NFP analysis, a weighted professional organizations, such as the American Congress logistic regression with a random intercept for county was of Obstetricians and Gynecologists’ (ACOG), that provid- used to estimate the relationship between program participa- ers should screen all women for IPV at periodic intervals, tion and the primary outcome, controlling for the variabil- including during obstetric care (at the first prenatal visit, ity in the outcome across counties. The presence of abuse at least once per trimester, and at the postpartum checkup) or injury prior to enrollment was included as an adjust- (“ACOG Committee Opinion No. 518: Intimate partner vio- ment covariate in PAT and EHS final outcome models (not lence” 2012). applicable for NFP modeling given prenatal enrollment) to For each set of primary analyses described above, we ran account for baseline injury risk. two additional models—one with a dichotomous covariate 1 3 Maternal and Child Health Journal (2018) 22 (Suppl 1):S79–S91 S83 for presence of maternal involvement with CPS prior to Results childbirth and another with a covariate for presence of IPV within child’s date of conception through the first month Quantitative Data of life. Cohort Demographics Presentation of Results The entropy balanced NFP cohort included 8736 clients enrolled in 22 NFP programs between 2008 and 2014 matched Logistic regression results were expressed as odds ratios to 165,033 comparisons. The propensity score matched PAT (with 95% confidence intervals) and standardized marginal cohort included 851 clients enrolled in nine PAT programs probabilities. All analyses were conducted using SAS ver- matched to 2929 comparisons; EHS cohort included 866 cli- sion 9.4, Stata version 14.2 and R. Stata’s ebalance package ents enrolled in seven EHS programs propensity score matched was used for entropy balancing and R’s MatchIt for PSM. to 3100 comparisons (Table 1). All statistical tests were two-sided and used an alpha = 0.05 Across all models, the majority of clients were unmarried as the threshold for statistical significance. and non-Hispanic white race/ethnicity. In terms of model- specific differences, compared to PAT and EHS client, NFP clients were most likely to be under the age of 18, most likely Qualitative Data to be Hispanic, and least likely to smoke prior to pregnancy (Table 1). Setting and Participants Abuse Episodes Among Clients and Comparison Women 11 of the 38 PA MIECHV-funded programs were selected for the qualitative study, chosen to supply a representative Across all models, children of clients were significantly more sample of agencies, based on program size, location, and likely to experience an abuse episode than comparisons: NFP model type, including NFP, PAT, EHS, and Healthy Families OR: 1.32, 95% CI [1.08, 1.62]; PAT OR: 4.11, 95% CI [1.60, America (HFA, which, due to data constraints, could only be 10.55]; EHS OR: 3.15, 95% CI [1.41, 7.06] (Table 2). Within included in the qualitative study). Program staff were inter - NFP, 1.4% of client children (n = 120) and 1.0% of comparison viewed during day-long site visits; enrolled clients, recruited children (n = 1488) sustained an abuse injury within 24 months with flyers and help from program staff, were interviewed of life; 1.1% (n = 9) of PAT-enrolled children and 0.3% of PAT over the phone. Participants were verbally consented before comparison children (n = 9) sustained an abuse injury within participating in interviews lasting between 30 and 60 min. 24 months of enrollment; and 1.3% of EHS program-enrolled Clients were sent a $20 gift card in appreciation for their children (n = 11) and 0.4% EHS comparison children (n = 14) time. Interviews took place between 2013 and 2015. sustained an abuse injury within 24 months of enrollment. Distribution of Injury Types among Client Children Measures and Analysis with an Abuse Episode The interdisciplinary project team worked with home vis- The most frequent injury types among home visited children iting leadership to develop three distinct interview guides who experienced an abuse episode included superficial inju- for program administrators, home visitors, and clients, each ries and dislocation, fracture or crush injuries. Burns were the including questions on specific program outcomes (e.g. child least prevalent injury type in aggregate. While poisonings were maltreatment; See Online Appendix C for questions that infrequent among NFP children with abuse episodes (2.0%), elicited content related to child maltreatment.). De-identi- one in five children in PAT with abuse episodes experienced fied transcripts were imported into NVivo 10 for coding and poisoning while dislocations, fractures and crush injuries analysis. We used a modified Grounded Theory approach to occurred with half the frequency among PAT children than coding (Glaser and Strauss 1967), including a priori codes NFP children with abuse (23.5 versus 59.7%) (Table 3). relating to quantitative metrics included in the evaluation. Using a constant comparative approach, coders met regularly Sensitivity Analyses to review memos and coding comparison queries to discuss and refine node definitions and the application of codes. The prevalence of maternal childhood CPS involvement Discrepancies were resolved through group consensus. A (prior to pregnancy) was equitable between clients and thematic analysis was conducted on all interview content comparisons, with greater than one-third of mothers related to child maltreatment. 1 3 S84 Maternal and Child Health Journal (2018) 22 (Suppl 1):S79–S91 1 3 Table 1 Characteristics of clients of nurse–family partnership (NFP), parents as teachers (PAT), and early head start (EHS), and comparison women, 2008–2014 NFP PAT EHS Comparison women Clients Comparison women Clients Comparison women Clients N = 165,033 % N = 8736 % N = 2929 % N = 851 % N = 3100 % N = 866 % Age, < 18 years 15,601 20.1 1955 22.4 183 6.5 51 6.0 185 5.8 54 6.2 Race/ethnicity  White 67,240 47.8 3028 47.4 2252 77.3 657 77.9 1506 50.8 420 49.1  Black 32,837 22.3 1451 22.7 251 8.4 72 8.5 987 30.5 275 32.1  Hispanic 25,232 28.2 1801 28.2 365 12.7 107 12.7 483 15.6 139 16.2  Other 5751 1.8 112 1.8 50 1.6 7 0.8 91 3.0 22 2.6 Unmarried 131,010 89.9 7857 89.9 2206 74.8 584 68.6 2480 78.3 657 75.9 Education, less than high school 37,307 37.9 3347 38.3 825 28.3 258 30.3 1123 36.3 322 37.2 TANF receipt 54,399 43.2 3770 43.2 1546 51.9 411 48.3 1676 52.5 446 51.5 Foodstamp receipt 55,753 49.0 4280 49.0 1882 63.3 517 60.8 1899 60.2 495 57.2 Depression in proximity to pregnancy 3820 6.0 527 6.0 287 10.6 113 13.3 162 5.3 69 8.0 Substance abuse in proximity to pregnancy 3599 3.2 280 3.2 166 5.9 46 5.4 128 4.1 40 4.6 Medicaid eligibility in proximity to pregnancy 55,664 40.3 3534 40.5 982 32.8 266 31.3 924 29.8 181 20.9 Smoking, prior to pregnancy 46,168 30.7 2684 30.7 1361 46.3 382 44.9 1321 42.5 350 40.4 Table reports unweighted Ns and weighted proportions Maternal and Child Health Journal (2018) 22 (Suppl 1):S79–S91 S85 Table 2 Marginally standardized probabilities and odds of child Qualitative Data abuse among comparison women and home visiting clients enrolled in nurse–family partnership (NFP), parents as teachers (PAT), and A total of 150 interviews were conducted with program early head start (EHS) administrators (N = 25), agency staff (N = 49), and home HV program Com- HV clients (%) OR (95% CI) p Value visiting clients (N = 76). The sample represents all four parisons MIECHV-eligible evidence-based programs in PA, with (%) greater representation from NFP (35% of staff and 47% NFP 1.0 1.4 1.32 (1.08, 1.62) 0.008 of clients). Participants from urban and rural sites were PAT 0.3 1.1 4.11 (1.60, 10.55) 0.003 equally represented (Table 4). EHS 0.4 1.3 3.15 (1.41, 7.06) 0.005 Our thematic analysis of the interview data related to child maltreatment identified two primary domains: (1) how programs and staff react when there is a suspicion Table 3 Marginally standardized probabilities of injury types among of child maltreatment; and (2) how programs and sta ff children with an abuse episode by home visiting program address and clients engage with child maltreatment pre- vention strategies. HV program Marginally stand- ardized probability (%) NFP  Superficial injury 38.1  Dislocation, fracture, crush 59.7 Table 4 Demographics of interview participants  Poisoning 2.0 Clients (N = 76) Staff (N = 74)  Burns 3.8 PAT Program % %  Superficial injury 30.4  NFP 47 35  Dislocation, fracture, crush 23.5  PAT 24 22  Poisoning 19.4  EHS 14 24  Burns –  HFA 14 20 EHS Urbanicity  Superficial injury 32.0  Urban 51 59  Dislocation, fracture, crush 65.3 Sex  Poisoning –  Female 93 99  Burns – Race  White 57 82 NFP nurse family partnership, PAT parents as teachers, EHS early head start  Black/African American 33 9 Due to small sample size, OR and marginally standardized probabil-  Other 11 5 ity cannot subsequently be calculated Ethnicity  Non-Hispanic 95 95 Age having CPS involvement prior to their first pregnancy. The  18 and under 4 – results of sensitivity analyses demonstrate a lack of con-  19–22 21 – founding by CPS involvement on the relationship between  23+ 75 – program enrollment and abuse (Online Appendix D). Employment Roughly 1% of NFP and PAT clients and comparisons  Unemployed 57 – were diagnosed with IPV in the pregnancy period (Online Marital status Appendix E, Table 1). In sensitivity analyses, maternal  Single 51 – IPV exposure was not found to be a confounder of the rela-  Married/partnered 45 – tionship of program enrollment and abuse (Online Appen-  Separated or divorced 4 – dix E). However, across both programs, mothers with diag- Education nosed IPV were significantly more likely to have a child  High school or less 50 – with an abuse episode than those without diagnosed IPV  Some college 37 – (NFP: 2.7 versus 1.2%, p = 0.027; PAT: 6.0 versus 0.6%,  College or higher 13 – p = 0.027) (Online Appendix E, Table 2). Total 100 100 1 3 S86 Maternal and Child Health Journal (2018) 22 (Suppl 1):S79–S91 Program and Staff Responses to Suspicion or Incidents [T]he program talks about strategies with kids having of Child Maltreatment temper tantrums and stuff. Not only things that you can do for the kiddo, but things as a parent. Maybe you In a few instances during interviews, clients and home visi- need to step away, take a few breaths, collect yourself tors described instances of potential child maltreatment, and then go back and address the issue. EHS Admin- shedding light on how the program responds in such cir- istrator, 1002 cumstances (Table 5). Parents valued this guidance and support, perceiving Of the cases describing incidents concerning for abuse improvements in managing their reactions to frustrating and neglect, most involved caregivers other than the client child behaviors. Clients attributed personal changes to the as the perpetrator, including partners, child siblings, and general encouragement from home visitors to attend to their grandparents. Program responses tended to focus on sup- own personal needs and more specific support related to porting the client after the abuse had occurred with parent- stress. ing support, referrals, and connections to resources. Some home visitors described relying on CPS to manage child I get stressed out easy and she’s helped me learn maltreatment issues. how to deal with stressful situations, when it comes down with my daughter when she starts getting in her, [I]f there is any indication of violence in the home “give me, wanna wanna” modes […] and I get really that would always be referred to [CPS], if that child stressed about it. She helps me learn how to deal with was being exposed to that. A lot of times, you can just the stress and teaches me how to deal with my daugh- see it while you’re there. […] sometimes they don’t ter […]. […S]he told me, it’s important that you do even try to hide it. Or they’ll […] say something that this for yourself. Also, it’s important you do this for the other person has been doing or whatever. So we you daughter. I need to take a little down time. HFA don’t get into that too much, other than to say a healthy Client, 1004 environment is what our goal is here and whatever you need to do to get that. HFA/PAT Home Visitor, 801 Some clients described how learning new discipline strat- egies gave them the opportunity to break a cycle of abuse. Home visitors qualified their relationship with CPS in many ways, including their role as mandated reporters. We’ve talked about discipline, not so much of abuse. Given the negative feelings often associated with CPS, […] I think it’s effective to make sure that my child some home visitors described being careful about how they understands that I know what they want and kind of explained their relationship with CPS to clients. having different methods instead of going directly to a punishment. […]My parents disciplined me by beating Approach to and Engagement with Child Maltreatment me and punishing me for extremely long periods of Prevention time […]. […] Those things that I didn’t really know about prior to being involved in the program because Home visitors described focusing on content related to posi- those are things that weren’t taught to me as a child. tive parenting, secure attachment, and stress management PAT Client, 7020 as the curricular drivers of child abuse prevention. Many When asked how they talked with their home visitor staff referenced the program training they received, which about protecting their child from abuse, a number of clients taught them that supporting the development of a strong did not remember the subject being addressed. Many clients bond between a parent and child could reduce the chances acknowledged that discussing child maltreatment should be of maltreatment. part of the curriculum covered by their home visitor. Implied We stress with [parents] attachment and bonding. in the absence of direct discussions of child abuse was the Because if a parent does attachment and bonding, described assumption that the home visitor knew there was we’ve learned through all of our trainings with the not abuse occurring within the household. HFA, they’re gonna be less likely to abuse their child I don’t remember talking about that. I would assume […], and that is what everything circles back to. HFA that she knows that we are not abusive at all here. But Home Visitor, 103 I know that she shouldn’t assume, because you don’t Home visitors focused with parents on approaches and know what happens when someone leaves. PAT Cli- frameworks for disciplining children. The described cur- ent, 2004 ricula addressed strategies directed both at the children, to From the client perspective, very few barriers to broach- manage their behavior, and at the adults, to manage associ- ing the topic of abuse were described. One parent postulated ated stress. 