No Financial Disclosure or Conflict of Interest Validation of Mayo Clinic Risk Adjustment Model for In-Hospital Mortality following Percutaneous Coronary Interventions using the National Cardiovascular Data Registry Mandeep Singh; Eric D. Peterson*; Sarah Milford-Beland*; John S. Rumsfeld,# John A. Spertus** Mayo Clinic, Rochester, DCRI* (S.M-B, E.P.), Mid America Heart Institute** (J.A.S.), Denver VA Medical Center# (J.S.R.) No Financial Disclosure or Conflict of Interest CP1087969 Si ngh, M KK 11-12-2002
BACKGROUND Predictive models can assist patients and clinicians in decision-making and informed consent. Existing PCI risk models include angiographic variables limiting routine clinical use. Mayo Clinic Risk Score (MCRS) for in-hospital mortality is based on pre-procedural clinical and non-invasive assessment. MCRS can potentially serve as a risk assessment aid to patients/physicians before coronary angiography for PCI. CP1087969 Si ngh, M KK 11-12-2002
Background External validation of the MCRS is lacking The NCDR cath-PCI registry presents an ideal opportunity to validate the MCRS Study population: Index PCI for 309,351 patients in NCDR participating hospital between January 2004 and March 2006. Outcome: In-hospital mortality during the hospital admission following PCI. CP1087969 Si ngh, M KK 11-12-2002
Mayo Clinic Risk Score (MCRS) Mortality 80 70 60 50 40 30 20 10 5 4 3 2 1 0.5 0.1 Points Score Age (yr) See below ____ Creatinine (mg/dL) See below ____ LV ejection See below ____ fraction (%) Preprocedural shock 9 ____ MI within 24 hours 4 ____ CHF on presentation 3 ____ (without AMI or shock) Peripheral 2 ____ vascular disease Total score ____ C-index=0.90 Estimated risk of death (%) Risk score Age (yr) Creatinine (mg/dL) LV ejection fraction (%) 20 30 40 50 60 70 80 90 1 2 3 4 5 6 7 8 9 10 11 20 40 60 80 2 1 1 2 3 4 5 1 1 2 3 4 5 6 4 3 2 1 CP1087969 Si ngh, M KK 11-12-2002 CP1246788-1
Statistical Methods Using the MCRS equation, predicted probabilities of death were calculated for each patient in the NCDR population. Patients with the same predicted mortality score were grouped together, and within each group, the observed (O) mortality rate was calculated. The O vs. E (expected) mortality rates for these groups were plotted and we used H-L method for calibration Model discrimination was assessed using ROC, or c-statistic, for the entire population and within pre-specified subgroups.
Statistical Methods (Cont.) The analysis was refined to include recalibration of the MCRS equation using the ACC population For this recalibrated model, patients with the same predicted mortality score were again grouped together. O vs. E mortality rates were plotted. Calibration: Hosmer-Lemeshow method. Internal validation of the new model using NCDR PCI patients April 2006, March 2007.
Patient Characteristics by In-Hospital Mortality in the NCDR Variable Number (%) Mortality p Age <60Y 114,844 (37.12) 0.60 <0.0001 ≥80Y 34383 (11.11) 3.22 Congestive heart failure Yes 27003 (8.73) 5.25 <0.0001 No 282,321 (91.26) 0.84 Acute Myocardial infarction Yes 68116 (22.02) 3.44 <0.0001 No 241,128 (77.95) 0.60 Peripheral vascular disease Yes 36568 (11.82) 2.18 <0.0001 No 272,768 (88.17) 1.10 Cardiogenic shock Yes 6314 (2.04) 24.83 <0.0001 No 303,007 (97.95) 0.73 Renal failure Yes 16323 (5.28) 3.89 <0.0001 No 293,012 (94.72) 1.08
Frequency (%) Risk Score Frequency of the Risk, based on the MCRS of Patients Undergoing PCI % Frequency (%) Risk Score
Discrimination of the MCRS Group N MCRS (Min- Max) C-index Overall 309,351 0-25 0.884 Shock/ AMI 69920 4-25 0.873 Age <40 5627 1-21 0.938 Age 65+ 151517 0-25 0.858 CHF 27003 3-25 0.82 Creatinine <0.7 10491 1-20 0.797 Creatinine >1.2 66839 1-25 0.875 Multivessel Dx 150579 0-25 0.87 Female 104110 0-24 0.872 Diabetes 98081 0-24 0.878 CP1087969 Si ngh, M KK 11-12-2002 CP1246782-7
Observed versus expected in-hospital mortality using the original MCRS prediction equation
Observed Mortality (%) Predicted Mortality (%) O vs. E in-hospital mortality with recalibrated quadratic MCRS, internal validation sample (433,045) Observed Mortality (%) 0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 50% 55% 60% 65% 70% Predicted Mortality (%) O=5,177; E=5,310 deaths (difference 2.5 per 100) c index= 0.885
Summary and Conclusions External validation of the MCRS using NCDR confirms its broader applicability. The MCRS has high discrimination for in-hospital mortality using 7 clinical/non-invasive variables. Most variables can be obtained at the time of first visit. This may help the operator to individualize the risk of procedural death from PCI, and to counsel patients at the time of PCI. External validation of the new, recalibrated MCRS model is, however, required. CP1087969 Si ngh, M KK 11-12-2002