Assessing the social determinants of health care costs for Medicaid‐enrolled adolescents in Washington State using administrative data

Abstract

Objective

The study used administrative data to identify the social determinants that have the greatest impact on Medicaid expenditures in adolescence.

Data Sources

Data were compiled using the Washington State Department of Social and Health Services Integrated Client Databases, which link data from state systems including Medicaid claims and social services receipt.

Study Design

Medical system and behavioral health service costs of over 180 000 Medicaid‐enrolled adolescents aged 12‐17 were measured using integrated administrative data from Washington State. Social determinants of health, including child maltreatment and parent risk factors, were also measured. Two‐stage regression models were used to identify factors associated with increased health care utilization and costs.

Principal Findings

Regression models revealed that the factors most predictive of higher health care costs were child abuse, child neglect, and instability in out‐of‐home placements related to foster care. Other social determinants of health, such as parent risk factors, were not associated with health care costs. Child maltreatment and placement instability impacted health care costs primarily through large increases in behavioral health utilization and costs.

Conclusions

Prevention and early interventions for children and families to decrease child maltreatment and increase foster care placement stability could reduce overall health care costs.

1. INTRODUCTION

Social determinants, or “the conditions in which people are born, grow, [and] live,” have significant impacts on health and health care costs.1 The family environment during childhood is a primary social determinant of health. Family adversity during childhood contributes to lifelong poor physical and mental health2, 3, 4, 5, 6 and elevated medical costs.7, 8, 9, 10 Identification of the childhood experiences common in high‐cost populations can help shape interventions and policies designed to reduce health spending. In 2015, U.S. national health expenditures equaled $3.2 trillion, almost 18% of the U.S. Gross Domestic Product11 and Medicaid spending alone comprised $545 billion, or 17% of total expenditures. The current study examines medical costs of over 180 000 Medicaid‐enrolled adolescents with the goal of identifying childhood adversities and social conditions most associated with increased health care utilization and costs. Using administrative data to link records for children with those of their biological parents confers a unique opportunity to assess social determinants and to evaluate impacts on health care costs net of other factors.

Historically, children age 0‐18 account for approximately 25% of total Medicaid spending11 and previous research links childhood adversity to increased Medicaid expenditures. A 2013 study demonstrated that maltreated Medicaid‐enrolled children cost $2600 more per year than children who were not maltreated and estimated that excess cost due to maltreatment accounted for 9% of all child Medicaid expenditures.7 Adverse Childhood Experiences (ACEs) studies investigate a broader set of adversities including maltreatment as well as family environment factors such as parent substance abuse, incarceration, and domestic violence. ACEs studies emphasize that individuals who experience a single ACE are likely to experience another ACE12 and that the likelihood of negative health outcomes increases as the number of ACEs increases.4, 5, 10, 13 Child adversity, or ACEs, are important social determinants of health for young people, because they shape the conditions in which they are born, grow up, and live. Earlier identification of and intervention with children experiencing adversity could reduce health care costs. Since a small subset of the population accounts for a large percentage of total medical costs,14, 15 targeting a small number children, those experiencing childhood adversity, could have large economic impacts.

The current study used data from the Washington State Department of Social and Health Services Integrated Client Databases16 to assess the association between health care costs and prior childhood adversity and to identify adversities that were the most impactful on health care costs. The utility of administrative data has been demonstrated and reviewed.4, 17, 18, 19 Using administrative data, child risk factors can be determined without reliance on retrospective self‐report or caregiver report, and this method has been shown to replicate findings from studies that used self‐report or national surveys.4 Administrative datasets are large, providing high statistical power, and are longitudinal, with historical data from multiple decades. The ability to link service records for children to those of their parents and to integrate data from a variety of state systems delivers a comprehensive profile of childhood adversity. This information, combined with Medicaid claims, service encounters, and expenditures, provides a unique opportunity to identify childhood adversities associated with high‐cost adolescents.

2. METHODS

2.1. Study cohort

The study cohort consisted of children insured by Washington State Medicaid for at least 1 month between 7/1/2014 and 6/30/2015 who were between the ages of 12 and 17 years old during their first month of coverage and who did not die during that period. Children eligible for inclusion had at least one biological parent identified through state administrative records. Of 233 054 individuals in the cohort, 181 176 (77.7%) were linked to at least one parent and were therefore included in the study cohort. Of the cohort, 30 483 (16.8%) children were linked to one biological parent, and 150 693 (83.2%) were linked to both biological parents.