1 3 Maternal and Child Health Journal (2018) 22 (Suppl 1):S79–S91 S87 Table 5 Qualitative interview data representing concern for child maltreatment Perpetrator Program response Excerpt from interview data Both parents Adapt program curriculum to specific child need “I do have one family that we’re working mainly on gross motor because he’s not – when he runs he doesn’t bend his knees quite as often. And that’s due to they’re kind of putting him in the playpen and not letting him out because he’ll make a mess and they don’t feel like cleaning up the mess, obvi- ously. So he’s kind of slowly developing gross motor because he’s being confined. So I’m trying to delicately word – in that maybe it’s because he’s in the playpen that his gross motor skills aren’t developing. So when I go over, we kind of do more activities towards that, and work that in with what I already have planned.” EHS Home Visitor, 402 Parent (non-client) Provide informational support “[I]f [my son] looks different for some reason or he comes back Adapt sessions to ask about relationship with Father from his dad’s and I have a visit with her. And she’s like, CPS referral he has another head contusion from his dad. [...] She really tells me – well, did you take him to the emergency room? Make sure you watch it. […] She gives me paperwork on the head contusion. What to look for – if he sleeps too much or – stuff like that. […] She gave me – I forgot what it’s called. Something Domestics – Children and Youth. But, I mean, it’s not like – I don’t think he’s harming my son. It’s just the fact that he isn’t very good at watching him. So – oh, yeah, I’ve thought about it and definitely think that next time that he does come home with a head contusion, then – or if he took him to the emergency room where he has a huge boonie on his head. Usually if he does have a boonie on his head, that’s a thick bruise or swollen, I take him straight to the emergency room right when I pick him up. We don’t go home. We don’t go eat. We go to the emergency room, just because I want to have that documented as soon as possible – and if anything is really like more internally wrong. But I would definitely – she’s given me stuff about Children and Youth.” NFP Client, Sibling (non-client) Adapt program curriculum to specific child need “[My son] was really aggressive towards [my daughter…]. [My son] was basically the only child that I was dealing with one-on-one. And then got pregnant having my daughter. And he started feeling neglected, started to act out more, so where he would start trying to hit her and do certain little things to her and stuff like that. And with the program, they basically helped me focus on him and have time with him and also bond with the baby. Also, help him bond with the baby, like play with – show him ways to play and stuff. [...] Basically, redirecting him to do something else, like to move him away from her, pull him aside and play with just him, like one-on- one with him.” PAT Client, 8004 1 3 S88 Maternal and Child Health Journal (2018) 22 (Suppl 1):S79–S91 Table 5 (continued) Perpetrator Program response Excerpt from interview data Sibling (non-client) Provide resource “I have two stepchildren and the one, their mother has filled Provide informational and emotional support their head with a lot of awful things, and he’s very abusive and we have a lot of problems. [...O]ne day she came and [my son] had a bruise on his cheek. And she said, how did that happen? And I explained to her that he had been over playing by the baby gate and his brother come out and just hit him in the face with the baby gate. And [...] then she had seen that I had a bruise on my face where he had hit me. And she had said, this has got to stop, this is what we’re gonna do. And she brought in [a therapist] to talk to me. And then we went from there and started informing all these agencies. And she said it might now help you, but at least it’s out there that if something major happens, you attempted to try to get the help you needed and the mother closed it down, but you attempted. And she said it’s the same thing if [Son] goes to school and they find a bruise on him, they’re gonna come after you and it may not be you, it could have been him, but you’re, you know, you have at least explained the situation and have tried to get help.” EHS Client, 6001 Grandparent (non-client) Provide informational and emotional support “I went through a domestic violence case between my mom and then through the girls’ dad. […My Home visitor]actually brought paperwork, and she physically worked with me of how to do things and that, or like if she wasn’t here, I could call her for advice or give her a text message for advice. It was like a full process. It helped by steps and physical help sometimes. [...] Like she showed me how – because my [daughter], she had fractured ribs from my mom. And [my home visitor] actually helped me of ways to hold her that’ll help her ease her pain, and then – It was really nice. Without it, I probably would have went insane with all the crying. [...] I asked my mom to watch my two kids at her house while I went to the hospital because I was really, really sick. And my current boyfriend at the time, he was working. And I had like nobody else to watch my kids, and I didn’t want them to get sick and see me suffer type of thing. And then we brought her back home and she wouldn’t stop crying, and we couldn’t figure out why, then we ended up taking her to the hospital, and CYS showed up at my door at 4:00 in the morning asking if we knew what was wrong with my daughter. And I broke. I bursted into tears, because I didn’t know what’s going on.” EHS Client, 10003 Fiancé (non-client) Unknown “So one time, we thought [my daughter] did break her leg whenever her and [my fiancé] were wrestling. But we went to [Hospital 1]. They made us wait two hours in the waiting room, so I left. And they called CYS on me, and CYS had came. And they just wanted to see that [my daughter] wasn’t afraid of [my fiancé] because the hospital was saying that the father was abusing her. They closed us out that day knowing that, you know, nothing bad was going on. So we took her to [Hospital 2] because I wasn’t waiting there if my daughter may have had – like, a one-year old may end up having a bro- ken leg. I’m not wasting any time, so I changed the hospital. I went up to [Hospital 2] instead. Here, she didn’t have a bro- ken leg. [...] It was, like, a fracture in her thigh bone. [...] So, like, I had told, you know, CYS, you know, I left [Hospital 1]. I went to [Hospital 2] because they were making my daughter wait in the waiting room for two hours.” EHS Client, 4005 1 3 Maternal and Child Health Journal (2018) 22 (Suppl 1):S79–S91 S89 that using videos to broach the subject of child maltreatment including fractures, dislocations, and crush injuries, the like- was the only way the program could address it without put- lihood of a family to seek healthcare is likely less dependent ting parents on the defensive. on the presence of home visitor. Other considerations for the observed increased rate of abuse-related injury among home I think that the videos and the conversations are really visited families include unidentified confounding that does all they can do without having somebody be like, not account for the referral bias driving higher risk clients you’re intruding or PC or social warrior. That’s a sen- into home visiting services. To address this concern, this sitive subject for a lot of people, how you’re gonna study included two sensitivity analyses to account for poten- discipline your kids or whatever. NFP Client, 5002 tial confounding by the two strongest psychosocial risk fac- Some home visitors discussed difficulties encouraging tors associated with child maltreatment: maternal childhood families to alter behaviors. The set of norms a parent devel- involvement with CPS or IPV. Despite a high prevalence of ops around how adults should interact with children from maternal childhood CPS involvement among clients, this years of personal experience is difficult to contend with risk factor was found to be balanced between clients and when misaligned with home visitor communication. comparison women and did not confound findings. Mater - nal IPV in pregnancy was also not an identified confounder Challenges are we’re not there all the time to monitor in this study, though diagnosed rates of IPV were low and them. We can refer them, we can’t force them to go. I likely represented underascertainment of true IPV risks mean, there’s a deep history in how people discipline within the cohort. Among mothers with diagnosed IPV in their children and it’s hard to break that within a cou- pregnancy, rates of abuse were higher even after adjustment ple of months. NFP/PAT Home Visitor, 710 for program enrollment, indicating that these families may Unless a parent is personally invested in changing their be a high risk subgroup of home visited clientele. patterns, it is difficult for the home visitor to champion A strength of this study is the provision of qualitative change alone. When parents are open to using new strate- contextual support for the development of hypotheses of why gies, true influence depends on the degree to which these home visiting programs may struggle in reducing rates of strategies replace problematic practices. child abuse. The instances of child abuse assessed in inter- views described events where harm occurred to the child I learned that there’s no wrong or right way to teach outside of the client’s oversight while under the supervi- your child behaving. […] Like say if she would think sion of a non-client caregiver. Acknowledging the limita- I’m teaching her, like I pop her or whatever, that could tion that interviewees may be more likely to discuss inci- be okay, to pop her, but you know, also you got to like dents in which they were not the perpetrator, the number start taking stuff away and then teaching her no, and of examples including other caregivers is still notable. The telling her don’t do stuff. NFP Client, 7015 data are reflective of other information described by program sites suggesting that the vast majority of serious and abuse- related injuries take place while the child is in the care of a Discussion non-client caregiver, often an intimate partner. As programs encourage clients to engage in educational and professional This study demonstrates quantifiably increased risk of early advancements, it is important to address resultant childcare childhood abuse-related injury among children of MIECHV needs. Additionally, the role of non-clients in child maltreat- clients compared to non-program enrolled comparison chil- ment events makes evident the importance of delivering pro- dren across three home visiting programs (NFP, EHS, and gram curriculum to as many of the caregivers in contact with PAT) implemented in a large state. Among client children children as possible. Further support for this hypothesis is experiencing abuse, elevated rates of fracture, dislocation, demonstrated by the quantitative sensitivity analysis of the and crush injuries were seen in comparison to client children significant impact of the presence of IPV on child abuse. experiencing non-abuse related injuries. The curriculum related to child maltreatment, as well as Null findings on child abuse prevention to home visit- home visitors’ delivery of it also plays a role in achieving ing programs are not new to the field. Others have cited intended program outcomes. Recent evidence suggests that surveillance bias among enrolled families as a reason for home visiting programs may struggle to achieve positive the lackluster program ee ff ct on child maltreatment (Gomby child maltreatment outcomes due to ineffectual program et al. 1999; Olds et al. 1995). While surveillance bias may delivery. Much of home visiting curricula is prepared to help be a contributor to increased observation of abuse-related parents decrease the effects of adverse childhood experi - injury among client families, the bias is likely most relevant ences (ACEs) on their children, even without a reduction in to minor injuries where healthcare seeking behavior was actual ACEs (McKelvey et al. 