Maximizing Interpretability and Cost-Effectiveness of Surgical Site Infection (SSI) Predictive Models Using Feature-Specific Regularized Logistic Regression on Preoperative Temporal Data

Table 4

Upper half of the total number of features selected (above 50%) and their coefficient’s signs in 100 repetitions of 10-fold cross-validation for FLM.

FLM variables		Sign
Leukocytes_nmbr_test_M	100	+
Sodium_nmbr_test_M	100	+
CRP_mean_M	99	+
Thrombocytes_prop_nmbr_test_M	99	+
Hemoglobin_prop_nmbr_test_S	97	−
Hemoglobin_prop_nmbr_test_M	97	+
Albumin_slope_L	96	−
Leukocytes_prop_nmbr_test_L	96	+
Albumin_mean_M	93	−
Amylase_prop_nmbr_test_L	93	+
Bilirubin.total_mean_M	91	+
CRP_prop_nmbr_test_S	88	−
Creatinine_slope_M	77	+
ALP_high_abn_prop_M	73	−
CRP_slope_L	73	+
Potassium_high_abn_prop_M	69	+
Creatinine_prop_nmbr_test_M	67	+
Hemoglobin_nmbr_test_M	56	+
Glucose_nmbr_test_M	55	+
ASAT_low_abn_prop_M	53	+
Albumin_slope_M	51	+