Scoring systems/statistics (ICU specific (APACHE, SOFA, MODS, ICU CAM, SMR…
SMR - the SMR is observed hospita or unitl mortality / expected hospital or unit mortality (as calculated by admission illness severity score e.g. APACHE or SOFA. Changes to SMR can occur as a result of increased mortality or decrease in admission illness severity scores. Increased mortality - change in care protocols, change in admission or discharge protocols (more palliative admissions/fewer discharge to pall. care hospice), change in hospital or unit staff (e.g. new intake of nurses/doctors). Decrease in predicted mortality - changes to severity score encoding (typically GCS calculation may underestimate severity), lead time bias (patients stabilised on ward prior to ICU admission so physiological variables improved, healthy worker effect - increases in low acuity case admission to the ward e.g. healthy post-ops.
Statistics and data analysis
Meta-analysis (the analysis of analyses - a form of quantitative systematic review that is used to weigh the available evidence from RCTs and other studies based on number of patients, effect size and statistical tests of agreement with other trials. It is NOT simply a collation and combination of data from multiple studies which would be subject to Simpsons paradox. Rather it seeks to combine results from multiple studies and increase statistical power but control for differences by quantifying and analysing inconsistencies of results across studies. Problems include - publication bias - this is assessed with the use of a funnel plot which plots study results according to size and standard error. There should be decreasing SE in larger studies (the peak of the funnel) with an even spread of SE in the smaller studies. Skew in the distribution of the smaller studies indicates a publication bias.