It is well known that the majority of people with emotional disorders consult and are treated by primary care physicians and not by psychiatrists (1, 2). It is also known that the emotional disorders of many patients are not detected by their physicians (3) and that such misdiagnosis may lead to insufficient access to appropriate care. By improving the ability of primary care physicians to identify emotional disorders, the care of patients may be improved. Some evidence suggests that patients who are correctly identified by their physicians as having emotional disorders receive more and better treatment than patients whose emotional disorders remain undiagnosed (4).
The ability of physicians to identify emotional disorders has been shown to be related to their clinical interview skills (5, 6). There is evidence that interview skills can be improved (7, 8) and that such improvement is associated with improvement in the physician's ability to identify emotional disorders (9).
The main objective of this research was to study the relationship between the ability of primary care physicians to identify emotional problems and the clinical interview skills they use. A second objective was to determine whether the association between interview skills and the ability to identify emotional disorders was maintained after the effects of other factors related to the latter variable were controlled for. To achieve this aim, the following factors were measured: physicians' education, physicians' clinical training and professional experience, social characteristics of the physicians and patients, and the severity of the patients' somatic and mental illnesses. Furthermore, we believe that physicians' attitudes toward patients may influence physicians' communication behaviors (10) and, therefore, their ability to identify distress; thus, the evaluation of attitudes was included in the research.
The offices of the 10 primary care physicians studied were selected with the aim of obtaining maximum variation in the physicians' ability to identify emotional distress. All the doctors from the Ciudad Jardin health center and two from the La Florida health center were included; both are in Alicante, Spain.
All consecutive clinical office visits of patients over age 16 were included, except those of patients with previous psychiatric histories (including previous psychopharmacological or psychological treatment) known to their doctors. The inclusion of patients was stopped after each physician had seen at least 15 patients, including five patients identified as psychiatric cases by the Present State Examination—CATEGO program (11, 12).
After a complete description of the study was given to the subjects, written informed consent was obtained.
Evaluation of Patients' Clinical Status and Case Definition
Patients were evaluated with the Present State Examination (PSE) (11). The PSE is a semistructured interview that gathers information on 140 psychiatric signs and symptoms, covering the 4 weeks before the interview. This information is processed by a computer program called CATEGO. The results for each patient are given in terms of a score on an 8-point scale of caseness, or "index of definition" of psychiatric disorder (12); the ranks range from "no symptoms" (level 1), through "threshold case" (level 5), to "definite case" (levels 6 to 8). The index of definition assigned by the computer program to each patient is arrived at by taking into account the number, severity, and specificity of the signs and symptoms recorded. In this study caseness was defined as a level 4 or higher CATEGO index of definition (12). This level has been shown to have a better correspondence to the DSM-III system as used in primary care settings than does level 5 (13). The concordance (kappa) between level 4 and DSM-III-R caseness was 0.8.
Two psychiatrists were trained in the use of the interview. Their interobserver reliability was as follows: for PSE nonpsychotic items, average kappa=0.83; for deterioration of social functioning, kappa=0.85; and for the presence of psychosocial stressors, kappa=0.95.
Rating of Physicians' Psychodiagnostic Ability
At the end of each office visit the doctor rated the degree of the patient's mental disorder on a 6-point scale (5) ranging from "no disturbance" (score=0) to "severe psychiatric illness" (score=5). This assessment was contrasted with that conducted by one of the two psychiatrists, who administered the PSE at the end of the office visit. Spearman correlation coefficients were obtained for the association between the total score on the PSE and the score on the physician's scale. The physician's psychodiagnostic ability was classified in three levels (good, medium, and poor) according to the correlation coefficients obtained.
The doctors had previously been informed of the aim of the study, but they were not given feedback on their performance until the study was over.
Rating of Clinical Interview Skills
Except for eye contact, the interview skills were rated from video recordings. Eye contact was registered at the time of the interview by two trained observers. The raters were blind to the results of the psychopathological assessment of the patient and to the physician's psychodiagnostic ability. The results presented here are taken from the exploratory part of the interview.
In order to get the physicians used to the presence of the camera, all office visits were videotaped starting 2 days before the research began. Apart from the presence of the observer and the camera, no other changes were introduced into the physician's daily routine.
