Background
GPs are usually confronted with a patient population presenting (unspecific) symptoms and unselected health care problems [
1] and patients with medically unexplained symptoms (MUS) are numerous in primary care [
2,
3]. Up to 20% of patients visiting the GP suffer from somatic symptoms without a sufficient organic explanation [
4,
5]. There are many terms used apart from MUS, like “bodily distress syndrome” [
6], “functional symptoms” and “somatic symptom distress” [
7,
8] and critique about the ambiguity and the pejorative connotation of the term MUS is often expressed [
9]. Anyway, this term was chosen as it is widely used in research literature and corresponds to the definition of somatoform disorders in the International Classification of Diseases, 10. Revision (ICD-10, [
10]) and Diagnostic and Statistical Manual of Mental Disorders (DSM) IV [
11], albeit the discussion is ongoing and fueled by a great diverseness of research literature [
12,
13]. Not all patients with (severe) MUS will eventually be diagnosed with a psychiatric somatization disorder (ICD-10 F45.x), as preferences for diagnoses differ according to the specialty [
14]; medical specialties prefer functional somatic syndromes [
6] like fibromyalgia, chronic fatigue syndrome and chronic benign pain syndrome [
15]. There is evidence that coding might be based upon the specialty (e.g., somatic or psychiatric) of the treating physician [
16] and might differ between primary and secondary care [
17]. The classification of somatoform disorders has been revised in the past [
8] and will experience further revisions with the publication of the ICD-11 [
18].
In health care systems all over the world it is mandatory to record a patient’s condition (symptoms or diseases) by coding it according to different coding systems, for example, via the ICD-10 [
10] or the International Classification of Primary Care (ICPC 2-PLUS, [
19]) both in primary and secondary care. It is widely considered indispensable to code with high validity and integrity to deliver high quality, efficient and safe care for patients [
13,
20]. The underlying assumption that better coding of information results in better care for patients [
21‐
23] is not challenged very often despite not being clearly proven. Reliable data on incidence and prevalence of diseases are also important for research purposes. Researcher need reliable data on the morbidity of a population to, for example, design and evaluate health care interventions, evaluate quality of care and monitor changes in longitudinal studies [
24‐
26]. Other purposes are epidemiology, outcome studies and the identification of suitable individuals to participate in clinical trials [
27]. In contrast, studies show that the reliability of coded diseases in primary [
24,
28‐
30] and secondary [
26,
31] health care settings is often poor, including for example the phenomenon of “billing diagnoses” [
24] which have their origin in the German Medicinal Products Directive and are used to legitimate certain prescriptions. Furthermore epidemiological data and routine data on functional and somatoform disorders are not congruent with each other [
16].
In Germany it is mandatory to code all reasons for consultations and diagnoses for which a contracted physician performs diagnostic or treatment services in an ambulatory setting according to the ICD-10-GM (German Modification, [
32]). As a result, diagnostic or treating efforts cannot be reimbursed without a legitimate diagnosis. Four-digit-coding (e.g., F45.9) is sufficient for German general practitioners (GPs), but it is mandatory to distinguish between “tentative diagnosis”, “(asymptomatic) condition after a confirmed diagnosis”, “excluded diagnosis” and “confirmed diagnosis”. In Austria the ICD-10 is also used in primary care. Other countries like for example Belgium, Denmark, France and many more use the ICD-10 in secondary care, but the ICPC, a system to classify reasons for encounters, problems/diagnosis managed and interventions performed, in primary care [
33].
The ICD-10 allows “double-coding” according to etiology and localization of diseases. That means, for example, the same clinical phenomenology without distinct medical explanation (e.g., MUS) can be coded using organ-related codes as a functional disorder (e.g., K58.0 irritable bowel syndrome) or as mental disorders using codes from chapter F on mental and behavioural disorders (e.g., F45.32 somatoform autonomic dysfunction lower gastrointestinal tract) [
16]. Patients with MUS having primarily physical rather than psychological symptoms usually consult general practitioners rather than mental health care specialists first [
8], however, somatoform categories are used less often by general practitioners [
8] and other clinicians (underassignment, p. 143 [
34]). Diagnoses not adequately adhering to guidelines or diagnostic frameworks may hinder communication between primary and specialized care [
17] and lead to inadequate treatment [
34]. Despite that, one may assume that patients do receive treatment even without a diagnosis of a somatoform disorder, since the management of unspecific symptoms is a core competency of general practitioners [
1].
