Skip to main content

Table 5 Accuracy of underlying cause of death in CPRD primary care data compared to the death registry gold standard, for the 1022 individuals with cause of death recorded in both sources. For coronary deaths not recorded as coronary in CPRD, the most common causes in CPRD were I469 ‘Cardiac arrest’, I500 ‘Congestive heart failure’ and I501 ‘Left ventricular failure’. For stroke deaths not recorded as stroke in CPRD, the most common causes in CPRD were ‘J180 Bronchopneumonia, unspecified’, ‘J189 Pneumonia, unspecified’ and ‘F03X Unspecified dementia’

From: Natural language processing for disease phenotyping in UK primary care records for research: a pilot study in myocardial infarction and death

Source of cause of death record in CPRD

Free text

Coded

Number of deaths

381

641

Same underlying cause

184 (48.3%)

293 (45.7%)

Same 2-character ICD-10 code for underlying cause

222 (58.3%)

371 (57.9%)

Same ICD-10 chapter for underlying cause

278 (73.0%)

463 (72.2%)

Coronary deaths (ICD-10 I20–I25, N = 163):

 Sensitivity, %

65.3 (50.4, 78.3)

68.4 (59.1, 76.8)

 Specificity, %

97.9 (95.7, 99.1)

98.3 (96.8, 99.2)

Cerebrovascular deaths (ICD-10 F01, I60–I69, N = 101):

 Sensitivity, %

66.7 (51.6, 79.6)

58.5 (44.1, 71.9)

 Specificity, %

98.5 (96.5, 99.5)

97.8 (96.2, 98.8)

Cancer deaths (ICD-10 C00–C97, N = 268):

 Sensitivity, %

93.0 (86.1, 97.1)

80.4 (73.5, 86.1)

 Specificity, %

95.7 (92.7, 97.8)

98.5 (97.0, 99.4)