Recruitment into psychiatry resident programs, as compared with most other medical fields, has been declining over the last two decades (1). This is not unique to the western world, but is also true for developing countries (2). The difficulty in recruiting psychiatrists from the medical workforce poses a problem for all healthcare systems, as psychiatric disorders rank high in the health morbidity of all countries (3). Attitudes toward psychiatry may affect recruitment, although this is by no means a simple relationship (4). These attitudes may also have an impact on the treatment of mentally ill patients by other doctors, especially general practitioners. It may also be appropriate to measure the attitude toward psychiatry among medical students as an outcome measure of the effectiveness of undergraduate training in psychiatry, because the stigmatization of mental illness among doctors is a real concern (5). There is therefore a need to be able to measure the attitude toward psychiatry among medical students and junior doctors in training.

There has been a limited amount of research into the attitudes of medical students toward a prospective career in psychiatry. Our literature search did not reveal any globally-accepted tool used to elicit medical students' attitudes toward psychiatry. Some studies have used their own scales, whereas others have used one of two established questionnaires (e.g., Muga and Hagali (6)). Burra et al. (7) developed the Attitudes Toward Psychiatry–30 scale (ATP–30). They proposed eight attitudinal objects: psychiatric patients, psychiatric illness, psychiatrists, psychiatric knowledge, psychiatric career choice, psychiatric treatment, psychiatric institutions, and psychiatric teaching. They designed their questionnaire around these attitudinal objects, but, on factor analysis, did not find that dimensions correlated around these attitudinal objects, nor did they find other dimensional subscales emerging from the data.

The Attitudes to Psychiatry Scale (APS) was a questionnaire developed by Balon et al. (4) in the United States. Initially used in that country, it was subsequently applied in other international sites, including Barcelona, Spain, (8) and Kumasi, Ghana (9). The APS consists of 29 questions (modified per app1) and was developed using some questions from an earlier study (10). The original study by Nielson and Eaton had developed 12 questions from “prior work and conventional wisdom.” A further 17 questions were added by Balon, based on his judgment, and the whole questionnaire was clustered around five dimensions: 1) overall merits of psychiatry; 2) efficacy; 3) role and function of the psychiatrist; 4) career and personal reward; and 5) medical school factors. The response to each question is on a 4-point Likert scale. The authors did not describe a factor analysis to examine whether the questions in each domain were correlated, or whether each of the 29 questions added to the scale's usefulness. Given that Burra's ATP–30 scale had not demonstrated dimensional subscales on factor analysis, we were interested in examining whether subscales emerged from the APS scale.

We sought to determine whether the questions on the APS correlate to form dimensional subscales, on the basis of principal-components analysis, and whether these dimensions are related to those proposed by Balon et al. in their original article.

We gathered data on the APS scale, a self-administered questionnaire modified per app1. One question was deemed insensitive, in an international context, and was removed at the beginning of this study (original #19: “Psychiatry is a discipline filled with international medical graduates whose skills are of low quality.”) We recruited 549 respondents (225 men (M), 289 women (W), [35, no reply on Gender]) medical students across four geographical regions including Ghana (N=94; M:W: 43:39), and three different medical colleges located in north (N=51: M:W: 24:27), west (N=203; M:W: 76:113), and central (N=201; M:W: 82:110) India. The sample included 218 in their first year (all from India) and 331 from their final year (India and Ghana). Original Questions 23–27 and 29 were completed by final-year students only as they related to medical school factors.

The questionnaire was completed by medical students in a variety of settings and at different stages of their medical training. This is a cross-sectional survey, and each student completed the questionnaire only once. Written ethics approval was collected at each geographic site in keeping and satisfying the local ethics requirements and procedures.

Factor analysis was used to identify the dimensions captured in the questionnaire. The method used was principal-components analysis with Varimax rotation, chosen since this allows the factors identified to be nonorthogonal and may correlate. The robustness of the factor solutions is tested across subsets of the data, such as by year of study or medical school attended or by repeated random sampling. Cronbach α was used to identify any redundancy in the most highly weighted items from the principal-components analysis. We examined the 28 questions independently of the five dimensions defined by Balon in the original APS questionnaire, looking for questions that cluster together to form emerging dimensions. We then compared these with the original dimensions chosen by Balon.

The questionnaire results were examined to detect any differences between different medical schools, sex, and year-group at medical school. There were no significant differences by these variables, and therefore all the responses to the questionnaire could be analyzed together.