In order to rate the physician's skills, an observer analyzed the video recordings by using the Physician's Skills Observation Scale. This scale is based on a similar one developed by Goldberg et al. (14). Modifications were introduced on the basis of our experience and descriptions made by others (15, 16). The verbal and nonverbal behaviors that may facilitate or inhibit the patient's expression and the detection of his or her problems by the physician were operatively defined. The Physician's Skills Observation Scale includes nonverbal behaviors (eye contact, posture, note taking, facilitations), interview initiation skills and initial orientation of the patient, type of question (open versus closed, leading questions, and questions with psychological content), verbal facilitations, verbal interruptions, and empathic statements. Nonverbal skills displayed while the physician and patient were speaking were assessed in relation to each utterance. Eye contact was measured on a 4-point frequency scale, and the average score was examined. Posture was measured on a 3-point scale, and the average was determined. An overall score of 10 points was used for interview initiation skills and initial orientation of the patient. The presence of empathic statements during the exploration was measured dichotomously (0=no, 1=yes). The score used for type of questions was the percentage of each type in relation to the total number of questions. The score used for the other skills was the percentage of each type in relation to the total number of utterances. Before skills were analyzed, the instrument was tested by means of trials, and its reliability was improved until kappa coefficients of >0.75 were achieved. The agreement in the number of utterances reached an intraclass correlation coefficient of 0.98. The coefficient for the intraclass correlation between the two observers' scores for eye contact during the office visit was 0.90.
In order to study the influence of interview skills on the ability to identify mental illness independent of other factors, two groups of variables were assessed, some physician related and others patient related. The physician variables included age, sex, and others related to the physicians' training and professional experience: grade obtained in undergraduate studies in psychiatry (1=pass, 4=honors), percentage of time assigned to studying psychiatry and medical psychology in the past year (1=0%, 6=≥25%), participation in training activities in the past year (measured as number of activities, maximum=4), years of clinical experience, and number of office visits per month during the research period. Furthermore, in order to explore the attitude of the physician toward each of the patients interviewed, each doctor completed a Semantic Differential (17) after each office visit. The Semantic Differential was composed of seven scales for rating the patient: pleasant–unpleasant, attractive–repulsive, friendly–unfriendly, interesting–boring, he/she makes me feel at ease–he/she makes me feel anxious, he/she cheers me up–he/she makes me sad, I care for him/her–I don't care for him/her. The factorization of the scores on the scales produced one single factor that accounted for 82.5% of the total variance and corresponded to the evaluative dimension. A "negative attitude" rating for the patient was expressed as the number of scales with scores higher than 4. The patient's age, sex, socioeconomic level as determined with the British classification of the General Register (18), marital status, severity of mental illness (total score on the PSE), and severity of somatic illness (score on item 2 of the PSE) were also registered.
For the study of the physician's psychodiagnostic ability and its association with social and clinical factors, the whole study group was used (N=233). For the study of the association between psychodiagnostic ability and clinical interview skills, all office visits were recorded on videotape, but only a subgroup of 10 visits per physician were analyzed. These were the recordings of the first five patients seen by each doctor who were designated as psychiatric cases (CATEGO level 4) and five recordings chosen at random from the rest of the noncases. Of these 100 recordings, four were excluded because the physicians thought they knew the patients' psychiatric status from external sources of information. In two cases it was not possible to rate eye contact.
Univariate analysis and multiple linear regression analysis (forward stepwise selection; alpha to enter=0.15, alpha to remove=0.20) were conducted. The value of alpha to enter for the stepwise procedure was chosen to be 0.15 on the basis of Monte Carlo studies of stepwise linear regression (19). These authors have shown that the choice of alpha=0.05 is too strict, often excluding important variables from the model.
In the multivariate analysis, the dependent variable of physician's psychodiagnostic ability was rated on a 3-point scale (0=poor, 1=medium, 2=good).
Social and Clinical Characteristics
The study group was collected over 1 year. Sampling took 1 month for eight of the physicians and 2 months for the remaining two. Fifteen patients (6.0%) who fulfilled the inclusion criteria refused to take part in the study. In total, 233 patients were studied. The average age was 46.4 years (SD=16.3); 75.1% of the patients (N=175) were women, 78.1% were married (N=182), and 81.5% were from socioeconomic levels IV to VI (N=190). As for the 10 physicians, five were men and five were women; their average age was 35.9 years (SD=4.0), and their average number of years of clinical experience was 9.0 (SD=4.0).
