Skip to main content
. Author manuscript; available in PMC: 2018 Apr 24.
Published in final edited form as: Proc SPIE Int Soc Opt Eng. 2018 Mar 7;10577:105770J. doi: 10.1117/12.2293954

Table 4.

The ten most important features selected from entire EHR variables using Lasso-LR

Ranking Features EHR component Values the “most harmful” cases (N=395) the “least harmful” cases (N=399) p-value
1 History of tobacco use symptoms present 189 (47.85%) 29 (7.27%) p<0.001
Not present 206 (52.15%) 370 (92.73%)
2 Tobacco use disorder symptoms present 189 (47.85%) 29 (7.27%) p<0.001
Not present 206 (52.15%) 370 (92.73%)
3 Mammography, screening, bilateral (two view film study of each breast) procedures present 116 (29.37%) 229 (57.39%) p<0.001
Not present 279 (70.63%) 170 (42.61%)
4 Other nonoperative measurements and examinations diagnoses present 191 (48.35%) 63 (15.79%) p<0.001
Not present 204 (51.65%) 336 (84.21%)
5 Prescription(s) generated and transmitted via a qualified eRx system procedure present 93 (23.54%) 0 (0%) p<0.001
Not present 302 (76.46%) 399 (100%)
6 Normal non-HDL cholesterol laboratory results present 90 (22.78%) 1 (0.25%) p<0.001
Not present 305 (77.22%) 398 (99.75%)
7 Hydrocodone acetaminophen medications present 105 (26.58%) 21 (5.26%) p<0.001
Not present 290 (73.42%) 378 (94.74%)
8 Cause of death symptoms present 190 (48.10%) 278 (69.67%) p<0.001
Not present 205 (51.90%) 121 (30.33%)
9 Computer-Aided Detection (CAD) mammography procedures present 72 (18.23%) 3 (0.75%) p<0.001
Not present 323 (81.77%) 396 (99.25%)
10 Gynecological examination diagnoses present 95 (24.05%) 173 (43.36%) p<0.001
Not present 300 (75.95%) 226 (56.64%)