Child and parent administrative data were linked to identify both child and parent risk factors that could impact health care costs. Parent risk factors were identified based on biological parents regardless of custody status. Data were compiled using the Washington State Department of Social and Health Services Integrated Client Databases,16 which links data from various state systems on individual characteristics, services, outcomes, and costs. Data sources used in this study included the Department of Social and Health Services (child welfare, economic services, developmental disabilities, mental health, and substance abuse), Health Care Authority (Medicaid), State Patrol (arrests), Administrative Office of the Courts (filings, adjudications, or convictions), Department of Corrections (prison incarcerations), and the Department of Health (birth and death records). The study was granted an Institutional Review Board exempt status because it relied on deidentified retrospective records.

2.2. Measures

2.2.1. Independent variables

Independent variables were categorized as maltreatment history, parent risk factors, demographics, medical coverage, or child social and health risk factors (full descriptions in Table S1). Maltreatment history factors included sexual abuse, physical abuse, neglect, and out‐of‐home placements. Sexual abuse, physical abuse, and neglect were defined as any allegation with the indicated type of maltreatment that generated an intake, regardless of whether that intake led to an investigation or substantiation. During the study period, Washington State was rolling out a differential response system, where low‐risk families referred to state Child Protective Services could be connected to services with no investigation or findings. Therefore, limiting to cases where abuse/neglect was investigated or where abuse or neglect was substantiated was not advisable. Prior research has found that unsubstantiated cases often have similar long‐term health impacts as substantiated cases.20 Recent abuse or neglect (in 2014) was distinguished from historical physical abuse, sexual abuse, and neglect (lifetime, excluding 2014) to discern the impact of temporality of maltreatment. Further, age at first maltreatment allegation and first out‐of‐home placement were included to measure possible age differences in impacts of maltreatment and out‐of‐home placement on costs. The out‐of‐home placement variable included foster care, kinship care, and congregate care placements. Extreme placement instability, defined as five or more lifetime placements, was also included as an adverse experience.

Parent risk factors included criminal behavior, domestic violence, disability, medical risk,21 homelessness, poverty (food assistance (SNAP)—moderately low income; cash assistance (TANF)—very low income), mental health condition, substance use disorder, and death. Parent risk factors were assessed for the 5 years leading up to the study year (2010‐2014), except where noted otherwise. The parent risk factors correspond to the original ACEs household dysfunction factors, with the addition of parental disability and medical risk. The maltreatment history variables and parent risk factors together measure the social conditions under which children grew up, or family‐level social determinants of health. These variables and their relationship to costs were the central focus of this analysis.

Child social and health risk factors included being a parent, medical risk,21 medical complexity,22 juvenile justice involvement, mental health need, and substance use. Child social and health risk factors were included to control for costly adolescent experiences in order to isolate impacts of child maltreatment and parent risk factors net of these variables. Demographic variables included age, sex, race, and limited English proficiency of the parent. Medical coverage controls included months of Medicaid coverage and months of third‐party coverage. Medicaid coverage months were important to include to control for duration of access to Medicaid‐paid services in the study period. Third‐party coverage typically reflects enrollment in employer‐based health insurance by families with incomes low enough to qualify for Medicaid. Under Medicaid rules, third‐party coverage is the first payer for covered services and therefore is associated with lower Medicaid‐paid health service costs.

2.2.2. Dependent variable

Health care costs were defined as federal and state funds used to provide health services in the study year (2015) and were divided into medical system and behavioral health. In Washington State during this period, behavioral health services were provided through behavioral health organizations contracting with service providers while medical services were administered by a separate medical system. Therefore, this analysis separately attended to Medicaid costs in the medical system and those that accrued to the currently separate behavioral health service system.

In Washington State, medical assistance includes a mental health benefit for children that covers services received in a medical setting, including primary care. Because these services were delivered in a medical setting, they were included in the measure of medical system costs. More functionally impaired children were served through the behavioral health service system and these services were included in behavioral health costs. Medical costs also included prescriptions such as psychotropic medications. Due to psychotropic medication and some primary care mental health services falling under medical system costs, associations between risk factors and behavioral health costs, broadly defined, may be an underestimate. There are substantive reasons to examine medical and behavioral health costs separately, including prior research on child adversity's impact on both adolescent physical and mental health.4, 23

Medical system costs were calculated from the provider amount paid for fee‐for‐service claims, the managed care amounts paid to reported provider for managed care encounters, and other fees derived from the Medicaid reporting system in Washington State. Medical costs included the following types of costs: outpatient hospital emergency department (ED), outpatient hospital non‐ED, inpatient, prescription drug, ambulatory care, and other costs (eg, transportation, medical equipment). Behavioral health costs were calculated based on payments made to behavioral health service providers and include outpatient mental health services, substance use disorder treatment, behavioral rehabilitation services, and inpatient mental health treatment.

