A Sundholm1,2, S Burkill3, S Bahmanyar3, A I M Nilsson Remahl1,2. 1. Department of Clinical Neuroscience, Karolinska Institutet, Solna, Sweden. 2. Department of Neurology, Karolinska University Hospital, Stockholm, Sweden. 3. Department of Medicine, Centre for Pharmacoepidemiology, Karolinska Institutet, Solna, Sweden.
Abstract
OBJECTIVE: Idiopathic intracranial hypertension (IIH) is often misdiagnosed. This can cause problems if conducting register-based studies. The study purpose was to produce algorithms that better identify patients with correct diagnosis of IIH in the Swedish National Patient Register (NPR). METHODS: Patients with ICD-10 code G93.2 for IIH registered in the NPR (2006-2013, Stockholm County) were included and diagnosis validated by medical record reviews. Patients were randomized into two groups: one used to produce the algorithm (n = 105) and one for validation (n = 102). We tested variables possible to extract from registries and used forward stepwise logistic regression which provided a predicted probability of correct diagnosis for each patient. RESULTS: We included 207 patients of which 135 had confirmed IIH. This gave a positive predictive value of 65.2% (CI: 58.4-71.4). The algorithm produced with variables extracted from registries, that is, age, number of times with diagnosis code G93.2 recorded (>2 times), and acetazolamide treatment, predicted the diagnosis correctly 88.2% (CI: 80.3-93.3) of the time. Excluding treatment data from the algorithm did not change the prediction notably, 86.3% (CI: 78.1-91.7). CONCLUSION: We produced two algorithms that with improved accuracy predict whether an IIH diagnosis in the NPR is correct. This can be a useful tool when performing register-based studies.
OBJECTIVE:Idiopathic intracranial hypertension (IIH) is often misdiagnosed. This can cause problems if conducting register-based studies. The study purpose was to produce algorithms that better identify patients with correct diagnosis of IIH in the Swedish National Patient Register (NPR). METHODS:Patients with ICD-10 code G93.2 for IIH registered in the NPR (2006-2013, Stockholm County) were included and diagnosis validated by medical record reviews. Patients were randomized into two groups: one used to produce the algorithm (n = 105) and one for validation (n = 102). We tested variables possible to extract from registries and used forward stepwise logistic regression which provided a predicted probability of correct diagnosis for each patient. RESULTS: We included 207 patients of which 135 had confirmed IIH. This gave a positive predictive value of 65.2% (CI: 58.4-71.4). The algorithm produced with variables extracted from registries, that is, age, number of times with diagnosis code G93.2 recorded (>2 times), and acetazolamide treatment, predicted the diagnosis correctly 88.2% (CI: 80.3-93.3) of the time. Excluding treatment data from the algorithm did not change the prediction notably, 86.3% (CI: 78.1-91.7). CONCLUSION: We produced two algorithms that with improved accuracy predict whether an IIH diagnosis in the NPR is correct. This can be a useful tool when performing register-based studies.
Authors: Anna Sundholm; Sarah Burkill; Elisabet Waldenlind; Shahram Bahmanyar; A Ingela M Nilsson Remahl Journal: Cephalalgia Date: 2020-05-25 Impact factor: 6.292
Authors: Ali G Hamedani; Dylan P Thibault; Karen E Revere; John Y K Lee; M Sean Grady; Allison W Willis; Grant T Liu Journal: JAMA Netw Open Date: 2020-12-01
Authors: Anna Sundholm; Sarah Burkill; Elisabet Waldenlind; Shahram Bahmanyar; A Ingela M Nilsson Remahl Journal: Cephalalgia Date: 2021-08-18 Impact factor: 6.292