2016). However, home visi- optional. In the case of more serious and emergent injuries, tors themselves have not reported high efficacy in guiding 1 3 S90 Maternal and Child Health Journal (2018) 22 (Suppl 1):S79–S91 discussions with parents related to sensitive topics such as analyses with likely confounders, minimizes this concern maltreatment and violence (Duggan et al. 2007). Casillas and is a widely accepted technique for estimating causality et al. notes that child maltreatment outcomes in particular are when randomization is not feasible or appropriate. Moreo- negatively affected by low home visitor efficacy and fidelity ver, the standardized approach to outcome ascertainment in service provision (Casillas et al. 2016). Our qualitative using administrative rather than self-reported data coupled data demonstrate variations in whether and how home visi- with the rigorous quasi-experimental design facilitate stand- tors discussed child abuse with clients, such that a number ardized interpretation of findings across three home visiting of clients did not recollect the topic ever being raised. This models and numerous diverse implementing sites. While our finding may have been driven by role issues, as mandatory analyses were unable to take program dosage and the per- reporters and assumed affiliations with CPS were barriers to petrator of abuse incidents into account in a standardized direct discussions of maltreatment. Home visitors also strug- fashion as this information was unavailable in administra- gled against normative factors shaping behavior change, as tive data files, this information was gleaned from qualita- well as those more directly related to maltreatment such as tive interviews when it was addressed by home visitors or corporal punishment. A recent evaluation of PAT program in clients. The study is not able to account for surveillance Connecticut demonstrated significant reductions in substan - bias; however, the study used an intention to treat design tiated child maltreatment; however, this finding was driven to observe children for two full years following enrollment by reductions in neglect with no demonstrated impact on while a significant proportion of families are not retained in child abuse. Such findings further highlight the need for services for this duration (Burwick et al. 2014). Lastly, given additional supports around abuse specifically, but also pro- the sensitive nature of this topic, the qualitative interview vide support for qualitative findings in this study that suggest data is likely impacted by social desirability bias. more home visitor comfort in the domains of parenting that may be more directly related to neglect-related maltreatment (Chaiyachati et al. 2018). Conclusion Despite hearing an openness to discussing issues of child maltreatment from more clients than those noting challenges, Home visiting programs have a stated objective of prevent- overall the qualitative data illustrate little direct engagement ing child abuse-related injury among high-risk families. around child maltreatment. Interviews with clients and staff The success of abuse prevention depends on the strength demonstrate that indirect approaches, such as activities to of the curriculum, the fidelity of delivery, and whether it alter parenting styles or decrease maternal stress, are present reaches the people in caregiving roles. Given that home visi- and more likely to be delivered with fidelity. However, it is tors depend greatly on other community agencies and public possible that focusing on supporting positive parenting and systems, it is naïve to expect home visiting to achieve strong stress reduction is not sufficient to effectively reduce the risk outcomes without a well-integrated and -resourced service of abuse incidents among the subset of home visited families network. Programs and clients would benefit from curricu- most at-risk for maltreatment. lum that more directly and proactively addresses child mal- Finally, the qualitative data describe limited reactive treatment, as well as knowledge of and access to quality responses to situations concerning for neglect and referral childcare options. to outside resources—mainly CPS. The data highlights the dependence of home visiting success on the system of ser- Acknowledgements This evaluation was supported by a grant from the vices in which the program resides. The importance of the Department of Human Services, Commonwealth of Pennsylvania. We thank the Pennsylvania Department of Human Services, Pennsylvania quality, access, and connectedness of CPS, childcare, and Department of Health, and the MIECHV local implementing agencies healthcare to home visiting services towards the prevention across Pennsylvania for contributing data for this study. We are also of child maltreatment prevention cannot be overstated. The grateful to the Office of Child Development and Early Learning of the ability of home visitors to transition from reactivity to pro- Pennsylvania Department of Human Services for their consultation on Pennsylvania MIECHV programming and partnership in identifying activity around child abuse is dependent on effective referral evaluation priorities for the Commonwealth. relationships, trust in the CPS system, and timely and afford- able access to services that mitigate abuse risks, including Open Access This article is distributed under the terms of the Crea- childcare and healthcare. tive Commons Attribution 4.0 International License (http://creat iveco This study has several limitations related to study design mmons.or g/licenses/b y/4.0/), which permits unrestricted use, distribu- tion, and reproduction in any medium, provided you give appropriate and data availability and reliability. The observational study credit to the original author(s) and the source, provide a link to the design is subject to bias when estimating program effects; Creative Commons license, and indicate if changes were made. however the use of propensity score and entropy matching, which mimic randomization and control for measured dif- ferences between clients and comparisons and sensitivity 1 3 Maternal and Child Health Journal (2018) 22 (Suppl 1):S79–S91 S91 Howard, K. S., & Brooks-Gunn, J. (2009). The role of home-visiting References programs in preventing child abuse and neglect. The Future of Children, 19(2), 119–146. ACOG Committee Opinion No. 518. (2012). Intimate partner vio- Institute of Medicine, & National Research Council. (2014). New lence. Obstetrics Gynecology, 119(2 Pt 1), 412–417. https ://doi. directions in child abuse and neglect research. Washington, DC: org/10.1097/AOG.0b013 e3182 49ff7 4. National Academies Press. Administration for Children and Families. https ://www.acf.hhs.gov/ Jasinski, J. L. (2004). Pregnancy and domestic violence: A review of sites /defau lt/files /cb/cm201 5.pdf. Accessed 1 Mar 2018 the literature. Trauma, Violence, and Abuse, 5(1), 47–64. Agran, P. F., Anderson, C., Winn, D., Trent, R., Walton-Haynes, L., & Leech, N. L., & Onwuegbuzie, A. J. (2009). A typology of mixed meth- Thayer, S. (2003). Rates of pediatric injuries by 3-month intervals ods research designs. Quality and Quantity, 43(2), 265–275. for children 0 to 3 years of age. Pediatrics, 111(6), e683–e692. Matone, M., O’Reilly, A. L., Luan, X., Localio, R., & Rubin, D. M. https ://doi.org/10.1542/peds.111.6.e683. (2012). Home visitation program effectiveness and the influence Berlin, L. J., Appleyard, K., & Dodge, K. A. (2011). Intergenerational of community behavioral norms: A propensity score matched continuity in child maltreatment: Mediating mechanisms and analysis of prenatal smoking cessation. BMC Public Health, 12, implications for prevention. Child Development, 82(1), 162–176. 1016. https ://doi.org/10.1186/1471-2458-12-1016. https ://doi.org/10.1111/j.1467-8624.2010.01547 .x. Matone, M., O’Reilly, A. L. R., Luan, X., Localio, A. R., & Rubin, Burwick, A., Zaveri, H., Shang, L., Boller, K., Daro, D., Strong, D. D. M. (2012). Emergency department visits and hospitaliza- (2014). Costs of early childhood home visiting: An analysis of tions for injuries among infants and children following statewide programs implemented in the supporting evidence-based home implementation of a home visitation model. Maternal and Child visiting to prevent child maltreatment initiative. Health Journal, 16(9), 1754–1761. https: //doi.org/10.1007/s1099 Casillas, K. L., Fauchier, A., Derkash, B. T., & Garrido, E. F. (2016). 5-011-0921-7. Implementation of evidence-based home visiting programs aimed McKelvey, L. M., Whiteside-Mansell, L., Conners-Burrow, N. A., at reducing child maltreatment: A meta-analytic review. Child Swindle, T., & Fitzgerald, S. (2016). Assessing adverse expe- Abuse & Neglect, 53, 64–80. https ://doi.or g/10.1016/j.c hiab riences from infancy through early childhood in home visiting u.2015.10.009. programs. Child Abuse & Neglect, 51, 295–302. https ://doi. Chaiyachati, B. H., Gaither, J. R., Hughes, M., Foley-Schain, K., & org/10.1016/j.chiab u.2015.09.008. Leventhal, J. M. (2018). Preventing child maltreatment: Exami- Merrick, M. T., & Latzman, N. E. (2014). Child maltreatment: A public nation of an established statewide home-visiting program. Child health overview and prevention considerations. Online Journal Abuse & Neglect, 79, 476–484. https ://doi.org/10.1016/j.chiab of Issues in Nursing, 19(1), 1–1. https ://doi.org/10.3912/OJIN. u.2018.02.019. Vol19 No01M an02. Children’s Bureau. (2016). Child maltreatment 2015. Washington, DC: Nelson, H. D., Selph, S., Bougastsos, C., & Blazina, I. (2013). Behav- Administration on Children Youth and Families. ioral interventions and counseling to prevent child abuse and Deutsch, S. A., Lynch, A., Zlotnik, S., Matone, M., Kreider, A. R., neglect: A systematic review to update the U.S. Preventative Ser- & Noonan, K. (2015). Mental health, behavioral and develop- vices Task Force recommendation. Annals of Internal Medicine, mental issues for youth in foster care. Current Problems in Pedi- 158(3), 179–190. atric and Adolescent Health Care, 45(10), 292–297. https ://doi. Olds, D., Henderson, C. R. Jr., & Kitzman, H. (1994). Does prenatal org/10.1016/j.cpped s.2015.08.003. and infancy nurse home visitation have enduring effects on quali- Duggan, A., Caldera, D., Rodriguez, K., Burrell, L., Rohde, C., & ties of parental caregiving and child health at 25 to 50 months of Crowne, S. S. (2007). Impact of a statewide home visiting pro- life? Pediatrics, 93(1), 89–98. gram to prevent child abuse. Child Abuse & Neglect, 31(8), 801– Olds, D., Henderson, C. R. Jr., Kitzman, H., & Cole, R. (1995). Effects 827. https ://doi.org/10.1016/j.chiab u.2006.06.011. of prenatal and infancy nurse home visitation on surveillance of Dusetzina, S. B., Tyree, S., Meyer, A. M., Meyer, A., Green, L., & Car- child maltreatment. Pediatrics, 95(3), 365–372. penter, W. R. (2014). Linking data for health services research: A Patient Protection and Affordable Care Act. § 2951 (2010). framework and instructional guide. Rockville, MD: Agency for Rubin, D. M., Curtis, M. L., & Matone, M. (2014). Child abuse pre- Healthcare Research and Quality. vention and child home visitation: Making sure we get it right. Eckenrode, J., Ganzel, B., Henderson, C. R. Jr., Smith, E., Olds, D. JAMA Pediatrics, 168(1), 5–6. https://doi.or g/10.1001/jamapediat L., Powers, J., … Sidora, K. (2000). Preventing child abuse and rics.2013.3865. neglect with a program of nurse home visitation: The limiting Schatz, J. N., & Lounds, J. J. (2007). Child maltreatment: Precursors effects of domestic violence. JAMA, 284(11), 1385–1391. of developmental delays. In J. G. Borkowski, J. R. Farris, T. L. Frioux, S., Wood, J. N., Fakeye, O., Luan, X., Localio, R., & Rubin, Whitman, S. S. Carothers, K. Weed & D. A. Keogh (Eds.), Risk D. M. (2014). Longitudinal association of county-level economic and resilience: Adolescent mothers and their children grow up indicators and child maltreatment incidents. Maternal and Child (pp. 125–150). Mahwah, NJ: Lawrence Erlbaum. Health Journal, 18(9), 2202–2208. https: //doi.org/10.1007/s1099 Stuart, E. A. (2010). Matching methods for causal inference: A review 5-014-1469-0. and a look forward. Statistical Science: A Review Journal of the Glaser, B. G., & Strauss, A. L. (1967). The discovery of grounded Institute of Mathematical Statistics, 25(1), 1. theory: Strategies for qualitative research. New York: Aldine de Wood, J. N. (2010). Disparities in the evaluation and diagnosis of abuse Gruyter. among infants with traumatic brain injury. Pediatrics, 126(3), Gomby, D. S., Culross, P. L., & Behrman, R. E. (1999). Home visiting: 408–414. Recent program evaluations: Analysis and recommendations. The Wood, J. N., Medina, S. P., Feudtner, C., Luan, X., Localio, R., Future of Children, 9(1), 4–26. Fieldston, E. S., & Rubin, D. M. (2012). Local macroeco- Hainmueller, J., & Xu, Y. (2013). Ebalance: A Stata package for nomic trends and hospital admissions for child abuse, 2000– entropy balancing. Journal of Statistical Software, 54(7). 2009. Pediatrics, 130(2), e358–e364. https ://doi.or g/10.1542/ Hellmuth, J. C., Gordan, K. C., & Stuart, G. L. (2013). Risk factors peds.2011-3755. for intimate partner violence during pregnancy and postpartum. Archives of Women’s Mental Health, 16, 19–27. 1 3 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Maternal and Child Health Journal Springer Journals

A Mixed Methods Evaluation of Early Childhood Abuse Prevention Within Evidence-Based Home Visiting Programs

Free
13 pages

Loading next page...
