| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
ORIGINAL ARTICLES |
From the Department of Psychiatry, Dalhousie University, Halifax, Canada.
Address correspondence and reprint requests to Stephen Randolph Kisely, MD, Departments of Psychiatry, Community Health and Epidemiology, Dalhousie University, Room 425, Centre for Clinical Research, 5790 University Ave., Halifax, NS B3H 1V7, Canada. E-mail: Stephen.Kisely{at}cdha.nshealth.ca.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODRESULTSDISCUSSIONNOTESREFERENCES |
|---|
Methods: A total of 1252 psychiatric cases were recruited using a 2-stage design from 5447 subjects presenting for primary care in 14 countries. Cases were assessed at the time of screening and 1 year subsequently. Information on physical, psychiatric, and social status was obtained using the Composite International Diagnostic Instrument adapted for use in primary care (CIDI-PHC) and the Groningen Social Disability Schedule (GSDS). Assessments of psychiatric morbidity were also obtained from the patients’ family practitioners.
Results: Medically explained somatic symptoms were strongly related to psychiatric outcome 1 year later. Whereas just over one half of patients (614 of 1078) with 4 or less medically explained symptoms had recovered from a psychiatric disorder, the percentage recovery fell to 38% (67 of 174) in those with 5 or more medically explained symptoms. Patients with 5 or more medically explained symptoms had a 70% increase in risk of remaining a psychiatric case 1 year later after controlling for demographics, country, initial severity of psychiatric disorder, medically unexplained somatic symptoms, and social disability.
Conclusion: Physical ill health is independently associated with psychologic outcome 1 year after a patient has been seen. The needs of these patients should receive greater attention.
Key Words: medically explained symptoms • psychiatric recovery • primary care
Abbreviations: PPGHC = Psychological Problems in General Health Care; MES = medically explained symptoms; MUS = medically unexplained symptoms; GHQ-12 = 12-item General Health Questionnaire; CIDI-PHC = Composite International Diagnostic Interview–Primary Health Care; GSDS = Groningen Social Disability Schedule; FP = family physician; SPHERE = Somatic and Psychological Health Report.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODRESULTSDISCUSSIONNOTESREFERENCES |
|---|
In terms of outcome, the findings are more equivocal. In 1 study that investigated possible predictors of psychiatric outcome 1 year later, including the role of medical illness, only initial severity of psychiatric symptoms and quality of social life remained significantly associated with outcome on discriminant function analysis (8). The psychiatrists who made the ratings of medical illness did so on clinical grounds and not according to the number of medically explained symptoms. It may be that a certain number of medically explained symptoms (MES) are required to have an effect on psychiatric outcome. In our previous study of 7 family practices in Manchester (Great Britain), we found that medically explained somatic symptoms were strongly related to psychologic outcome 1 year later, but only over a certain threshold (9). Whereas just over one half of patients with no MES had recovered from a psychiatric disorder, the percentage recovery fell to 41% in those with 1 to 4 medically explained symptoms and only 21% in patients with 5 or more medically explained symptoms. This effect persisted when a correction was made for initial severity of psychiatric disorder and the presence of associated social disability.
A number of factors limit the generalizability of these findings. Although research has shown that comorbid disorders occur frequently, and across the entire lifespan, most studies report on medical comorbidity in smaller clinical samples. Often the samples have been restricted by age, diagnosis, or by the clinic setting under study (10). The prevalence of psychologic symptoms in primary care settings has been shown to vary by as much as 9-fold across different locations (7), with the result that existing research may not be applicable to all settings.