General practice is characterized by a (long-term) social interaction between a patient and the treating general practitioner and characterized by uncertain diagnoses [
35] while the act of coding implies a diagnostic certainty, which is hardly obtained with MUS or somatoform disorders. Scobie et al. [
36] found that mental and psychological morbidities were coded less often as GPs reported them to be difficult to code. In addition, many health care systems seem to be biased; tolerating the under- and misdiagnosis of mental disorders, while deeming it unacceptable to overlook a physical diagnosis [
36]. Not all GPs deem the classification of diseases as (very) important in respect to the care of patients [
26]. Conditions with obvious diagnostic features have a higher rate of recording than conditions with more subjective criteria [
37]. There is evidence that GPs rarely code somatization syndromes in daily practice [
8,
38]. Not only patients fear stigmatization through psychiatric diagnoses [
8,
39], but GPs could also refrain from recording “stigmatizing” diagnosis to protect the patient from suffering negative societal and economic consequences [
8,
39].
The reasons for not recording MUS/somatoform disorders are not clear yet [
34] and need to be evaluated in qualitative research designs. There is a lack of studies on German GPs’ coding behavior especially in respect to MUS or somatoform disorders. Assuming that correct classification and coding of symptoms and diseases are an important stepping stone for adequate management and treatment, and essential for research purposes, insights into these processes could help identifying problematic areas and possible solutions. Therefore, our study aims to explore German GPs’ views on and approaches to coding and reasons for non-recording of MUS or somatoform disorders.
Methods
This study is part of the project “Identification of barriers and difficulties involved in the process of diagnosing somatic symptom disorders in primary care”. The project incorporates three studies, 1) focus groups with GPs, 2) interviews with affected patients and their treating GPs and 3) a survey with GPs [
40]. This article presents results from the focus group study. The project was funded by the German Research Foundation (DFG; funding numbers SCHE 1689/5–1, AL 1459/5–1 and LO 766/13–1) and registered at the German Clinical Trial Register under the ID: DRKS00009736.
Ethics
Ethics approval was obtained from the Ethics Committee of the Hamburg Medical Association (Germany) on April 7th, 2015 (approval number PV4763). All participants received oral and written information, had the possibility to ask further questions and gave their informed consent for the focus groups to be recorded, the recordings to be transcribed verbatim and the results of the analysis to be published anonymously.
Researcher characteristics
Focus groups were moderated by TZ, CJ and NJP. Data was analyzed by CJ and NJP. TZ is a male trained psychologist and postdoctorate researcher (exemplary research areas: mental disorders in primary care, quantitative research). He is experienced in moderating workshops and group discussions. CJ is a female GP registrar and postdoctorate researcher. Referring to research she was new to the field of psychosomatics, but has additional qualifications in basic psychosomatic care (“Psychosomatische Grundversorgung”). She was a first-time moderator and analyzer, but received training on moderating focus groups and qualitative content analysis prior to data collection and analysis. NJP is a female trained psychologist and postdoctorate researcher (exemplary research areas: potentially inadequate medication in primary care, rehabilitation, quality indicators), with comprehensive experience in conducting focus groups and interviews as well as qualitative data analysis. ML is a male trained psychologist and postdoctorate researcher, BL is a full professor of Medicine, board certified both in Psychosomatic Medicine and Internal Medicine and trained as psychologist and psychotherapist, MS is a full professor of Medicine, board certified in General Medicine.
Recruitment (pilot group)
We invited 20 members of the German College of General Practitioners and Family Physicians (DEGAM - Deutsche Gesellschaft für Allgemeinmedizin und Familienmedizin; general practitioners also working in research with varying degrees of specialized training and years of practice from all over Germany) via e-mail to participate in the pilot focus group. The e-mailed invitation included a short study description and an invitation, as well as detailed study information granting an allowance.
