Adherence to direct or vitamin K antagonist oral anticoagulants in patients with atrial fibrillation: a long-term observational study.

Our objectives were to measure long-term adherence to oral anticoagulants (OACs) in patients with atrial fibrillation (AF) and to identify patient factors associated with adherence. Using linked, population-based administrative data from British Columbia, Canada, an incident cohort of adults prescribed OACs for AF was identified. We calculated the proportion of days covered (PDC) as a time-dependent covariate for each 90-day window from OAC initiation until the end of follow-up. Associations between patient attributes and adherence were assessed using generalized mixed effect linear regression models. 30,264 patients were included. Mean PDC was 0.69 (SD 0.28) over a median follow-up of 6.7 years. 54% of patients were non-adherent (PDC < 0.8). After controlling for confounders, factors positively associated with adherence were number of drug class switches, history of stroke or transient ischemic attack, history of vascular disease, time since initiation, and age. Age > 75 years at initiation, polypharmacy (among VKA users only), and receiving DOAC (vs. VKA) were negatively associated with adherence. PDC decreased over time for VKA users and increased for DOAC users. Over half of AF patients studied were, on average, nonadherent to OAC therapy and missed 32% of their doses. Several patient factors were associated with higher or lower adherence, and adherence to VKA declined during therapy while DOAC adherence increased slightly over time. To min im ize the risk stroke, adherence-supporting interventions are needed for all patients with AF, particularly those aged > 75 years, those with prior stroke or vascular disease, VKA users with polypharmacy, and DOAC recipients.

The Random Effects Warfarin Days' Supply (REWarDS) Model: Development and Validation of a Novel Method for Estimating Exposure to Warfarin Using Administrative Data.

Warfarin's complex dosing is a significant barrier to measurement of its exposure in observational studies using population databases. Using population-based administrative data (1996-2019) from British Columbia, Canada, we developed a method based on statistical modeling (Random Effects Warfarin Days' Supply (REWarDS)) that involves fitting a random-effects linear regression model to patients' cumulative dosage over time for estimation of warfarin exposure. Model parameters included a minimal universally available set of variables from prescription records for estimation of patients' individualized average daily doses of warfarin. REWarDS estimates were validated against a reference standard (manual calculation of the daily dose using the free-text administration instructions entered by the dispensing pharmacist) and compared with alternative methods (fixed window, fixed tablet, defined daily dose, and reverse wait time distribution) using Pearson's correlation coefficient (r), the intraclass correlation coefficient, and the root mean squared error. REWarDS-estimated days' supply showed strong correlation and agreement with the reference standard (r = 0.90 (95% confidence interval (CI): 0.90, 0.90); intraclass correlation coefficient = 0.95 (95% CI: 0.94, 0.95); root mean squared error = 8.24 days) and performed better than all of the alternative methods. REWarDS-estimated days' supply was valid and more accurate than estimates from all other available methods. REWarDS is expected to confer optimal precision in studies measuring warfarin exposure using administrative data.

Interrogation of the Multiple Sclerosis Prodrome Using High-Dimensional Health Data.

BACKGROUND: There is growing evidence of a prodromal period in multiple sclerosis (MS). A better understanding of the prodrome may facilitate prompt recognition and treatment of MS as well as narrowing of the etiologically relevant -period when searching for MS risk factors. OBJECTIVES: To explore and further delineate the MS prodrome, we used statistical learning techniques to examine associations of physician-generated diagnostic codes and prescription medication classes in the 5 years before the first demyelinating-related claim for MS cases and matched population controls. METHODS: In this matched cohort study, we accessed data from linked health administrative hospital, physician, and prescription databases from British Columbia, Canada, between 1996 and 2013. We focused on 7 medication classes previously identified as associated with the MS prodrome: urinary anti-spasmodics, glucocorticoids, muscle relaxants, anti-epileptics, dopa-derivatives, benzodiazepine, and antivertigo preparations. Diagnostic codes associated with the use of each medication class were first identified using LASSO logistic regression analyses in two-thirds of the cohort and then validated using multivariate logistic regressions in the remaining cohort. RESULTS: Our analyses included 4,862 MS cases and 22,649 controls. Although the identified diagnostic codes showed fair to good predictive performance in 6 medication classes (C-index = 0.712-0.858), these codes failed to fully explain the higher usage of these medications by the MS cases. Compared to controls of the same age, sex, and diagnostic codes, MS cases had higher odds of filling a prescription for antivertigo preparations (adjusted OR [aOR] 2.48; 95% CI 1.92-3.19), anti-epileptics (aOR 2.34; 1.90-2.90), glucocorticoids (aOR 1.76; 1.52-2.03), urinary anti-spasmodics (aOR 1.72; 1.20-2.46), and muscle relaxants (aOR 1.33; 1.13-1.56). CONCLUSIONS: We observed markedly higher use of specific medications in MS cases in the 5 years before the first demyelinating claim. The overrepresentation of specific medications in MS cases, which was not fully explained by the physician diagnoses, may represent a signature of the MS prodrome.

Adjusting for differential misclassification in matched case-control studies utilizing health administrative data.

