Structured psychiatric diagnoses and self-rated symptoms in primary care patients on sick leave for common mental disorders: a clinical study

Background: To improve the quality of health care provided to primary care patients with mental disorders, it is crucial to understand more about the mental symptoms that underlie diagnoses on sick leave certificates. This study therefore aimed to: 1) investigate whether diagnoses on sick leave certificates corresponded to the results of a structured psychiatric interview and to self-rated symptom severity and 2) investigate the association between length of sick leave and the diagnoses on sick leave certificates, the diagnoses made in structured psychiatric interviews, and self-rated symptom severity. Methods: The study used data from 480 patients in SAFARI, a study on sick leave in patients with common mental disorders. At baseline, background variables were gathered and structured psychiatric interviews (M.I.N.I.) were performed. Severity of depression and adjustment disorder was assessed via self-rating scales. Data on sick leave were gathered at baseline and at 12 months from the Swedish Social Insurance Agency and patients’ medical records. Results: The diagnostic criteria for depression were fulfilled by a total of 76% of patients with a sick-leave diagnosis of adjustment disorder, 67% with a sick-leave diagnosis of anxiety, and 65% with a sick-leave diagnosis of depression (p=0.04). There was no significant difference in mean net sick leave days between those with a sick-leave certificate diagnosis of adjustment disorder (mean days 119.9), anxiety disorder (107.2), or depression (137.1). However, those with depression diagnosed via structured interview had a shorter mean net sick leave (112.3) than those who did not fulfil the depression criteria (155.9). Symptom severity was strongly associated with net sick leave days; those who rated their depression or adjustment disorder symptoms as more severe had longer net sick leave. Conclusions: Many patients with sick-leave certificate diagnoses of adjustment and anxiety disorders have ongoing depression. Longer sick leave duration was observed in those with adjustment disorder and more severe self-reported symptoms, both of which are appropriate according to Swedish guidelines.

The GP writes a sick leave certificate that includes a main diagnosis, a description of the patient's functional impairment, and a recommended length and degree (percentage of full-time work: 25%, 50%, 75%, or 100%) of sick leave. An administrator at the Swedish Social Insurance Agency assesses and approves or disapproves the recommended sick leave (14).
Research shows that sick leave can facilitate the recovery of people with common mental disorders such as severe depression or stress-induced exhaustion disorder (4,15). However, it can be unhelpful or even harmful for people with mental disorders such as social anxiety and phobias (16,17). The correct diagnosis helps ensure that patients receive the appropriate treatment and the appropriate length and degree of sick leave. It is thus crucial to investigate the association between mental symptoms and the psychiatric diagnoses on sick leave certificates.
Methods used in primary care to diagnose mental disorders, their comorbidity, and their severity are not standardized, and as a consequence, the diagnostic procedure may vary widely between clinics and between practitioners (18)(19)(20). According to a meta-analysis of studies from primary care (21), the inclusion of a structured psychiatric interview (M.I.N.I. Neuropsychiatric Interview) (22) in the clinical assessment of mental disorders can increase the accuracy of differential diagnoses. M.I.N.I. increases both the sensitivity and the specificity of differential diagnoses (21). Taking depression as an example, clinical assessment alone detects approximately 50% of all cases, a number that rises to 90% when M.I.N.I. is included in the assessment (21). However, M.I.N.I. does not assess the severity of disorders. Instead, this is typically assessed with additional symptom rating scales and/or measures of functional impairment (23).
Previous studies show that sociodemographic factors such as sex, educational level, and socioeconomic status (5,6,15,(24)(25)(26)(27)(28) correlate with the length of sick leave. Moreover, research shows that common mental disorders are a risk factor for sick leave, both sick leave for mental and for somatic disorders (2,8,(29)(30)(31)(32). Studies that examine which factors influence the length of sick leave in patients with common mental disorders differ widely in methodology and focus. In summary, some have found an association between specific psychiatric diagnoses such as depression (33) and anxiety (1) and long-term sick leave. Others have found that the severity of symptoms of anxiety and distress is the most important factor associated with longer sick leave (1,16,29,34,35). This study provides an opportunity to examine the results of independent structured psychiatric interviews and self-rated symptom severity assessments in a population of patients on sick leave for common mental disorders.

