Late-life mood and anxiety disorders are common and often co-occurring among community-dwelling adults (1). Although these disorders are treatable (2–7), it is estimated that over 50% of older adults symptomatic for a clinical diagnosis do not use mental health services (8–10). Yet little is known about why, despite symptoms of mood and anxiety disorders, these older adults typically do not seek services. It is commonly held that poor use is due to stigma associated with mental health care and poor coordination of care (11); however, these assumptions are often made from community-based samples including older adults who do not have a mental illness (12). With consideration of the current and projected growth of the older segment of the population and because mood and anxiety disorders are highly associated with poor health outcomes (13–17), the impact of untreated late-life mood and anxiety disorders has major public health implications. Thus understanding the factors that influence low use of mental health services by those with greatest need is vital.
Most prior research on mental health service use by older adults with psychiatric diagnoses has been with clinical populations (18,19). Because these studies are limited to individuals whose mental disorders become known only in the course of seeking care, nonutilizers are often excluded from study. Although population-based studies are important resources for the examination of use and nonuse, few have clinical measures of mood and anxiety disorders and none that we are aware of have determined key predictors among those with psychiatric conditions (8–10,20–26). In contrast, the National Comorbidity Survey Replication (NCS-R) data examined in this study are nationally representative, with clinically based measures of mood and anxiety disorders as well as a broad range of potential predictors of nonuse.
The primary purpose of our study was to determine the prevalence and key factors associated with not using mental health services in a national sample of older Americans meeting criteria for DSM-IV mood and anxiety disorders. This study contributes to understanding unmet need and barriers to care in the United States among the most affected older adults.
The NCS-R is a nationally representative survey of 9,282 noninstitutionalized participants ages 18 years and older in the coterminous United States (27–29). Respondents were selected from a multistage clustered area probability sample of households. Face-to-face interviews were conducted in respondents' homes between February 2001 and April 2003. The response rate was 70.9% (29). A detailed description of the NCS-R sampling and weighting procedures is provided elsewhere (28).
In this study, we examined a subpopulation of 348 NCS-R adults ages 55 and older who met DSM-IV (30) criteria for prevalent (12-month) mood disorders (major depressive disorder, dysthymia, and bipolar disorders types I and II) and anxiety disorders (panic disorder, agoraphobia without panic, specific phobia, social phobia, generalized anxiety disorder, and posttraumatic stress disorder) using the World Mental Health (WMH) Composite International Diagnostic Interview (CIDI) (31). Data were obtained from the Inter-university Consortium for Political and Social Research (32). Appropriate methods for subpopulation examination of complex sample survey data were implemented for all analyses. The institutional review boards of the University of California, San Francisco, and the San Francisco Veterans Affairs Medical Center approved this study.
To study factors associated with not using mental health services, we used as our conceptual framework the Andersen behavioral model of health services use (33). This model posits three major factors—predisposing, enabling, and need factors.
The predisposing factors examined were age, gender, education (completed up to 11 or ≥12 years), race-ethnicity (non-Hispanic white and black, Hispanic, and other), and attitudes and mental health beliefs. Attitudes toward mental health care were assessed with three questions, concerning willingness to seek professional help (if respondent had a serious emotional problem), comfort level in talking about personal problems with a health care professional, and stigma (respondent would feel embarrassed if friends knew that he or she was getting professional help). Belief in benefit of mental health care was determined if respondents stated that they thought ≥50% of people who see a professional for serious emotional problems are helped. Similar definitions were used in previous NCS-R studies (34).
The enabling factors included marital status (married or cohabitating; divorced, separated, widowed, or never married); income, defined by the poverty index (the ratio of household income to poverty threshold used in the 2001 U.S. census and adjusted for household size; ratios were categorized as low [≤1.5 times poverty line], middle [>1.5–6.0 times], and high [>6.0 times]) (35–37); and health insurance status, defined by private, public, or military sources.
The need factors included comorbid general medical conditions, disability, and severity of mood and anxiety disorders. The general medical conditions examined were relevant to an older age sample with mood and anxiety disorders and included major comorbidities related to cardiovascular disease (stroke, heart attack, or heart disease), diabetes mellitus (38), and chronic pain (arthritis, rheumatism, or chronic from back or neck problems) (39).
Disability was defined by five domains (out of role, self-care, mobility, cognition, and social) of the World Health Organization Disability Assessment Schedule (WHO-DAS) (40,41). “Out of role” was a measure of number of days during the past 30 days when the respondent was completely unable to work or carry out normal activities because of physical or mental health problems. The other domains were a product of frequency (number of days) and severity of problems (none, mild, moderate, or severe) that respondents reported experiencing in the past 30 days. Thus these were self-reported measures of perceived impairment (for example, cognition was defined as perceived difficulty in concentration, memory, understanding, or ability to think clearly). All five scales were normalized to have values from 0 to 100, where higher scores indicated worse functioning.