Our Manchester study formed part of the World Health Organization Collaborative Study of Psychological Disorders in General Health Care Settings (PPGHC), which is described more fully elsewhere (11). The prevalence, management, and outcome of common psychologic disorders in primary care patients were investigated in 15 centers from 14 countries. The aim of the present study was to determine whether there was a similar association between the number of medically explained symptoms and recovery from psychiatric illness 1 year later when we included all 15 centers. This would suggest that the association is present across a wide range of cultures and health systems and not confined to British family practice. We also aimed to determine the relative importance of MES to psychiatric outcome in relation to other possible determinants of psychiatric morbidity such as sociodemographic variables, comorbid medical diagnoses, measures of social disability, and medically unexplained symptoms (MUS).
|
METHOD |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODRESULTSDISCUSSIONNOTESREFERENCES |
|---|
A 2-stage sampling strategy was used. Initially, consecutive adult primary care attenders aged between 16 and 65 were screened using the General Health Questionnaire (GHQ-12). A stratified random sample, consisting of 10% of those with low GHQ scores, 35% of those with scores in the medium range, and 100% of those with high GHQ scores, were selected for a diagnostic assessment that included the following measures: the primary care version of the Composite International Diagnostic Interview (CIDI) termed the CIDI-PHC and the Groningen Social Disability Schedule (GSDS), a semistructured assessment of occupational disability. Subjects were also asked about the presence of specific physical diagnoses using a checklist. For each patient selected for this diagnostic assessment, the treating family physician (FP) completed a brief questionnaire regarding the presence of physical and psychologic illness. The interview also included a specific checklist of 20 chronic medical conditions (11).
All patients found to have significant psychologic problems, along with a 20% random sample of "noncases," were contacted for follow-up assessments after 12 months. Subjects were contacted by telephone, when possible, or by letter to arrange an appointment. If subjects failed to keep a previously booked appointment, 2 further attempts were made to arrange another assessment. The 12-month assessment consisted of a repeat administration of the CIDI-PHC.
Trained research workers carried out both assessments. Good interrater reliability had previously been demonstrated in a reliability study (12). Written informed consent was obtained after the purpose and requirements of the study has been fully explained. Ethical approval for the study in each center was obtained from the relevant governing body (12)
CIDI-PHC has been derived from version 1 of CIDI, which covers a wide range of substance use and psychiatric disorders but not adjustment reactions. The primary care version contains sections on somatization, anxiety, depression, hypochondriasis, and neurasthenia. It does not cover low-prevalence disorders such as schizophrenia, other psychosis, or bipolar affective disorder, which were not the focus of this study. The use of the CIDI-PHC in physically ill patients has been described elsewhere (6). The advantage of this instrument is that it allows the assessment of both medically explained and unexplained somatic symptoms in the preceding 4 weeks. For each symptom elicited, subjects were asked for details of diagnosis, investigations, and treatment in a highly structured way using a flowchart to establish the presence of an organic illness (13).
Somatic symptoms were only coded as being the result of a physical illness or injury if a doctor had been consulted for the symptom and if that doctor had given the patient a definite diagnosis, or if there had been any abnormalities reported on examination or on investigations. In the absence of a definite diagnosis, or abnormalities on examination or investigation, the symptom was coded as being nonorganic in etiology. Symptoms not leading to healthcare use, the taking of medication more than once, or marked interference with normal activities were excluded. Ratings of MES were therefore made very cautiously, with the default being that they were medically unexplained.
In addition, somatic symptoms suggestive of a physical illness were only considered to be medically explained if the complaint had always been the result of the stated physical condition; otherwise, it was recorded as being nonmedically explained. Finally, a medically qualified member of the interviewing team also reviewed ratings.
Using the same computer algorithm that was derived from the CIDI, it was possible to calculate the overall medically explained and unexplained symptom scores as well as to classify patients based on symptom counts for both explained and nonexplained symptoms: subthreshold symptoms (0–4) or moderate to severe symptoms (5 symptoms and above).
Analysis
The prevalence of psychiatric disorder was determined by computer algorithm using International Classification of Diseases, 10th Revision (ICD-10) criteria. Patients with alcohol dependence alone were not included as psychiatric cases. The presence of significant physical illness was defined using the same computer algorithm.
We assessed physical ill health in 3 ways: medically explained symptom count as measured by the CIDI, FP rating of the presence of physical illness, and the presence of chronic physical disease as reported by participants.
Patients who were not successfully followed up were compared with the sample that were assessed at 1-year follow up in terms of sociodemographic characteristics, severity of psychiatric disorder, and medically explained symptoms using the chi-squared or t test as appropriate.
