Epidemiological studies have shown that mental disorders are prevalent (
1,
2,
3,
4 ). Only a small proportion of persons with mental disorders seek professional help (
5,
6 ). Population-based surveys indicate that young people between 15 and 25 years old have the highest rate of mental disorders, around 30% (
7,
8 ). However, only 25% to 35% of young people with mental health problems seek professional help (
9,
10,
11,
12 ).
There are two dominant theoretical models of help seeking for mental disorders: Goldberg and Huxley's 1980 model (
13 ) and the model described by Andersen and Newman in 1973 (
14 ), which was revisited by Andersen in 1995 (
15 ). Goldberg and Huxley conceptualized the pathway to psychiatric care as a progression through different filters. The first filter is the decision by the affected individual to seek help. The second is the recognition of a psychiatric disorder by the primary care provider, and the third filter is this provider's decision to refer the patient to specialized care (
13 ). However, if the health care system, as in Sweden, permits self-referral directly to specialty mental health care, the second and third filter can be omitted. Giving individuals the option to self-refer may increase the rate of help seeking and reduce barriers for groups who experience access problems. However, it may also lead to self-referral by those with symptoms but no diagnosable disorders.
To our knowledge the consequences of permitting self-referral to specialty mental health care have not been studied. The behavioral model of health service use proposes that use is a function of a person's predisposition to use services, factors that enable or impede use, and the person's need for services (
14,
15 ). Interventions that improve mental health literacy, such as community awareness campaigns and school-based interventions (
16 ), increase appropriate help seeking. The combination of public education about mental disorders and the option to self-refer to mental health specialists has not been studied. It is not clear whether in the absence of a structural filter, a higher rate of direct help seeking would result in overutilization of specialty care.
The aim of this study was to determine whether patients who refer themselves to specialized mental health care differ from those who are referred by nonpsychiatrist professionals. We hypothesized that self-referred patients would have fewer psychiatric disorders than those referred by nonpsychiatrist professionals.
Methods
This naturalistic, longitudinal study was conducted at Flogsta Outpatient Clinic in the Department of Psychiatry of Uppsala Hospital between October 1, 2002, and September 30, 2004. Uppsala, Sweden, is a city with a population of nearly 180,000—32% of which is under the age of 25. In addition to the 180,000 residents, Uppsala has a large population of university students—about 40,000. All patients from 18 to 25 years of age who were living in the catchment area and who visited the clinic were consecutively included in the study during the first study year (until September 30, 2003). During this enrollment year, 217 patients (6% of the catchment area population in this age group) came for an initial appointment and were invited to participate in the study.
Of the 217 patients, 200 (92%) agreed to participate. No significant differences were found between participants and nonparticipants in gender distribution, diagnoses, and proportion of self-referrals. The mean age of those who agreed to participate was significantly higher (22.4±1.9 years, compared with 21.5±1.7 years for nonparticipants; t=2.21, df=215, p=.04).
After receiving a complete description of the study, participants provided written informed consent. The Ethics Committee of Uppsala University Hospital approved this study.
Procedure
An initial diagnostic assessment based on
DSM-IV criteria (
17 ) was conducted over three patient visits—two with a psychiatrist and one with a social worker. A clinical interview was conducted during the first visit, and a structured diagnostic interview was conducted during the second. Psychosocial and environmental problems were assessed during the third visit. The visits were followed by a team conference at which all available information was presented and axis I, III, IV, and V diagnoses were established. Patients were then provided with appropriate treatment. Axis II diagnoses were assessed by the psychiatrist after axis I symptoms were resolved. A total of 188 participants (94%) underwent personality disorder assessment.
