info:eu-repo/semantics/doctoralThesis
Predictive Model of Adverse Drug Events in older inpatients.
Modelo Predictivo de Eventos Adversos a Medicamentos en Personas Mayores hospitalizadas
Autor
Sandoval Quijada, Tamara Andrea
Institución
Resumen
Introduction: Adverse Drug Events (ADEs) represent a problem that affects the quality of healthcare and the achievement of therapeutic objectives in Older People (OP). It is important to prevent them and identify those factors that contribute to their appearance and theory predict its occurrence. The ADEs are defined as any injury or harm due to the appropriate or inappropriate use of medications. The ADEs that occur in the context of the appropriate use of the drug, that is, at the appropriate doses to modify any biological function, is known as Adverse Drug Reactions (ADRs). Other ADEs that occur as a result of the selection, use, and/or inappropriate monitoring of drugs are included in the definitions of Medication Errors (ME) or Medication-Related Problems (PRM). Even though ADEs constitute a high burden of morbidity in OP, to date there are two scales (GerontoNet and BADRI) created to predict ADR in geriatric patients. Their clinical use is limited by their low performance, possibly due to the use of indicators of low geriatric validity. None of these scales is oriented to comprehensively predict ADEs in OP, that is, including ME, ADR, and PRM. Therefore, until this work, there would be a gap in the development of this matter, given that published clinical support tools have not been described, which allow identifying the factors associated with the occurrence and prediction of ADEs in hospitalized OP. Objective: To develop a model to predict ADEs in older inpatients. Methods: Through a retrospective cohort study, the data of patients treated in the FONIS SA14ID0141 project, entitled "Impact of the clinical pharmacist in the post- discharge prevention of adverse events to drugs in older adults: Randomized Clinical Trial" was analyzed and identified those antecedents sociodemographic, socioeconomic, morbid, functional, clinical and pharmacotherapeutic with the potential to be used to create a predictive model of ADE in OP during hospitalization. The factors and interactions significantly associated with the occurrence of ADE in this group of patients were detected through bivariate and multivariate logistic regressions. Then, those that, together, presented the best performance to predict EAM were selected. The performance of the model was evaluated with the pseudo R2 tests, the goodness of fit (Hosmer-Lemeshow), and ROC curves (Receiver Operating Characteristic Curve). The validity of the model and its ability to detect the occurrence of ADEs were tested by determining sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV). The stability of the model was studied by calculating the sensitivity and specificity obtained after iteratively applying it to random and representative samples of the total number of selected patients. Results: The sample evaluated consisted of 613 hospitalized OP. The average age was 73.2 ± 8.7 years and 54.3% were women. Of the nine predictive models that met the inclusion of factors and interactions significantly associated with the occurrence of ADE in OP during hospitalization, the one that included the variables significantly associated with the occurrence of ADE and that had the highest clinical applicability and performance was selected. according to the correlation coefficient. The selected predictive model included clinical, functional, and pharmacotherapeutic assessments at the patient's hospital admission. These evaluations were the functionality in the Basic Activities of Daily Living according to the Barthel Index (version 2015), the use of 8 or more medications, the use of medications with anticholinergic load according to the Burden scale, the use of potentially inappropriate medications according to STOPP Criteria (2015 version) and the burden of disease estimated by the Charlson Comorbidity Index. The performance of the selected model showed a sensitivity of 63.60%, a specificity of 58.62%, a PPV of 57.69%, and an NPV of 64.48%. Also, the stability of the model was evaluated and presented a sensitivity of 58.46% and a specificity of 65.19%. Conclusions: The predictive model developed provides new evidence on the existence and application of factors significantly associated with the occurrence and prediction of ADE in OP during hospitalization. The application of this model in a selected sample of hospitalized OP allowed the early identification of 6 out of 10 OP who suffered ADE, considering as a cut-off point a prevalence of ADE during hospitalization of 47.3%. The tool created has a promising performance that must be evaluated through future multicenter and prospective studies, which allow determining its performance and external validity. Once the created model has been validated, it could be used by clinical staff to early identify OP at risk of suffering from ADE during hospitalization and previously adjust drug treatment, thereby improving the quality of health care and patient safety. La tesis doctoral aún no se ha publicado completamente en alguna revista científica.