Concordance between administrative health data and medical records for diabetes status in coronary heart disease patients: a retrospective linked data study

Table 4 Concordance measures for the recording of diabetes in hospital discharge data compared with medical records in Indigenous (n = 525) and non-Indigenous (n = 1733) coronary heart disease patients

Lookback period		Observed Agreement,%	Kappa,%	Sensitivity,%	Specificity,%	Positive predictive value,%	Negative predictive value,%	Underestimation (−) /overestimation (+) (%)^*
Index admission	Indigenous	90.3	80.6	84.7	97.4	97.7	83.3^‡	−13.3
Index admission	Non-Indigenous	93.0	81.4	79.6	98.3	94.9	92.3	−16.1
1 year	Indigenous	91.4	82.8	88.8	94.8^†	95.6	86.8^‡	−7.1
1 year	Non-Indigenous	94.1	85.0	84.8	97.8	93.7	94.2	−9.5
2 years	Indigenous	92.2	84.2	90.5^†	94.3^†	95.4	88.6^‡	−5.1
2 years	Non-Indigenous	93.8	84.5	85.4	97.2	92.3	94.4	−7.5
5 years	Indigenous	92.9	85.7	92.9^†	93.0^†	94.5	91.1^†	−1.7
5 years	Non-Indigenous	93.8	84.7	87.2	96.4	90.7	95.0	−3.9
10 years	Indigenous	92.9	85.7	93.6^†	92.2^†	93.9	91.8	−0.3
10 years	Non-Indigenous	93.6	84.3	87.2	96.2	90.2	95.0	−3.3
15 years	Indigenous	92.9	85.7	93.6^†	92.2^†	93.9	91.8	−0.3
15 years	Non-Indigenous	93.6	84.1	87.2	96.1	90.0	95.0	−3.1

Concordance measures calculated using the ICD-10 sample only.
*Calculated from (Sensitivity/PPV – 1) x 100. Negative values represent the percentage underestimation and positive values the percentage overestimation of diabetes recording in hospital discharge data compared with medical records.
P-values are from comparisons between Indigenous and non-Indigenous patients for each lookback period. †p < 0.05, ‡p < 0.0001.

ISSN: 1471-2288