A single-center abdominal-surgery study identified diazepam-use history as a delirium predictor, but prediction performance declined in a later validation cohort.
Huang and colleagues retrospectively examined patients undergoing elective abdominal surgery at a Shanghai hospital. After screening, the development cohort included 838 patients, of whom 91, or 10.9%, developed postoperative delirium. A later cohort of 346 patients was used for validation; 23 developed delirium.
The investigators screened preoperative and intraoperative characteristics, then built a four-factor model. It included age, diazepam-use history, cerebrovascular-disease history, and intraoperative fluid balance. Diazepam history was therefore one element of a broader clinical prediction, not an isolated explanation of delirium.
Diazepam-use history appeared in 19 patients without delirium and nine patients with delirium in the development cohort. The multivariable model reported an odds ratio of 3.386 for that history, with a p value of 0.007. The variable described a recorded history rather than a measured drug concentration or a standardized perioperative dose.
An adjusted association can help identify patients who merit closer assessment. It cannot show that diazepam caused every delirium episode, or that stopping diazepam would reduce risk by the same numerical amount. The observational study did not test that intervention.
The model’s area under the receiver-operating-characteristic curve was 0.703 in the fitting sample and averaged 0.684 during cross-validation. In the later validation cohort, it was 0.634, with a 95% confidence interval of 0.511–0.758. At the selected fitting-sample threshold, sensitivity was 0.556 and specificity 0.754.
Those figures describe a tool with substantial uncertainty and missed cases. The later cohort came from the same center, so the paper’s external-validation terminology should not be mistaken for confirmation across different hospitals or healthcare systems.
The authors acknowledged retrospective design, single-center sampling, and absent long-term follow-up. For a case review, diazepam history should prompt examination of indication, duration, timing, dose, sleep disturbance, co-medication, and other delirium risks. The study supplies a prediction association; it does not supply a concentration threshold or an individual causal verdict.
Original analysis of the supplied journal article, identified as PMC13020371.