Study Summary: Development of a Claims-Based Algorithm to Identify Patients With Potentially Undiagnosed Chronic Migraine
Pavlovic JM, Yu JS, Silberstein SD, et al. Development of a claims-based algorithm to identify potentially undiagnosed chronic migraine patients. Cephalalgia. 2019;39(4):465-476. doi:10.1177/0333102418825373.
Chronic migraine (CM) affects an estimated 1.4% to 2.2% of the global population but is underdiagnosed and undertreated. Chronic migraine (≥15 headache days with ≥8 migraine days a month) is associated with a lower health-related quality of life and greater impact on activities of daily living compared with episodic migraine (<15 headache days a month).1
The current gold standard for diagnosing CM is a diagnostic interview by a headache specialist. Screening tools such as the Identify Chronic Migraine (ID-CM) tool may also be used to detect patients with CM. A claims-based algorithm may be useful to help identify patients at the healthcare system level. Thus, the authors sought to develop an algorithm based on administrative claims data to identify patients with potentially undiagnosed CM.1
This observational study was conducted using data from medical and pharmacy claims and a patient survey that was carried out at a large medical group. The study included adult patients with a claim that included an International Classification of Diseases, Ninth or Tenth Revision diagnostic code for migraine but no diagnostic code for CM in the 12 months prior to screening. The study excluded patients with a head injury, surgical procedure on the head or neck, or illicit drug use during the screening period or a migraine-related onabotulinumtoxinA claim during the 12-month enrollment period.1
Patients were screened with the ID-CM tool. Those identified as having CM underwent a semistructured diagnostic telephone interview with a physician trained by a headache specialist. The semistructured diagnostic interview, considered the gold standard for a CM diagnosis, was based on the International Classification of Headache Disorders, Third Edition, and modified Silberstein-Lipton criteria for CM.1
Claims data over the 12-month enrollment period were evaluated for more than 40 possible CM predictors, including demographic factors, healthcare resource use, comorbidities, and medication claims. Clinically relevant variables that significantly predicted CM status on bivariate analyses were identified and included in a logistic regression model.1
Of the 108 patients in the final sample, 64 (59.3%) had CM and 44 (40.7%) did not have CM, based on the results of the semistructured diagnostic interviews. The CM group included significantly more women compared with those who were identified as non-CM (96.9% women vs 84.1%; P = .018). Based on claims data from the enrollment period, patients with CM were significantly more likely than those without CM to have hypertension (P = .004). Patients with CM also had significantly more claims for nonsteroidal anti-inflammatory drugs (P = .07) and opioids (P = .01).1
The investigators identified 4 significant predictors of CM: female sex, 15 or more claims for acute migraine treatment (including opioids) within 12 months, claims for at least 2 unique classes of migraine preventive medication within 12 months, and at least 24 healthcare visits within 12 months. Compared with men, women were 9 times more likely to have CM (odds ratio [OR], 9.17; 95% CI, 1.26-66.50). Patients with 15 or more acute migraine treatment claims within 12 months (including opioids) were almost 6 times more likely than those with fewer than 15 claims to have CM (OR, 5.87; 95% CI, 1.34-25.63). Patients with claims for 2 or more classes of migraine preventives were more than 4 times as likely as those with no preventive medication claims to have CM (OR, 4.39; 95% CI, 1.19-16.22). CM was almost 3 times as probable among patients with 24 or more healthcare visits as among those with fewer than 24 visits (OR, 2.80; 95% CI, 1.08-7.25).1
The investigators compared the results of the claims-based algorithm with those of the semistructured diagnostic interviews, with a probability cutoff of 0.55. The claims-based algorithm had a sensitivity of 78.1%, specificity of 72.7%, positive predictive value of 80.7%, and negative predictive value of 69.7% for detecting CM.1
The claims-based algorithm developed by the authors may potentially be used as a screening tool to identify patients with potentially undiagnosed CM; however, CM should be formally diagnosed by a healthcare practitioner who has evaluated the patient.1
The authors described several limitations. Because most patients in the study population were women, the algorithm might be less applicable for detecting CM in men. Data on potentially relevant variables (such as number of monthly headache days) that were not available in the claims database were not included in the algorithm. Not all possible migraine treatments were included. The study population was relatively small and had a high prevalence of CM compared with the general population. The algorithm should be evaluated further in studies with different patient populations.1
The authors concluded that the claims-based algorithm they developed has suitable sensitivity and specificity in detecting patients with potentially undiagnosed CM. The algorithm may be used in healthcare settings to improve diagnosis and treatment of patients whose CM might not otherwise be identified.1
1. Pavlovic JM, Yu JS, Silberstein SD, et al. Development of a claims-based algorithm to identify potentially undiagnosed chronic migraine patients. Cephalalgia. 2019;39(4):465-476. doi:10.1177/0333102418825373.