In epidemiological studies of secondary data sources, lack of accurate disease classifications often requires investigators to rely on diagnostic codes generated by physicians or hospital systems to identify case and control groups, resulting in a less-than-perfect assessment of the disease under investigation. Moreover, because of differences in coding practices by physicians, it is hard to determine the factors that affect the chance of an incorrectly assigned disease status. What results is a dilemma where assumptions of non-differential misclassification are questionable but, at the same time, necessary to proceed with statistical analyses. This paper develops an approach to adjust exposure-disease association estimates for disease misclassification, without the need of simplifying non-differentiality assumptions, or prior information about a complicated classification mechanism. We propose to leverage rich temporal information on disease-specific healthcare utilization to estimate each participant's probability of being a true case and to use these estimates as weights in a Bayesian analysis of matched case-control data. The approach is applied to data from a recent observational study into the early symptoms of multiple sclerosis (MS), where MS cases were identified from Canadian health administrative databases and matched to population controls that are assumed to be correctly classified. A comparison of our results with those from non-differentially adjusted analyses reveals conflicting inferences and highlights that ill-suited assumptions of non-differential misclassification can exacerbate biases in association estimates.

Bayesian analysis of pair-matched case-control studies subject to outcome misclassification.

We examine the impact of nondifferential outcome misclassification on odds ratios estimated from pair-matched case-control studies and propose a Bayesian model to adjust these estimates for misclassification bias. The model relies on access to a validation subgroup with confirmed outcome status for all case-control pairs as well as prior knowledge about the positive and negative predictive value of the classification mechanism. We illustrate the model's performance on simulated data and apply it to a database study examining the presence of ten morbidities in the prodromal phase of multiple sclerosis.

Mining healthcare data for markers of the multiple sclerosis prodrome.

BACKGROUND: Previous studies suggest the existence of a prodromal period in multiple sclerosis, but little is known about the phenotypic characteristics. This study aims to characterize the multiple sclerosis (MS) prodrome using data mining analytics in the healthcare setting. METHODS: We identified people with MS and matched general population controls using health administrative data in two Canadian provinces (British Columbia and Saskatchewan). Using a training dataset (66.6% of British Columbia's cohort), L1 penalized logistic regression models were fitted to predict MS from physician and hospital encounters (via International Classification of Diseases [ICD] codes) and prescriptions filled (as drug classes) during the five years before the MS case's first demyelinating event. Internal and external validation of identified predictors was performed using logistic regression on the remaining British Columbia (33.4%) and Saskatchewan data. Adjusted odds ratios (aORs) and Area under the Curve (AUC) metrics for the models' predictive performance were reported. RESULTS: We identified 8,669 MS cases and 40,867 controls. Good predictive performance was observed for physician data (internal/external validation AUC = 0.81/0.79). Physician-generated ICD codes that were associated with MS and validated in both provinces included disorders of the central and peripheral nervous system, disorders of the eye, and cerebrovascular disease (aOR = 1.3-7.0). Overall, hospital and prescription data showed very poor and poor predictive performance (internal/external validation AUCs = 0.54/0.55 and 0.66/0.61, respectively). However, hospitalizations related to the urinary system or spinal cord diseases, or prescriptions for urinary antispasmodics or anti-vertigo preparations, were associated with 2 to 3-fold higher odds of MS (aOR = 2.3-3.3). CONCLUSIONS: Findings provide insight into the clinical characteristics of the MS prodrome. Diagnostic codes from physician encounters were capable of differentiating between MS cases and controls.

Health-care use before a first demyelinating event suggestive of a multiple sclerosis prodrome: a matched cohort study.

BACKGROUND: Degenerative processes in neurodegenerative diseases can start years before clinical manifestation. We aimed to establish whether a multiple sclerosis prodromal period exists by examining patterns of health-care use before a first demyelinating event. METHODS: In this matched cohort study, we used data from linked health administrative and clinical databases from four Canadian provinces (British Columbia, Saskatchewan, Manitoba, and Nova Scotia) to compare hospital, physician, and prescription use data from people with multiple sclerosis and matched general population controls in the 5 years before the first demyelinating disease claim (health administrative index date) or clinically reported symptom onset (clinical index date). Rate ratios (RRs) were estimated using negative binomial regression and combined across provinces using random effect models. The primary outcome was all-cause use of health care during each of the 5 years before the health administrative or clinical index date. FINDINGS: The health administrative cohort included 14 428 multiple sclerosis cases and 72 059 matched controls for whom data were available between April, 1984, and April, 2014. Annual health-care use increased steadily between 5 years and 1 year before the first demyelinating disease claim in people with multiple sclerosis compared with controls (from RR 1·26 [95% CI 1·16-1·36] to 1·78 [1·50-2·10] for hospital admissions; from 1·24 [1·16-1·32] to 1·88 [1·72-2·07] for physician claims; and from 1·23 [1·06-1·41] to 1·49 [1·41-1·59] for prescriptions, assessed as drug classes). Similar patterns for physician claims and prescriptions were observed in the cohort with available clinical symptom onset (3202 individuals with multiple sclerosis and 16 006 controls), although the differences in use in each of the 5 years mostly did not reach statistical significance. INTERPRETATION: More frequent use of health care in patients with multiple sclerosis than in controls in the 5 years before a first demyelinating event, according to health administrative data, suggests the existence of a measurable multiple sclerosis prodrome. These findings have clinical and research implications, including the establishment of an earlier window of opportunity to identify and potentially treat multiple sclerosis. FUNDING: National Multiple Sclerosis Society.