Demographic, socioeconomic, and health correlates of unmet need for mental health treatment in the United States, 2002–16: evidence from the national surveys on drug use and health

Background Unmet need for mental health services remains high in the United States and is disproportionately concentrated in some groups. The scale and nature of these disparities have not been fully elucidated and bear further scrutiny. As such, in this study, we examine the demographic, socioeconomic, and health correlates of unmet need for mental health treatment as well as the reasons for unmet need. Methods We draw upon the National Survey for Drug Use and Health (NSDUH) from 2002 to 16 for adults aged 18 and over in the United States (n = 579,017). Using multivariable logistic regression, we simultaneously model the demographic, socioeconomic, and health correlates of unmet need for mental health treatment from 2002 to 16. We also analyse the reasons for unmet need expressed by these populations, reasons which include cost, perceived stigma, minimisation of symptoms, low perceived effectiveness of treatment, and structural barriers. Results Major characteristics associated with increased odds of unmet need include past year substance abuse or dependence (other than hallucinogens and sedatives), fair, poor, or very poor health, being female, and an educational attainment of college or higher. With respect to reasons for unmet need, cost was most often cited, followed by perceived stigma, structural barriers, and minimisation. Characteristics associated with increased odds of indicating cost as a reason for unmet need include: being uninsured or aged 26–35. Minimisation and low perceived effectiveness are mentioned by high-income persons as reasons for unmet need. College-educated persons and women had higher odds of citing structural barriers as a reason for unmet need. Conclusions The correlates and causes of unmet need highlight the intersectionality of individual health needs with implications on addressing inequities in mental health policy and practice.


Introduction
Mental disorders and substance use disorders are major contributors to years lived with disability in the United States (3,536,895.4 years lived with disability [YLDs] in 2016, or 1095.45 YLDs per 100,000 population) [1]. They are common, with data from the National Survey on Drug Use and Health (2010-2012) finding that 18.4% of adults had a mental illness and 8.6% reported substance abuse/dependence, while 2.2% had both [2]. Federal policymakers have responded with a variety of legislative approaches, notably the Mental Health Parity and Addiction Equity Act of 2008 (MHPAEA), while the Patient Protection and Affordable Care Act of 2010 (ACA) includes specific provisions for mental health [3]. Yet, despite these policies, there are large and persisting disparities in access to and receipt of mental health services. Individual studies have identified a range of underserved populations, which include certain ethnic minorities [4] and those lacking insurance [5]. Moreover, difficulties in accessing adequate mental health treatment have been documented among individuals with substance use disorders [6]. These previous studies signal an urgent need to address disparities in mental health treatment.
However, describing those whose needs for care are unmet is only a first step. It is also necessary to understand the reasons, if an appropriate policy response is to be developed. Possible explanations may lie within the affected individual or within the health care system.
In this study, we seek to characterise the demographic, socioeconomic, and health correlates of unmet need for mental health care among Americans, using data from the period 2002 to 2016 and to understand the reasons their needs are not being met. To our knowledge, this is the first attempt to simultaneously analyse these correlates at once. With respect to the reasons for unmet need, we draw upon and extend existing analyses which have reviewed these causes which include: cost, perceived stigma, minimisation of symptoms, low perceived effectiveness of treatment, and structural barriers [5].
Education was coded as the highest level reached from among elementary school, middle school, high school or college and higher. Employment was coded as full-time employed, part-time employed, unemployed, or other (defined as those not in the labour force such as students, retirees, or disabled individuals). Annual household income was coded as less than $20,000, between $20,000 and $49,999, between $50,000 and $74,999, and $75,000 or greater. A dichotomous variable was created to indicate whether the respondent was a recipient of a government assistance program (i.e. Supplemental Security Income [SSI], food stamps, cash assistance, and/or non-cash assistance). Insurance provider was coded as privately insured, insured by Medicare, insured by Medicaid, insured by Tricare or Veterans Administration (VA), uninsured, or other.
Self-rated health was dichotomised into two categories: those reporting excellent, very good, or good self-rated health and those reporting fair, poor, or very poor selfrated health. Dichotomous variables indicating past-year substance abuse or dependence were coded for each of: alcohol, pain relievers, cocaine, hallucinogens, heroin, inhalants, marijuana, sedatives, stimulants, and tranquilizers.
Statistical analyses were performed in Stata 14. We conducted bivariate descriptive analysis of our primary dichotomous variable of interest, unmet need for mental health treatment, and multiple maximum-likelihood logit regression with weighted least squares on social, economic, and health correlates, with and without adjustment, to assess the demographic, socioeconomic, and health correlates of unmet mental health treatment both individually and simultaneously adjusting for all other correlates. Further analyses of our secondary dichotomous variables of interest indicating reason(s) for unmet need were also conducted using multiple maximum-likelihood logit regressions with weighted least squares on social, economic, and health correlates. Given the NSDUH's complex sampling design, all analyses were weighted using analytical weights provided by SAMHSA with each annual dataset.

Results
Descriptive characteristics of our sample are shown in Table 2. From 2002 to 16, between 36,000-43,000 adults aged 18 or above were surveyed annually, yielding a total study population of 579,017. Table 3 shows the reasons respondents gave for past-year unmet need for mental health treatment from 2002 to 16.

