(BSPOT10) Impact of longitudinal data completeness on risk adjustment and outcome evaluation in hospitals database

Background: Data discontinuity due to patients receiving care outside index hospital could result in information bias in baseline comorbidity assessment and discontinuity of follow-up.

Objectives: To evaluate the impact of longitudinal data completeness on the risk evaluation of AF recurrence post ablation between paroxysmal (PAF) and persistent (PsAF) AF patients through data linkage.

Methods: Patients undergoing catheter ablation for atrial fibrillation (AF) from 2018 to 2020 were extracted from the Premier Healthcare Database (PHD) and linked with Komodo claims data through tokenization. AF recurrence was defined as the composite of repeat ablation, AF-related hospitalization, or direct cardioversion. AF recurrences within 1-year post-index and baseline Elixhauser comorbidities within 1 year pre-index were identified in the hospital database and linked database, respectively. Propensity score matching was used to match PAF and PsAF patients on patients’ demographic and Elixhauser comorbidities, and hospitals’ characteristics in hospital and linked databases respectively. The risk of AF recurrence between PAF and PsAF was analyzed using logistic regression with matched data. Sensitivity analysis was conducted to compare the risk of each component of AF recurrence.

Results: In the hospital dataset, there were 255 (11.5%) cases of AF recurrence among PAF patients and 608 (19.9%) among PsAF patients, which were lower than the linked dataset with 14.8% for PAF and 25.2% for PsAF patients. As expected, the linked dataset had a higher proportion of Elixhauser comorbidities than hospital database with the median difference of 4.8% (interquartile range: 1.75% - 9.6%) and with the larger difference in hypertension (23.3%), valvular disease (23.3%), and obesity (15.9%). Using the hospital dataset, PsAF patients had a higher risk (odds ratio (OR): 1.73 (95% confidence intervals (CI): 1.43 – 2.08)) of AF recurrence than PAF patients, which was consistent with the results (OR (95% CI): 1.72 (1.45-2.05)) from the linked dataset. The results for each component of AF recurrence were consistent between the hospital and linked dataset.

Conclusions: Results from Hospital data showed that PsAF had higher risk of AF recurrence than PAF patients. The results were consistent with results using linked data and aligns with findings in the literature.