Skip to main content

Table 3 Evaluation of reasons for coding false positives concepts by type of variable on the test set

From: Evaluation of natural language processing from emergency department computerized medical records for intra-hospital syndromic surveillance

 

Chief complaint

Observation

Specialists' notes

Procedures

Prescriptions

Total

 

N*

%

95% CI

N*

%

95% CI

N*

%

95% CI

N*

%

95% CI

N*

%

95% CI

N*

%

95% CI

Temporality of event not recognized (antecedents)

0

-

-

157

35.8

31.3-40.5

0

-

-

0

-

-

0

-

-

157

35.4

31.0-40.1

False disambiguation of concept

2

66.7

9.4-99.2

253

57.8

53.0-62.4

1

100

-

1

100

-

0

-

-

257

58.0

53.3-62.7

Negation not detected

0

0

-

19

4.3

2.6-6.7

0

-

-

0

-

-

0

-

-

19

4.3

2.6-6.6

False disambiguation of abbreviation or acronym

1

33.3

0.8-90.6

6

1.4

0.5-3.0

0

-

-

0

-

-

0

-

-

7

1.6

0.6-3.2

Other UrgIndex error

0

-

-

3

0.7

0.1-2.0

0

-

-

0

-

-

0

-

-

3

0.7

0.1-2.0

Total number of false positive concepts

3

100

-

438

100

-

1

100

-

1

100

-

0

 

-

443

100

-

  1. UrgIndex- Processing of emergency services medical narrative records (DMU) of the University of Lyon's Hôpital de la Croix-Rousse, Lyon, France
  2. *N = Number of evaluated concepts