Unmet need
Unadjusted and adjusted odds of reporting unmet need among those perceiving a need for mental health treatment in the past year according to a range of demographic, socioeconomic, and health characteristics are shown in Table 4. Factors increasing the odds of reported unmet need included: being female, attaining an educational level of college or higher, receiving government assistance, reporting fair, poor, or very poor health, and being insured by Medicaid, Tricare or VA, or not having health insurance. In addition, with the exception of hallucinogens and sedatives, past-year abuse or dependence on any substance increased the odds of unmet need. On the other hand, attributes of decreased odds of unmet need included: age over 34, being married or widowed, and having a household income of $50,000 or more. Table 5 shows the results of adjusted logistic regression of individual demographic, socioeconomic, and health characteristics associated reported reasons for unmet need among respondents indicating past-year unmet need for mental health treatment.

Reasons for unmet need
Cost was more likely to be cited as a reason for unmet need by subjects between 26 and 49 years of age, those reporting fair, poor, or very poor self-rated health, and those with who reported being uninsured. Those living in suburban or rural areas were more likely to indicate perceived stigma and minimisation as reasons for unmet need. Respondents with fair, poor, or very poor self-rated health or who reported past year abuse or dependence on either alcohol or pain relievers were also more likely to cite perceived stigma as a reason for unmet need. Respondents with an annual household income of $50,  000-$74,999 were more likely to indicate low perceived effectiveness of treatment as a reason for unmet need. In addition, respondents aged 50 and over, females, non-Hispanic Asian respondents, and respondents with at least some high school education were more likely to cite structural barriers as a reason for unmet need. Several groups had higher odds of reporting reasons other than those shown above as a cause for unmet need: respondents aged 26-49, non-Hispanic mixed respondents, those not working full-time as well as respondents insured by Tricare or VA or those reporting tranquilizer abuse Notably, those with at least some high school education showed much higher odds of reporting a reason not listed above as a cause of unmet need than those with an elementary school education.

Discussion
Our analyses have elucidated the major characteristics associated with increased odds of unmet need, which include: past year substance abuse or dependence (other than hallucinogens and sedatives), fair, poor, or very poor health, being female, and an educational attainment   of college or higher. With respect to reasons for unmet need, cost was most often cited, followed by perceived stigma, structural barriers, and minimisation. Characteristics associated with increased odds of indicating cost as a reason for unmet need include: being uninsured or aged 26-35. Minimisation and low perceived effectiveness are mentioned by high-income persons as reasons for unmet need. College-educated persons and women had higher odds of citing structural barriers as a reason for unmet need. Our study has some limitations such as using household survey data to assess unmet need as well as demographic, socioeconomic, and health characteristics of the sample population. Perceived unmet need is also a self-reported variable which was not validated using psychiatric diagnostic information; consequently, underreporting or over-reporting of perceived unmet need would affect the accuracy of prevalence estimates for unmet need. Indeed, perceived unmet need is subjective, based on sociocultural factors such as patient expectations and Allin and Masseria suggest that analyses of unmet need are contingent upon the specific phrasing of questions [22]. Small samples for specific subpopulations in this study limit the ability to identify specific patterns of unmet need in these populations. Moreover, the NSDUH excludes individuals with no fixed household address and those living in institutional premises, such as prisons, precluding conclusions on unmet need among these vulnerable populations. In addition, given the repeated, cross-sectional nature of the NSDUH, we are not able to conduct analyses of individuals over time and, therefore, we are unable to determine causality between unmet need and the demographic, socioeconomic, and health correlates under examination. Indeed, there are a number of other potential reasons for and contributors to unmet need which are not examined in this study which bear further scrutiny and study. Nevertheless, the NSDUH has been shown to provide comparable findings to other validated health studies such as the National Comorbidity Survey Replication (NCS-R) and is the only source of data in the United States which provides information on unmet need for a nationally representative sample of adults living in the United States [23]. Our results are consistent with recent and established literature which identify disparities in expressed unmet need based upon age [24], gender [25,26], economic disadvantage [27], urban/rural status [28], health insurer [29], illicit substance use [30].

Conclusion
This study extends our understanding of disparities in mental health treatment by not only considering demographic, socioeconomic, and health characteristics of those expressing unmet need from 2002 to 16 but also identifying how these correlate with reasons for unmet need. Cost was a major cause of unmet need among respondents aged 26-49 and those who were uninsured; though the ACA has attempted to reduce the number of uninsured young adults, provisions only guarantee continued enrolment of dependent children until age 25 and, consequently, this may partly explain this observed pattern of expressed unmet need due to cost among these  subgroups [31]. Of the reasons for unmet need examined in this study, only perceived stigma and minimisation appear to increase the odds of expressed unmet need among those living in suburban or rural areas, controlling for other demographic, socioeconomic, and health characteristics, highlighting a potential opportunity to develop health promotion interventions for these subpopulations to address unmet need consistent with established literature [32][33][34][35]. That the odds of indicating structural barriers as a reason for unmet need were not statistically significantly higher among those living in suburban or rural areas is somewhat surprising, given existing research which indicate that availability of adequate mental health care is a cause for concern [34,36]. This, in turn, suggests that further research is needed to fully understand the availability or lack thereof of mental health treatment in suburban or rural areas which can inform new initiatives focused on access, such as telemedicine approaches [37,38]. In addition, the large number of subpopulations expressing unmet need for reasons not specified in this study (i.e. cost, perceived stigma, minimisation, low perceived effectiveness of treatment, and structural barriers) suggests a need to investigate the reasons why these subpopulations express unmet need to address the causal factors underlying why these subpopulations do not receive adequate mental health care