Prevalence of survey-based criteria for fibromyalgia and impact on hospital burden: a 7 year follow-up study from an outpatient clinic

Objectives To evaluate the prevalence of survey-based criteria for fibromyalgia (FM) among newly referred patients in a rheumatic outpatient clinic, and to compare the use of secondary healthcare services between survey-based FM and non-FM cases. Method Newly referred patients to an outpatient clinic were screened for the fulfilment of the 2011 FM survey criteria during a 6 month period in 2013 in this observational cohort study. Demographic data were obtained at baseline. Patients’ medical files were evaluated and comparisons between groups were made regarding the use of hospital healthcare facilities during the 7 year observation period. Results Out of 300 invited patients, 248 (83%) completed the questionnaire; 90 patients (36%) fulfilled survey-based criteria for FM at enrolment. FM cases were primarily women (80% vs 54% of non-FM cases), and received more medications (median 4 vs 3 drugs) and public economic support (62% vs 20%). At the 7 year follow-up, crude analyses showed that FM cases had a higher number of hospital courses (median 10 vs 8) and had undergone more invasive procedures (78% vs 60%). Neurologists (42% vs 28%), gastroenterologists (30% vs 13%), endocrinologists (40% vs 21%), pain specialists (13% vs 3%), psychiatrists (20% vs 7%), and abdominal surgeons (43% vs 30%) were consulted more often by FM than by non-FM cases. Conclusion Fulfilment of FM survey criteria among newly referred patients to a rheumatic outpatient clinic is frequent. Our study findings show that FM continues to present a challenge for healthcare professionals as well as for patients. RESEARCH HIGHLIGHTS ● Fulfilment of FM survey criteria among newly referred patients to a rheumatic outpatient clinic is frequent. ● The burden on the secondary healthcare system for these patients is significant. ● This study suggests the need for increased awareness about the diagnosis of FM among certain medical and surgical specialties.


RESEARCH HIGHLIGHTS
• Fulfilment of FM survey criteria among newly referred patients to a rheumatic outpatient clinic is frequent.
• The burden on the secondary healthcare system for these patients is significant.
• This study suggests the need for increased awareness about the diagnosis of FM among certain medical and surgical specialties.
Fibromyalgia (FM) has been subject to controversy since the development of classification and diagnostic criteria (1-3).Many clinicians do not recognize FM as either a disease or a disorder (4)(5)(6), and several studies have shown that FM remains underdiagnosed.For instance, one study demonstrated that up to 75% of FM cases remain undiagnosed (7), and epidemiological studies from Japan and Germany have documented that only 2.5% and 14%, respectively, of patients satisfying survey criteria for FM have been given this diagnosis (8,9).Furthermore, a survey conducted among 172 general practitioners demonstrated that only 55% were familiar with the diagnostic criteria for FM (10).Currently, the diagnosis of FM is given if patient-reported symptoms meet the predefined criteria for FM, in the absence of other identifiable explanations.The criteria are based on the cluster of symptoms that characterize the syndrome according to clinical studies and expert consensus, whereas objective clinical findings are missing.Nevertheless, the World Health Organization has recognized FM as a specific subgroup entity of patients with chronic widespread pain (11).
The most common symptom leading to referral to rheumatic outpatient clinics is pain, along with the suspicion of osteoarthritis or systemic inflammatory rheumatic disease (12).However, refraining from considering FM as a possible differential diagnosis when evaluating the origin of pain in these patients often leads to extensive examinations in the healthcare system.Thus, numerous studies have documented that patients fulfilling criteria for FM have substantial healthcare costs compared to the general population (7,(13)(14)(15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29).These data show that FM patients have an increased number of specialist consultations, yet no study has quantified the extent to which FM burdens the different specific medical specialities in the secondary healthcare sector.
The prevalence of FM has been widely investigated throughout the different continents, ranging from 0.2% to 6% in the general population depending on the criteria used for identification (9,(30)(31)(32)(33)(34)(35)(36).This is consistent with findings in Denmark and the neighbouring countries, where the prevalence has been found to range from 0.66% to 4.8% (37)(38)(39).Furthermore, prevalence studies of FM have been conducted in many subpopulations (40)(41)(42)(43)(44); however, to our knowledge, no prevalence study of FM exists concerning newly referred patients to rheumatic outpatient clinics.
Thus, the aim of this study was to evaluate the prevalence of FM, as defined by the fulfilment of the 2011 survey-based criteria for FM (45), and its impact on the hospital burden among newly referred patients to a rheumatic hospital clinic.Our primary objective was to compare the use of secondary healthcare services [i.e.burden, defined as the number of different courses (i.e. a period where the patient is evaluated by a specific specialist)] in the hospital sector between FM-and non-FM cases, during 7 years.Secondary objectives included comparing the number of diagnoses, number of admissions to hospital, number of invasive procedures and consultations in different medical specialities, number of doctors' notes in patients' medical journals, other rheumatic diseases, and use of medications between the groups over the 7 year follow-up period.