Assessments
Psychiatric and personality disorders were assessed, respectively, with the Structured Clinical Interview for DSM-IV axis I disorders, clinical version (SCID-I-CV) (
18 ) and the Structured Clinical Interview for DSM-IV axis II, personality disorders (SCID-II) (
19 ). The SCID-I-CV provides comprehensive assessment of most diagnoses. However, it includes only screening questions for some diagnoses—agoraphobia, social phobia, specific phobia, generalized anxiety disorder, anorexia nervosa, bulimia nervosa, and the somatoform disorders. If participants screened positive on these questions, a checklist of all
DSM-IV criteria for the disorder in question was used. For those who screened positive for the two eating disorders or for social phobia, supplementary information was gathered with the Eating Disorders Examination-Questionnaire (
20 ) and with the Social Phobia Scale and the Social Interaction Anxiety Scale (
21 ).
Two of the authors (MR and AR) performed SCID interviews; their training on the SCID involved assessment of patients on audio- and videotapes and of patients in clinical practice. Interrater reliability was measured for eight randomly selected SCID-I-CV interviews and six randomly selected SCID-II interviews (Kappa coefficients of 1.0 and .89, respectively). All staff members were trained and experienced in use of the Global Assessment of Functioning (
22 ) (GAF).
Age at onset of previous and current disorders was determined by patient self-report during the initial clinical interview and when the SCID-I-CV interview was reviewed with the patient. A life chart was used to facilitate recall for patients who had difficulties. Two questions were asked during the first visit to obtain information on previous contacts with the health care system for psychiatric symptoms: "Have you ever had any previous contact with the health care system because of mental problems?" Those who answered affirmatively were asked, "From whom did you receive help for your mental problems?" If available, medical records were checked for previous psychiatric symptoms and past contacts. Also, when the SCID-I-CV indicated that a patient had a particular diagnosis, the interviewer asked about previous health care use in regard to those symptoms.
Previous use was defined as a visit to a mental health professional because of a psychiatric problem before the patient visited the study clinic. Four categories of previous visits were used: psychiatry, child and adolescent psychiatry, general practitioner, and school or other mental health services, including social services and other mental health care providers. Current mental health care use was assessed as the number of visits during the study period.
Analysis plan
To compare differences between groups, we used Student's t test for continuous data and the chi square test for categorical data. When expected frequencies were small, Fisher's exact test was used. A conditional logistic regression was estimated to evaluate factors that distinguished self-referred patients from those referred by nonpsychiatrist professionals. Factors with a univariate p value of <.10 were entered into the model, and the level of significance was .05. Two-tailed tests were used. SPSS, version 13, was used for all analyses.
Results
Among the 200 patients, 161 (80%) were women and 39 (20%) were men. Most participants were born in Sweden (N=180, 90%) and living by themselves (N=154, 77%), and most were students (N=126, 63%). Forty-two patients (21%) lived with their parents, and only four (2%) were living in an institution. Forty (20%) were working, one person was in job training, and 22 (11%) were unemployed, of whom 20 (10%) were receiving short-term disability and two (1%) had no job.
Table 1 presents data on demographic characteristics of self-referred patients and those referred by nonpsychiatrist professionals. No statistically significant differences were found between groups.
Among the 200 patients, 49 (25%) were self-referred and 151 (75%) were referred by nonpsychiatrist professionals. Most of the self-referrals (N=32, 65%) had previous contacts with professionals in regard to psychiatric symptoms, and 28 (57%) had previous contacts with a psychiatrist. Among these 28 patients, two had visited a psychiatric emergency unit in the past month. However, the emergency unit did not provide them with a letter of referral. Seven of the self-referrals who had seen a psychiatrist during the past year had formally completed their treatment and had not seen a psychiatrist in the month before self-referral. Therefore, all self-referrals had decided, without guidance from their professional contacts, to seek specialized care.
Of those who were not self-referred, 73 (48%) were referred by psychiatric services, 35 (23%) were referred by their general practitioner, 31 (21%) came through school health services, and 12 (8%) were referred by other mental health professionals.