The association between significant physical illness and continued psychiatric caseness was assessed using descriptive statistics such as odds ratio, chi-squared test, and the Student t test. Statistical significance was tested using p values and 95% confidence intervals. Possible explanations and confounding variables in any association were examined using the same tests of significance.
Logistic regression analysis was used to further examine those variables, including medically explained somatic symptoms, that had been previously identified as being associated with psychiatric outcome using univariate analysis. All these variables were entered into the model.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODRESULTSDISCUSSIONNOTESREFERENCES |
|---|
The patients who were lost to follow up did not differ from subjects who completed the follow-up interview in terms of the following: age, sex, duration of symptoms, medically explained somatic symptoms, diagnosis by the treating family physician of physical disorder, or the presence of self-reported chronic physical illness (data available at http://www.geocities.com/skisely/somatic.html). They were significantly more likely to be single (43% vs. 39%, chi-square = 6.3, df = 1, p = .01), not in full-time work (60% vs. 53%, chi-square = 16.9, df = 1, p <.001), and diagnosed by the treating FP as having a psychiatric disorder (39% vs. 34%, chi-square = 10.2, df = 1, p = .001). They also had completed less years of schooling (means of 8.6 vs. 9.8, t test = –7.39, df = 4416, p <.001) and reported a greater number of psychiatric symptoms (means of 21.8 vs. 15.5, t test = 13.8, df = 4442, p <.001).
The Association Between Physical Ill Health and Psychiatric Recovery 1 Year Later
Of the 3201 subjects who were successfully followed up, 1252 had been psychiatric cases at baseline assessment. Three fourths (n = 907) were female and 43% (n = 532) not married. Of these, 681 (54%) had recovered from their psychiatric symptoms.
The presence of MES was significantly associated with lower rates of psychiatric recovery. Whereas just over one half of patients (614 of 1078) with 4 or less MES had recovered from a psychiatric disorder, the percentage recovery fell to 38% (67 of 174) in those with 5 or more MES (chi-square = 19.8, df = 1, p <.0001). This was also seen for the 2 most common ICD-10 diagnoses in the sample: depression (60% of cases) and generalized anxiety (35% of cases). Seventy percent of patients with 4 or less MES recovered from depression 1 year later (433 of 628) as opposed to 52% (53 of 101) of those with 5 or more MES (chi-square = 9.9, df = 1, p = .002). Seventy-five of patients with 4 or less MES recovered from generalized anxiety 1 year later (278 of 373) as opposed to 52% (39 of 68) of those with 5 or more MES (chi-square = 7.6, df = 1, p = .006).
We found similar but less marked patterns in the case of the other 2 measures of physical ill-health, which did not always reach statistical significance. In the case of FP-treated physical caseness, the difference in recovery only reached statistical significance for depressive illness and anxiety disorder. Only 62% recovered from depression 1 year later (241 of 383) as opposed to 71% (235 of 332; 241 of 383) of those who were not physical cases (chi-square = 4.5, df = 1, p = .03). Sixty-seven percent recovered from generalized anxiety 1 year later (164 of 245) as opposed to 78% (141 of 180) of those who were not physical cases (chi-square = 6.1, df = 1, p = .01).
In the case of self-reported chronic physical illness, only 52% recovered psychiatrically 1 year later (449 of 857), as opposed to 60% (232 of 395) of those with no self-reported illness (chi-square = 4.1, df = 1, p = .04). Similar trends were seen for depression and anxiety but they did not reach statistical significance.