Study design and prespecification
This study was designed as a single-centre observational cohort study with consecutive enrolment and time-true registration of all data in patients' files.The prespecified statistical analysis plan is available as online supplementary material.

Setting and participants
Patients who had been newly referred (for any cause) to an outpatient rheumatic clinic at the regional hospital of Southwest Denmark, Esbjerg, were screened for FM by self-administered questionnaires based on the 2011 survey criteria for FM (45) over 6 months (March to October) in 2013.A diagnosis based on these criteria depends on the subjective reporting of the presence and severity of symptoms, including emotional distress, cognitive disturbance, fatigue, sleep disturbance, and related somatic symptoms scored on a 0-3 scale, along with the report of multifocal pain.The term multifocal pain is derived from the Widespread Pain Index (WPI), based on patients' report on the number of areas in which they have had pain over the past week (2,3).
The questionnaire was translated into Danish.Three rheumatologists (PRLH, SC, and Leif Ejstrup) approved the Danish translation of the questionnaire.Subsequently, the questionnaire was tested on a few patients by cognitive debriefing.Only minor language revisions were performed after the debriefing.Questions were added regarding smoking status, working status, marital status, and social security number.The last of these was done to identify patients in the electronic journal system at follow-up.The questionnaires were posted to the patients prior to their scheduled appointment in the outpatient clinic.The patients were told to place their answers in a closed box, to which only the investigators had access.
In 2020, questionnaire data were transferred to the project database.Patients' medical files were identified by the social security number noted by the study subjects in the original questionnaire.Relevant data (all data regarding each patient's medical history, obtained at any hospital in the region during the observational period) from the patients' electronic medical files were transferred to the database.The total number of current medications at the time of follow-up was retrieved through the online medication facility.Every prescribing doctor is responsible for updating the online medication facility when changing or prescribing medicine.If any inconsistencies were noted between the electronic medical record and the currently prescribed medicine, e.g.prescription of antidepressants but no medical record of depression, the patient would be considered with the diagnosis of depression in the database.This was done because the electronic journal system only includes incidents in the secondary hospital sector of the healthcare system and does not include any action taken by the general practitioners.If patients received medicines for more than one indication, an individual evaluation of the medical journal was made.For example, pregabalin can be prescribed for relieving pain or for treating epilepsy.However, if the specific electronic medical file did not indicate any diagnosis of epilepsy, it was assumed that pregabalin was given for pain relief, and patients would be categorized as having chronic pain and not epilepsy.
Based on the answers to the questionnaires, patients were divided into two groups depending on the fulfilment of the survey-based criteria for FM, and comparisons were made regarding the use of the secondary healthcare services during the observation period.All data were prospectively collected in patients' medical files from the date of questionnaire fulfilment; however, as described above, data were transferred to the final database simultaneously at the defined 7 year follow-up assessment.
Variables, data sources, and measurements Study participants are described with respect to age at inclusion, sex, smoking status, working status, marital status, WPI, and Symptom Severity Score (SSS), including scores on individual items in the SSS, based on their answers to the questionnaire.The numbers of diagnoses, medications, and pain medications at inclusion, along with the presence of rheumatic inflammatory diseases, were obtained separately for subjects with and without FM through journal assessment, as previously described.
During the following 7 years, all variables of interest were prospectively documented in the patients' medical journals by their treating physicians in the secondary sector of the public healthcare system.The identified outcome variables were as follows: number of different courses at a hospital (i.e. a period ranging from one consultation to an infinite number depending on the specific condition, where the patient is evaluated by any specific specialist, e.g. a cardiologist), number of chronic diseases, description of surgical or diagnostic procedures performed at any hospitals in the region, and a list of the specific medical specialities consulted.Furthermore, the number of doctors' notes in the journal, admissions to any hospital in the region, and the number of medications prescribed, including pain medications, were noted at the 7 year follow-up.