2.3. Analysis

The study population was described including summary statistics on the independent and dependent variables. Bivariate comparisons identified the prevalence of select risk factors in the top decile relative to the bottom nine deciles of the cohort based on total medical system or behavioral health costs. Regression modeling was conducted to isolate the relationship between medical system and behavioral health costs and our measures of interest.

Because both behavioral health costs and medical system costs were zero for a nontrivial proportion of the population of interest, a two‐step model was used to predict costs. In the first step, two separate logistic regression models predicting behavioral and medical system costs, respectively, were fit to assess which factors contributed to the likelihood of receiving care. In the second step, two ordinary least squares (OLS) models predicting behavioral and medical system costs for the subset of youth with nonzero costs were run using the same predictor variables as in the first step. The second step was also estimated using generalized linear models (GLM) with the gamma distribution and log link function. The substantive conclusions were not different in the GLM models, so the results from the OLS regression are presented here because the resultant coefficients are more easily interpretable as changes in dollars associated with each factor. We note that OLS regression models are most commonly used in health care risk‐adjustment processes because they predict mean expenditures more accurately for all risk and demographic subgroups, which renders results more readily interpretable in health care policy contexts.24 All costs are listed as 2015 dollars, rounded to the nearest tens or hundreds of dollars in the text and signify the predicted increase or decrease in costs per individual for a given risk factor net of other factors in the model.

We combined the results from the first and second step in order to estimate the overall impact on behavioral and medical system costs of each predictive factor of interest. In order to calculate the overall impact, the predicted probability of having any expenditures from the first step was multiplied by the predicted value of expenditures from the second step for a subject with and without the focal risk factor, and then the average difference for all subjects was calculated to show the overall impact. A confidence interval of the overall impact was estimated by bootstrapping with 1000 replicates. All statistical analyses were run in SAS Statistical Software version 9.4 (SAS Institute Inc., Cary, NC, USA).

3. RESULTS

Descriptive statistics for the cohort are presented in Table 1. The 181 176 adolescents were 49.2% male and 50.8% female. Participant age ranged from 12 to 17, with an average age of 14.4. On average, participants were Medicaid insured for 11.0 months of the year‐long outcome period with an average of 1.2 months of third‐party liability coverage. Racial composition was 42.8% non‐Hispanic white, 29.2% of Hispanic origin, 25.6% non‐Hispanic minority, and 2.3% unknown. More than one‐third (36.3%) of the cohort had a mental health condition, 40.9% were reported to state CPS for neglect or abuse during their lifetime, and 78.8% experienced at least one family risk factor.

Table 1.

Demographic characteristics and prevalence of risk factors within the population

Variable	No.	%	Variable	No.	%
Total N	181 176	100%	Race or ethnicity
Gender			Hispanic	52 820	29.2%
Female	89 165	49.2%	Non‐Hispanic White	77 587	42.8%
Male	92 011	50.8%	Non‐Hispanic Black	9279	5.1%
Age			Non‐Hispanic Indian	3119	1.7%
Average (SD)	14.4 (1.7)		Non‐Hispanic Asian	4834	2.7%
Average months of coverage in SFY2015			Non‐Hispanic Pacific Islander	2031	1.1%
Medicaid (SD)	11.0 (2.5)		Multiple races (non‐Hispanic)	27 239	15.0%
Third‐party liability (SD)	1.2 (3.4)		Unknown race	4258	2.3%
Biological parents identified
One	30 483	16.8%	Limited English (parent)	30 101	16.6%
Two	150 693	83.2%
Maltreatment
Maltreatment type			No. out‐of‐home placements, continued
Any abuse or neglect	74 072	40.9%	3‐5 placements	4458	28.1%
Physical abuse	31 343	17.3%	6‐8 placements	1070	6.7%
Sexual abuse	12 536	6.9%	9 +  placements	534	3.4%
Neglect	66 896	36.9%	Age at first placement (for ≥1 placement)
Any out‐of‐home placement	15 862	8.8%	0‐3	7876	49.7%
No. out‐of‐home placements (for ≥1 placement)			4‐11	6270	39.5%
1‐2 placements	9800	61.8%	12+	1716	10.8%
Child social and health risk factors
Mental health condition	65 702	36.3%	PMCA
Substance use disorder	10 388	5.7%	Nonchronic	126 604	69.9%
Arrest or conviction	6387	3.5%	Chronic noncomplex	38 167	21.1%
Has a biological child	1872	1.0%	Chronic complex	16 405	9.1%
Parent risk factors
Domestic violence	30 224	16.7%	Mental health condition	71 761	39.6%
Low income			Substance use disorder	44 142	24.4%
Cash or food assistance (proxy for low income)	128 842	71.1%	Arrest or conviction	57 737	31.9%
Food assistance only (proxy for moderate low income)	80 990	44.7%	Homeless	41 417	22.9%
Cash assistance (proxy for extremely low income)	47 852	26.4%	Number of parent risk factors (ACEs)
Medical			0	38 472	21.2%
High medical risk score (CDPS)	42 049	23.2%	1‐3	88 610	48.9%
Disability	24 693	13.6%	4‐6	43 442	24.0%
Died	5647	3.1%	7+	10 652	5.9%

