Mental Health Care Services for Children With Special Health Care Needs and Their Family Members: Prevalence and Correlates of Unmet Needs

MATERIALS AND METHODS

We use the NS-CSHCN, which was fielded from April 2000 to October 2002, to estimate the prevalence of unmet mental health care needs among CSHCN (1–17 years of age) and their families.^{27, 37, 38} This survey uses the State and Local Area Integrated Telephone Survey (SLAITS), a random-digit-dialing telephone survey that is conducted jointly by the National Center for Health Statistics and the Centers for Disease Control and Prevention and sponsored by the Maternal and Child Health Bureau (Health Resources and Services Administration) and the Office of the Assistant Secretary for Planning and Evaluation of the Department of Health and Human Services. The NS-CSHCN initially sampled and interviewed 372174 children (living in 196888 households) in the United States; data were collected in 3 phases. First, demographic information on children <18 years of age was collected. This was followed by a Foundation for Accountability screener to identify CSHCN.^{39, 40} A child with a condition that lasted or was expected to last at least 1 year and matched at least 1 of 5 criteria (using or needing more medical care, mental health services, or educational services than are provided for other children of the same age; using or needing prescription medication; having a limitation in the ability to do things most children of the same age do; using or needing special therapy; or using or needing EDB treatment or counseling) was classified as having a special health care need. The Foundation for Accountability screener identified 48690 CSHCN, 38866 of whom (1 per household) were later interviewed in depth. Using questions from the screener, we identified children as having an EDB condition lasting for >12 months (ie, has a special need arising from an EDB condition) if the following 2 questions were answered in the affirmative: “Does your child have any kind of emotional, developmental, or behavioral problem for which he needs treatment or counseling?” and “Has that child's emotional, developmental or behavioral problem lasted or is it expected to last 12 months or longer?” We define “unmet need” as reporting a need for mental health care services in the past 12 months and reporting not having received all needed services. CSHCN and/or their families who reported no need were considered to have no unmet need. Although this definition is sensitive, it is not specific: the definition of “unmet need” will also include those who are still receiving care and those on waiting lists. However, these situations are unlikely to influence our results. For both CSHCN and other family members the prevalence rates of reporting still receiving ongoing care and/or difficulty obtaining an appointment are <0.25%.

We operationalize enabling factors (income, insurance, and usual source of care) as follows: Income is operationalized as the family's poverty status, which was collapsed into 3 categories representing <100% of the federal poverty level (FPL), 100% to 185% FPL, and >185% FPL. Following the scheme used by Honberg et al,⁴¹ we classified children as having private insurance coverage (composed of private plans or military coverage), public coverage (Medicaid, State Children's Health Insurance Program, Title V coverage, or other public coverage), both private and public, or uninsured at the time of interview. We exclude children with “other” or unknown coverage (n = 146). A child's actual source of care was obtained for those children reporting having a usual source of care. The categories were doctor's office, clinic, and other (which included sources such as a school or hospital emergency department). Children who did not have a usual source of care were identified as such.

Need factors (condition severity, condition stability, and the presence of EDB conditions) are operationalized as follows: Severity was originally measured in the survey on a 0 to 10 Likert scale, which we collapsed into 4 categories following the classification system used by Mayer et al³¹ in their analysis of unmet needs for routine and specialty care. Stability was recorded in the survey as “needs change all the time; needs change only once in a while; needs are usually stable; none of the above.” We exclude “none of the above” (n = 89) from our analyses. We also include a variable capturing the presence of an EDB condition for the child in the needs factor. Predisposing factors (age, maternal education, and race/ethnicity) are operationalized as follows: Child's age was categorized to reflect 5 developmental stages: 1 to 4, 5 to 9, 10 to 12, and 13 to 18 years of age; we dropped children <1 year old from our analyses (n = 486). Maternal education was collapsed into 3 categories representing less than high school, high school (completed), or more than high school education. Race and ethnicity were considered as separate variables, with race being represented by 3 categories (white, black, and other) and ethnicity being coded as Hispanic (vs non-Hispanic). As mentioned above, we include family unmet need in the child regressions and child unmet need in the family regressions in an effort to tap the family health dimension of the Andersen framework.³⁶

To capture environmental and health systems factors, we include an indicator for residence in a metropolitan area, and we also merged per-capita rates of general pediatricians, pediatric specialists, and pediatric psychiatrists from the area resource file on the basis of state and metropolitan statistical area (MSA) status.⁴² The area resource file is a county-specific health resources information system containing data on health facilities, health professions, measures of resource scarcity, health status, economic activity, health training programs, and socioeconomic and environmental characteristics. Per the method used by Mayer et al,³¹ we created 3 variables to capture local availability of relevant health care resources (the number of pediatricians per 10000 children <18 years old, the number of pediatric specialists per 10000 children <18 years old, and the number of pediatric psychiatrists per 10000 children <18 years old) that are matched to the NS-CSHCN respondents on state and MSA/non-MSA status.