 
/lp/springer_journal/a-mixed-methods-evaluation-of-early-childhood-abuse-prevention-within-APbBH9ZF6M
Publisher
Springer Journals
Copyright
Copyright © 2018 by The Author(s)
Subject
Medicine & Public Health; Public Health; Sociology, general; Population Economics; Pediatrics; Gynecology; Maternal and Child Health
ISSN
1092-7875
eISSN
1573-6628
D.O.I.
10.1007/s10995-018-2530-1
Publisher site
See Article on Publisher Site

Abstract

Objectives In this large scale, mixed methods evaluation, we determined the impact and context of early childhood home visit- ing on rates of child abuse-related injury. Methods Entropy-balanced and propensity score matched retrospective cohort analy- sis comparing children of Pennsylvania Nurse–Family Partnership (NFP), Parents As Teachers (PAT), and Early Head Start (EHS) enrollees and children of Pennsylvania Medicaid eligible women from 2008 to 2014. Abuse-related injury episodes were identified in medical assistance claims with ICD-9 codes. Weighted frequencies and logistic regression odds of injury within 24 months are presented. In-depth interviews with staff and clients (n = 150) from 11 programs were analyzed using a modified grounded theory approach. Results The odds of a healthcare encounter for early childhood abuse among clients were significantly greater than comparison children (NFP: 1.32, 95% CI [1.08, 1.62]; PAT: 4.11, 95% CI [1.60, 10.55]; EHS: 3.15, 95% CI [1.41, 7.06]). Qualitative data illustrated the circumstances of and program response to client issues related to child maltreatment, highlighting the role of non-client caregivers. All stakeholders described curricular content aimed at prevention (e.g. positive parenting) with little time dedicated to addressing current or past abuse. Clients who reported a lack of abuse-related content supposed their home visitor’s assumption of an absence of risk in their home, but were supportive of the introduction of abuse-related content. Approach, acceptance, and available resources were mediators of successfully addressing abuse. Conclusions for Practice Home visiting aims to prevent child abuse among high-risk families. Adequate home visitor capacity to proactively assess abuse risk, deliver effective preventive curriculum with fidelity to caregivers, and access appropriate resources is necessary. Keywords Home visiting · Maternal and child health · Child maltreatment · Mixed methods research Significance Statement What is already known on this topic? Reducing child mal- treatment is a public health priority. At present, home visit- ing represents the primary available prevention strategy for Electronic supplementary material The online version of this article (https ://doi.org/10.1007/s1099 5-018-2530-1) contains supplementary material, which is available to authorized users. * M. Matone Department of Family Medicine and Community Health, matonem@email.chop.edu University of Pennsylvania, Philadelphia, PA, USA Center for Public Health Initiatives, University PolicyLab, The Children’s Hospital of Philadelphia, of Pennsylvania, Philadelphia, PA, USA Philadelphia, PA, USA Leonard Davis Institute of Health Economics, University Division of General Pediatrics, The Children’s Hospital of Pennsylvania, Philadelphia, PA, USA of Philadelphia, Philadelphia, PA, USA Roberts Center for Pediatric Research, 2716 South Street, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA 19146, USA Philadelphia, PA, USA Mixed Methods Research Lab, University of Pennsylvania, Philadelphia, PA, USA Vol.:(0123456789) 1 3 S80 Maternal and Child Health Journal (2018) 22 (Suppl 1):S79–S91 child maltreatment; however, evaluations of home visiting visitation programs found two programs reduced long-term success on this outcome have been varied. child maltreatment reports or death contrasted by four other What does this study add? Our mixed methods evaluation randomized trials showing no associated effect of home pairs a large administrative dataset that allows us to measure visitation on child maltreatment reports (Nelson et al. 2013; child abuse outcomes reliably across multiple home visiting Rubin et al. 2014). programs with qualitative interviews with key stakehold- As home visiting programs have scaled in the context ers to demonstrate and explore home visiting’s impact on of the federal Maternal, Infant, and Early Childhood Home child maltreatment. We find no evidence of positive pro- Visiting program (MIECHV) (“Patient Protection and gram effects but identify explanatory implementation factors Affordable Care Act” 2010), it is important to refine our potentially limiting program effectiveness. understanding of their role and mechanism, and opportuni- ties for maltreatment prevention. In this large scale, mixed methods evaluation, we aimed to estimate the effectiveness Introduction of three MIECHV-funded home visiting models in Penn- sylvania (PA) on early childhood maltreatment ascertained Child maltreatment is a serious public health problem (Chil- from clinical diagnoses in emergency department and inpa- dren’s Bureau 2016) resulting in innumerable short and tient healthcare encounters. A concurrent qualitative anal- long-term health consequences including trauma, adverse ysis of interviews with program staff and clients explores physical and mental/behavioral health, changes to brain the content of and response to curriculum related to child architecture and development, challenges to educational maltreatment. achievement, and reduced social-emotional functioning and relational attachment (Deutsch et al. 2015; Merrick and Latzman 2014). Unfortunately, prevention is challenging to Data and Methods achieve in public health programming, in part due to the level of support needed to overcome risk factors and bolster This study was performed within the Commonwealth of protective factors for families and communities (Agran et al. Pennsylvania (PA) MIECHV evaluation. The study followed 2003; Frioux et al. 2014; Wood et al. 2012). Early childhood a partially mixed, concurrent, equal status design, in which maltreatment prevention programs (i.e. home visitation) qualitative and quantitative data were analyzed separately intervene at a critical period when children are most at risk and mixed at the stage of interpretation (Leech and Onwueg- (Schatz and Lounds 2007) and promote prevention though buzie 2009). The study was approved by PA’s Department of strengthening protective factors within a family and con- Human Services with human subjects approval by the Chil- necting families with community services. Programs teach dren’s Hospital of Philadelphia’s Institutional Review Board. approaches to child rearing, decreasing parental stress, pro- vide guidance on reducing childhood hazard exposures, and Quantitative Data serve a monitoring function for identifying and responding to maltreatment (Gomby et al. 1999). Analytic Sample Inconclusive and at times conflicting clinical trials and post-implementation studies regarding home visiting’s Data was obtained for clients enrolled in MIECHV funded impact on child maltreatment rates have highlighted the need PA nurse–family partnership (NFP, n = 22), parents as for further attention and evaluation (Institute of Medicine teachers (PAT, n = 9), or early head start (EHS, n = 7) from and National Research Council 2014; Rubin et al. 2014). The 2008 to 2014. Clients were matched to local-area non-client varied success in reducing child maltreatment necessitates women (comparisons) who (1) had similarly aged children further investigation into the conditions under which home identified in birth certificate files and (2) resided in the same visiting programs can achieve prevention, and for whom local implementing agency catchment area (i.e., county or (Howard and Brooks-Gunn 2009). multi-county service area). Inclusion criteria for clients Early results from home visiting evaluations suggested and comparisons were as follows: (1) child affiliated with efficacy in decreasing child maltreatment. Notably, a ran - MIECHV program enrollment was identifiable in PA birth domized controlled trial of the Nurse–Family Partnership certificate files and (2) child affiliated with MIECHV pro- (NFP) resulted in 80% fewer injury and ingestion-related gram enrollment had enrollment in the state medical assis- doctor visits in intervention group (though non-significant tance program (Medicaid) during the outcome observation at 2 years) (Olds et al. 1994). However, following wide- period. scale replication of evidence-based home visiting programs, Clients and potential comparisons were identified in a data are limited in supporting the effectiveness in prevent- multisource administrative data file linked using an itera- ing abuse (Matone et al. 2012). A 2013 review of home tive deterministic approach reliant on unique identifiers 1 3 Maternal and Child Health Journal (2018) 22 (Suppl 1):S79–S91 S81 constructed from social security numbers, names, and dates particular catchment were matched to comparison women of birth that included program enrollment, vital statistics living in the same catchment. Entropy balancing and PSM (birth and death), welfare eligibility, and medical assistance were performed within catchments, and then the samples claim files (Dusetzina et al. 2014). were aggregated. Quasi-experimental Design Description of PSM for EHS and PAT The primary analysis examined if the prevalence and rate of PSM is a matching technique for observational data that child abuse episodes significantly differed between program mimics a randomized control trial by creating pairs (or sets) clients and comparison women for NFP, EHS, and PAT pro- from clients and comparison women with similar values of grams separately. Two primary quasi-experimental methods the propensity score (Stuart 2010). Multivariable logistic were used for causal inference related to program effect on regression models estimate the probability of program par- child abuse: entropy balancing for NFP and propensity score ticipation using available maternal sociodemographic and matching (PSM) for EHS and PAT. Both analytic approaches clinical characteristics—from birth certificate: mother’s are widely used for obtaining covariate balance in observa- age at birth (continuous), race/ethnicity (white/black/His- tional data, but neither approach was suitable for all three panic/other), maternal education (< high school/high school program model analyses for the following reasons. or greater), gestational age (continuous), smoking prior to First, PSM has a disadvantage of dropping subjects una- pregnancy (y/n); from welfare eligibility files: receipt of ble to be matched to counterparts, which creates a biased Temporary Assistance for Needy Families (TANF) or sup- sample. In the case of this study, PSM did not retain a gen- plemental nutrition program prior to or during the first tri- eralizable subset of the clients in the NFP analysis (specifi- mester of pregnancy (y/n); from medical assistance claims: cally, young mothers in rural areas were disproportionately Medicaid eligibility (y/n), maternal diagnosis of substance dropped in attempted PSM). Entropy balancing retained all abuse, depression and/or bipolar disorder in the immediate cases and was the approach used for NFP. While PSM was preconception period or first trimester of pregnancy (y/n). not the optimal approach for NFP, it was chosen for the EHS A separate logistic regression model was performed within and PAT analyses because it allowed for standardized fol- each of the catchment areas for each local implementing low-up time in the outcome observation windows of matched agency. Our matching approach used both caliper and exact sets of clients and comparisons. This is a critical analytic matching on covariates to produce matched sets. Any nearest design feature for programs without standardized enroll- neighbor within a caliper of 0.05 was considered a match ment at a point in time. Unlike NFP, which