Skip to main content

Table 2 Results of the NLP pipeline applied to all clinical notes of 2020

From: Information extraction from free text for aiding transdiagnostic psychiatry: constructing NLP pipelines tailored to clinicians’ needs

Theme

Mean number of sentences concerning theme per trajectory (sd)

Mean number of sentences with relevant change in theme per trajectory (sd)

Examples of sentences translated from Dutch marked as correct

Classification accuracy of pipeline

Precision

Recall

F1-score

Symptom reduction

103 (99.8)

8.0(10.2)

“Nervousness increased over the course of the day”,

“The patient appears drowsier than before”

0.988

0.857

0.461

0.6

Social functioning

131 (121)

4.0 (5.1)

“Friendly, more interaction than yesterday”

0.997

0.750

1.00

0.857

General well-being

119 (126)

6.4 (8.9)

“This afternoon, the patient felt less well”,

“Had less energy”

0.951

0.833

0.250

0.385

Patient experience

164 (159)

9.4 (11.0)

“Says that it is going well, has the idea that it is going better and better”

0.993

0.333

1.00

0.5