Variables in this data set were missing data at varying rates, ranging from 0% for insurance to 3.4% for maternal education and 9.4% for family income. Almost 16% of the cases were missing some data. To preserve as many cases as possible and minimize potential internal validity bias resulting from deleting cases with missing data (eg, those children missing income were more likely to have public versus private insurance and lower maternal education and less likely to rank in intermediate condition severity),^43–46 we imputed multiple values for all missing data by using the sequential regression approach developed by Raghunathan et al⁴⁷ as implemented by the IVEware macro for SAS statistical software.^{48, 49} Because the variables in the NS-CSHCN are categorical and there are logical restrictions for some variables (eg, the question about having a usual source of care must be first affirmatively endorsed to report the source of care), the sequential regression approach is more appropriate than the more common multivariate normal approach.^{46, 50} This approach is also appropriate for use with the NS-CSHCN data because the variables do not follow the multivariate normal distribution. Conventional analyses (computing descriptive statistics and estimating logistic-regression models) were conducted using the multiple imputed data sets, and standard procedures for combining descriptive statistics, regression coefficients, and SE estimates from the multiple imputed data were then used.^{43, 44} Descriptive statistics and logistic-regression models were estimated, adjusting for the complex sampling design of the NS-CSHCN by using both Stata 8.2⁵¹ and SAS 9⁴⁹ software, and unless noted otherwise, all estimates were weighted to be nationally representative.

Because of the strong correlation between the health and health care experiences of all family members, we need to be concerned about 2 sources of clustering or correlation that may affect our estimates: family composition and the impact of unobservable factors on both child and family unmet needs. To address the first concern we added variables capturing the presence and/or the number of children with and without SHCN in the family to the basic regression models. We also limited the analyses to those families with only 1 child (which is the target CSHCN) and to those families with only 1 CSHCN (but with potentially other children without SHCN). To address the second concern we estimated variants of the basic regression equations using seemingly unrelated probit equations.^{51, 52} This method accounts for the correlation between child unmet needs and other family unmet needs that result from common unobservable factors. Our main regression results are robust to these 2 types of sensitivity analyses.

After excluding children <1 year old, those with equivocal condition stability, and those with other or unknown insurance coverage, the final sample size is 36512 (93.9% of the original sample [N = 38866]).

RESULTS

Approximately 30% of CSHCN have some type of EDB condition, representing ∼2.6 million children in the United States. Table 1 displays the distribution of reported needs for mental health care services. Overall, 25.7% of CSHCN in the United States and 13.3% of their other family members had some need for mental health care (column 1). CSHCN with EDB conditions were more likely to have a need for mental health care than those CSHCN without EDB conditions (row 1, columns 2 and 3). Other family members of CSHCN with EDB conditions were also more likely to have some need for mental health care than other family members of CSHCN without EBD conditions (row 2, columns 2 and 3). Of those with reported needs for mental health care, 18.4% of the CSHCN and 21.3% of their other family members reported that their needs were not met (data not shown). Overall (when considering those with and without a reported need for mental health care services, as defined in “Methods”), 4.7% of CSHCN and 2.8% of other family members reported unmet mental health care needs, representing ∼416000 CSHCN and 249000 families.