ACEs, Adverse Childhood Experiences; No., number; PMCA, Pediatric Medical Complexity Algorithm.

On average, each cohort member's yearly medical costs were $1688, while median costs were $605 indicating a skewed distribution of medical costs with a small number of very costly children. A majority of youth had low or no medical system costs: 13.4% of individuals had no medical costs for the year. An average of $632 was spent on behavioral health costs per cohort member, while the median was $0 due to nearly 87% of the cohort receiving no behavioral health services in the year.

To characterize risk factors associated with high costs, the percentage of individuals who experienced parent and child risk factors in the top decile and bottom nine deciles for medical costs and for behavioral health costs were compared (Table 2). The adolescents with the highest health care costs were 1.4‐2.0 times more likely to have experienced at least five parent risk factors than the lower deciles, and 1.4‐1.9 times more likely to have experienced abuse or neglect or 5 +  out‐of‐home placements. In particular, high‐cost adolescents were more likely to have experienced recent maltreatment: Those in the top decile for medical costs were 2.0 times more likely to have experienced recent maltreatment and those in the top decile for behavioral health costs were 3.9 times more likely to have experienced recent maltreatment.

Table 2.

Prevalence of child adversities in the top decile relative to the rest of the population for medical and behavioral health costs

	Medical cost			Behavioral health cost
	Bottom 90 percent	Top 10 percent	Relative risk	Bottom 90 percent	Top 10 percent	Relative risk
Total N	163 059	18 117	‐	162 416	18 760	‐
Total cost	$121 746 543	$184 091 206	‐	$1 271 699	$113 183 929	‐
Cost per adolescent	$747	$10 161	‐	$8	$6033	‐
Risk factors
5+ parental ACEs risk factors	19.1%	29.0%	1.5	18.3%	35.4%	1.9
Any abuse or neglect in lifetime	39.3%	55.2%	1.4	37.8%	68.0%	1.8
Any abuse or neglect in prior year	7.5%	15.0%	2.0	6.3%	24.8%	3.9
5 +  placements (among children ever placed out of home)	13.8%	21.4%	1.5	12.6%	23.5%	1.9

Given that a variety of child and parent risk factors were more frequently experienced by the adolescents with the highest health care costs, two‐step regression modeling was used to isolate risk factors associated with increased health care usage and costs while controlling for other variables. First‐stage logistic regression modeling (Table 3) showed that youth with histories of abuse, neglect, or out‐of‐home placement were more likely to have costs in the behavioral health system relative to similar children without maltreatment or foster care history. Recent (prior year) abuse or neglect more than doubled a young person's odds of behavioral health system costs (OR = 2.39), while any lifetime physical abuse and neglect were associated with increased odds of behavioral health costs of 12% and 18%, respectively. Recent out‐of‐home placements (at age 12 or older) were associated with increased likelihood of behavioral health costs by about 34%. Youth who experienced extreme placement instability (5+ placements) were also more likely to experience behavioral health system costs (OR = 1.44).

Table 3.

First step models predicting any costs (relative to no costs) to the behavioral health system and medical system using logistic regression

