Development and internal validation of the Simplified Mortality Score for the Intensive Care Unit (SMS‐ICU)

Abstract

The SMS-ICU predicted 90-day mortality with reasonable and stable performance. If performance remains adequate after external validation, the SMS-ICU could prove a valuable tool for ICU clinicians and researchers because of its simplicity and expected very low number of missing values.

Development and internal validation of the Simplified Mortality Score for the Intensive Care Unit (SMS‐ICU)

“…We congratulate the authors on the development and validation of a simplified critical care prognostic model using a combined (predominantly) European clinical trial data set and the adherence to the TRIPOD statement. 1,2 Work undertaken by our groups in resourcelimited settings has demonstrated challenges in the development and application of relatively simple prognostic models such as APACHE II. [3][4][5][6] Missing predictor variables, lack of post-ICU outcome measures, poor clinical uptake of scores due to perceived complexity and poor adherence to guidelines such as TRIPOD transcend healthcare settings.…”

A brief comment about predictive models for mortality in intensive care units

“…Editor, Granholm and colleagues recently published an interesting paper in the Acta Anaesthesiologica Scandinavica concerning the construction and internal validation of a scoring system for 90day death after admission to the intensive care unit (ICU). 1 Nevertheless, we have a few observations concerning the development of the model used, as well as a comment they make about a similar study undertaken by us. 2 The sample used by Granholm and colleagues to develop their model includes data from different sources, including observational studies and clinical trials.…”

“…The authors also mention that our study was undertaken in just one centre. 1,2 Although we mentioned that this was, to a certain extent, a limitation of our study (requiring external validation), it nevertheless had the advantage that the study sample was completely homogenous.…”

“…3 The authors also state that our model is likely underpowered and had a high risk of overfitting. 1 To achieve valid results of the sample size, based on the ratio of events-per-variable (EPV), it should be equal to or greater than ten, 3 as Granholm and colleagues also mention. The EPV in our study was 15.…”

“…The EPV in our study was 15. 1 Finally, although the outcomes of the two models are different, they both have clinical usefulness.…”