Table 2 reports the population estimates of the characteristics of the CSHCN meeting our sample selection criteria along with estimates of the proportion of unmet needs for mental health care services by those demographic characteristics. Family income is inversely related to unmet needs for mental health care services for CSHCN. Children living in families with household incomes <100% FPL are >3 times more likely to have unmet needs than those CSHCN living in families with household incomes >185% FPL. Uninsured CSHCN are more likely to have reported that their needs were not met than insured children, and CSHCN lacking a usual source of care are also more likely to report unmet mental health care needs than those with an identified doctor's office. CSHCN whose usual source of care is a clinic report similar rates of unmet mental health care needs as those without a usual source of care. The severity of a child's condition is directly related to the probability of reporting unmet mental health care needs for CSHCN, and those CSHCN with unstable conditions are also more likely to have unmet needs than those with stable conditions. CSHCN with EDB conditions are more likely to report unmet need than for those CSHCN without EDB conditions. Older CSHCN experience higher rates of unmet needs than do younger children, and CSHCN of more educated mothers experience lower rates of unmet need than CSHCN of less-educated mothers. Black CSHCN experience the highest rates of unmet need of all racial/ethnic groups. There is a strong correlation between child and other family members' unmet needs: 3.4% of children had unmet needs if other family members did not have unmet needs, yet 51.0% of children had unmet needs if other family members did have unmet needs. Except for the racial categories, the patterns of unmet need for mental health care services among other family members are similar to the patterns for CSHCN.

Table 3 displays the results of the 2 logistic-regression models for CSHCN unmet mental health care needs and for other family member's unmet mental health care needs. After simultaneously adjusting for child health and demographic factors, we find that access to higher levels of resources (enabling factors) and higher levels of need factors are associated with less unmet need for mental health care services for CSHCN, which largely reflect the unadjusted findings presented in Table 2. Even after simultaneously adjusting for income, insurance, and usual source of care, each of these factors are independently and statistically significantly associated with access to mental health care services for CSHCN. Compared to the highest income group, to those with private insurance coverage, and to those with a doctor's office at their usual source of care, CSHCN living in the poorest homes, those who are uninsured, and those with no usual source of care are more likely to report unmet needs for mental health care services. More severely affected children are more likely to report unmet needs than less severely affected children, as are children with EDB conditions (condition stability is not independently associated with unmet needs in this model). Maternal education is not independently associated with unmet needs in this model. Relative to white CSHCN, black children are more likely to experience unmet needs, Hispanic children are less likely to experience unmet needs than non-Hispanic children, and CSHCN living in metropolitan areas are more likely to experience unmet needs than those in nonmetropolitan areas. CSHCN living in families in which other family members experience unmet needs are very likely to experience unmet needs for mental health care services.

Many of the child factors, to some degree, are shared family factors (eg, income, metropolitan status, insurance coverage, and race/ethnicity), but as stated above, some of the other child factors may not be theoretically applicable to other family members. We find, after simultaneously adjusting for all of the same factors, that some differences emerge. Enabling factors (family income, the target child's insurance coverage, and the child's usual source of care) are not associated with unmet needs for other family members. On the other hand, all of the target child's need factors are associated with unmet needs for other family members: family members of more severely affected children and family members of CSHCN with EDB conditions tend to have higher unmet needs for mental health care services, with the odds ratios (ORs) for the most severely affected children (OR: 4.5) almost twice that for CSHCN unmet needs (OR: 2.6). However, relative to usually stable conditions, family members of CSHCN with unstable conditions (changing all of the time) are less likely to experience unmet needs for mental health care services. Family members of CSHCN who experienced unmet needs are also much more likely to experience unmet needs for mental health care services.

DISCUSSION

This study presents the first set of estimates of the prevalence of reported need and unmet need for mental health care services for both CSHCN and their families using the nationally representative NS-CSHCN. Approximately 26% of CSHCN and ∼13% of their other family members reported a need for mental health care services. Overall, 4.7% of children and 2.8% of their other family members reported that their needs for mental health care services were not met, representing ∼416000 children and at least 249000 members of their families in the United States. Given the American Academy of Pediatrics statements on the use of medical homes for CSHCN,^28–30 it is disconcerting that such significant proportions of CSHCN and their families continue to report not receiving needed care. Our estimated rates of need and unmet need for the children in the NS-CSHCN are approximately equivalent to rates reported by others.^{12, 32–34}

In multiple logistic-regression models we found that the factors that are most strongly related to unmet need for CSHCN are poverty, lack of insurance coverage, lack of a usual source of medical care, condition severity, having an EDB condition, and family mental health care unmet needs. Older children and black children are also more likely to experience unmet needs, but Hispanic children are less likely to report experiencing unmet needs. Many of these factors tend to be associated with lower socioeconomic status, and using the same data, others have found that these factors are also related to unmet need for routine and specialty medical care,³¹ therapy services, eyeglasses or vision care, hearing aids and communication devices or hearing care, and mobility devices,⁵³ as well as being related to delayed or forgone care.⁵⁴ In addition to examining the prevalence and correlates of unmet mental health care needs for CSHCN, we also investigated unmet mental health care needs for other family members. As reviewed above, the literature reports that health and health care experiences of family members tend to be correlated.^{23, 25} In addition, the stresses of caring for a child with special needs (which may or may not include EDB conditions) may be related to mental health care needs for caregivers and siblings. In multiple logistic-regression models we found that if the target child's condition was more severe or if the child had an EDB condition or had unmet mental health care needs, then other family members were likely to have unmet needs.

