Researchers evaluated biopsy-free, gradient-boosting risk scoring models intended to reduce bone marrow biopsy use among cytopenic patients in Thailand. In a cohort of 285 patients, they tested four computational models based on previously published frameworks and evaluated performance using discrimination metrics and decision thresholds. Two models achieved decision thresholds with 100% negative predictive value, identifying subgroups with no evidence of myeloid malignancy where invasive diagnostic procedures might be safely deferred pending confirmation. The study highlights how data-mined clinical variables—demographics, routine labs, and in some models mutation information—can be combined to create practical triage tools, particularly in settings where biopsies carry burden or limited access. For hematology workflows, the near-term translation will depend on prospective validation, calibration across sites, and how often deferred patients convert to diagnosed malignancy during follow-up.