	Behavioral health system		Medical system
	Coefficient	Odds‐ratio	Coefficient	Odds‐ratio
Intercept	−3.49		−2.15
Maltreatment history
Any lifetime physical abuse (excluding prior year)	0.12	1.12***	−0.01	0.99
Any lifetime sexual abuse (excluding prior year)	0.05	1.06	0.09	1.09*
Any lifetime neglect (excluding prior year)	0.17	1.18***	−0.14	0.87**
Any abuse or neglect in prior year	0.87	2.39***	−0.10	0.91*
First out‐of‐home placement at age 0‐3	0.01	1.01	−0.33	0.72***
First out‐of‐home placement at age 4‐11	−0.04	0.96	−0.28	0.76***
First out‐of‐home placement at age 12 or older	0.29	1.34***	−0.12	0.89
5 or more out‐of‐home placements	0.36	1.44***	−0.08	0.92
First abuse or neglect at age 0‐3	0.11	1.12*	0.09	1.09
First abuse or neglect at age 4‐11	0.15	1.17***	0.03	1.03
First abuse or neglect at age 12‐17	0.21	1.24***	−0.04	0.96
Parent risk factors
Parent has mental health condition	0.11	1.12***	0.08	1.09***
Parent has a substance use disorder	0.09	1.09***	0.00	1.00
Parent was homeless	0.07	1.07***	−0.04	0.97
Parent was involved in criminal justice system	0.01	1.01	0.03	1.03
Parent was disabled	0.11	1.11***	0.02	1.02
Parent was involved in domestic violence	0.02	1.02	−0.02	0.99
Parent used food assistance, not cash	−0.09	0.92***	−0.09	0.91***
Parent used cash assistance	−0.14	0.87***	−0.23	0.80***
Parent has disabling condition	−0.03	0.97	0.12	1.12***
Parent died	0.15	1.16***	−0.07	0.93
Child and family demographics
Age in years	−0.05	0.95***	0.02	1.02***
Female	0.26	1.30***	0.41	1.50***
Hispanic	0.13	1.14***	0.27	1.32***
Non‐Hispanic Black	0.06	1.06	−0.11	0.90**
Non‐Hispanic Indian	−0.37	0.69***	0.33	1.40***
Non‐Hispanic Asian	−0.14	0.87***	−0.05	0.95
Non‐Hispanic Pacific Islander	−0.64	0.53***	−0.23	0.80***
Multiple races (non‐Hispanic)	0.35	1.42***	0.05	1.05*
Reference is non‐Hispanic White
Parent was limited English proficient	−0.18	0.83***	0.34	1.41***
Medical coverage
Count of Medicaid eligibility months	0.07	1.07***	0.30	1.35***
Count of third‐party liability coverage months	0.00	1.00	−0.08	0.93***
Child social and health risk factors
Chronic noncomplex health conditions	0.59	1.81***	0.61	1.84***
Chronic complex health conditions	0.65	1.92***	0.90	2.45***
Reference is nonchronic noncomplex conditions
Child has mental health condition	1.23	3.42***	0.42	1.52***
Child has substance use disorder	0.92	2.50***	−0.19	0.83***
Child is involved in criminal justice system	0.73	2.08***	−0.27	0.77***
Child is a parent	−0.68	0.50***	0.48	1.62***
Child medical risk score (CDPS)	−0.01	0.99	1.51	4.54***

*** P < 0.001; ** P < 0.01; * P < 0.05.

History of neglect, abuse, or out‐of‐home placement generally had no association with the likelihood of incurring medical costs or was associated with lower likelihood. Any lifetime neglect was associated with lower likelihood of medical costs (OR = 0.87), as were out‐of‐home placements starting before the age of 12 (age 0‐3: OR = 0.72; age 4‐11: OR = 0.76), and abuse or neglect in the prior year (OR = 0.91). However, youth with a history of sexual abuse had a slightly higher likelihood of incurring medical system costs (OR = 1.09) relative to similar children without these risk factors. No cost differences were found by age at first abuse or neglect allegation.

Most parent risk factors showed no discernable pattern of associations with the likelihood of incurring behavioral health and medical system costs, with the exception of parent mental health need, which was associated with increased likelihood of costs in both systems by 9%‐12%, and food assistance and cash assistance use (proxy for low and very low income), which was associated with decreased likelihood of costs in both systems by 8%‐20%.

Second‐step regression modeling (Table 4) for children with nonzero systems costs showed that child maltreatment was associated with greater costs, especially to the behavioral health system. Abuse and neglect within the year prior to this study and historical physical and sexual abuse were associated with increased behavioral health system costs. Recent abuse was associated with annual per person behavioral health cost increases of $3600, lifetime physical abuse was associated with cost increases of $840, and lifetime sexual abuse was associated with cost increases of $1600 net of other factors. A history of neglect alone did not significantly predict behavioral health costs, net of other factors.

Table 4.

