Table 4. The C-statistics and concordance index (CI) with 95% confidence intervals using 10-fold cross validation for mortality and cardiovascular disease hospitalization among newly diagnosed hypertension by data sources and ‘look back’ intervals.
C-statistics | CI | |
---|---|---|
Mortality | ||
Claims only | ||
6 months | 0.859(0.853–0.865) | 0.635(0.630–0.641) |
1 year | 0.861(0.855–0.867) | 0.636(0.631–0.642) |
2 years | 0.863(0.859–0.868) | 0.641(0.635–0.646) |
3 years | 0.863(0.858–0.868) | 0.642(0.637–0.647) |
Hospital discharge abstract only | ||
6 months | 0.881(0.876–0.886) | 0.632(0.627–0.636) |
1 year | 0.891(0.885–0.897) | 0.636(0.631–0.640) |
2 years | 0.890(0.885–0.896) | 0.639(0.635–0.643) |
3 years | 0.892(0.886–0.897) | 0.639(0.634–0.645) |
Claims and Hospital discharge abstract | ||
6 months | 0.892(0.887–0.897) | 0.642(0.637–0.647) |
1 year | 0.911(0.906–0.915) | 0.646(0.640–0.651) |
2 years | 0.912(0.906–0.917) | 0.645(0.639–0.650) |
3 years | 0.912(0.907–0.917) | 0.647(0.641–0.652) |
Cardiovascular disease hospitalization | ||
Claims only | ||
6 months | 0.733(0.728–0.738) | 0.547(0.541–0.552) |
1 year | 0.735(0.730–0.739) | 0.549(0.543–0.554) |
2 years | 0.738(0.732–0.744) | 0.552(0.546–0.557) |
3 years | 0.739(0.732–0.745) | 0.556(0.550–0.562) |
Hospital discharge abstract only | ||
6 months | 0.739(0.733–0.744) | 0.544(0.538–0.550) |
1 year | 0.741(0.735–0.746) | 0.546(0.541–0.552) |
2 years | 0.747(0.741–0.751) | 0.552(0.547–0.557) |
3 years | 0.749(0.744–0.755) | 0.554(0.549–0.560) |
Claims and Hospital discharge abstract | ||
6 months | 0.756(0.751–0.761) | 0.580(0.574–0.586) |
1 year | 0.768(0.763–0.773) | 0.611(0.606–0.616) |
2 years | 0.769(0.763–0.774) | 0.612(0.607–0.618) |
3 years | 0.770(0.764–0.775) | 0.614(0.609–0.619) |
The C-statistics were estimated using a logistic regression model while adjusting for age group, sex and 17 Charlson conditions. The confidence intervals were estimated using bootstrap method.
CI was estimated using Cox's proportional regression model while adjusting for age group, sex, and 17 Charlson conditions. The confidence intervals were estimated using bootstrap method.