Severity of mood and anxiety disorders was defined as serious, moderate, or mild in accordance with the WHO World Mental Health Survey Consortium (42). Serious disorders were defined as one of the following: meeting criteria for bipolar type I disorder, attempting suicide in conjunction with any other prevalent WMH-CIDI or DSM-IV disorder, or reporting at least two areas of role functioning with severe role impairment (score ≥7) resulting from the mental disorder in the disorder-specific Sheehan Disability Scales (43). Respondents' disorders were classified as moderate if interference was rated as at least moderate (score 4–6) in any Sheehan Disability Scales domain, and all others were classified as mild.
Dependent variable: nonuse of mental health services
The health services section of the NCS-R asked participants about receiving treatment for emotional problems (that is, “emotions, nerves, mental health, or use of alcohol or drugs”) (35,37,44). Classification of nonuse of mental health services was based on participants' reporting negatively—that is, reporting not seeing either a specialty mental health provider (psychiatrist, psychologist, other mental health professional, social worker, or counselor in a mental health specialty setting; overnight hospital stay; or use of a mental health hotline) or a general medical provider (primary care physician, other general practitioner or family doctor, nurse, occupational therapist, or other nonspecialty mental health professional) in the prior 12 months.
To produce nationally representative estimates, we used clustering and weighting techniques to reduce systematic bias and imprecision imbedded in the complex sampling design. Thus percentages and means were weighted, and statistical differences of nonuse among predisposing, enabling, and need factors were initially estimated on the basis of unadjusted weighted logistic regression analyses. The standard errors were determined from a recalculation of variance with the Taylor- series linear approximation method (45).
Multivariable logistic regression analyses estimated the relationship of the combined predisposing, enabling, and need factors with the odds of not using mental health services. To obtain the most parsimonious model, we selected important factors determined by a priori criteria. Factors were eligible for inclusion in the multivariable model if they were associated with nonuse in bivariate analyses (with p≤.20). To be included in the final model, eligible factors were systematically added to the model and then removed if they did not maintain a p value of ≤.10. Odds ratios and 95% confidence intervals were estimated, along with design-corrected likelihood ratio statistics and Wald chi square tests.
All analyses were performed with SAS Survey procedures, version 9.1.3.
Table 1 presents weighted sample distributions of the NCS-R subpopulation of older adults with prevalent DSM-IV mood and anxiety disorders by predisposing, enabling, and need characteristics. The average age of this subpopulation was 64 years. A majority were female and non-Hispanic white and had a high school education or higher, middle income, and health insurance (approximately 50% received Medicare), whereas less than half were married or cohabitating. In general, attitudes toward mental health care were positive: 88% reported their willingness to see a mental health professional for emotional problems, over 85% reported that they would be comfortable discussing personal problems with a health care professional, and almost 60% believed that such help would be beneficial. Still, 34% said they would feel embarrassed about getting professional help. In addition, although over 80% of respondents had chronic pain, disability scores were low in most domains, and most mood and anxiety disorders were reported as mild to moderate.
Probability of not using mental health services
Overall, 71.3% of respondents did not use mental health services. Table 2 shows that there was a high prevalence of nonuse for most individual disorders (>50%) with the highest (65%–80%) for specific phobia, social phobia, and generalized anxiety disorder. Interestingly, there was a high rate of comorbid mood and anxiety disorders, and prevalence of nonuse among these respondents was 50%.
We also examined predisposing, enabling, and need factors in nonuse of mental health services [four figures illustrating these findings are available in an online supplement to this article at ps.psychiatryonline.org]. Approximately 80% of individuals who were from racial-ethnic minority groups did not use mental health services [eFigure 1; ps.psychiatryonline.org]. In addition, 80% prevalence of nonuse of mental health services was found for respondents who would feel stigmatized if they sought professional help and also those who were not comfortable with the possibility of discussing personal problems with a professional. Approximately 70% of individuals who were non-Hispanic white did not use mental health services, and 70% prevalence of nonuse was also estimated for those who reported no stigma as well as those who were comfortable discussing problems. Although the pattern was similar for the willingness factor, the association was not significant according to the study criterion (p>.20).
