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

Table 3 Concordance measures for the recording of diabetes in hospital discharge data compared with medical records, in the sample of coronary heart disease patients

	Observed agreement,%		Kappa,%		Sensitivity,%		Specificity,%		Positive predictive value,%		Negative predictive value,%		Underestimation (−) / overestimation (+)^*
Lookback period	ICD-9	ICD-10	ICD-9	ICD-10	ICD-9	ICD-10	ICD-9	ICD-10	ICD-9	ICD-10	ICD-9	ICD-10	ICD-9	ICD-10
Index admission	97.0	92.3	91.2	82.5	91.1	81.5^‡	98.7	98.2	95.3	96.0	97.4	90.8^‡	−4.4	−15.1^†
1 year	96.6	93.4	90.2	85.3	91.6	86.3^†	98.1	97.3	93.3	94.4	97.6	93.0^‡	−1.8	−8.6^†
2 years	96.6	93.4	90.3	85.4	92.1	87.3^†	97.9	96.7	92.8	93.5	97.8	93.4^‡	−0.7	−6.6^†
5 years	96.5	93.6	90.0	85.9	92.4	89.3	97.7	95.9^†	92.1	92.2	97.8	94.4^‡	+0.3	−3.1^†
10 years	96.5	93.5	90.0	85.6	92.6	89.6	97.6	95.6^†	91.9	91.6	97.8	94.5^‡	+0.8	−2.2^†
15 years	96.4	93.4	89.7	85.5	92.6	89.6	97.5	95.5^†	91.4	91.5	97.8	94.5^‡	+1.3	−2.1^†

* 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 comparison of ICD-9 and ICD-10 for each lookback period. †p < 0.05. ‡p < 0.0001.
ICD, International Classification of Diseases.

ISSN: 1471-2288