We asked physicians to estimate the percentage of adolescent patients they saw in the last year who had specific mental disorders. In addition to mental disorders, we asked about the percentage of adolescent patients who engaged in risky behaviors such as excessive alcohol use. The estimates given by the physicians roughly resemble the reported prevalence of these mental health conditions in the adolescent population (Table 30.9).

Depression was the most commonly seen mental disorder, followed by anxiety disorders and alcohol abuse. On average, physicians said that in the last year about 16% of their adolescent patients were depressed. Anxiety disorders and alcohol abuse were the next most commonly seen disorders, with physicians saying that, on average, about 10% of their patients had these conditions. About 9% of adolescent patients were thought to have an eating disorder. And physicians reported that about 5% of their patients, on average, were suffering from bipolar disorders and a little over 1% from schizophrenia.

These estimates are not far off from reported national prevalence of mental health conditions in adolescents. Approximately 10% to 15% of adolescents exhibit some signs of depression, although the percentage of adolescents who meet the criteria for a full-fledged diagnosis of depression is closer to 5% to 8%. Similarly, while only 1% of adolescents meet the full criteria for bipolar disorders, nearly 6% of adolescents present with many of the classic symptoms of the con dition. Additionally, 13% of the adolescent population is thought to have an anxiety disorder, whereas 1% is believed to have schizophrenia (National Institutes of Mental Health, 2000a, 2000b); U.S. Department of Health and Human Services, 1999). Estimates of alcohol dependence are lower than the estimated rate found here; however, alcohol abuse is quite common among adolescents.

We conducted regression analyses to determine significant predictors of prevalence estimates. Table 30.9 shows the changes in prevalence rates attributable to several differences in provider characteristics with other differences held constant. Not shown are differences in provider gender, age, rural vs. urban location, reported weekly adolescent patient load, and income level of patient population. It is clear that the self-reported diagnostic skill of the provider is a consistent predictor of estimated prevalence. This was measured by taking the mean response to the questions concerning self-assessed diagnostic ability (Table 30.4) for the five mental disorders (alpha = .79). For alcohol abuse, we used the reported diagnostic knowledge for alcohol abuse. The only condition not predicted by self-assessed skill was the prevalence of eating disorders. Screening for mental disorders was also a predictor of anxiety and schizophrenia.

It is important to note that diagnostic skill was assessed on a four-point scale. The distribution of this scale was such that the difference between those who were most confident in their diagnostic skill (the top 15%) and those who were least confident (the lowest 15%) produced a difference of about 5 percentage points in reported prevalence of depression, anxiety, and alcohol abuse.

Concern about stigma was a predictor of diagnosis for alcohol abuse but was not systematically related to any of the other disorders. Belief in treatment efficacy (not shown) was also not related to any diagnosis rates. Family physicians and general practitioners also gave higher prevalence estimates for all the conditions but bipolar disorder. Prevalence estimates did not differ according to the proportion of the patient population that had private insurance or that was served by managed care.

We cannot confirm that the estimated prevalence rates are valid; however, the predictors are consistent with the hypothesis that providers with better diagnostic skills (or those who screen for disorder) are more likely to identify mental disorders. Furthermore, by holding constant other provider and practice characteristics that might affect prevalence rates, we can be more confident that the differences are a reflection of diagnostic skill and not patient populations.

We also asked providers to estimate the percentage of patients with symptoms of each mental condition they had referred to other professionals in the past year. Although referral rates did not differ dramatically across diagnoses (Table 30.10), the rates did differ considerably within each diagnosis. Roughly 30% to 45% of providers said they referred most adolescent patients with the given symptoms to other professionals. About 4 in 10 physicians said they referred most of their adolescent patients with symptoms of schizophrenia (44%) or alcohol abuse (38%). Somewhat fewer referred most of their patients with symptoms of bipolar disorders (35%), anxiety disorders (33%), depression (32%), or eating disorders (28%).

At the same time, many physicians referred 25% or fewer of their patients with symptoms of

a mental health problem. Roughly 4 in 10 did not make a referral for most of their adolescent patients with signs of eating disorders (46%), alcohol abuse (46%), anxiety disorders (43%), schizophrenia (41%), bipolar disorders (40%), or depression (38%).

A regression analysis of referral rates indicated that types of practice and insurance were the major predictors (Table 30.10). Providers with most of their patients in managed care plans were more likely to refer patients with anxiety, bipolar, and alcohol problems. Providers with patients who were mostly privately insured were more likely to refer patients with schizophrenia and eating disorders. Providers who screened their patients for mental disorders were also more likely to refer them for treatment of depression and eating disorders. Diagnostic knowledge, stigma, belief in treatment efficacy, or type of provider were not significantly related to referral rates.

Although not shown in Table 30.10, our analysis also revealed that providers who were more likely to refer patients to mental health specialists were more likely to rate the treatment resources in their community as inadequate. This was especially true for the referral of anxiety disorders and bipolar disorder. Evaluations of treatment resources were not related to practice characteristics such as managed care or private insurance status.

