Answers are grounded in this paper and cite their sources.
Cross-sectional study assesses agreement of algorithm and specialists in antibiotic therapy, suggesting algorithms can support clinical decisions.
Variability among specialists within the same specialty limits studies that use clinical judgment as a gold standard and can lead to inconsistent decisions. This study assessed whether agreement between a rules-based clinical algorithm and infectious disease specialists could be comparable to the agreement observed between the specialists themselves, taken as a reference for acceptable variability in daily practice. Cross-sectional study with 87 cases formed by simulated clinical scenarios and real urine culture antibiograms. A deterministic algorithm was created by two infectious disease specialists not involved in analyzing these cases for the study. Two other independent infectious disease specialists (E1 and E2) and a deterministic algorithm classified, blindly, the best antibiotic therapy in each scenario. Agreement proportions were calculated for three pairs: algorithm–E1, algorithm–E2, and E1–E2. Ninety-five percent confidence intervals (95% CI) were estimated using the Wilson method; CI overlap indicated no statistically significant difference. Finally, a chi-square test was used to assess whether there were statistically significant differences in agreement E1–E2 compared to algorithm–E1 and algorithm–E2, considering p<0.05 significant. Agreement between E1 and E2 was 77.0% (95% CI 67.1%–84.6%). The algorithm agreed with E1 in 78.2% of cases (95% CI 68.4%–85.6%) and with E2 in 90.8% (95% CI 82.9%–95.3%), overlapping the benchmark, so the hypothesis that agreement between specialists is similar to each specialist’s agreement with the algorithm could not be rejected. Comparing algorithm–E1 to E1–E2, p=0.856, indicating no statistical difference. In contrast, agreement algorithm–E2 was statistically different from agreement E1–E2 (p=0.013). It was not possible to refute similarity of agreement between the algorithm and one of the specialists, and the likelihood that the higher agreement between the algorithm and the second specialist was due to chance is low compared with the agreement between the specialists themselves. In other words, these findings suggest that algorithms may operate within the existing standards of variability in clinical prescribing, reinforcing their potential as decision-support tools for antibiotic therapy at a level similar to specialists.
No takes yet. Share an insight, caveat, or question.
Marchetti et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: