Browsing by Issue Date, starting with "2025-01-01"
Now showing 1 - 3 of 3
Results Per Page
Sort Options
- High Positive End-Expiratory Pressure (PEEP) with Recruitment Maneuvers versus Low PEEP during General Anesthesia for Surgery: A Bayesian Individual Patient Data Meta-analysis of Three Randomized Clinical TrialsPublication . Mazzinari, Guido; Zampieri, Fernando G.; Ball, Lorenzo; Campos, Niklas S.; Bluth, Thomas; Hemmes, Sabrine N. T.; Ferrando, Carlos; Librero, Julian; Soro, Marina; Pelosi, Paolo; Abreu, Marcelo Gama de; Schultz, Marcus J.; Neto, Ary Serpa; Nunes, Catarina S.Background: The influence of high positive end-expiratory pressure (PEEP) with recruitment maneuvers on the occurrence of postoperative pulmonary complications after surgery is still not definitively established. Bayesian anal ysis can help to gain further insights from the available data and provide a probabilistic framework that is easier to interpret. The objective was to estimate the posterior probability that the use of high PEEP with recruitment maneuvers is associated with reduced postoperative pulmonary complications in patients with intermediate-to-high risk under neutral, pessimistic, and opti mistic expectations regarding the treatment effect. Methods: Multilevel Bayesian logistic regression analysis was performed on individual patient data from three randomized clinical trials carried out on sur gical patients at intermediate to high risk for postoperative pulmonary com plications. The main outcome was the occurrence of postoperative pulmonary complications in the early postoperative period. This study examined the effect of high PEEP with recruitment maneuvers versus low PEEP ventilation. Priors were chosen to reflect neutral, pessimistic, and optimistic expectations of the treatment effect. Results: Using a neutral, pessimistic, or optimistic prior, the posterior mean odds ratio for high PEEP with recruitment maneuvers compared to low PEEP was 0.85 (95% credible interval, 0.71 to 1.02), 0.87 (0.72 to 1.04), and 0.86 (0.71 to 1.02), respectively. Regardless of prior beliefs, the posterior proba bility of experiencing a beneficial effect exceeded 90%. Subgroup analysis indicated a more pronounced effect in patients who underwent laparoscopy (odds ratio, 0.67 [0.50 to 0.87]) and those at high risk for postoperative pul monary complications (odds ratio, 0.80 [0.53 to 1.13]). Sensitivity analysis, considering severe postoperative pulmonary complications only or applying a different heterogeneity prior, yielded consistent results. Conclusions: High PEEP with recruitment maneuvers demonstrated a moderate reduction in the probability of postoperative pulmonary complica tion occurrence, with a high posterior probability of benefit observed consis tently across various prior beliefs, particularly among patients who underwent laparoscopy.
- Face-to-face interactions estimated using mobile phone data to support contact tracing operationsPublication . Cumbane, Silvino; Gidófalvi, Gyözö; Cossa, Osvaldo; Madivadua Júnior, Afonso; Branco, Frederico; Sousa, NunoUnderstanding people’s face-to-face interactions is crucial for effective infectious disease management. Traditional contact tracing, often relying on interviews or smartphone applications, faces limitations such as incomplete recall, low adoption rates, and privacy concerns. This study proposes utilizing anonymized Call Detail Records (CDRs) as a substitute for in-person meetings. We assume that when two individuals engage in a phone call connected to the same cell tower, they are likely to meet shortly thereafter. Testing this assumption, we evaluated two hypotheses. The first hypothesis—that such co-located interactions occur in a workplace setting—achieved 83% agreement, which is considered a strong indication of reliability. The second hypothesis—that calls made during these co-location events are shorter than usual—achieved 86% agreement, suggesting an almost perfect reliability level. These results demonstrate that CDR-based co-location events can serve as a reliable substitute for in-person interactions and thus hold significant potential for enhancing contact tracing and supporting public health efforts.
- Mitigating false negatives in imbalanced datasets: an ensemble approachPublication . Cavique, Luís; Vasconcelos, MarceloImbalanced datasets present a challenge in machine learning, especially in binary classification scenarios where one class significantly outweighs the other. This imbalance often leads to models favoring the majority class, resulting in inadequate predictions for the minority class, specifically in false negatives. In response to this issue, this work introduces the MinFNR ensemble algorithm, designed to minimize False Negative Rates (FNR) in imbalanced datasets. The new approach strategically combines data-level, algorithmic-level, and hybrid-level approaches to enhance overall predictive capabilities while minimizing computational resources using the Set Covering Problem (SCP) formulation. Through a comprehensive evaluation of diverse datasets, MinFNR consistently outperforms individual algorithms, showing its potential for applications where the cost of false negatives is substantial, such as fraud detection and medical diagnosis. This work also contributes to ongoing efforts to improve the reliability and effectiveness of machine learning algorithms in real imbalanced scenarios.