During the study period the physicians received an average of 31 office visits per day. The mean number of visits performed during the research period did not significantly differ from the number of visits performed in the previous 6 months. The mean duration of the exploratory part of the interview was 3 minutes.
Of the 233 patients, 50 (21.5%) were given a PSE-CATEGO index of definition of 4 or higher. Of the 233 patients, 62 (26.6%) fulfilled DSM-III-R diagnostic criteria: 45.2% for depressive disorders (N=28), 30.6% for anxiety disorders (N=19), 14.5% for adjustment disorders (N=9), 8.1% for somatoform disorders (N=5), and 1.6% for a manic state (N=1).
Physicians' Psychodiagnostic Ability
Three physicians accurately diagnosed psychopathology, i.e., they had ratings that correlated highly (statistically significant Spearman correlation coefficients of >0.4) with those of the psychiatrists using the PSE. For another three physicians, the Spearman coefficients were between 0.0 and 0.4 and p>0.05. Four physicians were considered to have low psychodiagnostic ability, with Spearman correlation coefficients less than 0.0. These three groups showed significant differences in the rate of psychiatric cases identified (PSE-CATEGO index of definition >3 and score on the physician scale >0). Those with good psychodiagnostic ability detected 93.3% of the cases (14 of 15), those with medium psychodiagnostic ability identified 33.3% (five of 15), and those with poor psychodiagnostic ability identified 10.0% (two of 20) (χ2=25.1, df=2, p<0.001).
Clinical Interview Skills and Psychodiagnostic Ability
T1 shows the association between psychodiagnostic ability and clinical interview skills. The majority of skills showed a statistically significant relationship with the physician's ability to diagnose mental illness. Of the 29 interviews for which the physicians were rated as having good psychodiagnostic ability, five (17.2%) contained empathic statements; none of the interviews associated with medium or poor diagnostic ability contained empathic statements (χ2=12.2, df=2, p=0.002). The Physician's Skills Observation Scale includes a number of skills that were not analyzed in this study because the great majority of physicians did not use them. In order to determine which skills were independently and significantly associated with the physician's ability to detect emotional disorders, a multiple linear regression model was fitted, and the following 14 independent variables were initially included in the model: eye contact, note taking, posture (while the physician spoke and while the patient spoke for each of the preceding three variables), facilitations (verbal and nonverbal), nonconstructive verbal interruptions, open versus closed questions, leading questions, questions with psychological content, the presence of empathic statements, and interview initiation skills and initial orientation of the patient. T2 presents the results of the multiple linear regression analysis. The stepwise procedure selected four variables with statistically significant regression coefficients: eye contact and posture while the patient spoke, the absence of nonconstructive verbal interruptions, and the presence of questions with a psychological content. The model specified with these four variables explained 50.9% of the variance and produced statistically significant regression coefficients for all four.
Other Factors and Psychodiagnostic Ability
T3 presents the association of psychodiagnostic ability with the time spent in exploration during the office visit (in seconds) and the rating of negative attitudes toward the patient. Both variables showed statistically significant associations with psychodiagnostic ability.
Psychodiagnostic ability was found to be independent of physician variables (sex, age, percentage of time devoted to studying psychiatry and medical psychology in the past year, participation in training activities in the past year, grade obtained in undergraduate studies in psychiatry, years of clinical experience, and number of office visits per month during the research period) and patient variables (sex, age, socioeconomic level, marital status, PSE total score, and severity of somatic illness).
In order to determine whether the four skills of the physicians that were associated independently and significantly with the physicians' psychodiagnostic ability would maintain such an association after exploration time and the negative attitude rating were controlled for, a stepwise multiple linear regression analysis that included the six variables was carried out. T4 shows the result of this analysis. The stepwise procedure selected five variables associated independently and significantly with psychodiagnostic ability: greater eye contact and a more open and face-to-face posture while the patient was speaking, the absence of nonconstructive verbal interruptions, more time spent on the exploratory part of the interview, and the absence of negative attitudes toward the patient. A model specified with these five variables explained 56.2% of the variance and produced statistically significant regression coefficients for all the variables included.