The consistent evidence that child and other family members' unmet needs are highly intercorrelated suggests either reciprocal causal feedbacks (children's special health care needs status affects the mental health of other family members, which in turn can further impact children's health) or common related factors, including genetic factors, or both. However, as reported above, our results are robust to a number of different analytic approaches that accounted for such family clustering and correlations. These findings are also consistent with other work that has shown that stressful family environments, including parental illness, are negatively related to parents' self-reported efforts to obtain care for their children.⁵⁵ Regardless of the underlying causal mechanism, it is clear that families caring for CSHCN with more severe problems or with unmet mental health care needs themselves tend to have high rates of unmet mental health care needs.

Although these results are important, there are limitations to our analyses. They are based on self-reported cross-sectional data, which makes it difficult to clearly assess the causal relationships between the enabling, need, predisposing, and environmental factors and unmet mental health care needs. Assessing the impact of the effects of other family members is also not straightforward, because the decisions to enroll children in mental health care services are dynamically linked to parent's mental health and well-being.⁵⁶ In addition, parentally reported need (and therefore unmet need) may reflect parentally perceived needs versus medically evaluated needs, which, because of various mechanisms such as culturally and socioeconomically related values and perceived stigma, may bias our results in unknown directions. For example, in some studies, black parents report higher levels of negative expectations about the need for and the efficacy of mental health care for their children than white parents^{57, 58} (although minority children are often more likely to be identified as requiring mental health care services⁵⁹), thus possibly downwardly biasing either the numerator (use of services) or the denominator (need for services), or both.

Although we attempted to control for the environmental/health care system characteristics by adding measures of availability of pediatricians and pediatric specialists, these measures may not capture all of the complexity of this factor. In particular, aspects of insufficient supply of mental health care services to children may be related to health system policies, such as low reimbursement rates (especially for children covered by public insurance programs), that constrain physician and even family behavior.⁶⁰

Despite these limitations, using this rather large and nationally representative sample of CSHCN, we present the first set of estimates of these associations that show that both CSHCN and their families do experience mental health care needs and that some level of these needs go unmet. Given that even after controlling for need and predisposing factors, demographic and enabling factors such as poverty, insurance, source of care, and race are statistically significantly associated with unmet needs, indicating some degree of inequity in access to needed mental health care service for CSHCN and their families. These factors are strongly related to socioeconomic status, indicating that some families of lower socioeconomic status are disproportionately reporting unmet needs and that broader policies to identify and connect these families with needed services is warranted. Each of the enabling factors (income, insurance, usual source of care) in our analyses actually provide starting points for possible policy interventions to increase access to care such as antipoverty and insurance-expansion efforts.

Primary care physicians who provide care to CSHCN may have difficulty referring children and their families to mental health care services. In a sample of Boston-area primary care physicians caring for CSHCN, only 18% of physicians reported satisfaction with their ability to access outpatient mental health services for their patients, and only 12% were satisfied with their ability to access outpatient mental health care services for their patients' families.⁶¹ Furthermore, only ∼30% reported being satisfied with their ability to address their patients' psychological and emotional needs, and 77% reported that they needed more access to psychological consultants. Expanded access to and better implementation of medical homes may also help to address the problem of unmet mental health care needs for CSHCN and their families.²⁷

CONCLUSIONS

These results imply that still more can be done to improve access to needed mental health care services for CSHCN and their families. The correlations between child and other family members' unmet need further imply that broader policies to identify and connect these children and their families with needed services are warranted but that child- and family-centered approaches are not likely, alone, to be sufficient to meet the needs of CSHCN and their families. Other interventions such as antipoverty, insurance-expansion, and medical home–implementation efforts may be needed as well. Disadvantaged families and children may especially benefit from these changes.