uniformly enrolls (up to a maximum of four comparison women per client). clients into the program prior to a child’s birth, EHS and Matching was conducted exactly on catchment area, infant PAT programs do not uniformly enroll at a particular age. year of birth, and maternal age (< 18 years of age at birth or Therefore, for each client, PSM allowed for the identification 18 and older). A threshold of 2.5 absolute percentage points of comparisons with similarly aged children at the time of was used to determine balance within each catchment area program enrollment. The analysis then standardized obser- model. Interaction terms were added to the propensity score vation periods for outcome ascertainment within matched model when needed to achieve balance. Analytic weights sets of clients and comparisons using the client’s child’s age were developed within matched sets; each comparison of enrollment and length of time in the program as the refer- woman was given a weight equal to the inverse of the num- ence point (e.g., if client enrolled child at 3 months and was ber of comparison women matched to that client and each observed through month 27, all comparison children for that client was given a weight of 1. These weights were applied client are observed for months 3 through 27). This level of to outcome modeling. modeling flexibility is not possible with entropy balancing, but was also not necessary for NFP given the requirement Description of Entropy Balancing for NFP of prenatal enrollment, which serves as a standardization (i.e., all client and comparison children begin observation Entropy matching is a multivariable weighting technique at birth). that creates a balanced sample by reweighting the control Both entropy balancing and PSM were performed within group (in this case the comparison women) to have the same local implementing agency catchment areas (Matone et al. covariate distribution as the treatment group (i.e., the cli- 2012) to address the possibility that there is confounding ents) using the above described maternal sociodemographic by geography (i.e., the outcomes might vary across sites at and clinical characteristics. In this approach, specifications a community level beyond maternal-level characteristics). for each covariate can be applied as to whether exact bal- Catchments included each implementing agency’s county ance between the two groups should be achieved on the first and contiguous counties. Clients enrolled in a program in a moment (mean), second moment (standard deviation), or 1 3 S82 Maternal and Child Health Journal (2018) 22 (Suppl 1):S79–S91 higher moments (Hainmueller and Xu 2013). As is intended Sensitivity Analyses with this methodology, there is automatic balance created between the samples after conducting entropy balancing, so Two sensitivity analyses were conducted to test if the rela- no additional balance checks or model adjustment to create tionship found between program participation and abuse balance was necessary. Covariates used in entropy balancing outcome was robust to potential confounders that could not were the same as included in PSM. be included as covariates in the PSM or entropy balancing model. We tested for confounding between program par- Abuse and Injury Episode Creation ticipation and abuse by maternal psychosocial risk factors by separately including each risk factor in the primary out- The primary outcome for this study was the presence of an come model and examining if the estimated odds ratio effect abuse episode or high risk injury episode (composite meas- for the program participation, adjusting for the risk factor, ure) with a secondary outcome that identified the presence changed by 10% or more. The risk factors ascertained from of any injury episode. Outcome measures were derived from the literature to be confounders are (1) maternal previous child Medicaid claims. Episodes were created to conserva- involvement with child protective services (CPS) before tively count unique instances of abuse and injury recogniz- pregnancy and (2) intimate partner violence (IPV) measured ing that multiple claims/encounters may exist for a single after conception (Berlin et al. 2011; Eckenrode et al. 2000). event. The methodology of collapsing claim encounters to To identify clients involved in CPS, NFP clients and create episodes is described in Matone et al. 2012 and fur- comparisons residing in Philadelphia were linked to county ther in Online Appendix A. child welfare records via first name, last name, date of birth, Abuse episodes were those in which an ICD-9 code indi- and gender. Child welfare systems are administered at the cated child abuse (995.50-5, 995.59), as well as high risk county-level in PA; Philadelphia represents the largest injuries (HRI), specific types of severe injuries considered county in the state and produced a sample large enough for highly suspicious for abuse without the presence of a medi- sensitivity analysis. For any clients and comparisons suc- cal diagnosis of abuse in the medical record. These episodes cessfully linked to child welfare records, dates of protective feature injuries that include fractures of the femur, radius, service were provided. We identified mothers with childhood ulna, tibia, fibula, humerus, ribs, or traumatic brain injuries CPS involvement (prior to pregnancy). within the first 24 months of life (without the presence of Regarding the second sensitivity analysis, IPV was iden- ICD-9 codes indicating an injury due to a motor vehicle tified in maternal medical encounters as an ICD-9 code of crash) (Wood 2010) (Online Appendix B). 995.8x during the observation window of child’s date of Injury outcomes included: superficial injuries, a compos- conception through the first month of life. Even though we ite of dislocation, fracture, and crush injuries, poisonings, identified IPV in some clients after program enrollment and burns identified through ICD-9 codes. occurred or after the program services could have started, we The observation window for episodes were claims during deemed this time window as most meaningful for measuring 0 to 24 months of life for NFP cohort and, for EHS and PAT IPV for two reasons: (1) to reflect a baseline risk proximal to cohorts, 24 months post-enrollment, up to 6 years of life. program enrollment and (2) to increase likelihood of ascer- Right-censoring of episodes occurred for children whose tainment in medical assistance files given increased health observation periods exceeded the study end period of 2014. seeking during pregnancy and increased risk period for IPV. While rates of IPV during pregnancy vary depending on Outcome Models the samples studied and measures used, the prevalence of IPV during pregnancy is elevated compared to women of The primary exposure was NFP, PAT, or EHS program par- non-reproductive age and may be increased compared to ticipation. For EHS and PAT analyses, a weighted condi- non-pregnant patients (Hellmuth et al. 2013; Jasinski 2004). tional logistic regression model was used to examine the Less biased screening (i.e. more universal screening) may unadjusted association between program participation and occur during the prenatal period due to recommendations by the primary outcome. For the NFP analysis, a weighted professional organizations, such as the American Congress logistic regression with a random intercept for county was of Obstetricians and Gynecologists’ (ACOG), that provid- used to estimate the relationship between program participa- ers should screen all women for IPV at periodic intervals, tion and the primary outcome, controlling for the variabil- including during obstetric care (at the first prenatal visit, ity in the outcome across counties. The presence of abuse at least once per trimester, and at the postpartum checkup) or injury prior to enrollment was included as an adjust- (“ACOG Committee Opinion No. 518: Intimate partner vio- ment covariate in PAT and EHS final outcome models (not lence” 2012). applicable for NFP modeling given prenatal enrollment) to For each set of primary analyses described above, we ran account for baseline injury risk. two additional models—one with a dichotomous covariate 1 3 Maternal and Child Health Journal (2018) 22 (Suppl 1):S79–S91 S83 for presence of maternal involvement with CPS prior to Results childbirth and another with a covariate for presence of IPV within child’s date of conception through the first month Quantitative Data of life. Cohort Demographics Presentation of Results The entropy balanced NFP cohort included 8736 clients enrolled in 22 NFP programs between 2008 and 2014 matched Logistic regression results were expressed as odds ratios to 165,033 comparisons. The propensity score matched PAT (with 95% confidence intervals) and standardized marginal cohort included 851 clients enrolled in nine PAT programs probabilities. All analyses were conducted using SAS ver- matched to 2929 comparisons; EHS cohort included 866 cli- sion 9.4, Stata version 14.2 and R. Stata’s ebalance package ents enrolled in seven EHS programs propensity score matched was used for entropy balancing and R’s MatchIt for PSM. to 3100 comparisons (Table 1). All statistical tests were two-sided and used an alpha = 0.05 Across all models, the majority of clients were unmarried as the threshold for statistical significance. and non-Hispanic white race/ethnicity. In terms of model- specific differences, compared to PAT and EHS client, NFP clients were most likely to be under the age of 18, most likely Qualitative Data to be Hispanic, and least likely to smoke prior to pregnancy (Table 1). Setting and Participants Abuse Episodes Among Clients and Comparison Women 11 of the 38 PA MIECHV-funded programs were selected for the qualitative study, chosen to supply a representative Across all models, children of clients were significantly more sample of agencies, based on program size, location, and likely to experience an abuse episode than comparisons: NFP model type, including NFP, PAT, EHS, and Healthy Families OR: 1.32, 95% CI [1.08, 1.62]; PAT OR: 4.11, 95% CI [1.60, America (HFA, which, due to data constraints, could only be 10.55]; EHS OR: 3.15, 95% CI [1.41, 7.06] (Table 2). Within included in the qualitative study). Program staff were inter - NFP, 1.4% of client children (n = 120) and 1.0% of comparison viewed during day-long site visits; enrolled clients, recruited children (n = 1488) sustained an abuse injury within 24 months with flyers and help from program staff, were interviewed of life; 1.1% (n = 9) of PAT-enrolled children and 0.3% of PAT over the phone. Participants were verbally consented before comparison children (n = 9) sustained an abuse injury within participating in interviews lasting between 30 and 60 min. 24 months of enrollment; and 1.3% of EHS program-enrolled Clients were sent a $20 gift card in appreciation for their children (n = 11) and 0.4% EHS comparison children (n = 14) time. Interviews took place between 2013 and 2015. sustained an abuse injury within 24 months of enrollment. Distribution of Injury Types among Client Children Measures and Analysis with an Abuse Episode The interdisciplinary project team worked with home vis- The most frequent injury types among home visited children iting leadership to develop three distinct interview guides who experienced an abuse episode included superficial inju- for program administrators, home visitors, and clients, each ries and dislocation, fracture or crush injuries. Burns were the including questions on specific program outcomes (e.g. child least prevalent injury type in aggregate. While poisonings were maltreatment; See Online Appendix C for questions that infrequent among NFP children with abuse episodes (2.0%), elicited content related to child maltreatment.). De-identi- one in five children in PAT with abuse episodes experienced fied transcripts were imported into NVivo 10 for coding and poisoning while dislocations, fractures and crush injuries analysis. We used a modified Grounded Theory approach to occurred with half the frequency among PAT children than coding (Glaser and Strauss 1967), including a priori codes NFP children with abuse (23.5 versus 59.7%) (Table 3). relating to quantitative metrics included in the evaluation. Using a constant comparative approach, coders met regularly Sensitivity Analyses to review memos and coding comparison queries to discuss and refine node definitions and the application of codes. The prevalence of maternal childhood CPS involvement Discrepancies were resolved through group consensus. A (prior to pregnancy) was equitable between clients and thematic analysis was conducted on all interview content comparisons, with greater than one-third of mothers related to child maltreatment. 