Table 1 Result on the MIMIC-IV datasets on the test set with the weighted BCE loss and a batch size of 20. The fusion models use the MAGBERT mechanism. The table also presents the average AUROC along with its confidence interval (CI).

From: Affordable and real-time antimicrobial resistance prediction from multimodal electronic health records

Data and Settings

Models

AUPR

AUROC

\(\overline{\text {AUROC}}\)

CI

Gentamicin

\(T = 4, dt=1\)

learning rate: \(1e-4\)

Line

0.0886

0.5027

0.5980

(0.5439, 0.6521)

LSTM

0.1136

0.5852

Star

0.1274

0.5893

Encoder

0.1151

0.5547

BERT

0.1644

0.5807

BertLstm

0.1445

0.6077

BertStar

0.1631

0.6874

BertEncoder

0.1249

0.6224

LstmBert

0.1422

0.6156

StarBert

0.1379

0.6495

EncoderBert

0.1364

0.5824

Gentamicin

\(T=3, dt= 1\)

learning rate: \(5e-5\)

LSTM

0.0935

0.4659

0.5388

(0.4818, 0.5958)

Star

0.1041

0.5173

BERT

0.1513

0.4927

BertLstm

0.0935

0.4659

BertStar

0.1271

0.6103

LstmBert

0.1395

0.5770

StarBert

0.1223

0.5883

BertEncoder

0.1474

0.6145

EncoderBert

0.1041

0.5173

P. aeruginosa

\(T=3, dt=1\)

learning rate: \(1e-5\)

LSTM

0.2166

0.5000

0.5775

(0.5500, 0.6050)

BERT

0.2120

0.6000

BertLstm

0.2405

0.5800

BertStar

0.2407

0.6300

Back to article page