Background/Purpose: Diagnosing FM is difficult because there is no specific laboratory test to confirm the disorder. Billing practices dictate that physicians assign ICD-9-CM (Corresponding International Classification of Diseases, Ninth Revision, Clinical Modification) codes based on the procedure for which they use to treat patients.  Since FM pain encompasses many regions of the body and associated problems, physicians commonly do not assign 729.1.   Instead, patients may receive an ICD-9-CM code based on a procedure performed for pain relief, or some other diagnosis commonly associated with a symptom of FM. One of the objectives of this study was to develop a method to identify FM patients who do not meet the inclusion criteria of the 729.1 diagnosis assigned at least twice in a span greater than 12 months apart.  

Methods: The FM predictive model was developed using different techniques.  Among the 9,758 patients, there were 183,540 patient touch points, defined as a distinct time for which information could be obtained from the patient (Physician Health Assessment (PHA) questionnaire, office visits, medication prescriptions, and surgeries).  Over 150 possible predictors were entered into a random forest as a second step for variable reduction. In place of the FM ICD-9-CM code, patients were required to have some type of procedure or visit within the same time frame. A logistic regression model identified the four predictors: musculoskeletal procedures, total unique medications, total unique diagnoses, and days between touch points. The model was validated by using 10-fold cross validation. Propensity score 1:1 matching based on gender, age, length of treatment, and physician was then used to identify a control group of non-malignant chronic pain but without FM group.  A two-sample t-test was used to assess differences between cases and controls for patient variables including total number of procedures, diagnoses, medications, and days between touch points. 

Results: There were 15 diagnoses highly associated with FM that were distinguished by moderate to large effect sizes (Cohen’s d > 0.5).  The diagnoses included chronic pain syndrome, latex allergy, muscle spasm, fasciitis, cervicalgia, thoracic pain, shoulder pain, rheumatoid arthritis, cervical disorders, cystitis, cervical degeneration, anxiety, joint pain, lumbago, and cervical radiculitis. There was a significant association between FM patients and the region where they receive procedures.  Specifically, FM patients are more than four times as likely to receive a procedure in the shoulder (3.55, 5.67) or neck (3.50, 5.02) as non-malignant chronic pain patients.  Overall, FM patients are more likely to receive more procedures encompassing six of the seven regions used for analysis: shoulder, neck, arm, chest, knee, thoracic area (all p < 0.05).  

Conclusion: The predictive model allows a larger selection of FM patients and, by doing so, may assist in accurately targeting more patients for appropriate treatment.  This project gives new insights into the diagnosis, co-morbidities, and treatments associated with FM patients in a pain management setting.

« Back to 2015 ACR/ARHP Annual Meeting

ACR Meeting Abstracts - https://acrabstracts.org/abstract/assessing-alternative-selection-criteria-for-fibromyalgia-patients-within-a-multicenter-chronic-pain-claims-database/