Aims
This study had two aims. The first was to investigate whether diagnoses on sick leave certificates corresponded to the results of a structured psychiatric interview and to self-rated symptom severity.
The second was to investigate the association between length of sick leave and the diagnoses on sick leave certificates, the diagnoses made in structured psychiatric interviews, and self-rated symptom severity.

Methods
The data in this study were gathered between 2012 and 2017 in the SAFARI project, which has been described in detail elsewhere (36)(37)(38). Briefly, SAFARI investigated sick leave in people with common mental disorders. The current study included SAFARI data collected for two substudies: an RCT in the region of Stockholm (37) and an observational study in the region of Västra Götaland (38). These two regions include rural and urban areas and areas of varying socioeconomic status.

Study population
Our analyses included data from 300 patients from the region of Stockholm and 180 from the region of Västra Götaland. Figure 1 shows the recruitment of participants to the two substudies. Data used in the current study were collected between 2013 and 2016. In the region of Stockholm, participants were recruited from registers at the Swedish Social Insurance Agency and via advertisements in the press. In the region of Västra Götaland, participants were recruited at 28 primary health care centers. Inclusion and exclusion criteria are shown in Table 1.  1 Generalized anxiety disorder was an exclusion criteria only in the region of Västra Götaland. No other anxiety disorders were exclusion criteria in either the region of Stockholm or the region of Västra Götaland. 2 Ongoing psychotherapy was an exclusion criteria only in the region of Stockholm.

Data collection
Data collection is described in greater detail in previous publications (36)(37)(38). In summary, background variables and responses to self-rated symptom severity scales were gathered at baseline.
Structured psychiatric interviews were performed by specially trained health care personnel. Baseline assessments occurred between 14 days and 3 months after the physician completed the sick-leave certificate. Data on length and degree (percentage of full-time work) of sick leave were collected at 12 months.

Variables and instruments
Background variables included self-reported data on employment status, marital status, and level of education.
Treatment data. In both regions, data on prescriptions for antidepressants were gathered via selfreport. In the region of Västra Götaland, prescription data were also gathered from patients' medical records. In the region of Stockholm, treatment also included psychotherapy (acceptance and commitment therapy or ACT) and work interventions, which were part of the RCT.

Statistical analyses
Continuous variables were described with mean, median, and standard deviation. Categorical variables were described using numbers and percentages. Comparisons between groups were performed using student's t-tests for continuous variables and chi-square tests for categorical variables.
Regression analysis was used to determine factors associated with the number of net sick-leave days per year while controlling for confounding. The variable "net sick-leave days" per year was analyzed as a count outcome. To account for model overdispersion, i.e., greater variability than would be expected from a Poisson distribution, negative binomial regression models (44) were used to estimate sick-leave rate ratios (RRs) with 95% confidence intervals. All models were adjusted for age, sex, education, and treatment (antidepressants, psychotherapy, and/or work interventions). Statistical tests were two-tailed, and p-values <0.05 were considered statistically significant. Analyses were carried out with SAS 9.4 (SAS Institute, Inc.).