The low levels of confidence in recognizing depression that were observed in this study are consistent with another recent survey of primary care pediatricians (Olson et al., 2001). Although the present survey indicates that providers feel most capable of identifying depression (about 50%), their confidence in identifying other disorders is even lower. For example, only 25% reported being very capable of identifying drug abuse in their adolescent patients. In the absence of effective screening procedures, primary care providers will continue to be unable to recognize mental and substance abuse disorders in their adolescent patients.

Providers' estimated prevalence of mental disorders in their patient populations indicated that although the rates are in line with national estimates, variation across providers was strongly related to diagnostic knowledge. Providers who were more confident in their ability to diagnose mental conditions reported higher prevalence of disorders in their patients. In addition, those who regularly screened for mental disorders were also more likely to recognize some of the conditions. The importance of appropriate diagnosis is also underlined in the finding that providers who screened their patients were more likely to refer them for treatment of depression and eating disorder.

The inability to detect mental disorders in adolescents is a serious flaw in the primary care system. No matter how well intentioned providers may be, there is little that can be done to help adolescents in need of treatment if they are not first identified. Although providers recognize the importance of referring their patients to mental health specialists and of including parents in the treatment of mental disorders, these intentions do not come into play without adequate resources and abilities to detect patients at risk for mental and substance abuse disorders.

Pediatricians as a group were even less likely to report mental health problems in their pa

tients than family doctors or general practitioners. In view of the much higher representation of pediatricians in the adolescent primary care system, there is a large opportunity to increase the ability to detect mental disorders in primary care. The Bright Futures and KySS programs already recommend screening and referral for adolescent mental disorders and have materials in place that could advance the adoption of better screening practices. Research would also be valuable to determine if computer-assisted screening mechanisms such as that developed by the Columbia Teen Screen program (McGuire & Flynn, 2003) could be adopted in primary care.

Increased knowledge and ability to diagnose mental health problems may also reduce concerns about stigma, a factor that could also impede appropriate referral and treatment. Providers with greater confidence in their understanding of mental disorders were also less likely to express concerns about the stigma of mental disorder. Although concerns about stigma were only related to the estimated prevalence of alcohol abuse, substance abuse is highly comorbid with other disorders and its association with stigma is likely to reflect a failure to identify these disorders as well (American Academy of Pediatrics, 2000).

Recognition of mental disorder is only a first step in delivering appropriate care. Providers must also make decisions about appropriate treatment. Estimates of rates of treatment referral indicated that about half of patients with major mental disorders are referred on average. It is difficult to evaluate the optimal level of referral without greater knowledge of the diagnostic criteria used by providers. However, given these reported rates of referral, it is not surprising that less than half of adolescents with serious mental conditions are estimated to receive appropriate mental health services (Costello et al., 1998; Horwitz et al., 1992; Sturm et al., 2003).

Practice characteristics were the strongest predictors of reported referral rates. In particular, providers whose patient population was primar ily served by managed care were more likely to refer those patients for anxiety disorder, bipolar disorder, and alcohol abuse. This finding may reflect the influence of increased deployment of mental health carve-outs in managed care. This approach is designed to contain costs for mental health treatment, but it could also encourage referral for those who are appropriately diagnosed with mental conditions (Conti et al., 2004). Unfortunately, the present research could not assess the quality of the referrals that are being made in managed care. There is evidence to suggest that the barriers to successful referral are actually greater in pediatric managed-care systems (Walders, Childs, Comer, Kelleher, & Drotar, 2003). Hence, many have expressed concerns that mental health carve-outs will encourage referral to less than optimal care providers (Glied & Neufield, 2001; Jellinek & Little, 1998; Kelleher, Scholle, Feldman, & Nace, 1999). Research to determine the fate of adolescents with mental disorders in managed care is clearly a high priority.

Providers whose patient population was primarily served by private insurance were more likely to refer them for treatment of some disorders (schizophrenia and eating disorder). This finding suggests that patients in the public sector may be receiving even less appropriate treatment for these conditions.

An important concern in the findings is the unfavorable evaluation that most providers report of treatment resources in their communities. Indeed, providers who were more likely to refer patients to mental health specialists were also more likely to evaluate those services as inadequate. This finding suggests that better screening will only solve a part of the mental health service delivery problem for adolescents. Better treatment resources will also be needed (cf. Walders et al., 2003). In sum, this survey indicates that appropriate screening and diagnosis of adolescents in primary care is a critical step in advancing the mental health of youth. At the same time, how

ever, increased referral to mental health specialists may have the potential to make providers even more aware of the limitations in the mental health treatment system for adolescents.

As part of its Sunnylands Adolescent Mental Health Initiative, the Annenberg Public Policy Center commissioned Princeton Survey Research Associates International (PSRAI) to conduct the survey. The fieldwork was conducted by Princeton Data Source, LLC, and Braun Research, Inc., from September 29, 2003, to January 23, 2004. In total, 727 interviews were conducted with a nationally representative sample of pediatricians, family practitioners, and general practitioners who regularly treat adolescents between the ages of 10 and 18. The interviews took approximately 25 min to complete. The maximum margin of sampling error for results based on the full sample is ±4%.