Bias
A prerequisite for a clinical diagnosis of FM is that the patient does not have a disorder that would otherwise explain the pain.In fact, the misdiagnosis of FM is most likely to occur in the setting of early undiagnosed rheumatic disease, which may appear as a more ill-defined pain syndrome in the early stages.In the clinical setting, the 2011 survey criteria may support an FM diagnosis, but a complete physical examination is still mandatory to consolidate the diagnosis or identify features that may point to some other condition having a presentation similar to FM.Thus, the 2011 survey criteria, which primarily were developed for use in large-scale epidemiological studies and allow assessment based on self-report, may potentially be influenced by the presence of any rheumatic inflammatory disease, leading to the misclassification of patients.To address this issue, the number of the different physician-derived rheumatic diagnoses in the population under study is reported in each of the two groups.

Power and sample size considerations
Based on the period from March to October 2013, we anticipated that more than 300 patients would be evaluated for the first time (assuming at least 50 new referrals per month).Assuming that 25% have FM (this represents an assumed propensity of having FM, as a one-to-three group allocation), we would have a very good statistical power (96%) to detect a difference between groups of five different courses in the hospital sector, assuming that the standard deviation is 10 different courses in the hospital sector.Because of uncertainty regarding the response rate to the questionnaires, a total number of 300 participants was the anticipated sample size.However, following all the assumptions above, even a sample size of 200 individuals would be sufficient to detect this apparently important difference (five different courses in the hospital sector) with 86% statistical power.

Quantitative variables
The primary analyses are based on the intention-tosurvey (ITS) population.The ITS principle uses an analogy of an intention-to-treat population that asserts the effect of the implied consequence of having an FM phenotype, independent of competing comorbidities, prescribed medicine, subsequent treatment adherence, and withdrawal of consent.Accordingly, participants 'allocated' to a (survey-based) phenotype group (X FM and X Non-FM , respectively) were followed up, assessed, and analysed as members of that group, irrespective of what happened over the 7 year follow-up period.The ITS population includes all enrolled patients assessed (survey-based) for FM at baseline.

Statistical methods
Categorical data are summarized by numbers and percentages.Descriptive statistics for continuous data are summarized and reported as medians with the corresponding interquartile range (IQR), as well as the model-agnostic difference between medians with an approximate 95% confidence interval (CI) based on the Wilcoxon rank sum test.Also, to assess the balance diagnostics comparing medians and proportions at baseline, we used standardized differences.All outcome analyses are derived from patients' medical files from baseline to up to 7 year follow-up.Missing data from patients who had moved to another region or patients who died during the follow-up period were considered lost to follow-up; to handle these missing data, multiple imputation (using five imputed data sets) was used for the purpose of sensitivity, respecting the ITS principle, to the robustness of the overall findings.
To statistically model the difference between groups at follow-up while enabling us to adjust for potentially confounding variables, we applied general linear models.Inferring (potentially causal) effects and impacts based on the baseline phenotype (i.e.survey-based FM) on an important outcome (i.e.hospital burden) in an observational study requires adjustment using observed baseline variables, thereby aiming to deconfound the inference to avoid bias.The following true pre-exposure variables were considered a priori as potential confounders: age and sex at baseline.For the purpose of sensitivity analysis, improved confounding variable balance between the exposed (X FM ) and control (X Non-FM ) groups was approached by adjusting observations from each group based on the propensity score, which, in this case, is the probability that a patient has FM (X FM ) given all the observed baseline covariates.

Ethics
Before initiation of the study, the local ethics committee was contacted, and they concluded that the study was observational and thereby non-notifiable according to Danish law (S-20222000-100).Study subjects were fully informed regarding the study, including its purpose and plans.The managing physicians were aware of the ongoing study but had no access to the results.Data collection was approved by the local data protection agency and study data were pseudonymized.Permission to access journal material was granted by the region of Southern Denmark, and the study was registered accordingly (ID number: 20/45390).

Participants
Three-hundred newly referred patients for rheumatic evaluation were invited to participate in the study by questionnaire.Thirty individuals abstained from filling in the questionnaire and 22 refrained from filling in their social security number, leaving 248 individuals for further investigation.Ninety patients (36%) fulfilled the 2011 survey criteria for FM, while 158 (64%) did not.For the FM group, four individuals moved to another region, leaving 86 patients for complete investigation.Four patients died during the follow-up period in this group.For the non-FM group, seven patients moved to another region, leaving 151 for complete evaluation, and 17 died during the followup period (Figure 1).