The main finding of this research was the close association between the physician's interview skills and the physician's ability to identify emotional disorders in patients. The skills associated independently and significantly with ability to identify disorders were eye contact and an open and face-to-face posture while the patient was speaking, the absence of nonconstructive verbal interruptions, and asking questions with a psychological content. The similarities between the pioneering findings of Goldberg and colleagues (5, 6, 14) and our own findings remain despite differences in methods of analysis and cultural differences between the two settings. In their study, the physicians with good ability to identify psychiatric illness showed better active listening behaviors (visual contact, posture, facilitations) and used more open questions and questions with psychological content than did general practitioners with poor ability to recognize psychiatric distress.
Second, we found that the association between skills and psychodiagnostic ability was maintained when other factors related to psychodiagnostic ability were controlled for. In two studies by Goldberg et al. (2, 5, 6), factors other than skills showed independent associations with the ability to identify mental illness (personality factors, academic qualifications, and clinical knowledge). The findings of the present study cannot be compared with the results of the multivariable analysis of Goldberg et al. owing to important differences in the variables studied, in the procedures of multivariable analysis used, and in the nature of the independent variables (in the studies by Goldberg et al., factorial scores instead of individual variables were used).
Third, the inverse relationship between negative attitudes toward the patient and the ability to identify emotional disorders was significant. The preexisting attitudes of the physician may create distance between the physician and the patient and, therefore, inhibit the patient's communication (10). This is a new and interesting finding and should be explored further in the future.
Fourth, in this study we did not find an association between ability to diagnose emotional disorders and two variables related to traditional medical education: the grades awarded in psychiatric subjects during undergraduate medical studies and the amount of time devoted to continuing medical education in psychiatry and medical psychology. This may be due to the fact that the teaching of interview skills to trainees and to general practitioners is seldom practiced in Spain. Alternatively, the lack of a significant association between training and ability to diagnose emotional disorders could also be attributed to the lack of power related to the size of the study group in this study. It has been shown in controlled studies that training may increase the psychiatric knowledge of practitioners working in primary care (20), and this may in turn lead to improved care of patients with mental disorders who are treated in primary care settings (21). Psychiatric knowledge may be a necessary condition, although not a sufficient one, for the proper care of mentally ill patients in general practice (20).
Last, we should point out that the findings of this research underline the need for training physicians in specific interview skills, in order to improve their ability to identify mental disorders in their practices.
The relationship between physicians' skills during office visits and their psychodiagnostic ability will be explored and analyzed in a separate study.
The limitations of this study include the fact that our design did not allow for determining the causal direction of the association between clinical interview skills and psychodiagnostic ability. Experimental studies are required to settle causality.
Also, these findings applied to the first office visit. The possibility that the physician's psychodiagnostic ability improves in subsequent office visits should be considered (22). A longitudinal evaluation could address the analysis of time-scale changes and could also help to identify the physician's skills associated with the patient's compliance, but this type of design might seriously compromise the viability of the study.
Another less relevant although important limitation of the current study is Hawthorn's effect. In observational studies it is difficult not to distort the behavior of the people being observed with regard to their knowledge of the research aims and to the presence of an observer. This effect might have altered the physicians' overall psychodiagnostic ability, but it seems likely that it would have had less effect on the relationship between psychodiagnostic ability and clinical skills.
Presented at II Congreso Nacional de la Sociedad Española de Epidemiología Psiquiátrica, Pamplona, Spain, Sept. 27–29, 1993. Received April 18, 1997; revisions received Aug. 22 and Oct. 8, 1997; accepted Oct. 24, 1997. From the Department of Internal Medicine and Psychiatry, University of Alicante, Alicante, Spain, and the Department of Internal Medicine and Psychiatry, University of Valencia, Valencia, Spain. Address reprint requests to Dr. Girón, Departamento de Medicina y Psiquiatría, Facultad de Medicina, Universidad Miguel Hernández, Carretera Alicante-Valencia s/n, 03550 San Juan (Alicante), Spain. The authors thank Prof. David Goldberg for providing the Scale for Rating Doctor's Behavior and the Scale for Rating General Practice Problem Based Consultation Behaviours.