Skip to main content

Table 1 Accuracy of automated extraction of data elements

From: Development and validation of a novel model for characterizing migraine outcomes within real-world data

 

Manual annotation

EHR structured data using traditional approaches

EHR unstructured data using advanced approaches

Features

Encounter occurrence

Patient-level occurrence

Encounter occurrence

Recall

Precision

F1 score

Encounter occurrence

Recall

Precision

F1 score

Headache

1,996

1,001

737

33.1%

87.7%

48.1%

1,963

97.8%

91.9%

94.8%

Migraine

1,314

809

95

58.5%

82.6%

68.5%

1,233

93.5%

95.1%

94.3%

Headache severity*

 Headache (mild)

104

92

0

0.0%

NA

NA

66

74.0%

81.8%

77.7%

 Headache (severe)

236

194

0

0.0%

NA

NA

197

81.1%

65.5%

72.5%

Severe headache descriptors

 Throbbing

132

107

0

0.0%

NA

NA

87

75.3%

80.5%

77.8%

 Stabbing

53

39

0

0.0%

NA

NA

36

82.1%

88.9%

85.3%

 Pulsating

86

65

0

0.0%

NA

NA

53

70.6%

90.6%

79.3%

Associated symptoms

 Nausea

1,195

757

39

2.5%

75.0%

4.9%

1,144

95.1%

92.1%

93.6%

 Vomiting

886

607

12

1.0%

56.3%

2.0%

864

97.2%

89.2%

93.1%

 Photophobia

457

335

0

0.0%

NA

NA

429

86.6%

87.7%

87.1%

 Phonophobia

319

232

0

0.0%

NA

NA

281

87.5%

96.8%

91.9%

  1. *Sample sizes for moderate headache, disabling headache, debilitating headache, and piercing headache were below requirements for inclusion (< 20 occurrences)
  2. EHR Electronic health record, NA Not applicable