1 3 S84 Maternal and Child Health Journal (2018) 22 (Suppl 1):S79–S91 1 3 Table 1 Characteristics of clients of nurse–family partnership (NFP), parents as teachers (PAT), and early head start (EHS), and comparison women, 2008–2014 NFP PAT EHS Comparison women Clients Comparison women Clients Comparison women Clients N = 165,033 % N = 8736 % N = 2929 % N = 851 % N = 3100 % N = 866 % Age, < 18 years 15,601 20.1 1955 22.4 183 6.5 51 6.0 185 5.8 54 6.2 Race/ethnicity  White 67,240 47.8 3028 47.4 2252 77.3 657 77.9 1506 50.8 420 49.1  Black 32,837 22.3 1451 22.7 251 8.4 72 8.5 987 30.5 275 32.1  Hispanic 25,232 28.2 1801 28.2 365 12.7 107 12.7 483 15.6 139 16.2  Other 5751 1.8 112 1.8 50 1.6 7 0.8 91 3.0 22 2.6 Unmarried 131,010 89.9 7857 89.9 2206 74.8 584 68.6 2480 78.3 657 75.9 Education, less than high school 37,307 37.9 3347 38.3 825 28.3 258 30.3 1123 36.3 322 37.2 TANF receipt 54,399 43.2 3770 43.2 1546 51.9 411 48.3 1676 52.5 446 51.5 Foodstamp receipt 55,753 49.0 4280 49.0 1882 63.3 517 60.8 1899 60.2 495 57.2 Depression in proximity to pregnancy 3820 6.0 527 6.0 287 10.6 113 13.3 162 5.3 69 8.0 Substance abuse in proximity to pregnancy 3599 3.2 280 3.2 166 5.9 46 5.4 128 4.1 40 4.6 Medicaid eligibility in proximity to pregnancy 55,664 40.3 3534 40.5 982 32.8 266 31.3 924 29.8 181 20.9 Smoking, prior to pregnancy 46,168 30.7 2684 30.7 1361 46.3 382 44.9 1321 42.5 350 40.4 Table reports unweighted Ns and weighted proportions Maternal and Child Health Journal (2018) 22 (Suppl 1):S79–S91 S85 Table 2 Marginally standardized probabilities and odds of child Qualitative Data abuse among comparison women and home visiting clients enrolled in nurse–family partnership (NFP), parents as teachers (PAT), and A total of 150 interviews were conducted with program early head start (EHS) administrators (N = 25), agency staff (N = 49), and home HV program Com- HV clients (%) OR (95% CI) p Value visiting clients (N = 76). The sample represents all four parisons MIECHV-eligible evidence-based programs in PA, with (%) greater representation from NFP (35% of staff and 47% NFP 1.0 1.4 1.32 (1.08, 1.62) 0.008 of clients). Participants from urban and rural sites were PAT 0.3 1.1 4.11 (1.60, 10.55) 0.003 equally represented (Table 4). EHS 0.4 1.3 3.15 (1.41, 7.06) 0.005 Our thematic analysis of the interview data related to child maltreatment identified two primary domains: (1) how programs and staff react when there is a suspicion Table 3 Marginally standardized probabilities of injury types among of child maltreatment; and (2) how programs and sta ff children with an abuse episode by home visiting program address and clients engage with child maltreatment pre- vention strategies. HV program Marginally stand- ardized probability (%) NFP  Superficial injury 38.1  Dislocation, fracture, crush 59.7 Table 4 Demographics of interview participants  Poisoning 2.0 Clients (N = 76) Staff (N = 74)  Burns 3.8 PAT Program % %  Superficial injury 30.4  NFP 47 35  Dislocation, fracture, crush 23.5  PAT 24 22  Poisoning 19.4  EHS 14 24  Burns –  HFA 14 20 EHS Urbanicity  Superficial injury 32.0  Urban 51 59  Dislocation, fracture, crush 65.3 Sex  Poisoning –  Female 93 99  Burns – Race  White 57 82 NFP nurse family partnership, PAT parents as teachers, EHS early head start  Black/African American 33 9 Due to small sample size, OR and marginally standardized probabil-  Other 11 5 ity cannot subsequently be calculated Ethnicity  Non-Hispanic 95 95 Age having CPS involvement prior to their first pregnancy. The  18 and under 4 – results of sensitivity analyses demonstrate a lack of con-  19–22 21 – founding by CPS involvement on the relationship between  23+ 75 – program enrollment and abuse (Online Appendix D). Employment Roughly 1% of NFP and PAT clients and comparisons  Unemployed 57 – were diagnosed with IPV in the pregnancy period (Online Marital status Appendix E, Table 1). In sensitivity analyses, maternal  Single 51 – IPV exposure was not found to be a confounder of the rela-  Married/partnered 45 – tionship of program enrollment and abuse (Online Appen-  Separated or divorced 4 – dix E). However, across both programs, mothers with diag- Education nosed IPV were significantly more likely to have a child  High school or less 50 – with an abuse episode than those without diagnosed IPV  Some college 37 – (NFP: 2.7 versus 1.2%, p = 0.027; PAT: 6.0 versus 0.6%,  College or higher 13 – p = 0.027) (Online Appendix E, Table 2). Total 100 100 1 3 S86 Maternal and Child Health Journal (2018) 22 (Suppl 1):S79–S91 Program and Staff Responses to Suspicion or Incidents [T]he program talks about strategies with kids having of Child Maltreatment temper tantrums and stuff. Not only things that you can do for the kiddo, but things as a parent. Maybe you In a few instances during interviews, clients and home visi- need to step away, take a few breaths, collect yourself tors described instances of potential child maltreatment, and then go back and address the issue. EHS Admin- shedding light on how the program responds in such cir- istrator, 1002 cumstances (Table 5). Parents valued this guidance and support, perceiving Of the cases describing incidents concerning for abuse improvements in managing their reactions to frustrating and neglect, most involved caregivers other than the client child behaviors. Clients attributed personal changes to the as the perpetrator, including partners, child siblings, and general encouragement from home visitors to attend to their grandparents. Program responses tended to focus on sup- own personal needs and more specific support related to porting the client after the abuse had occurred with parent- stress. ing support, referrals, and connections to resources. Some home visitors described relying on CPS to manage child I get stressed out easy and she’s helped me learn maltreatment issues. how to deal with stressful situations, when it comes down with my daughter when she starts getting in her, [I]f there is any indication of violence in the home “give me, wanna wanna” modes […] and I get really that would always be referred to [CPS], if that child stressed about it. She helps me learn how to deal with was being exposed to that. A lot of times, you can just the stress and teaches me how to deal with my daugh- see it while you’re there. […] sometimes they don’t ter […]. […S]he told me, it’s important that you do even try to hide it. Or they’ll […] say something that this for yourself. Also, it’s important you do this for the other person has been doing or whatever. So we you daughter. I need to take a little down time. HFA don’t get into that too much, other than to say a healthy Client, 1004 environment is what our goal is here and whatever you need to do to get that. HFA/PAT Home Visitor, 801 Some clients described how learning new discipline strat- egies gave them the opportunity to break a cycle of abuse. Home visitors qualified their relationship with CPS in many ways, including their role as mandated reporters. We’ve talked about discipline, not so much of abuse. Given the negative feelings often associated with CPS, […] I think it’s effective to make sure that my child some home visitors described being careful about how they understands that I know what they want and kind of explained their relationship with CPS to clients. having different methods instead of going directly to a punishment. […]My parents disciplined me by beating Approach to and Engagement with Child Maltreatment me and punishing me for extremely long periods of Prevention time […]. […] Those things that I didn’t really know about prior to being involved in the program because Home visitors described focusing on content related to posi- those are things that weren’t taught to me as a child. tive parenting, secure attachment, and stress management PAT Client, 7020 as the curricular drivers of child abuse prevention. Many When asked how they talked with their home visitor staff referenced the program training they received, which about protecting their child from abuse, a number of clients taught them that supporting the development of a strong did not remember the subject being addressed. Many clients bond between a parent and child could reduce the chances acknowledged that discussing child maltreatment should be of maltreatment. part of the curriculum covered by their home visitor. Implied We stress with [parents] attachment and bonding. in the absence of direct discussions of child abuse was the Because if a parent does attachment and bonding, described assumption that the home visitor knew there was we’ve learned through all of our trainings with the not abuse occurring within the household. HFA, they’re gonna be less likely to abuse their child I don’t remember talking about that. I would assume […], and that is what everything circles back to. HFA that she knows that we are not abusive at all here. But Home Visitor, 103 I know that she shouldn’t assume, because you don’t Home visitors focused with parents on approaches and know what happens when someone leaves. PAT Cli- frameworks for disciplining children. The described cur- ent, 2004 ricula addressed strategies directed both at the children, to From the client perspective, very few barriers to broach- manage their behavior, and at the adults, to manage associ- ing the topic of abuse were described. One parent postulated ated stress. 1 3 Maternal and Child Health Journal (2018) 22 (Suppl 1):S79–S91 S87 Table 5 Qualitative interview data representing concern for child maltreatment Perpetrator Program response Excerpt from interview data Both parents Adapt program curriculum to specific child need “I do have one family that we’re working mainly on gross motor because he’s not – when he runs he doesn’t bend his knees quite as often. And that’s due to they’re kind of putting him in the playpen and not letting him out because he’ll make a mess and they don’t feel like cleaning up the mess, obvi- ously. So he’s kind of slowly developing gross motor because he’s being confined. So I’m trying to delicately word – in that maybe it’s because he’s in the playpen that his gross motor skills aren’t developing. So when I go over, we kind of do more activities towards that, and work that in with what I already have planned.” EHS Home Visitor, 402 Parent (non-client) Provide informational support “[I]f [my son] looks different for some reason or he comes back Adapt sessions to ask about relationship with Father from his dad’s and I have a visit with her. And she’s like, CPS referral he has another head contusion from his dad. [...] She really tells me – well, did you take him to the emergency room? Make sure you watch it. […] She gives me paperwork on the head contusion. What to look for – if he sleeps too much or – stuff like that. […] She gave me – I forgot what it’s called. Something Domestics – Children and Youth. But, I mean, it’s not like – I don’t think he’s harming my son. It’s just the fact that he isn’t very good at watching him. So – oh, yeah, I’ve thought about it and definitely think that next time that he does come home with a head contusion, then – or if he took him to the emergency room where he has a huge