Patient clusters and cost trajectories in the Swiss Atrial Fibrillation cohort

Abstract

ObjectiveEvidence on long-term costs of atrial fibrillation (AF) and associated factors is scarce. As part of the Swiss-AF prospective cohort study, we aimed to characterise AF costs and their development over time, and to assess specific patient clusters and their cost trajectories.MethodsSwiss-AF enrolled 2415 patients with variable duration of AF between 2014 and 2017. Patient clusters were identified using hierarchical cluster analysis of baseline characteristics. Ongoing yearly follow-ups include health insurance clinical and claims data. An algorithm was developed to adjudicate costs to AF and related complications.ResultsA subpopulation of 1024 Swiss-AF patients with available claims data was followed up for a median (IQR) of 3.24 (1.09) years. Average yearly AF-adjudicated costs amounted to SFr5679 (€5163), remaining stable across the observation period. AF-adjudicated costs consisted mainly of inpatient and outpatient AF treatment costs (SFr4078; €3707), followed by costs of bleeding (SFr696; €633) and heart failure (SFr494; €449). Hierarchical analysis identified three patient clusters: cardiovascular (CV; N=253 with claims), isolated-symptomatic (IS; N=586) and severely morbid without cardiovascular disease (SM; N=185). The CV cluster and SM cluster depicted similarly high costs across all cost outcomes; IS patients accrued the lowest costs.ConclusionOur results highlight three well-defined patient clusters with specific costs that could be used for stratification in both clinical and economic studies. Patient characteristics associated with adjudicated costs as well as cost trajectories may enable an early understanding of the magnitude of upcoming AF-related healthcare costs.

Patients on vitamin K treatment: is switching to direct-acting oral anticoagulation cost-effective? A target trial on a prospective cohort

“…This study is based on data from Swiss-AF, an ongoing prospective observational cohort study of AF patients across 14 clinical centres in Switzerland. 15 16 2415 patients with a history of documented AF and mostly aged >65 years were enrolled between April 2014 and August 2017. 228 patients were enrolled in the age range 45–64 years to enable the study of socioeconomic aspects in potentially professionally active patients.…”

Estimating the cost impact of atrial fibrillation using a prospective cohort study and population-based controls

“…To assess the cost impact of AF, the Swiss-AF patients were compared with the population-based controls, using different multivariable regression methods: (1) ordinary least square regression (OLS), (2) OLS-based two-part modelling, (3) generalised linear model (GLM)-based two-part modelling and (4) 1:1 nearest neighbour propensity score matching. Furthermore, (5) estimates were compared with AF costs estimated using a previously developed adjudication algorithm 24. In brief, the AF-adjudication algorithm combined clinical event data collected in Swiss-AF with health insurance claims, adjudicating each cost component as AF-related or non-AF related.…”

“…Third, the different regression-based estimates of AF costs were compared with the estimates of AF costs resulting from applying the AF adjudication algorithm to the Swiss-AF patients’ claims data 24…”

Longitudinal Changes in Health‐Related Quality of Life in Patients With Atrial Fibrillation

“…To identify clusters, a hierarchical cluster analysis was run, as previously described. 12 , 16 Hierarchical clustering is a multivariate statistical approach to create a classification of observations using information on their characteristics. 17 It is performed in such a way that groups are as homogeneous as possible within their class and as dissimilar as possible between the classes.…”

“…Using data from a large prospective cohort study, we aimed to better capture and characterize this heterogeneity by classifying the studied AF population into more homogeneous clusters defined by patient and disease characteristics representing previously described phenotypes 12 and to present longitudinal trajectories of HRQoL over a 5‐year follow‐up period in both the full population and the identified clusters. We were also interested in the HRQoL impact of distinct, clinical events.…”