Recruitment (main study)
We followed a purposive sampling approach [
41]. To recruit participants for the five focus groups conducted in Hamburg, a random sample of general practitioners registered in the Hamburg metropolitan area (
N = 500) and adjacent counties of Lower Saxony and Schleswig-Holstein (Stade, Harburg, Pinneberg, Segeberg, Stormarn and Herzogtum Lauenburg,
N = 500) were invited via mail to participate in the focus groups. The sample was drawn from a register of 2313 GPs (Hamburg
N = 1343, adjacent counties
N = 970) maintained by the Institute of Primary Medical Care of the UKE. Invitation letters consisted of (1.) a short study description and invitation, (2.) a response sheet, (3.) detailed study information (communicating goals of the study), and (4.) a short demographic questionnaire. The invitation granted an allowance and four credit points for vocational training by the Medical Association of Hamburg.
Aiming at conducting five heterogeneous groups in respect to GPs’ gender, years of practice (< 10 years vs. > 10 years) and area of registration (HH or adjacent county) comprised of around eight participants per group, we determined the need to invite 1000 GPs expecting a response rate of 5%. We received feedback from 141 GPs, of which 81 were willing to participate in the study. 82 letters were returned to sender due to either an unknown new address (N = 74), the practice being closed (N = 6) or the addressee being deceased (N = 2).
Data collection (pilot group/main study)
The pilot focus group was moderated by TZ and NJP following a semi-structured guideline (see below). The discussion lasted 123 min. All other focus groups were moderated by CJ and NJP using the slightly revised semi-structured guideline. Discussions lasted between 107 and 114 min (Ø 111 min). The discussions were digitally recorded in full length and logged by a trained student research assistant.
Vignette development
Five different vignettes were used to provide a well-defined stimulus for the focus groups. The vignettes for focus groups were derived from anonymized and modified cases from the outpatient department of psychosomatics at the University Medical Center Hamburg-Eppendorf. The story was set in a primary care practice. The first half of the vignette outlined the relationship between the patient and the GP. The main complaints were presented as well as some background information on psychosocial circumstances. Part one also provided a résumé of former diagnoses, results of physical examinations, laboratory findings or other diagnostic results. The second half of the vignette plotted another consultation by the patient. The complaints would have evolved, mostly in an unfavorable way. The patient’s attitude towards the physician’s behaviour, e.g. suggesting a referral to a specialist or introducing a bio-psycho-social disease model, was described. Vignettes contained some possible strategies on dealing with somatization to maintain authenticity whilst avoiding narrowing the options of what to do. Vignettes were varied in different aspects. These included variation of sex, age group, main type of bodily complaints, ICD 10-codes that match best with the description given, fulfillment of the new diagnostic criterions of DSM-5 somatic symptom disorder and corresponding DSM-IV disorders, and the number of symptoms and diagnostic barriers that were implicitly embodied in the story.
Focus group guideline
The focus group guideline was structured into four main sections. Each group discussion started with the presentation of a patient vignette to be discussed. Vignettes are “short stories about hypothetical characters […], to whose situation the interviewee is invited to respond” (p. 105, [
42]). Vignettes can act as an icebreaker or warm-up exercise in focus groups to get the discussion between the participants started [
43]. After presenting the first part of the vignette, moderators stimulated discussion by asking: 1) Which problems/difficulties do you face at this point? 2) What would you do with this patient?. After the presentation of the second part of the vignette, moderators asked: 1) Which problems/difficulties do you face now? 2) What impedes diagnosing a somatoform disorder for this patient? 3) How do you recognize such a disorder? 4) How do you document such a disorder?.
The vignette discussion was followed by a discussion of barriers to and helpful tools for diagnosing somatoform disorders, the coding of diagnoses and the changes in the DSM-5 (somatic symptom disorder). See Additional file
1 for questions relevant for the present study. The semi-structured form of the guideline allowed for questions deviating from the guideline topics, if unexpected important aspects were mentioned or if the participants’ accounts needed more clarification. Piloting the guideline resulted only in minor changes in the wording of some questions.
Protection of data privacy
To ensure confidentiality, invitees’ replies with contact information were separated from pseudonymized questionnaires. In order to protect the focus group participants’ identity, all names were replaced by numbers and all potentially identifying details were changed during the transcription of the recordings.
Transcription of focus group recordings
All focus groups recordings were transcribed verbatim by a trained student research assistant following designated transcription rules. Transcripts were checked for accuracy by NJP.