Additional Predictors of Psychiatric Outcome 1 Year Later
Table 1 shows baseline sociodemographic and clinical factors that were investigated as possibly being associated with continued psychiatric caseness at 1 year. Psychiatric caseness did not include alcohol dependence. Initial severity of psychiatric symptoms was determined using the psychiatric symptom count at the initial interview. Scores were divided into low and high scores about the median score (20/21). This count included psychologic symptoms of psychiatric disorder but not MUS, which were considered separately. We also divided duration of presenting symptoms about the median score (8 months/9 months). As we described in the previous section, the 2 most common ICD-10 diagnoses in the sample were depression (60% of cases) and generalized anxiety (35% of cases). However, patients with these disorders had other comorbid psychiatric disorders such as panic disorder and dysthymia. These sometimes overlapping conditions are also listed in Table 1.
|
In terms of demographic factors, fewer years of education and not being in full-time employment were associated with lower rates of recovery at 1 year (Table 1). Sex and marital status were not associated with psychiatric outcome. Aside from physical ill health, the following clinical features as measured by standardized instruments were also associated with continued psychiatric morbidity 1 year later: more than 4 MES, more than 4 MUS, high initial psychiatric score, depression, dysthymia, generalized anxiety disorder, panic disorder, agoraphobia, somatoform disorder, and alcohol dependence (Table 1). These results were confirmed by similar results in terms of the presence of self-reported chronic physical disorder, and FP-assessed physical and psychiatric cases (Table 1).
Only 7 self-reported physical diagnoses were present in sufficient numbers to allow for statistical analysis: hypertension, diabetes, arthritis, cardiovascular disease, chronic obstructive pulmonary disease, peptic ulcer, and asthma. All but chronic obstructive pulmonary disease, asthma, and diabetes were associated with the persistence of psychiatric caseness at follow up (Table 1). Physical and social disability as measured by standardized instruments was also associated with subsequent psychiatric morbidity (Table 1).
The prescription of psychotropic medication was associated with remaining a psychiatric case at follow up, presumably because the unadjusted analysis did not take into account that the more severely psychiatrically ill would have been more likely to receive treatment. We found the same result when we specifically considered the prescription of antidepressants (Table 1). Only 6% of the sample was referred to a mental health professional and this made very little difference to outcome (Table 1).
There was considerable variation between centers in psychiatric outcome 1 year later (Table 2). Patients in Shanghai had the highest recovery rates, with only 22% remaining psychiatric cases at follow up. Subjects in Santiago recovered the least, with 62% remaining psychiatric cases 1 year later. These differences reached statistical significance (chi-square = 46.4, df = 14, p <.0001).
|
Because of the low number of participants who were followed up in some on the centers (Table 2), it was not possible to assess any differences between centers in the association between medically explained symptoms and psychiatric outcome. Where there were sufficient numbers to assess statistical significance, there was always an association between psychiatric outcome and the presence of more than 4 MES, this being most marked in Ankara (data available at: http://www.geocities.com/skisely/centdiff.html).
Multivariate Analysis
Adjusted odds ratios were calculated by fitting a logistic regression that included all variables that had previously been identified as being associated with psychiatric outcome. This included sociodemographic factors, measures of physical and psychologic health, disability, and prescription of psychotropic medication. We also controlled for study center given the significant differences in the number of psychiatric cases at 1-year follow up (Table 2).
We undertook 2 separate analyses. In 1, we entered the total number of medically explained and unexplained somatic symptoms as continuous variables. In the other, we used the dichotomous variables shown in Table 1.
Table 3 shows the variables that were independently associated with continuing psychiatric caseness. Severity of initial psychiatric symptoms emerged as the strongest predictor of continued psychiatric caseness (p = .0001).
|
Subjects with 5 or more medically explained somatic symptoms at baseline interview were 70% more likely to remain a psychiatric case 1 year later. Medically unexplained somatic symptoms showed a similar affect but only at the higher threshold of more than 6 symptoms (Table 3). Demographic variables associated with subsequent psychiatric morbidity included older age and less years of education (Table 3). The following ICD-10 diagnoses were also associated with a worse outcome: depression, dysthymia, generalized anxiety, agoraphobia, mixed depression and anxiety, and somatoform disorder.
Factors such as not being in employment, other ICD-10 diagnoses, and self-reported chronic physical illness were not significantly associated with outcome after controlling for the other variables included in the model. Within the group reporting chronic physical illness, none of the specific physical diagnoses remained significantly associated with psychiatric outcome when they were added into the model. Neither did duration of symptoms or the prescription of psychotropic medication.