boonie on his head. Usually if he does have a boonie on his head, that’s a thick bruise or swollen, I take him straight to the emergency room right when I pick him up. We don’t go home. We don’t go eat. We go to the emergency room, just because I want to have that documented as soon as possible – and if anything is really like more internally wrong. But I would definitely – she’s given me stuff about Children and Youth.” NFP Client, Sibling (non-client) Adapt program curriculum to specific child need “[My son] was really aggressive towards [my daughter…]. [My son] was basically the only child that I was dealing with one-on-one. And then got pregnant having my daughter. And he started feeling neglected, started to act out more, so where he would start trying to hit her and do certain little things to her and stuff like that. And with the program, they basically helped me focus on him and have time with him and also bond with the baby. Also, help him bond with the baby, like play with – show him ways to play and stuff. [...] Basically, redirecting him to do something else, like to move him away from her, pull him aside and play with just him, like one-on- one with him.” PAT Client, 8004 1 3 S88 Maternal and Child Health Journal (2018) 22 (Suppl 1):S79–S91 Table 5 (continued) Perpetrator Program response Excerpt from interview data Sibling (non-client) Provide resource “I have two stepchildren and the one, their mother has filled Provide informational and emotional support their head with a lot of awful things, and he’s very abusive and we have a lot of problems. [...O]ne day she came and [my son] had a bruise on his cheek. And she said, how did that happen? And I explained to her that he had been over playing by the baby gate and his brother come out and just hit him in the face with the baby gate. And [...] then she had seen that I had a bruise on my face where he had hit me. And she had said, this has got to stop, this is what we’re gonna do. And she brought in [a therapist] to talk to me. And then we went from there and started informing all these agencies. And she said it might now help you, but at least it’s out there that if something major happens, you attempted to try to get the help you needed and the mother closed it down, but you attempted. And she said it’s the same thing if [Son] goes to school and they find a bruise on him, they’re gonna come after you and it may not be you, it could have been him, but you’re, you know, you have at least explained the situation and have tried to get help.” EHS Client, 6001 Grandparent (non-client) Provide informational and emotional support “I went through a domestic violence case between my mom and then through the girls’ dad. […My Home visitor]actually brought paperwork, and she physically worked with me of how to do things and that, or like if she wasn’t here, I could call her for advice or give her a text message for advice. It was like a full process. It helped by steps and physical help sometimes. [...] Like she showed me how – because my [daughter], she had fractured ribs from my mom. And [my home visitor] actually helped me of ways to hold her that’ll help her ease her pain, and then – It was really nice. Without it, I probably would have went insane with all the crying. [...] I asked my mom to watch my two kids at her house while I went to the hospital because I was really, really sick. And my current boyfriend at the time, he was working. And I had like nobody else to watch my kids, and I didn’t want them to get sick and see me suffer type of thing. And then we brought her back home and she wouldn’t stop crying, and we couldn’t figure out why, then we ended up taking her to the hospital, and CYS showed up at my door at 4:00 in the morning asking if we knew what was wrong with my daughter. And I broke. I bursted into tears, because I didn’t know what’s going on.” EHS Client, 10003 Fiancé (non-client) Unknown “So one time, we thought [my daughter] did break her leg whenever her and [my fiancé] were wrestling. But we went to [Hospital 1]. They made us wait two hours in the waiting room, so I left. And they called CYS on me, and CYS had came. And they just wanted to see that [my daughter] wasn’t afraid of [my fiancé] because the hospital was saying that the father was abusing her. They closed us out that day knowing that, you know, nothing bad was going on. So we took her to [Hospital 2] because I wasn’t waiting there if my daughter may have had – like, a one-year old may end up having a bro- ken leg. I’m not wasting any time, so I changed the hospital. I went up to [Hospital 2] instead. Here, she didn’t have a bro- ken leg. [...] It was, like, a fracture in her thigh bone. [...] So, like, I had told, you know, CYS, you know, I left [Hospital 1]. I went to [Hospital 2] because they were making my daughter wait in the waiting room for two hours.” EHS Client, 4005 1 3 Maternal and Child Health Journal (2018) 22 (Suppl 1):S79–S91 S89 that using videos to broach the subject of child maltreatment including fractures, dislocations, and crush injuries, the like- was the only way the program could address it without put- lihood of a family to seek healthcare is likely less dependent ting parents on the defensive. on the presence of home visitor. Other considerations for the observed increased rate of abuse-related injury among home I think that the videos and the conversations are really visited families include unidentified confounding that does all they can do without having somebody be like, not account for the referral bias driving higher risk clients you’re intruding or PC or social warrior. That’s a sen- into home visiting services. To address this concern, this sitive subject for a lot of people, how you’re gonna study included two sensitivity analyses to account for poten- discipline your kids or whatever. NFP Client, 5002 tial confounding by the two strongest psychosocial risk fac- Some home visitors discussed difficulties encouraging tors associated with child maltreatment: maternal childhood families to alter behaviors. The set of norms a parent devel- involvement with CPS or IPV. Despite a high prevalence of ops around how adults should interact with children from maternal childhood CPS involvement among clients, this years of personal experience is difficult to contend with risk factor was found to be balanced between clients and when misaligned with home visitor communication. comparison women and did not confound findings. Mater - nal IPV in pregnancy was also not an identified confounder Challenges are we’re not there all the time to monitor in this study, though diagnosed rates of IPV were low and them. We can refer them, we can’t force them to go. I likely represented underascertainment of true IPV risks mean, there’s a deep history in how people discipline within the cohort. Among mothers with diagnosed IPV in their children and it’s hard to break that within a cou- pregnancy, rates of abuse were higher even after adjustment ple of months. NFP/PAT Home Visitor, 710 for program enrollment, indicating that these families may Unless a parent is personally invested in changing their be a high risk subgroup of home visited clientele. patterns, it is difficult for the home visitor to champion A strength of this study is the provision of qualitative change alone. When parents are open to using new strate- contextual support for the development of hypotheses of why gies, true influence depends on the degree to which these home visiting programs may struggle in reducing rates of strategies replace problematic practices. child abuse. The instances of child abuse assessed in inter- views described events where harm occurred to the child I learned that there’s no wrong or right way to teach outside of the client’s oversight while under the supervi- your child behaving. […] Like say if she would think sion of a non-client caregiver. Acknowledging the limita- I’m teaching her, like I pop her or whatever, that could tion that interviewees may be more likely to discuss inci- be okay, to pop her, but you know, also you got to like dents in which they were not the perpetrator, the number start taking stuff away and then teaching her no, and of examples including other caregivers is still notable. The telling her don’t do stuff. NFP Client, 7015 data are reflective of other information described by program sites suggesting that the vast majority of serious and abuse- related injuries take place while the child is in the care of a Discussion non-client caregiver, often an intimate partner. As programs encourage clients to engage in educational and professional This study demonstrates quantifiably increased risk of early advancements, it is important to address resultant childcare childhood abuse-related injury among children of MIECHV needs. Additionally, the role of non-clients in child maltreat- clients compared to non-program enrolled comparison chil- ment events makes evident the importance of delivering pro- dren across three home visiting programs (NFP, EHS, and gram curriculum to as many of the caregivers in contact with PAT) implemented in a large state. Among client children children as possible. Further support for this hypothesis is experiencing abuse, elevated rates of fracture, dislocation, demonstrated by the quantitative sensitivity analysis of the and crush injuries were seen in comparison to client children significant impact of the presence of IPV on child abuse. experiencing non-abuse related injuries. The curriculum related to child maltreatment, as well as Null findings on child abuse prevention to home visit- home visitors’ delivery of it also plays a role in achieving ing programs are not new to the field. Others have cited intended program outcomes. Recent evidence suggests that surveillance bias among enrolled families as a reason for home visiting programs may struggle to achieve positive the lackluster program ee ff ct on child maltreatment (Gomby child maltreatment outcomes due to ineffectual program et al. 1999; Olds et al. 1995). While surveillance bias may delivery. Much of home visiting curricula is prepared to help be a contributor to increased observation of abuse-related parents decrease the effects of adverse childhood experi - injury among client families, the bias is likely most relevant ences (ACEs) on their children, even without a reduction in to minor injuries where healthcare seeking behavior was actual ACEs (McKelvey et al. 2016). However, home visi- optional. In the case of more serious and emergent injuries, tors themselves have not reported high efficacy in guiding 1 3 S90 Maternal and Child Health Journal (2018) 22 (Suppl 1):S79–S91 discussions with parents related to sensitive topics such as analyses with likely confounders, minimizes this concern maltreatment and violence (Duggan et al. 2007). Casillas and is a widely accepted technique for estimating causality et al. notes that child maltreatment outcomes in particular are when randomization is not feasible or appropriate. Moreo- negatively affected by low home visitor efficacy and fidelity ver, the standardized approach to outcome ascertainment in service provision (Casillas et al. 2016). Our qualitative using administrative rather than self-reported data coupled data demonstrate variations in whether and how home visi- with the rigorous quasi-experimental design facilitate stand- tors discussed child abuse with clients, such that a number ardized interpretation of findings across three home visiting of clients did not recollect the topic ever being raised. This models and numerous diverse implementing sites. While our finding may have been driven by role issues, as mandatory analyses were unable to take program dosage and the per- reporters and assumed affiliations with CPS were barriers to petrator of abuse incidents into account in a standardized direct discussions of maltreatment. Home visitors also strug- fashion as this information was unavailable in