Second stage models predicting costs (in dollars) to the behavioral health system and medical system using OLS regression

	Behavioral health system	Medical system
	Coef. (SE)	Coef. (SE)
Intercept	2804 (1030)	−2311 (282)
Maltreatment history
Any lifetime physical abuse (excluding prior year)	844 (218)***	26 (85)
Any lifetime sexual abuse (excluding prior year)	1617 (267)***	336 (109)**
Any lifetime neglect (excluding prior year)	274 (345)	264 (131)*
Any abuse or neglect in prior year	3633 (357)***	466 (147)**
First out‐of‐home placement at age 0‐3	1146 (352)**	444 (140)**
First out‐of‐home placement at age 4‐11	1437 (355)***	1 (146)
First out‐of‐home placement at age 12 or older	3674 (489)***	−4 (260)
5 or more out‐of‐home placements	9826 (460)***	737 (232)**
First abuse or neglect at age 0‐3	−313 (440)	−232 (160)
First abuse or neglect at age 4‐11	−404 (406)	−139 (144)
First abuse or neglect at age 12‐17	−829 (432)	−280 (163)
Parent risk factors
Parent has mental health condition	29 (217)	10 (66)
Parent has a substance use disorder	−221 (219)	−141 (75)
Parent was homeless	619 (212)**	−54 (72)
Parent was involved in criminal justice system	−648 (203)**	−1 (67)
Parent was disabled	368 (228)	52 (83)
Parent was involved in domestic violence	−218 (214)	−184 (78)*
Parent used food assistance, not cash	−1292 (248)***	−175 (65)**
Parent used cash assistance	−1945 (297)***	−209 (86)*
Parent has disabling condition	−621 (212)**	141 (70)*
Parent died	987 (404)*	−75 (143)
Child and family demographics
Age in years	−243 (56)***	145 (15)***
Female	−182 (169)	−16 (50)
Hispanic	−650 (235)**	−37 (66)
Non‐Hispanic Black	415 (410)	−227 (119)
Non‐Hispanic Indian	984 (683)	723 (193)***
Non‐Hispanic Asian	−645 (807)	−158 (159)
Non‐Hispanic Pacific Islander	604 (1520)	−27 (252)
Multiple races (non‐Hispanic)	−105 (210)	−11 (75)
Reference is non‐Hispanic White
Parent was limited English proficient	−949 (329)**	−86 (78)
Medical coverage
Count of Medicaid eligibility months	295 (47)***	98 (14)***
Count of third‐party liability coverage months	89 (25)***	−91 (8)***
Child social and health risk factors
Chronic noncomplex health conditions	1024 (216)***	468 (68)***
Chronic complex health conditions	3115 (264)***	3053 (91)***
Reference is nonchronic noncomplex conditions
Child has mental health condition	636 (231)**	112 (62)
Child has substance use disorder	388 (239)	−210 (116)
Child is involved in criminal justice system	3438 (287)***	−136 (148)
Child is a parent	−3298 (703)***	3446 (239)***
Child medical risk score (CDPS)	479 (99)***	2869 (24)***

*** P < 0.001; ** P < 0.01; * P < 0.05.

For medical system costs, recent abuse and neglect were associated with a cost increase of $470, history of sexual abuse was associated with a cost increase of $340, and history of neglect was associated with a cost increase of $260. A history of physical abuse was not associated with increased medical costs, net of other controls.

History of out‐of‐home placement at any age was associated with higher behavioral health system costs, though the biggest monetary impact was found for extreme placement instability ($9800). Recent placement at age 12 or later was associated with a $3700 increase in behavioral health costs; in contrast, first placement at age 0‐3 increased behavioral health costs by $1100 and first placement at age 4‐11 increased behavioral health costs by $1400. For medical system costs, a first placement in early life (before age three) was associated with modest increases in costs ($440), as was placement instability ($740). No differences in behavioral health or medical costs were found by age at first abuse or neglect allegation.

The associations between parent risk factors and costs showed few consistent patterns. However, low‐income status was associated with lower costs, net of other factors in the model. While the bivariate comparisons indicated that costlier children had experienced more parent risk factors, when regression modeling controlled for maltreatment experiences and other child characteristics, largely mixed or nonsignificant associations were found for parent risk factors.

Overall estimates of the impacts of maltreatment and parent risk measures on behavioral health costs and medical costs were calculated based on the results of the two‐step models (Table 5). Placement instability had the largest overall association with increased behavioral health costs ($1700) and medical costs ($620). Prior year abuse and neglect had the next largest estimated association with increased behavioral health costs ($1100) and medical costs ($400). Experiencing a first out‐of‐home placement at age 12 or older was also associated with a large increase in behavioral health costs ($700), but not with medical costs. First out‐of‐home placement experiences before the age of 12 were associated with smaller increases in behavioral health costs ($150‐$170), while in the case of medical costs, increases of about $340 were only found for children who had a first placement prior to age 4. Maltreatment experiences also had associations with overall costs. Physical and sexual abuse allegation history was associated with increased behavioral health costs of $160 and $240, respectively, with neglect showing no association. In the case of medical costs, sexual abuse and neglect were associated with overall cost increases of $300 and $210, respectively, while historical physical abuse showed no association.