We found the same results when we entered the total number of medically explained and unexplained somatic symptoms as continuous variables (data available at: http://www.geocities/skisely/table4.html) and when using forward stepwise regression (data available at: http://www.geocities.com/skisely/stepsom.html). We also found the same results when we only considered the prescription of antidepressants rather than prescription of any psychotropic drug.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODRESULTSDISCUSSIONNOTESREFERENCES |
|---|
There are several limitations of the study that might affect the generalizability of our findings. One is that our study was restricted to patients aged between 18 and 65 years, and the results may not apply to those at the extremes of life who often have the highest rates of attendance in primary care. Another is the 1243 patients (27%) who were lost to follow up. This rate was similar to the 31% who were lost to follow up in our Manchester study (9). Although they were not significantly different from the sample who completed the 1-year assessments in terms of age and sex and the presence of physical ill health as measured by medically explained symptom count, FP ratings, or self-reported chronic conditions, they were significantly different in other areas. This included marital, educational, and employment status, and greater psychiatric morbidity at baseline assessment as determined by standardized instruments and treating FP. We attempted to control for some of these differences using multivariate analyses, but this would not entirely eliminate follow-up bias.
The severity of physical ill health was assessed by interview rather than by the use of diagnostic tests, and it is possible that the observed association between physical and psychologic morbidity, as determined by the research interview, could have been the result of the misclassification of physical and psychologic symptoms. Information on specific physical diagnoses was collected from subjects using a checklist.
On the other hand, somatic symptoms were only coded as being related to physical illness in the presence of a clear diagnosis, or reported abnormalities on examination or investigation. Ratings of medically explained symptoms were therefore made very cautiously, with the default being that they were medically unexplained. This method should have increased the likelihood that the medically explained group was relatively homogeneous, MES being the primary focus of interest in this study. It was the medically unexplained group that was more likely to contain subjects who had had a medically explained symptom at some stage, but who had also had the same symptom when this was not medically explained.
In addition, medical diagnoses were confirmed through medical review of the interview schedules and FP case notes, as well as discussion with the treating doctor in the case of any doubt. Nevertheless, there will inevitably be a blurred boundary between medically explained and unexplained symptoms. Symptoms such as sleep dysfunction, anorexia, anergia, and reduced libido can be explained by psychiatric disorder, by physical comorbidity such as leukemia or coronary heart disease, or by interplay of the 2 processes. Recent reports have suggested that systematic research tools for chart rating can help to distinguish between medically explained and unexplained symptoms (14).
Another limitation was that information on the presence of specific chronic diagnoses was only collected from subjects, and not from treating physicians. Neither were we able to investigate the effect of duration of untreated symptoms on psychiatric outcome.
This study did not consider low-prevalence disorders such as schizophrenia, other psychosis, or bipolar affective disorder. However, it is possible that some of the depression cases really had bipolar disorder.
Comparisons With Previous Studies
Previous studies have demonstrated an association between the number of somatic symptoms and psychiatric disorder, the threshold varying from 4 to 6 depending on study and gender (2–7). There are less data on the effect of somatic symptoms on recovery from psychiatric disorder. In a previous paper, restricted to just 1 center in the PPGHC study (9), we showed that 5 or more medically explained symptoms were associated with significantly lower rates of recovery. The present study extends this work to 15 centers in 14 countries and reports similar findings.
Other factors associated with a worse psychiatric outcome 1 year later included not being in employment, greater severity of initial psychiatric symptoms, depression, anxiety, a diagnosis of generalized anxiety, and social disability. Sociodemographic variables such as age did not affect subsequent psychiatric status.
The 2 other measures of physical ill health (FP ratings and self-reported chronic conditions) did not show an independent association with worse psychiatric outcome at 12-month follow up. This is consistent with another study in primary care that found that physician-rated physical illness did not predict psychiatric caseness 1 year later (8). One difference between that study and ours is that Mann et al. (8) used research psychiatrists rather than FPs to assess the presence of physical illness.