In eFigure 2 (ps.psychiatryonline.org), showing enabling characteristics, a greater proportion (80%–86%) of respondents who were married or cohabitating, who did not have health insurance, and who had a middle income did not use mental health services compared with respondents who were divorced, separated, widowed, or never married; insured; and who had low or high income (<70%). Being married or cohabitating was significantly associated with nonuse of mental health services (Wald χ2=11.2, df=1, p<.001). In contrast, not having health insurance and having middle income were not statistically significant (p>.05) but still important potential predictors of nonuse (p≤.20).
eFigure 3 (ps.psychiatryonline.org) shows that those without a history of cardiovascular conditions and no chronic pain or with a mild mood or anxiety disorder were highly prone to not using mental health services (cardiovascular condition, Wald χ2=4.7, df=1, p=.03; chronic pain, Wald χ2=8.2, df=1, p=.004; and severity of mood or anxiety disorder, Wald χ2=31.9, df=2, p<.001). In fact, almost 90% of those with mild disorders did not use mental health services compared with approximately 50% of those with serious disorders. eFigure 4 (ps.psychiatryonline.org) presents the mean disability scores of the five WHO-DAS domains by nonuse and use of mental health services. Lower scores corresponded to nonuse of mental health services. The out-of-role and cognition domains were statistically significant (p<.05) predictors of nonuse, and the mobility and social domains were potentially important predictors (p≤.10).
Table 3 presents the key predisposing, enabling, and need factors included in the final multivariable logistic regression model. The odds of nonuse of mental health services more than doubled among respondents who were black, Hispanic, or of a race or ethnicity other than non-Hispanic white; among those not comfortable with the idea of talking to a professional about personal problems; among individuals married or cohabitating; and among respondents with middle versus high income (odds ratios [ORs]=2.14, 2.50, 2.28, and 2.25, respectively). In addition, older adults with mild versus serious mood or anxiety disorders were almost five times more likely not to use services (OR=4.78). Similarly, respondents reporting no chronic pain had higher odds of nonuse (OR=1.84). Those with perceived cognitive impairment were significantly less likely to report nonuse (OR=.74). Of note, when all variables from Table 1 were combined into one model, nearly identical statistically significant results were found as presented in Table 3. Furthermore, when analyses were considered only among participants categorized as having serious mood or anxiety disorders (N=92), we found similar statistically significant (p<.05) predictors of nonuse as found in the overall model—that is, being married or cohabitating, being in a racial-ethnic minority group, and having low perceived cognitive impairment. However, instead of discomfort in seeking help, those with a serious mental illness reported a lack of belief in the benefits of professional help, which was related to not using services (OR=5.81, 95% confidence interval=2.14–15.78).
Among older NCS-R respondents with a 12-month anxiety or mood disorder, the vast majority did not use mental health services. Although the prevalence of nonuse was high across all predisposing, enabling, and need factors (>50%), not using services was most strongly related to racial-ethnic minority status, discomfort with seeking mental health care, marriage or cohabitation, middle income, mild mood or anxiety disorder, and no chronic pain or minor cognitive complaints. These findings suggest that low perceived need, moderate resources, and low motivation to use mental health care help to explain why services may not be sought, despite diagnosable mood and anxiety disorders.
Prior epidemiological studies have examined potential predictors of service use in the overall sample, pooling data of those with and without psychiatric disorders (8–10,20–26). Most of these studies have shown that the strongest predictors of use were recent diagnosis of a mental disorder and other medical conditions associated with need for care. For example, a study from the 2001 National Survey on Drug Use and Health found that the only variables associated with mental health services use by older adults were having at least one mental health syndrome and poor physical health (10). Although such studies of pooled samples are informative, they do not explain why the majority of older adults symptomatic for a mental disorder do not seek care.
In this study, we found that low levels of mental and physical complaints were particularly important predictors of nonuse of services. This included mild complaints of mood or anxiety disorders' interrupting daily functioning (severity of disorder) and minor or no complaints about pain or cognition. These findings suggest that such minor complaints should be taken seriously, because they may equate to low perceived need for care among older adults with diagnosable DSM-IV mood and anxiety disorders. Moreover, over 80% of respondents had chronic pain complaints, and most (approximately 70%) did not use mental health services, which suggests that many older adults with mood and anxiety disorders may present with somatic symptoms to their primary care physician and yet have little insight into psychological problems. Furthermore, use of mental health services by older adults with mild or moderate mood and anxiety disorders may have been low because these individuals received sufficient informal support and self-help; however, further investigation of this was beyond the scope of our study.
