Skip to main content
. 2023 Feb 7;380:e071058. doi: 10.1136/bmj-2022-071058

Table 7.

Example table showing results from model validation and subsequent updating161

Clinical model
(n=5677)
Parma
(n=1241)
Miami (n=821) Innsbruck (n=668) Rotterdam (n=471)
Prediction model performance
C statistic (95% CI) 0.79 0.78 (0.74 to 0.81) 0.80 (0.74 to 0.85) 0.78 (0.73 to 0.82) 0.81 (0.76 to 0.88)
Calibration-in-the-large 0.334 (P=0.001) −0.046 (P=0.76) 0.113 (P=0.33) 0.344 (P=0.01)
Recalibration slope −0.117 (P=0.14) 0.046 (P=0.75) 0.023 (P=0.85) 0.097 (P=0.48)
Model coefficients
Intercept β= −7.539 δ=1.010 (P=0.14) δ= −0.292 (P=0.77) δ= −0.355 (P=0.72) δ= −0.420 (P=0.72)
Age β=0.062 δ= −0.003 (P=0.72) δ= −0.015 (P=0.29) δ= −0.018 (P=0.18) δ= −0.013 (P=0.43)
Male sex β=1.332 δ= −0.015 (P=0.94) δ=0.199 (P=0.56) δ= 0.184 (P=0.50) δ= 0.161 (P=0.59)
Atypical chest pain β=0.633 δ= −0.470 (P=0.06) δ= −0.782 (P=0.05) δ= −0.569 (P=0.17) δ= −0.311 (P=0.53)
Typical chest pain β=1.998 δ= −0.615 (P=0.01) δ= −0.485 (P=0.44) δ= −0.839 (P=0.14) δ= −0.194 (P=0.70)
Diabetes β=0.828 δ=0.241 (P=0.32) δ=0.088 (P=0.79) δ= −0.043 (P=0.88) δ=0.038 (P=0.92)
Hypertension β=0.338 δ= 0.096 (P=0.59) δ= −0.433 (P=0.18) δ= −0.274 (P=0.30) δ=0.219 (P=0.45)
Dyslipidaemia β=0.422 δ= −0.131 (P=0.43) δ=0.008 (P=0.98) δ= −0.042 (P=0.89) δ= −0.117 (P=0.70)
Smoking β=0.461 δ= −0.121 (P=0.57) δ=0.449 (P=0.77) δ=0.022 (P=0.93) δ=0.222 (P=0.50)

β=estimated regression coefficient; δ=difference between the hospital-specific predictor effects and the predictor effects as estimated in the development data. P values above 0.05 support the validity of the developed prediction model in a specific hospital (Parma, Miami, Innsbruck, or Rotterdam).