Analysis of acid–base disorders in an ICU cohort using a computer script

Abstract

Background/aims: Acid–base status is important for understanding pathophysiology, making a diagnosis, planning effective treatment and monitoring progress of critically ill patients. Manual calculations are cumbersome, easily result in wrong conclusions. We wanted to develop an automated assessment of acid–base status. Methods: A simplified adaptive MATLAB script processing all available theory to date was created, evaluated and used on blood gas analyses drawn immediately after admission to ICU. The script was compared to golden standard, calculating manually by two experienced ICU physicians. Results: Results from the script correlated completely with detailed manual calculations of randomly chosen 100 blood gas results and it was able to deliver complex data on cohort level with advanced graphics. The initial blood gas analyses from 8875 admissions constituted the cohort, of which 4111 (46.3%) were normal. Respiratory acidosis was the primary disturbance in 2753 (31.0%) and metabolic acidosis in 464 (5.2%). Respiratory alkalosis was the primary disturbance in 1501 (17.0%) and metabolic alkalosis in 46 (0.5%). Of the disturbances 74.7% were mixed with two and 2.1% with three simultaneous disturbances. Acidoses were less compensated compared to alkaloses. Conclusions: Acid–base theories are developed on ideal models and not on critical care patients, they require inputs that might not be available, and therefore, estimations are needed. In our cohort, it was difficult to develop a working script based on Stewart, whereas Boston/Copenhagen worked better. Acidoses were more common and more deviated compared to alkaloses.