Even after adjusting for need factors, we found important effects of predisposing and enabling characteristics. However, although we found that respondents who were black or Hispanic or who were of a racial-ethnic background other than non-Hispanic white were more likely not to use mental health services, we did not find a significant effect of gender, age, or education, contrary to utilization studies that examined pooled samples with and without mental health disorders (9,20–26). Yet similar to other studies examining use, in the overall sample we found that respondents who were married or otherwise not living alone were more likely not to use mental health services. This finding suggests that older adults who are symptomatic for a mental disorder and married or cohabitating do not perceive a need for mental health care and may not receive support for seeking care or may view their significant other as a surrogate of care. In contrast, this finding may indicate the power of relationship loss or strife as a motivator for seeking care (37,46). In addition, having a middle income was an enabling factor associated with nonuse, suggesting that resources available for treatment are more prominent among older adults with high or low income.
Although most prior research has suggested that stigma is a significant contributing factor to older adults' not using mental health services (47–49), we found that discomfort discussing personal problems with a professional was a more predominant predictor among older adults with mood and anxiety disorders. This finding may be attributable to defining stigma as being embarrassed to tell others about seeking help, suggesting the possibility of seeking help but hiding it from others, and defining discomfort as being uncomfortable interacting with a health care professional. Thus the negativity of discomfort implicates difficulty in initial seeking of help and high potential for discontinuation if help is sought. In contrast, when we focused our analyses on those with severe mood or anxiety disorders, we found that it was not stigma or discomfort which predicted nonuse but instead a lack of belief that mental health care will be beneficial.
The strengths of this study include a nationally representative probability sample, current DSM diagnostic assessment, and a comprehensive list of potential predictors selected on the basis of the well-established Andersen behavioral model of health services use. This study helps to describe the low use of mental health services in a nationally representative sample of older Americans with DSM-IV mood and anxiety disorders. Our study is the first we are aware of to present key factors associated with not using mental health services among older community-dwelling U.S. adults with a mood or anxiety disorder.
Investigating patterns of nonuse of services among older adults has important implications for both policy and clinical practice, where findings support efforts at local and national levels to improve screening, awareness of need for care, and service availability for treatment of mood and anxiety disorders in late life. In particular, these results resonate with two overarching issues and recommendations identified by the New Freedom Commission: improving access and continuity and improving quality of mental health care (for example, with screening and prevention) (11). Also, depression screening has been recommended by the U.S. Preventive Services Task Force as a preventive service under Medicare (50). However, Medicare does not cover mental health screening as a benefit for most older adults (50,51). Ideally, health care reform will eventually increase such coverage, but currently it only expands payment of services already covered (52). However, such change is vital, because coverage of preventive services would promote collaborative care and service integration, which would significantly increase access to and quality of care (4,50,53) and improve the quality of life of older adults.
There were several limitations of this study. First, the NCS-R underrepresents homeless, institutionalized, and non-English-speaking older adults. Second, given issues of stigma, older adults with mental illness might have been less inclined to participate in a mental health survey. Third, even though the WMH-CIDI was shown to have good concordance with the Structured Clinical Interview for DSM-IV (29), it is still a lay-administered interview. Therefore, the WMH-CIDI may not correspond to cases identified in clinical settings despite the use of similar diagnostic criteria. Fourth, some older adults with mood and anxiety disorders who responded to the NCS-R may have been excluded from the study because of difficulty recalling symptoms. Fifth, we cannot validate self-reported use of mental health services. Finally, although less than half of the sample was 65 years or older, which may limit generalizability to elderly adults, the NCS-R was limited in its assessment of mental health disorders considered common among older adults (including depression not otherwise specified, dementia with depression, mood disorders secondary to medical disorders, adjustment disorders with depression, and bereavement). Thus, given the above limitations, the estimates herein are probably conservative.
The results of this study are disturbing, in that all of the study respondents had clinically diagnosable mood and anxiety disorders but most did not use mental health services. However, these findings help to inform and support important public health targets. First, screening and monitoring programs are urgently needed to improve recognition of mood and anxiety disorders among community-dwelling older adults. Second, increased efforts are needed to improve motivation to seek help through outreach services. Third, efforts to expand coverage of mental health services for older Americans through health care reform is a step in the right direction, but the need for more, such as coverage of preventive services, is imperative.
This work was supported by grants MH079093 and MH074717 from the National Institute of Mental Health and by grant AG031155 from the National Institute on Aging. The NCS-R was supported by grant U01-MH60220 from the National Institute of Mental Health, with supplemental support from the National Institute on Drug Abuse, the Substance Abuse and Mental Health Services Administration, grant 044708 by the Robert Wood Johnson Foundation, and the John W. Alden Trust. The views and opinions expressed in this report are those of the authors and should not be construed otherwise.
Dr. Yaffe reports being a consultant to Novartis, Inc., for reasons not related to this project and serves on the data safety and monitoring boards for Pfizer, Medivation, Inc., and the Citalopram for Agitation in Alzheimer's Disease trial for the National Institute of Mental Health. The other authors report no competing interests.