Descriptive data
Baseline characteristics are presented in Table 1 for FM and non-FM study subjects.Significant differences between groups were observed considering sex distribution, working status, and medicine consumption.The last of these was explained by the number of pain medications (2 vs 1).At inclusion, the median number of diagnoses for FM study subjects was two compared to one for non-FM study subjects, driven primarily by the diagnosis of chronic pain.In contrast, an overweighting of physician-derived rheumatic inflammatory diagnoses at baseline was observed for non-FM study subjects.No differences between groups regarding specific rheumatic diagnoses were observed; however, nine patients (10%) were formally given a physician-derived diagnosis of FM in the FM group compared to one patient (0.6%) in the non-FM group (Table 2).

Main results and primary analyses
Crude analyses for outcome data for primary, secondary, and other outcomes are shown in Table 3. FM cases had 25% more courses in the secondary hospital sector compared to non-FM cases during the 7 year period.Small but statistically significant differences were seen between groups regarding numbers of diagnoses.The difference was driven by the diagnoses of chronic pain [n = 41 (48%) vs n = 23 (15%)] and gastrointestinal reflux [n = 39 (45%) vs n = 40 (26%)], whereas no other discrepancies between groups concerning specific diagnoses, including osteoarthritis [n = 15 (17%) vs n = 20 (13%)], were observed.
Group differences concerning invasive procedures were also observed: 78% in the FM group versus 60% in the non-FM group had undergone any surgical procedures, which consisted primarily of gastroscopies, colonoscopies, and joint surgery.
During the 7 year period, more FM cases had consulted neurologists, gastroenterologists, endocrinologists, pain specialists, abdominal surgeons, and psychiatrists compared to non-FM study subjects (Table 4).

Sensitivity analyses and secondary analyses
When analyses were adjusted for sex, age, and missing data by multiple imputation, mean numbers and differences concerning the number of hospital courses for the FM and non-FM groups increased (Table 5).On average, FM cases had 13 courses versus nine courses for non-FM study subjects (−3.76, 95% CI −5.83 to −1.69, p < 0.001).Considering the numbers of diagnoses,

Key results
This study found that more than one-third of all newly referred patients to a rheumatic outpatient clinic fulfilled the 2011 survey-based criteria for FM.Furthermore, our study showed that patients fulfilling the 2011 criteria for FM had 25% more courses in the secondary hospital sector compared to non-FM cases during the 7 year follow-up period.In addition, we found differences between groups comparing the use of pain medication, whereas the proportion of patients with a specific rheumatic diagnosis, number of other diagnoses, and number of hospital admissions and doctors' notes in the medical journal were identical in the two groups.Finally, we observed that FM cases had undergone more surgical procedures and consulted specific medical specialities more often than non-FM study subjects.

Limitations
This study has several limitations.First, as it was an observational cohort study, proper randomization and balancing between groups could not be performed at baseline.Considering baseline data, differences between the groups exist.Given the observational character of the study, causal relationships should be interpreted with caution, as there is a risk of confounding.However, the adjusted analyses did not alter the overall conclusion regarding the primary endpoint.Secondly, the cohort was not unselected.Patients were referred by their general practitioner, and selection bias cannot be ruled out.At the time of the study, the local instructions for referring patients to the rheumatic outpatient clinic dictated that only patients with the suspicion of rheumatic inflammatory disease should be evaluated.Patients suspected of sole FM were, in general, not referred.This could influence our primary objective as well as other objectives.It could be speculated that the patients in this cohort represent a subset of FM patients who are more prone to be referred and to seek secondary care, compared to the vast majority of FM patients who are handled in primary care.However, our results are in accordance with those of other studies, indicating that this possibility is less likely.
Thirdly, the diagnosis of FM was based on questionnaire-derived survey criteria.The fulfilment of the survey criteria for FM without clinical consolidation could potentially lead to misclassification by the presence of any undiagnosed rheumatic inflammatory disease in patients referred for rheumatological evaluation.However, the fulfilment of FM criteria was not associated with an increased number of any rheumatic inflammatory diagnoses.
Finally, the number of diagnoses was based on patients' electronic medical files and currently prescribed medicine.As described in the Method section, any inconsistencies between the electronic medical record and the currently prescribed medicine would result in an individual evaluation regarding specific diagnoses, especially if patients received medicines for more than one indication.We cannot rule out any risk of misclassification using this approach; however, inconsistencies were only seen in a small number of patients.