administra- gled against normative factors shaping behavior change, as tive data files, this information was gleaned from qualita- well as those more directly related to maltreatment such as tive interviews when it was addressed by home visitors or corporal punishment. A recent evaluation of PAT program in clients. The study is not able to account for surveillance Connecticut demonstrated significant reductions in substan - bias; however, the study used an intention to treat design tiated child maltreatment; however, this finding was driven to observe children for two full years following enrollment by reductions in neglect with no demonstrated impact on while a significant proportion of families are not retained in child abuse. Such findings further highlight the need for services for this duration (Burwick et al. 2014). Lastly, given additional supports around abuse specifically, but also pro- the sensitive nature of this topic, the qualitative interview vide support for qualitative findings in this study that suggest data is likely impacted by social desirability bias. more home visitor comfort in the domains of parenting that may be more directly related to neglect-related maltreatment (Chaiyachati et al. 2018). Conclusion Despite hearing an openness to discussing issues of child maltreatment from more clients than those noting challenges, Home visiting programs have a stated objective of prevent- overall the qualitative data illustrate little direct engagement ing child abuse-related injury among high-risk families. around child maltreatment. Interviews with clients and staff The success of abuse prevention depends on the strength demonstrate that indirect approaches, such as activities to of the curriculum, the fidelity of delivery, and whether it alter parenting styles or decrease maternal stress, are present reaches the people in caregiving roles. Given that home visi- and more likely to be delivered with fidelity. However, it is tors depend greatly on other community agencies and public possible that focusing on supporting positive parenting and systems, it is naïve to expect home visiting to achieve strong stress reduction is not sufficient to effectively reduce the risk outcomes without a well-integrated and -resourced service of abuse incidents among the subset of home visited families network. Programs and clients would benefit from curricu- most at-risk for maltreatment. lum that more directly and proactively addresses child mal- Finally, the qualitative data describe limited reactive treatment, as well as knowledge of and access to quality responses to situations concerning for neglect and referral childcare options. to outside resources—mainly CPS. The data highlights the dependence of home visiting success on the system of ser- Acknowledgements This evaluation was supported by a grant from the vices in which the program resides. The importance of the Department of Human Services, Commonwealth of Pennsylvania. We thank the Pennsylvania Department of Human Services, Pennsylvania quality, access, and connectedness of CPS, childcare, and Department of Health, and the MIECHV local implementing agencies healthcare to home visiting services towards the prevention across Pennsylvania for contributing data for this study. We are also of child maltreatment prevention cannot be overstated. The grateful to the Office of Child Development and Early Learning of the ability of home visitors to transition from reactivity to pro- Pennsylvania Department of Human Services for their consultation on Pennsylvania MIECHV programming and partnership in identifying activity around child abuse is dependent on effective referral evaluation priorities for the Commonwealth. relationships, trust in the CPS system, and timely and afford- able access to services that mitigate abuse risks, including Open Access This article is distributed under the terms of the Crea- childcare and healthcare. tive Commons Attribution 4.0 International License (http://creat iveco This study has several limitations related to study design mmons.or g/licenses/b y/4.0/), which permits unrestricted use, distribu- tion, and reproduction in any medium, provided you give appropriate and data availability and reliability. The observational study credit to the original author(s) and the source, provide a link to the design is subject to bias when estimating program effects; Creative Commons license, and indicate if changes were made. however the use of propensity score and entropy matching, which mimic randomization and control for measured dif- ferences between clients and comparisons and sensitivity 1 3 Maternal and Child Health Journal (2018) 22 (Suppl 1):S79–S91 S91 Howard, K. S., & Brooks-Gunn, J. (2009). The role of home-visiting References programs in preventing child abuse and neglect. The Future of Children, 19(2), 119–146. ACOG Committee Opinion No. 518. (2012). Intimate partner vio- Institute of Medicine, & National Research Council. (2014). New lence. Obstetrics Gynecology, 119(2 Pt 1), 412–417. https ://doi. directions in child abuse and neglect research. Washington, DC: org/10.1097/AOG.0b013 e3182 49ff7 4. National Academies Press. Administration for Children and Families. https ://www.acf.hhs.gov/ Jasinski, J. L. (2004). Pregnancy and domestic violence: A review of sites /defau lt/files /cb/cm201 5.pdf. Accessed 1 Mar 2018 the literature. Trauma, Violence, and Abuse, 5(1), 47–64. Agran, P. F., Anderson, C., Winn, D., Trent, R., Walton-Haynes, L., & Leech, N. L., & Onwuegbuzie, A. J. (2009). A typology of mixed meth- Thayer, S. (2003). Rates of pediatric injuries by 3-month intervals ods research designs. Quality and Quantity, 43(2), 265–275. for children 0 to 3 years of age. Pediatrics, 111(6), e683–e692. Matone, M., O’Reilly, A. L., Luan, X., Localio, R., & Rubin, D. M. https ://doi.org/10.1542/peds.111.6.e683. (2012). Home visitation program effectiveness and the influence Berlin, L. J., Appleyard, K., & Dodge, K. A. (2011). Intergenerational of community behavioral norms: A propensity score matched continuity in child maltreatment: Mediating mechanisms and analysis of prenatal smoking cessation. BMC Public Health, 12, implications for prevention. Child Development, 82(1), 162–176. 1016. https ://doi.org/10.1186/1471-2458-12-1016. https ://doi.org/10.1111/j.1467-8624.2010.01547 .x. Matone, M., O’Reilly, A. L. R., Luan, X., Localio, A. R., & Rubin, Burwick, A., Zaveri, H., Shang, L., Boller, K., Daro, D., Strong, D. D. M. (2012). Emergency department visits and hospitaliza- (2014). Costs of early childhood home visiting: An analysis of tions for injuries among infants and children following statewide programs implemented in the supporting evidence-based home implementation of a home visitation model. Maternal and Child visiting to prevent child maltreatment initiative. Health Journal, 16(9), 1754–1761. https: //doi.org/10.1007/s1099 Casillas, K. L., Fauchier, A., Derkash, B. T., & Garrido, E. F. (2016). 5-011-0921-7. Implementation of evidence-based home visiting programs aimed McKelvey, L. M., Whiteside-Mansell, L., Conners-Burrow, N. A., at reducing child maltreatment: A meta-analytic review. Child Swindle, T., & Fitzgerald, S. (2016). Assessing adverse expe- Abuse & Neglect, 53, 64–80. https ://doi.or g/10.1016/j.c hiab riences from infancy through early childhood in home visiting u.2015.10.009. programs. Child Abuse & Neglect, 51, 295–302. https ://doi. Chaiyachati, B. H., Gaither, J. R., Hughes, M., Foley-Schain, K., & org/10.1016/j.chiab u.2015.09.008. Leventhal, J. M. (2018). Preventing child maltreatment: Exami- Merrick, M. T., & Latzman, N. E. (2014). Child maltreatment: A public nation of an established statewide home-visiting program. Child health overview and prevention considerations. Online Journal Abuse & Neglect, 79, 476–484. https ://doi.org/10.1016/j.chiab of Issues in Nursing, 19(1), 1–1. https ://doi.org/10.3912/OJIN. u.2018.02.019. Vol19 No01M an02. Children’s Bureau. (2016). Child maltreatment 2015. Washington, DC: Nelson, H. D., Selph, S., Bougastsos, C., & Blazina, I. (2013). Behav- Administration on Children Youth and Families. ioral interventions and counseling to prevent child abuse and Deutsch, S. A., Lynch, A., Zlotnik, S., Matone, M., Kreider, A. R., neglect: A systematic review to update the U.S. Preventative Ser- & Noonan, K. (2015). Mental health, behavioral and develop- vices Task Force recommendation. Annals of Internal Medicine, mental issues for youth in foster care. Current Problems in Pedi- 158(3), 179–190. atric and Adolescent Health Care, 45(10), 292–297. https ://doi. Olds, D., Henderson, C. R. Jr., & Kitzman, H. (1994). Does prenatal org/10.1016/j.cpped s.2015.08.003. and infancy nurse home visitation have enduring effects on quali- Duggan, A., Caldera, D., Rodriguez, K., Burrell, L., Rohde, C., & ties of parental caregiving and child health at 25 to 50 months of Crowne, S. S. (2007). Impact of a statewide home visiting pro- life? Pediatrics, 93(1), 89–98. gram to prevent child abuse. Child Abuse & Neglect, 31(8), 801– Olds, D., Henderson, C. R. Jr., Kitzman, H., & Cole, R. (1995). Effects 827. https ://doi.org/10.1016/j.chiab u.2006.06.011. of prenatal and infancy nurse home visitation on surveillance of Dusetzina, S. B., Tyree, S., Meyer, A. M., Meyer, A., Green, L., & Car- child maltreatment. Pediatrics, 95(3), 365–372. penter, W. R. (2014). Linking data for health services research: A Patient Protection and Affordable Care Act. § 2951 (2010). framework and instructional guide. Rockville, MD: Agency for Rubin, D. M., Curtis, M. L., & Matone, M. (2014). Child abuse pre- Healthcare Research and Quality. vention and child home visitation: Making sure we get it right. Eckenrode, J., Ganzel, B., Henderson, C. R. Jr., Smith, E., Olds, D. JAMA Pediatrics, 168(1), 5–6. https://doi.or g/10.1001/jamapediat L., Powers, J., … Sidora, K. (2000). Preventing child abuse and rics.2013.3865. neglect with a program of nurse home visitation: The limiting Schatz, J. N., & Lounds, J. J. (2007). Child maltreatment: Precursors effects of domestic violence. JAMA, 284(11), 1385–1391. of developmental delays. In J. G. Borkowski, J. R. Farris, T. L. Frioux, S., Wood, J. N., Fakeye, O., Luan, X., Localio, R., & Rubin, Whitman, S. S. Carothers, K. Weed & D. A. Keogh (Eds.), Risk D. M. (2014). Longitudinal association of county-level economic and resilience: Adolescent mothers and their children grow up indicators and child maltreatment incidents. Maternal and Child (pp. 125–150). Mahwah, NJ: Lawrence Erlbaum. Health Journal, 18(9), 2202–2208. https: //doi.org/10.1007/s1099 Stuart, E. A. (2010). Matching methods for causal inference: A review 5-014-1469-0. and a look forward. Statistical Science: A Review Journal of the Glaser, B. G., & Strauss, A. L. (1967). The discovery of grounded Institute of Mathematical Statistics, 25(1), 1. theory: Strategies for qualitative research. New York: Aldine de Wood, J. N. (2010). Disparities in the evaluation and diagnosis of abuse Gruyter. among infants with traumatic brain injury. Pediatrics, 126(3), Gomby, D. S., Culross, P. L., & Behrman, R. E. (1999). Home visiting: 408–414. Recent program evaluations: Analysis and recommendations. The Wood, J. N., Medina, S. P., Feudtner, C., Luan, X., Localio, R., Future of Children, 9(1), 4–26. Fieldston, E. S., & Rubin, D. M. (2012). Local macroeco- Hainmueller, J., & Xu, Y. (2013). Ebalance: A Stata package for nomic trends and hospital admissions for child abuse, 2000– entropy balancing. Journal of Statistical Software, 54(7). 2009. Pediatrics, 130(2), e358–e364. https ://doi.or g/10.1542/ Hellmuth, J. C., Gordan, K. C., & Stuart, G. L. (2013). Risk factors peds.2011-3755. for intimate partner violence during pregnancy and postpartum. Archives of Women’s Mental Health, 16, 19–27. 1 3

Journal

Maternal and Child Health JournalSpringer Journals

Published: May 31, 2018

References

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off