The analysis used retrospective data from 596 unique patients who had both a complete blood count and dengue NS1 antigen testing performed on the same day at Uttara Adhunik Medical College Hospital. Data collection spanned 19 August to 12 October 2023. One-half of the cohort were NS1 positive: investigators reported 306 positives, representing 51.3% of the sample. The study team explicitly treated these data as a single-center, retrospective dataset and did not perform prospective enrollment.
Investigators applied least absolute shrinkage and selection operator (LASSO) penalized logistic regression for predictor selection from routinely available demographic and hematological candidates. After LASSO identified a subset of predictors, the authors refit those selected variables using unpenalized logistic regression to derive an integer CBC score. The source did not provide the full list of candidate predictors considered prior to selection, nor did it provide the exact regression coefficients in the abstract.
The final parsimonious model retained three hematological parameters: platelet count (rescaled per 10 × 10^9/L), white blood cell count, and monocyte count. The article presents the resulting instrument as a three-parameter score intended for same-day triage among clinically suspected dengue patients. The source did not report additional laboratory or clinical items in the final score beyond those three CBC parameters.
Discriminative performance was very high in internal testing. The optimism-corrected area under the receiver operating characteristic curve (AUC) was reported as 0.9429. The study team derived a Youden-based cutoff of 11.825 on the integer score; performance at that threshold included sensitivity 0.833 (95% CI, 0.787–0.873) and specificity 0.966 (95% CI, 0.938–0.983). Positive predictive value was 0.962 (95% CI, 0.932–0.982) and negative predictive value 0.846 (95% CI, 0.802–0.883). Overall accuracy at the Youden cutoff was 0.898 (95% CI, 0.870–0.921). The authors noted that rounding the cutoff to 12 produced similar diagnostic performance.
The study included a benchmark analysis against platelet count as a single predictor. The model reported a modest improvement in discrimination compared with platelet count alone: AUC for the three-parameter model was reported at 0.944 versus 0.931 for platelet count alone, yielding a Delta AUC of 0.012 and a DeLong p-value of 0.00364. The source provided these comparative AUC figures and the statistical test result in the abstract.
Internal validation used bootstrap optimism correction with 400 resamples to adjust estimates for overfitting. Investigators reported the optimism-corrected AUC value noted above. The abstract did not include numerical calibration statistics, calibration plots, or decision-curve analysis in the available summary, and the full model coefficients were not listed in the abstract text.
The authors framed the score as a CBC-only triage adjunct intended to estimate the likelihood of same-day NS1 antigen positivity when confirmatory testing may be delayed or intermittently available in high-volume settings. They explicitly cautioned that the score is not a replacement for confirmatory dengue testing. Key limitations described in the source include the single-center, retrospective design and the absence of external validation; authors called for validation in independent cohorts before routine clinical use. The source did not report prospective implementation data, cost-effectiveness evaluation, or threshold-specific guidance for different clinical settings.
Limitations and unanswered questions that remain in the source: the abstract does not provide the complete list of candidate predictors initially considered, it does not present full model coefficients or a ready-to-use scoring sheet in the abstract, and external generalizability remains untested. These gaps define the next steps required before the score could be recommended for operational triage workflows.