Results
A total of 480 participants were included in this study, 300 from the region of Stockholm and 180 from the region of Västra Götaland. Baseline characteristics of the participants in the two regions are shown in Table 2. There were no statistically significant differences between the regions in type of psychiatric diagnoses on sick-leave certificates, severity of depression, or severity of adjustment disorder. Significant differences in age and educational level were observed between the two regions.
Participants from the region of Stockholm were 3 years older on average than those from the region of Västra Götaland (p<0.001); 60% in the region of Stockholm and 44% in the region of Västra Götaland had a university education (p<0.009). At the 3-month follow-up, 33% from the region of Stockholm were treated with antidepressants, whereas 50% from the region of Västra Götaland had such treatment (p<0.001). Women were in the majority in both regions (76% in the region of Stockholm and 81% in the region of Västra Götaland).  Table 3 shows the association between the sick-leave certificate diagnoses and the psychiatric diagnoses received after the structured psychiatric interview (M.I.N.I.). The psychiatric diagnoses (M.I.N.I.) differed significantly between the three sick-leave certificate diagnosis groups (adjustment disorder, anxiety, and depression). There were significant differences in the number of diagnoses and the percentage diagnosed with panic disorder, agoraphobia, post-traumatic stress disorder, and bulimia. A total of 65% of the people on sick leave for depression met the criteria for ongoing depression according to M.I.N.I., whereas a higher percentage of those on sick leave for adjustment disorder (76%) and those on sick leave for anxiety (67%) did so (p=0.041). There were no significant differences between the three diagnosis groups in history of depression, recurrent depression, or generalized anxiety disorder.  The analysis of the association between mean symptom severity scale scores and the three sick leave certificate diagnosis groups (data not shown) showed that the severity of the current depressive episode differed across the three groups (p<0.001). Those with a diagnosis of depression on their sick leave certificate had a mean MADRS-S score of 22.8 (SD 7.5), those with a diagnosis of anxiety disorder had a mean score of 21.4 (SD 9.5), and those with a diagnosis of adjustment disorder had a mean score of 19.6 (SD 7.5). The mean severity scores for adjustment disorders (KEDS scores) did not differ between the three diagnostic groups (adjustment disorder: 27.9, SD 8.1; anxiety disorder, 27.6, SD 10.1; depression, 29.1, SD 7.5) (p=0.322). Table 4 shows the association between mean net sick leave days and sick-leave certificate diagnoses, as well as mean net sick leave days and diagnoses made in the structured psychiatric interviews.
There was no significant difference in mean net sick leave days between those with a sick-leave certificate diagnosis of adjustment disorder (mean net sick leave days, 119.9), anxiety disorder Adjusting for age, sex, level of education, and treatment did not change the pattern of the results; crude and adjusted rate ratios (RRs) are shown in Table 4.

Summary of results
Many patients with sick-leave certificate diagnoses of adjustment and anxiety disorders had ongoing depression. The diagnostic criteria for depression were fulfilled by 76% of patients with a sick-leave diagnosis of adjustment disorder, 67% with a sick-leave diagnosis of anxiety, and 65% with a sickleave diagnosis of depression. Sick-leave certificate diagnoses of anxiety disorders seemed to be associated with diagnoses of anxiety disorders made in the structured psychiatric interviews (M.I.N.I).
There was no significant difference in net sick-leave days between people with sick-leave certificate diagnoses of adjustment disorder, anxiety, or depression. Depression diagnoses made in the structured psychiatric interviews were associated with shorter net sick leave than adjustment disorder diagnoses made in the interviews. High scores on either the MADRS-S or the KEDS symptom severity scale were associated with longer net sick leave.

Comparison of current study results with those of previous studies
The overlapping, dynamic, and sometimes chronic nature (12,17,32,(45)(46)(47)(48) of common mental disorders seen in primary care patients may underlie several of the findings of the current study.
Comorbidity could help explain why depression was found in so many patients on sick leave for adjustment and anxiety disorders. It could also help explain why previous depression, recurrent depression, and generalized anxiety disorder were found in similar levels in all three groups (those with sick leave certification for adjustment disorder, anxiety disorder, or depression).
The dynamic nature of common mental disorders may also have contributed to the relatively low number of patients on sick leave for depression who had ongoing depression according to the structured interview. Perhaps some patients had recovered from their depression by the time of the structured interview, which took place 14 days to 3 months after patients received sick leave certificates. The delay between sick-leave certification and the interviews may also help explain why large numbers of the patients on sick leave for adjustment and anxiety disorders had ongoing depression according to the interview. Patients on sick leave for these disorders might have developed depression during the interval between certification and the structured psychiatric interview. Sick leave itself is a risk factor for depression (2,15), especially in people with anxiety disorders (2,8,17). Burnout also predicts depressive symptoms (49).
The infrequent use of structured psychiatric interviews for mental disorders in primary care (18,19,21,47,50) could also help explain the differences we observed between the diagnoses on sick leave certificates and the diagnoses made in the structured interviews. A previous review found that clinical assessment alone detects approximately half of all cases of depression, a number that rises to 90% when a structured psychiatric interview (M.I.N.I.) is added to the diagnostic procedure (21,51). A study in primary care has found that in the absence of a structured psychiatric interview, some mental disorders (e.g., bulimia, obsessive-compulsive disorder) can go undetected or be mistaken for other disorders with similar clinical presentations (51).
Although diagnoses on sick leave certificates are supposed to guide the length and degree of sick leave, we found no association between the two. We found, however, that the diagnoses made in the structured psychiatric interviews were associated with the length and degree of sick leave. Patients who fulfilled the M.I.N.I. criteria for ongoing depression had significantly fewer net sick leave days than those who did not. Additionally, few patients in either region in the current study had sick-leave certificate diagnoses of anxiety disorders, which suggests that the GPs were following recommendations that short and preferably no sick leave should be given to patients with anxiety disorders (52). Finally, the structured psychiatric interviews showed that people with adjustment disorder had the longest sick leave. This is in accordance with Swedish recommendations, which state that long sick leave may be necessary for these patients (52).
More severely ill patients received longer sick leave, which is in keeping with national guidelines (52).
In all groups, participants who scored high on KEDS and/or MADRS-S had longer net sick leave than those who had low scores on these instruments. The same pattern has been found in previous studies (1,16,29,34,35,53), which suggests that symptom severity may be the factor with the greatest influence on length of sick leave.