One explanation is that our 3 different measures (medically explained symptom count, FP ratings, and self-reported chronic conditions) all measure different aspects of physical ill health and somatic distress, including the overall severity and type of medical illness. Our finding that symptom count had a stronger association with psychiatric outcome than either self-reported medical diagnosis or physician ratings of medical disease severity suggests that medical comorbidity acts through behavioral pathways like symptom count rather than biomedical pathways.
We were surprised by the lack of difference between medically explained and unexplained symptoms in predicting psychiatric outcome. Cross-sectional studies of inpatients and outpatients attending general hospital settings have suggested that patients with somatic symptoms for which no organic cause can be found report significantly more psychiatric morbidity than patients who have a clear organic cause for their complaint (15–19). This was confirmed in our cross-sectional comparison of medically explained and unexplained somatic symptoms from the 15-center PPGHC study (7).
It might have been expected that medically unexplained somatic symptoms would therefore be associated with a worse psychiatric outcome 1 year later than those with medically explained symptoms. In fact, although medically explained and unexplained symptoms were independently associated with psychiatric caseness, any difference between the effects of the 2 sets of symptoms disappeared. This mirrors the results of our smaller study from Manchester (9). Previous findings that the presence of MUS leads to greater subsequent psychiatric morbidity may be because these subjects were inpatients (19). It is therefore possible that MUS in general practice have a better psychiatric outcome. Another explanation might be the difficulty in separating out medically explained and unexplained symptoms in this population, because these may be intertwined conceptually and etiologically.
It may be important to distinguish between the diagnosis and prognosis of mental disorder. Medically unexplained somatic symptoms may show a specific association with anxiety or depressive disorder (15,20). In contrast, we find that somatic symptoms—whether medically explained or not—are associated with the persistence of mental disorder.
Implications of the Study
Patients with coexisting physical and psychiatric symptoms represent a significant source of continuing morbidity in general practice. It appears that it is the number of MES rather than their chronicity or the specific diagnosis that affects outcome. Furthermore, this effect is seen across a wide range of countries, health systems, and cultures. Screening and intervention should be particularly targeted at these patients. This might include the education of FPs and members of consultation–liaison psychiatric teams in primary care. Awareness could be increased through including physical and psychiatric comorbidities in national education programs for FPs such as the Somatic and Psychological HEalth Report (SPHERE) national mental health project in Australia (21). This aims to increase FPs’ identification, effective treatment, and management of common psychologic disorders such as depression, anxiety, and somatic distress. Aside from education, closer collaboration between FPs and psychiatrists through such initiatives as Shared Mental Health Care would also be indicated (22). Joint management plans can be fostered through psychiatrists providing consultation and treatment in the FP’s office to ensure that all biopsychosocial aspects of management are considered.
The growth of liaison psychiatry within general hospitals has reflected a growing recognition of the special problems faced by patients who have both physical and psychiatric illness. In contrast, consultation–liaison psychiatry in general practice has traditionally been concerned with psychiatric morbidity alone, and not focused on the problems of patients with psychiatric and physical comorbidity. As demographic changes in Europe, North America, and Japan will mean that an increasing proportion of the population will be elderly, the problems of physical and psychiatric comorbidity will become an increasingly important public health issue (9).
|
NOTES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODRESULTSDISCUSSIONNOTESREFERENCES |
|---|
This paper is based on the data and experience obtained during the WHO project on Psychological Problems in General Health Care, a project sponsored by the World Health Organization and the participating field research centers.
DOI:10.1097/01.psy.0000149280.88430.7c
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODRESULTSDISCUSSIONNOTESREFERENCES |
|---|
This article has been cited by other articles:
 |
|
 |
|
  C. M. DICKENS, L. McGOWAN, C. PERCIVAL, B. TOMENSON, L. COTTER, A. HEAGERTY, and F. H. CREED Contribution of depression and anxiety to impaired health-related quality of life following first myocardial infarction The British Journal of Psychiatry, October 1, 2006; 189(4): 367 - 372. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