Data analysis
The transcripts were analyzed using structuring qualitative content analysis [
44‐
46]. Qualitative content analysis extracts and preserves the essential content of the data, while significantly reducing the amount of data to be processed. It allows the integration of previous theoretical knowledge while being open to new findings. Deductive categories were derived using the review conducted by Murray et al. [
47], other relevant studies (e.g. [
6,
36,
48]) and the focus group guideline. However, due to the exploratory character of the study, a further focus was placed on inductive category development by summarizing content analysis [
45] derived from the material to ensure the openness of the analysis. Since the guideline pretest only resulted in minor changes to the focus group guideline and the pilot group’s participants did not differ significantly from main study’s participants (except for in their area of registration) we decided to analyze all transcripts using the same category system.
Both researchers (CJ/NJP) read all transcripts to ensure familiarity with the material and discussed their first impressions of the material. The transcripts were broken down into fragments of analysis, which were coded using deductively and inductively formed categories. These fragments could adopt different sizes ranging from part of a sentence to one or more paragraphs (according to the amount of data needed to understand the content and context of the relevant fragment). The transcripts were coded by CJ and NJP separately and the results were discussed after coding each transcript. All categories and codes were described extensively in code-memos which were supported by typical quotes. To ensure intersubjective reproducibility and comprehensibility, the results were presented to and discussed with the interdisciplinary workgroup “qualitative methods” (led by NJP) of the Department of General Practice / Primary Care and the co-authors (BL – professor, psychological therapist, MD; TZ and ML – trained psychologists/postdoctorate researchers and MS – professor, MD). Then, the entire material was subjected to a second round of coding. Data was managed using MAXQDA 11 (Verbi GmbH).
Conclusions
Our results are based on participants’ accounts of their clinical reasoning and behavior in daily practice. It cannot be ruled out that direct observation of consultations and coding behavior would highlight different aspects of coding of MUS or somatoform disorders and unearth other practices. Further studies (e.g., participatory observation in situ) are needed to enrich the understanding of this phenomenon.
Researchers in health services research often rely on administrative data, for example for grouping patients by their diagnoses, to study patterns and outcomes of disease or to study variations in access to care, and quality, cost and effectiveness of care [
61]. The abovementioned studies and our findings supply qualitative evidence for the commonly seen difference between routine and epidemiological data [
62]. Routine data created for billing or reimbursement purposes respectively, should be used with restraint and prudence [
24,
27,
55,
62]. For researchers this results in a need to validate routine diagnostic data where possible [
27].
From GPs’ points of view, ICD-10-coding does not seem to be a necessary requirement for treating patients (except with psychotherapy) and coding might be avoided to protect the patient from stigmas and other negative consequences. Our insights on GPs’ coding behavior do not allow conclusions about documentation, diagnostic and treatment quality.
The changes in the DSM-5 and the upcoming changes in the ICD-11 will bring the era of somatoform disorders - a diagnoses of exclusion - to an end [
26]. The term ‘somatoform disorder’ is changed to ‘somatic symptom disorder’ in the DSM-5. Now, the diagnosis can be considered by the GP as one single somatic symptom is present. The distinction between medically explained and unexplained symptoms has been given up (criterion A). The B criterion (‘positive psychological criteria’) requires one or more of the following to apply: health anxiety or disproportionate and persistent concern about the medical seriousness of one’s symptoms respectively excessive time and energy to be devoted to these symptoms or health concerns [
26]. Considering the already existing beta version of the ICD-11 one can assume that the ICD-11 will adopt these changes, even though it is not finally consented yet.
The described developments of substituting diagnoses of exclusion by the use of positive psychological criteria will supposedly change acceptance and handling of these diagnoses for both parties, GPs and patients. Nevertheless, the problem of “double-coding” and the abovementioned general unattractiveness of the ICD-system for many GPs due to its complexity will remain. Therefore the consequences of these changes for GPs’ diagnosing and coding behavior are not yet foreseeable. DSM-5’s somatic symptom disorder might be a concept that is easier applicable in primary care, whilst it is still challenged by implementation issues and the consequences of exact diagnostic coding of psychological disorders in this setting.