External validation, extension and recalibration of Braunwald's simple risk index in a community-based cohort of patients with both STEMI and NSTEMI

https://doi.org/10.1016/j.ijcard.2005.03.052Get rights and content

Abstract

Background

Using the simple risk index (SRI) that is based on age, heart rate and systolic blood pressure, we sought to evaluate the ability to predict outcome in AMI patients in a community-based population.

Methods and results

We identified and evaluated 3684 consecutive patients with an admission diagnosis of possible AMI, who attended between 1st September and 30th November 1995. Two thousand one hundred fifty three patients had confirmed evidence of WHO definition AMI, of whom 1656 survived to hospital discharge. We evaluated the ability of the SRI to predict mortality over 30 days using the score generated by the equation (heart rate × [age / 10]2) / systolic blood pressure. The SRI was a strong (c-statistic = 0.77 CI 0.74–0.79) predictor of 30-day mortality in both STEMI and all consecutive cases of WHO definition AMI. However, the model showed poor calibration when used on a community-based population with 30-day mortality being underestimated across all risk quintiles. Consequently we sought to recalibrate the quantitative aspects of the model for the total AMI population in the following way (Risk Index; 30-day mortality) (≤ 29.2; 9.2%), (29.3–37.8; 23.9%), (37.9–47.3; 34.6%), (47.4–61.5; 40.3%), (≥ 61.6; 65.5%).

Conclusion

We have externally validated the SRI in an unselected cohort of consecutive WHO definition AMI patients. However, the original model consistently underestimated the likelihood of death at 30 days regardless of the calculated risk score. We have therefore suggested corrections to the risk estimates, to allow its application in an unselected community cohort.

Section snippets

Background

Many methods for risk adjustment have been derived in an attempt to reliably compare care at different clinical centres and also at different points in time. With the exception of age and gender standardisation, most of the risk models have been derived from randomised control trial (RCT) populations and tend to be limited by complexity and also exclusion of patients with higher risk profiles. For hospitals to be accountable for the care they provide to AMI patients, they need to be able to

The EMMACE-1 project

The EMMACE-1 project was supported by a UK National Health Service (NHS) Research and Development grant being carried out on behalf of the Yorkshire Cardiology Working Group. We obtained ethical approval and the co-operation of all consultants and clinical audit departments in 20 adjacent hospitals. Further study details have been published previously [13], [14].

Patient population

Over a three-month period (1 September to 30 November 1995) 3684 potential cases of acute MI were identified in 20 adjacent hospitals

Results

We analysed 2153 consecutive patients with AMI admitted over a three-month period to 20 adjacent UK hospitals. Summary demographic and clinical details for both the STEMI cohort and all cases of AMI are shown in Table 1. The mean age was 71 years for all AMI and 70 years for STEMI. Associated mortality rates at 30 days were 22% for STEMI and 24% for all AMI. The SRI values (Table 2) at presentation for the total population ranged from 4.1 to 161.5 with a mean risk score of 33.0 (SD 18.3). It is

Discussion

The SRI was derived and validated in the subset of patients eligible for thrombolysis in a clinical trial. In our investigation we have externally validated the model in an unselected population of patients with STEMI and shown it to be a robust predictor of 30-day mortality. The c-statistic of 0.76 indicates good discrimination of the model. Moreover, we have also demonstrated its utility at predicting mortality for all cases of AMI, i.e. STEMI and NSTEMI. However, the SRI approach as

Acknowledgement

This project was commissioned and funded by the National Health Service Research and Development programme.

References (17)

There are more references available in the full text version of this article.

Cited by (0)

View full text
Copyright © 2005 Elsevier Ireland Ltd. All rights reserved.