Our analysis is summarized in Table 1. Five dimensions emerged from questions that correlated. On filtering the information, we found that there was good independent correlation (α >0.6) in four of the five dimensions emerging from the analysis. Some questions were external to the domains and were therefore excluded. We examined the questions that correlated to form the dimensions and named the theme of the questions that emerged (Table 2).

A dimension covering inspiration from medical school emerged as the strongest and most consistent (α=0.784). Dimensions looking at the stigma of psychiatry (α=0.696), the merits of psychiatry as scientific medicine (α=0.642), and effectiveness of treatment (α=0.640) also showed a satisfactory internal consistency. The weakest dimension appeared to be a group of questions measuring social factors influencing choice of psychiatry (α=0.50). These emerging dimensions are described and compared with the original groups of questions in Balon's original article in Table 2.

On examining the APS scale using principal-components analysis, we have found clusters of questions around four themes. These themes do seem intuitively relevant for attitudes to psychiatry. They are not the same as Balon et al. 's original dimensions (4), but there are clear similarities.

There are limitations to this study. First-year medical students did not complete original Questions 23–27 and 29 because this population would not have had the requisite exposure to psychiatry to answer these with appropriate knowledge. The fact that students were from two different countries may influence results, but the study is examining the internal validity of the scale, so this should not be a significant problem. The statistical methodology is not without its critics (11), as factor analysis can be confused with principal-components analysis, and criteria for retaining factors and the method of rotation can be disputed. Despite these criticisms, this methodology can work well and was felt to be the most appropriate for this study.

Burra et al. (7), in developing their ATP–30 scale, proposed eight attitudinal objects: 1) psychiatric patients; 2) psychiatric illness; 3) psychiatrists; 4) psychiatric knowledge; 5) psychiatric career choice; 6) psychiatric treatment; 7) psychiatric institutions; and 8) psychiatric teaching. They designed their questionnaire around these attitudinal objects, but found they did not lead to dimensional subscales that correlated. Also, a factor analysis of the 30-item version yielded no other dimensions. The dimensional subscales we have identified can be compared with some of the above eight attitudinal objects, but are perhaps more precise in their definition. The ATP–30 lends itself to a total score for attitude toward psychiatry, which the APS scale does not.

Previous qualitative research on attitudes that affect recruitment to psychiatry suggests that negative attitudes influencing recruitment to psychiatry include an aversive view of patients; a belief in the lack of effectiveness of treatment; a view that psychiatry lacks scientific credibility; and the antipsychiatry bias of family, friends, and the public (1). Doctors recruited to psychiatry are more likely to be from cities, of lower social class, politically liberal, responsive to abstract ideas, fond of complexity, and tolerant of ambiguity (12, 13). This study is consistent with this previous research in that it identifies consistency in student attitudes to stigma, scientific credibility, and effectiveness of treatment in the attitudes questionnaire. Experience in medical school also shows a consistent cluster of questions, but may reflect a positive experience of undergraduate psychiatry, rather than a positive attitude to the profession.

Using the questions in the emerging themes that correlated closely, we have suggested a modified questionnaire (see app1). We found two questions from the original scale to be very similar (Questions 8 and 27), and, therefore, omitted Question 8, as this omission does not affect the subset's Cronbach α value greatly (Table 2).

However, we do feel that there is further work needed before a scale can be used with confidence to measure attitudes toward psychiatry. There may be single questions in the original scale that have inherent value but do not correlate with a cluster of questions. We feel that Question 14 (“psychiatrists frequently abuse their legal power to hospitalize patients against their will”) may be in this category. There may be other aspects of attitudes toward psychiatry that have been overlooked, such as self-perceived abilities in psychiatry and enjoyment of the work. Further research is needed in using the dimensional questions identified in this study, along with additional single questions that may be of possible value, and present them to medical students and psychiatrists to measure their validity and reliability. Focus groups and Delphi groups may be beneficial in further refining the scale. The scale may also benefit in investigating whether a total score is a valid indicator of the overall attitude toward psychiatry. This might enable the scale to be used as a marker of the effectiveness of interventions to improve attitudes toward psychiatry. We believe that this study has provided a step in the further development of a robust scale.

We thank Dr. Karen Mattick, who helped us with constructing the paper.

At the time of submission the authors reported no competing interests.