Table 5.

Estimated overall associations of selected risk factors with behavioral health system and medical system costs

	Behavioral health system		Medical system
	Estimate	CI	Estimate	CI
Maltreatment history
Any lifetime physical abuse (excluding prior year)	159	[89, 232]	22	[−114, 170]
Any lifetime sexual abuse (excluding prior year)	238	[135, 335]	302	[87, 567]
Any lifetime neglect (excluding prior year)	99	[−1, 205]	213	[37, 422]
Any abuse or neglect in prior year	1104	[934, 1293]	390	[161, 633]
First out‐of‐home placement at age 0‐3	154	[26, 308]	335	[109, 575]
First out‐of‐home placement at age 4‐11	168	[48, 291]	−29	[−220, 169]
First out‐of‐home placement at age 12 or older	698	[410, 1001]	−15	[−462, 507]
5 or more out‐of‐home placements	1777	[1452, 2118]	624	[177, 1156]
First abuse or neglect at age 0‐3	−2	[−137, 120]	−193	[−467, 45]
First abuse or neglect at age 4‐11	1	[−115, 112]	−117	[−312, 60]
First abuse or neglect at age 12‐17	−42	[−180, 84]	−246	[−550, 15]
Parent risk factors
Parent has mental health condition	46	[−8, 101]	17	[−80, 115]
Parent has a substance use disorder	3	[−60, 65]	−122	[−239, −3]
Parent was homeless	109	[51, 168]	−50	[−156, 39]
Parent was involved in criminal justice system	−82	[−138, −29]	3	[−101, 117]
Parent was disabled	91	[29, 154]	47	[−92, 182]
Parent was involved in domestic violence	−20	[−75, 44]	−160	[−289, −37]
Parent used food assistance, not cash	−201	[−274, −129]	−161	[−280, −37]
Parent used cash assistance	−299	[−376, −221]	−202	[−280, −47]
Parent has disabling condition	−92	[−144, −41]	134	[15, 255]
Parent died	197	[49, 357]	−71	[−223, 118]

Parent risk factors did not exhibit the same strong association with increased behavioral health and medical costs. Some risk factors, like homelessness and parent death, were associated with modest increases in behavioral health costs, but had estimates near zero for medical costs. Other factors were associated with moderate decreases in costs or were not associated with costs at all. The only parental factor that showed a strong association with costs was poverty. Adolescents whose parents had used food or cash assistance had lower behavioral health and medical costs. Of all the adverse experiences examined in our models, maltreatment, out‐of‐home placement, and placement instability were the factors consistently associated with increased costs to the behavioral health and medical systems.

4. DISCUSSION

By using integrated administrative records, the association between parent and child risk factors and annual health care costs was assessed for a population of over 180 000 Medicaid‐insured adolescents without reliance on retrospective self‐report. Among the social determinants examined here, child maltreatment, out‐of‐home placement, and placement instability had the largest cost impact to the Medicaid system (Tables 3, 4 and 5). These experiences were associated with higher use of behavioral health services, even when controlling for pre‐existing behavioral health conditions. In contrast, parent risk factors, even though they influence the social conditions under which children grow up, had more modest, mixed effects on costs.

This study found that the adolescents with the highest health care costs tended to experience more parental ACEs compared to adolescents with lower expenses (Table 2), which mirrors prior research demonstrating that more adverse experiences lead to a greater likelihood of negative health outcomes.4, 5, 10, 13 However, two‐stage regression modeling (Tables 3 and 4) demonstrated that the experiences most impactful on costs were those related to child maltreatment (abuse, neglect, out‐of‐home placement, placement instability) as opposed to parent risk factors (parent domestic violence, criminal justice involvement, mental health need, and disability). The one pattern in parent risk factors across all models was lower utilization and costs among adolescents whose parents relied on cash or food assistance (proxies for low income). This finding could indicate less need for services because of connections to cash and food benefits or could indicate barriers to health care access for low‐income individuals. Although adolescents with high health care costs experienced a variety of childhood adversities, the factors most associated with increased health care spending were related to maltreatment. Given that approximately 40% of the studied youth had been reported for some form of abuse or neglect during their lifetime, programs targeting maltreatment, both through prevention and intervention, could have a large impact on costs.

