Table 2 Forty-two candidate predictors retained for model training.

From: Construction and validation of a risk prediction model for complications in patients with acute leukemia based on machine learning

Category

Predictor variables (n)

Demographics (4)

Age, sex, body-mass index (BMI), smoking status

Comorbidities (2)

Charlson comorbidity index, ECOG performance status

Baseline laboratory (25)*

Haemoglobin; platelet count; total WBC; absolute neutrophils, lymphocytes, monocytes; peripheral blast %; serum creatinine, urea, sodium, potassium, calcium; ALT, AST, bilirubin, albumin; PT-INR, aPTT, fibrinogen; C-reactive protein, procalcitonin, ferritin, fasting glucose, uric acid

Disease biology (7)

AML vs. ALL subtype, cytogenetic-risk tier, FLT3-ITD, NPM1, TP53, marrow cellularity, blast immunophenotype

Treatment logistics (4)

Induction-regimen class (anthracycline vs. non-anthracycline), central-line placement, antimicrobial prophylaxis, first-dose delay > 48 h

Back to article page