{
  "abstract": "Objective Atrial fibrillation in elderly patients increases the risk of thromboembolism, necessitating long-term anticoagulation. While non-vitamin K oral anticoagulants (NOACs) are generally preferred, appropriate dosing in older patients who are frail remains a challenge. This study aimed to evaluate the impact of NOAC underdosing and identify bleeding risk factors using artificial intelligence in a local elderly population.Methods A retrospective study was conducted that included 119 patients with atrial fibrillation who were treated with apixaban or rivaroxaban between October 2020 and May 2022. Patients were categorised based on whether NOAC prescriptions were in accordance with dosing recommendations. Bivariate analyses and univariable logistic regression were performed to assess associations with clinical outcomes. To identify bleeding risk factors, a combination of stepwise logistic regression, learning vector quantisation and variable permutation was used. These risk factors were then used to develop supervised machine learning models to predict bleeding risk, for interpretation purposes.Results Significant differences in bleeding and thrombotic events were observed between patients with guideline-concordant and underdosed prescriptions. Using univariable logistic regression, underdosing NOACs was associated with a lower risk of bleeding (OR 0.3) but a higher risk of thrombosis (OR 6.7). In the multivariable analysis, guideline adherence, sex and NOAC choice were identified as key predictors of bleeding events. Guideline-concordant prescriptions were independently associated with an increased bleeding risk.Conclusions Underdosing NOAC was associated with a reduced bleeding risk but at the cost of a markedly increased thrombosis risk. Guideline-concordant dosing was also associated with higher bleeding risk in the multivariable model. Overall, the results do not support systematic underdosing of NOACs in elderly patients. These findings were shared with local prescribers to reinforce appropriate dosing practices and to improve follow-up for patients identified as being at increased bleeding risk.",
  "authors": [
    {
      "affiliations": [
        "Pharmacy, Hôpital Privé Guillaume de Varye, Saint-Doulchard, France",
        "Pharmacy, Centre Hospitalier Intercommunal des Alpes du Sud, Gap, France"
      ],
      "name": "Dorian Protzenko"
    },
    {
      "affiliations": [
        "Centre Hospitalier Universitaire de Nice Hopital Pasteur, Nice, France"
      ],
      "name": "Cecile Berard"
    },
    {
      "affiliations": [
        "Geriatric Ward, CH des Escartons Briançon, Briançon, France"
      ],
      "name": "Vincent Hoang"
    },
    {
      "affiliations": [
        "Pharmacy, Assistance Publique Hopitaux de Marseille, Marseille, France"
      ],
      "name": "Guillaume Hache"
    }
  ],
  "title": "Evaluation of the impact of NOAC underdosing and exploration of bleeding risk factors in elderly patients with atrial fibrillation: artificial intelligence-based approach",
  "uid": "7157cd0e-bc7d-50e3-906d-ece6aa24a263"
}