Limitations and strengths
This study had a number of limitations. One was that differences in recruitment in the regions of Stockholm and Västra Götaland led to study populations that differed in age, educational level, and treatment with antidepressants. In Stockholm, participants were recruited primarily via invitations from the Swedish Social Insurance Agency. Approximately 10% of those contacted by letter responded that they were interested in participating, which means that this was a highly self-selected group. Additional participants were recruited via advertisements in the press, which may have resulted in another highly self-selected group. In the region of Västra Götaland, rehabilitation coordinators at primary health care centers asked patients who were on sick leave to participate in the study; 21% agreed to participate, and these people may also have differed from those who declined to participate.
A further limitation was the delay between the sick leave diagnoses and the structured psychiatric interviews. As noted previously, common mental disorders seem to overlap and affect one another over time (1,3,29,32,(45)(46)(47)54). It is possible that our results would have differed if all the structured psychiatric interviews had been conducted as part of or immediately after the sick-leave certificate diagnosis.
Finally, our study did not measure self-rated anxiety symptoms, which means that we were not able to compare anxiety symptoms in our study population with length of sick leave. In at least one other study, symptoms of anxiety were the most important predictor of length of sick leave (29).
A strength of this study was that the assessments with structured psychiatric interviews and symptom severity scales were performed by assessors other than the clinicians who performed the sick leave certification. This may have diminished assessment bias. Additionally, the study population was drawn from two regions of Sweden and included both rural and urban areas and areas of varying socioeconomic status, which could increase generalizability.

Conclusion
Many patients with sick-leave certificate diagnoses of adjustment and anxiety disorders had ongoing depression, which likely reflects the overlapping and dynamic nature of common mental disorders.
There was no difference in sick leave duration by the diagnoses on sick-leave certificates, possibly reflecting the rarity of structured psychiatric diagnoses for mental disorders in primary care. However, diagnoses made in structured psychiatric interviews and the severity of patients' symptoms were associated with net sick leave, meaning that regardless of the diagnoses on the sick-leave certificates, overall, the length and degree of sick leave prescribed by GPs were appropriate.

Consent for publication: Not applicable.
Availability of data and materials: The data used in these analyses were gathered for the Return to Work: Promoting Health and Productivity in Workers With Common Mental Disorders (SAFARI) Study, ClinicalTrials.gov Identifier: NCT01805583. For information about data availability and to contact the party responsible for the study in the region of Stockholm, see ClinicalTrials.gov, and for Västra Götaland, see https://www.researchweb.org/info/index.php/vgr/project/207801.

Competing interests:
The authors declare that they have no competing interests.

Funding:
The SAFARI study was funded by grants from Stockholm County, Region Västra Götaland, and the Swedish Social Insurance Agency. These funding bodies played no role in the design of the study; the collection, analysis, and interpretation of data; or in writing the manuscript.

Authors' contributions:
The study was conducted as part of SWH's doctoral project. The data used in the study were gathered as part of the SAFARI study. AN, CB, and JW supervised the work. SWH analyzed the data and prepared the draft manuscript with the support of all the other authors. All the authors approved the final manuscript.