A sizeable body of research evaluates short‐term7, 25, 26 and long‐term6, 9, 27 health care costs associated with childhood maltreatment or estimated total costs of child maltreatment.28, 29, 30, 31 Child maltreatment increases costs in multiple domains (health care, criminal justice, education, lifetime lost productivity, and social services), all of which contribute to its large economic burden.29, 30, 31, 32 In Washington State, the 2012 cost of child welfare services alone was over $500 million.28 The estimated lifetime costs of maltreatment for all new abuse and neglect cases in 2008 were $124 billion,31 which is comparable to the economic burden associated with health conditions such as diabetes. 33

Prior research demonstrates that childhood maltreatment negatively affects childhood mental health,4, 34 with associated costs estimated at over $1 billion annually.30 In this study, physical or sexual abuse increased annual behavioral health costs by approximately $150‐$250, and sexual abuse or neglect increased annual medical system costs by $200‐$300. Neglect or abuse within the prior year increased behavioral health costs by $1100 and medical costs by about $400. In this study, recent maltreatment was associated with larger cost increases than lifetime measures, which concurs with prior research demonstrating that recent adversity has a larger impact on adolescent health than earlier life adversity.10

Prior research suggests that as the number of recent foster care placements increases for children, the likelihood of high mental health service usage increases.35 The current study found that high placement instability had the largest association with increased costs of all social determinants examined and that it was associated with cost increases to both the behavioral health and medical system. Five or more out‐of‐home placements were associated with a $1700 increase in behavioral health costs and $620 increase in medical costs. Age at first placement had differing impacts on behavioral health and mental health system costs. A first out‐of‐home placement after age 11 was associated with a $700 increase in behavioral health costs, while the associations were smaller (around $150) for first placements prior to age 12. In contrast, first out‐of‐home placement before the age of 4 was associated with around $340 in medical costs, whereas later first out‐of‐home placement experiences were not associated with medical costs.

These findings suggest that preadolescent interventions targeting child maltreatment could have a large impact on health care costs associated with maltreatment and multiple foster care placements. Meta‐analyses assessing the efficacy of early interventions on child outcomes find that overall, early interventions benefit children and the economy.36, 37, 38, 39, 40, 41 In particular, programs designed to decrease child maltreatment have been shown to not only be successful,39, 42 but to provide net positive benefit to cost ratios.40 Prevention of abuse and neglect and interventions in response to maltreatment require a cross‐systems approach that connects multiple key policy sectors, including health care.43 For example, primary health care visits offer a point of contact for universal prevention efforts and screening of children for maltreatment risk factors, as well as an opportunity to make referrals for evidence‐based interventions in response to abuse and neglect.44

This study had a number of limitations. Adolescents were included only if they could be linked with at least one biological parent through administrative data, resulting in the exclusion of 22% of the potential cohort. Parent risk factors were based solely on biological parents, which did not account for risk factors potentially associated with other caregivers. Further, the frequency and/or severity of risk factors was not coded—for example, parent domestic violence signified only that one or more occurrence was documented and also would not register undocumented events. The current study only examined health care costs associated with child risk factors in adolescents. Future studies should expand to younger age groups and expand beyond health care costs (eg, criminal justice costs).

This study was also limited because it was observational in nature and established associations but not causality. There are few alternatives to observational studies when examining social determinants of health, given that maltreatment and parental risk factors cannot be randomly assigned to youth. The analyses included controls for a number of factors that drive costs, such as medical complexity and behavioral health conditions, but there remains a possibility that unmeasured factors affecting both maltreatment and costs could account at least partially for the identified associations.

This study used administrative data linking Medicaid‐insured adolescents to their biological parents to assess the impact of childhood social conditions (social determinants) on health care costs. A small subset of the studied population accounted for the majority of costs. Although these adolescents with high health care expenses tended to experience multiple risk factors, multivariate analyses indicated that the childhood adversities that most increased costs were abuse, neglect, out‐of‐home placement, and out‐of‐home placement instability. These results suggest that prevention and intervention efforts focused on children and families who are at risk for child welfare system involvement and for those children who have already experienced maltreatment could substantially reduce the health care costs in Medicaid‐insured adolescent populations.

Supporting information

 

 

Patton DA, Liu Q, Adelson JD, Lucenko BA. Assessing the social determinants of health care costs for Medicaid‐enrolled adolescents in Washington State using administrative data. Health Serv Res. 2019;54:52–63. 10.1111/1475-6773.13052