Skip to main content

Table 4 Summary of Medicare datasets

From: The effects of class rarity on the evaluation of supervised healthcare fraud detection models

Year

Part B

Part D

DMEPOS

Combined

Fraud

%Fraud

Fraud

%Fraud

Fraud

%Fraud

Fraud

%Fraud

(a) By year

2012

546

0.062

–

–

–

–

–

–

2013

403

0.044

465

0.069

323

0.110

229

0.090

2014

285

0.030

329

0.047

193

0.068

154

0.061

2015

175

0.018

224

0.031

119

0.041

90

0.035

2016

99

0.010

135

0.018

75

0.026

55

0.021

PCC

Part B

Part D

DMEPOS

Combined

Fraud

%Fraud

Fraud

%Fraud

Fraud

%Fraud

Fraud

%Fraud

(b) Class imbalance and rarity

All

1409

0.038

1018

0.048

635

0.074

437

0.062

1000

1000

0.027

–

–

–

–

–

–

400

400

0.011

400

0.019

400

0.046

–

–

200

200

0.005

200

0.010

200

0.023

200

0.026

100

–

–

100

0.005

100

0.012

100

0.013