Clinical and genetic factors associated with prognosis in critical care medicine

This thesis explores the heterogeneity of critical illness from predictive and genetic perspectives. Chapters 2 and 3 use data from the SICS-II cohort (n=976) to develop clinical prediction models. Chapter 2 presents a model predicting maximum acute kidney injury (AKI) stage within seven days of ICU admission, demonstrating robust performance. Chapter 3 develops mixed-effects models for next-day deterioration in overall and organ-specific (respiratory, cardiovascular, renal) function, using fewer predictors than APACHE IV. Chapters 4 and 5 present a systematic review (PROSPERO: CRD42021209744) of genetic associations in ICU patients. A new quality assessment tool was developed, revealing high risk of bias in most (97%) of the 59 included studies. Two robust associations were identified: FER with mortality and FLT1 with respiratory failure. Chapter 6 investigates genetic variants associated with serum procalcitonin (PCT) levels via a two-stage GWAS meta-analysis (n=12,448) and PheWAS (UK Biobank, n=457,418). Three loci were associated with PCT, and polygenic risk scores (PRS) linked PCT-related genetic variants to metabolic and inflammation-related traits, as well as cardiovascular, renal, and liver function.