Mood disorders and specific phobia were more common among self-referred patients (p<.05 for both), but no other statistically significant between-group differences in axis I disorders were found (
Table 2 ). Comorbidity of axis I disorders was pronounced in both patient groups (mean±SD of 2.2±1.3 diagnoses); 48 self-referred patients (98%) met criteria for more than one axis I disorder, compared with 145 (96%) referred by others (
Figure 1 ). One self-referred patient (2%) and six patients referred by others (4%) did not meet criteria for any current diagnosis. No significant difference between self-referred patients and the other group were found in number of current axis I disorders (2.3±1.3 and 2.2±1.3, respectively) or number of lifetime diagnoses (2.9±1.7 and 2.9±1.6). Self-reported age at onset of the first disorder also did not differ between groups (13.6±4.7 and 13.5±4.2 years, respectively).
Of the 188 patients who completed the assessment for axis II disorders, 53 (28%) had at least one personality disorder; 27 patients (14%) met criteria for more than one: only one disorder, 38 participants (19%); two disorders, 19 (10%), three disorders, six (3%), and four disorders, two (1%). No differences were found by referral status in the prevalence of personality disorders (
Table 2 ). The GAF rating at study entry also did not differ between groups (54.5±6.5 for self-referred and 54.7±8.0 for others). (Possible scores on the GAF range from 0 to 100, with higher scores indicating better functioning.)
Of the 200 patients in the sample, 155 (78%) had previous contacts with professionals to address psychiatric symptoms: child and adolescent psychiatry, 53 patients (34%); psychiatry, 58 (37%); general practitioner, 53 (34%); school health services, 57 (37%); and some other mental health care provider, 14 (9%). Among the 155 with previous contacts, 90 (58%) had one contact, 50 (32%) had two, and 15 (10%) had three. No significant differences by referral status were found in number of previous contacts (1.1±1.0 for self-referred and 1.2±.8 for others). As shown in
Table 3, a larger proportion of self-referred patients had no previous contacts with professionals in regard to psychiatric symptoms (35% and 18%). During the study period, there was no significant difference by referral status in the number of visits (16.5±12.7 visits for self-referred and 17.1±16.0 for others).
A logistic regression was performed with referral status as the dependent outcome variable. Variables with a univariate p value of <.10 were entered into the model as independent predictors—any mood disorder, any anxiety disorder, any eating disorder, and no previous contact with mental health services. Patients who did not have any previous contact with professionals in regard to psychiatric symptoms were significantly more likely to refer themselves (odds ratio [OR]=2.16; 95% confidence interval [CI]=1.02–4.55). In addition, having any mood disorder also increased the likelihood of self-referral (OR=2.36; CI=1.01–5.54). Having any anxiety disorder or any eating disorder was not a significant predictor of referral status.
Discussion
The young adult patients who visited the psychiatric clinic during the year of study enrollment accounted for 6% of the young adults in the catchment area—or about one of every 20 individuals in this age group. This percentage is similar to those found in studies in the Netherlands (
23 ) and Canada (
24 ), even though the option of self-referral to specialized care is not available in all health care systems. In Sweden, the public mental health service provides equal access to treatment, regardless of a person's ability to pay. Therefore, even though the Swedish system does not present financial barriers to treatment or require referral to specialized care from general practitioners, the rate of help seeking among young adults did not exceed rates in other health care systems (
4,
8,
9,
10,
25 ). Permitting direct self-referral did not seem to increase the proportion of individuals seeking specialty mental health care.
Young adults who self-referred and those referred by nonpsychiatrist professionals had few differences in clinical diagnoses. However, contrary to our expectations, the self-referred patients had more mood disorders (major depressive disorder, bipolar I and II disorder, dysthymia, and depression not otherwise specified)—84% compared with 69% of the other group. Our hypothesis, which stated that a lower threshold for help seeking would result in self-referral by patients who had symptoms but no diagnosable disorders was not supported. Self-referred patients were equally as likely or more likely than those referred by nonpsychiatrist professionals to have diagnosable mental disorders.
A larger proportion of the self-referred patients had not had previous contact with professionals in regard to psychiatric symptoms, which suggests that the option of self-referral lowers the threshold to specialized care. However, the level of morbidity among these patients did not indicate overutilization of care.
