Definition of baseline renal function in hospitalised patients with multiple preadmission measures of serum creatinine-A joint modelling approach.

Aim With the use of a joint model (JM) we investigated how different definitions of baseline serum creatinine (SCr) would affect the association between an acute increase in inpatients' SCr and 30-day mortality and whether this effect depends on premorbid SCr trajectory. Methods This was a retrospective study including adult patients admitted to a tertiary acute-care hospital in Porto, Portugal, between January 1, 2013, and December 31, 2015, who had at least two preadmission ambulatory and two inpatients SCr measurements. The baseline SCr was defined as the lowest (-min), the most recent (-last) or the median (-medi) value over the preadmission period. The JM combined a linear mixed model for repeated inpatient SCr relative to baseline value and a Cox proportional survival model. Preadmission SCr courses were identified using linear regression and subsequently clustered based upon a patient-specific slope. Preadmission SCr trajectories were described as median SCr courses within clusters. Results SCr trajectories were: "Stable" (78.0% of patients), "Decreasing" (11.3%) and "Increasing" (10.7%). Overall, an increase in inpatient SCr by 50% relative to baseline SCr-min raised the risk of 30-day mortality by 74%; the estimate was not different from hazard ratio (HR) obtained for SCr-last (1.78) and SCr-medi (1.71). We found no differences in HR across preadmission trajectories. Conclusion The increased risk of death associated with an abrupt rise in inpatient SCr depends neither on the definition of baseline SCr nor patients ' SCr trajectory before hospitalisation. Preadmission SCr-medi value may be the least biased estimate of the baseline renal function.