Interpretation and generalizability
Because of the observed high prevalence of FM among newly referred patients to a rheumatic outpatient clinic, paying increased attention to FM as a probable differential diagnosis is justified when evaluating patients for potential inflammatory rheumatic diseases.Similarly to the findings of other studies (7-9), our survey showed that only a minority of these patients were given a clinical diagnosis of FM based on the physician's assessment.We found no increased prevalence of inflammatory rheumatic diagnoses at baseline in the FM group compared to the non-FM group.However, 4% of RA patients fulfilled the 2011 survey criteria for FM at the time of diagnosis of RA, which is in accordance with the findings of Schelin et al (46).
The main finding of our study was in accordance with that of other studies (7,(13)(14)(15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29), showing that FM study subjects relative to non-FM study subjects burdened the secondary healthcare system significantly during the 7 year follow-up period.Crude analyses showed that FM study subjects had 25% more courses consulting medical specialists than non-FM study subjects.The adjusted analyses only strengthened these findings, showing a difference between groups of 44%.Previous studies investigating the burden of FM on healthcare facilities are primarily from countries outside Scandinavia (7,(13)(14)(15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29).Earlier studies in Denmark or Sweden have focused on chronic pain conditions, including FM among other chronic pain conditions, particularly chronic back pain (47).Similarly to our study, they reported chronic pain conditions to be associated with increased healthcare costs and decreased workability (48,49).Furthermore, one study concluded that patients with chronic pain conditions had 25% increased contacts with healthcare facilities (50).The overall results of our study in a Danish context seem to be very similar to those of other studies, in spite of the differences concerning the structure and underlying financing of the healthcare systems.
Our results showed that FM study subjects underwent significantly more invasive procedures than non-FM study subjects.Joint surgery and endoscopic procedures of the gastrointestinal tract were the procedures explaining the differences between groups.This is partly in accordance with the study by Wolfe et al (13), where FM patients also had increased surgery on the back and neck, carpal tunnel surgery, gynaecological surgery, and tonsillectomy.In our study, no differences were observed considering these procedures; however, our study differs on numerous factors, including the country, the organization of the healthcare system, the number of participants, and the year in which the study was performed.
As the first of its kind, this study pinpointed in which medical specialities the diagnosis of FM should receive increased awareness.Neurologists, gastroenterologists, endocrinologists, pain specialists, psychiatrists, and abdominal surgeons encountered FM study subjects more frequently than all other medical specialities.The finding of a minimal difference between FM study subjects, relative to non-FM study subjects, regarding the number of resulting diagnoses underlines that prolonged and excessive use of healthcare services with referral to multiple specialists should be avoided.

Conclusion
Fulfilment of the 2011 survey criteria for FM among newly referred patients to a rheumatic outpatient clinic is high, but only a few patients are given a diagnosis of FM.FM patients burden the hospital sector substantially and this study pinpoints for the first time the medical specialities in which awareness of FM should be increased.This is the first Scandinavian study investigating the subject, but our findings are similar to those of studies in other countries, including the USA and Canada.Our study findings demonstrate that FM continues to present a challenge for healthcare professionals as well as for patients.A timely and coordinated effort across medical specialities, including rheumatologists, could reduce diagnostic delay, facilitate management, and reduce healthcare utilization in this patient population.

Table 1 .
Patient characteristics at baseline (enrolment) representing the intention-to-survey population.
Data are shown as mean ± sd, n (%), or median (interquartile range).*Based on model-agnostic tests (Wilcoxon rank sum test) for continuous outcomes (median differences) and Wald z-tests for the estimated risk differences (proportion differences).SSS, Symptom Severity Score.

Table 3 .
Crude analyses of outcome measurements collected during the 7 year follow-up in the intention-to-survey population.
Data are shown as median (interquartile range) or n (%).*Based on model-agnostic tests (Wilcoxon rank sum test) for continuous outcomes (median differences) and Wald z-tests for the estimated risk differences (proportion differences).FM, fibromyalgia; CI, confidence interval.

Table 2 .
Number of rheumatic diagnoses in fibromyalgia (FM) and non-FM patients at baseline.

Table 4 .
Medical specialities consulted in the fibromyalgia (FM) and non-FM groups: crude analyses.

Table 5 .
Observed primary and key secondary outcomes up to 7 years after baseline.*†admissions,and pain medications prescribed, no differences were observed after adjusting for covariates.According to the results in Table2, FM apparently increased the risk of having an invasive procedure performed, corresponding to a crude odds ratio (OR) of 2.36 (95% CI 1.29 to 4.32).When this logistic regression model was further adjusted for sex and age at enrolment, the signal increased further to an OR = 2.61 (95% CI 1.38 to 4.95).