Community studies have found that persons who seek mental health treatment are those with a younger age at onset of the disorder (
26 ) and those with comorbid psychiatric disorders (
2,
9,
24,
27,
28 ). Other determinants of help seeking include being female (
24 ); having mood and anxiety disorders, especially if the disorders are comorbid (
29,
30 ); and being in a younger age group (
2 ). We did not find that the option of self-referral altered this pattern of help seeking.
The general practitioner in Goldberg and Huxley's model (
13 ) accounts for two filters between the patient and specialized care: first, recognition of the disorder, and second, the decision to refer the patient. However, because the self-referral option did not increase the rate of help seeking, filters other than the general practitioner could exist. It is obvious that sex, social factors, and specific diagnoses matter. According to Andersen and Newman (
14,
15 ), predisposing individual determinants of health service use can be demographic characteristics or factors related to social structure, such as level of education. Help seeking can also be related to knowledge and beliefs. In this sample, there was a preponderance of university students. This could reflect the social structure in a university city; however, within the catchment area, there were also socially disadvantaged sections of the city, where relatively more immigrants live. Very few patients came from these neighborhoods. It is probable that knowledge of mental disorders is higher among university students, increasing the likelihood of help seeking.
Previous contact with child and adolescent psychiatry should affect knowledge and beliefs about psychiatric services and mental disorders. The influence of previous contact with child and adolescent psychiatry on help-seeking behavior among 1,420 young adults was examined in a Swedish study in a rural area (
31 ). In that study, one-fourth of the individuals who had previously received care from child and adolescent psychiatry sought help as adults from psychiatrists—most often early in life, at age 19 or 20. However, only 8% of the participants who were receiving care from a psychiatrist had received previous care from child or adolescent psychiatry (
31 ). However, in our study a larger proportion—26% of participants in both referral groups—had received previous care from child or adolescent psychiatry. Lowering the threshold for help seeking did not increase the rate among persons with previous contact with child or adolescent psychiatry. Other social factors likely influence help seeking, such as specific symptoms, especially depression and suicidal thoughts (
32 ). The existence of these factors is supported by findings from a study by Engqvist and Rydelius (
31 ).
Strengths of the study reported here are the comprehensive diagnostic evaluation that participants received and a high level of participation from a group of psychiatric patients. Patients who sought help at the clinic had no knowledge of the ongoing study before they visited the clinic; such knowledge might have influenced patterns of referral. Examination took place in an ordinary clinical setting, and patients were not recruited for the study, which increases the likelihood that the sample is representative of other clinical samples. In addition, to our knowledge this is the first study of referral patterns in relation to psychopathology. Clinical epidemiological studies, like community-based studies, should be based on standardized evaluations; however, in many cases they are not. The largest clinical epidemiological study to use semistructured diagnostic interviews is the Rhode Island Methods to Improve Diagnostic Assessment and Services project (MIDAS) (
33 ). Using SCID interviews, they reported prevalence of axis I and II disorders among 2,300 psychiatric outpatients (
19,
34 ). The prevalence of different axis I disorders in this clinical sample is similar to the prevalence in the study by Zimmerman and colleagues (
33 ). Our sample is younger than the sample in that study and has a larger proportion of women; thus small differences exist. In our sample 68% of participants had comorbid axis I disorders, which is similar to the proportion in the study by Zimmerman and colleagues (
33,
35 ).
The most obvious weakness of our study involves the generalizability of our findings to other settings with older and less educated patients and in other countries with differing health care systems. In a university town, such as Uppsala, the level of education is high, which accounts for the large proportion of university students in our sample. Another weakness is the possibility of a heterogeneous comparison group—that is, patients in the comparison group were referred by several types of providers with a range of medical competency. Although most were referred by physicians, some were referred by nurses, psychologists, and social workers from school health services and social services. An additional weakness is that age at onset, previous diagnoses, and previous use of health care services for psychiatric symptoms were based on self-report (
36 ). It has been shown that patients' poor recall affects reports of lifetime diagnoses (
37 ). Finally, interrater reliability was calculated from a limited number of interviews. However, both interviewers received clinical training in the same department by the same tutors and both interviewers had used the SCID to reliably assess patients from audio and video recordings.