We present a semi-automated point-of-care (POC) sensor approach for the simultaneous and reagent-free determination of clinically relevant parameters in blood plasma. The portable sensor system performed direct mid-infrared (MIR) transmission measurements of blood plasma samples using a broadly tunable external-cavity quantum cascade laser source with high spectral power density. This enabled the use of a flow cell with a long path length (165 μm) which resulted in high signal-to-noise ratios and a rugged system, insensitive to clogging. Multivariate calibration models were built using well established Partial-Least-Squares (PLS) regression analysis. Selection of spectral pre-processing procedures was optimized by an automated evaluation algorithm. Several analytes, including glucose, lactate, triglycerides, cholesterol, total protein as well as albumin, were successfully quantified in routinely taken blood plasma samples from 67 critically ill patients. Although relying on a spectral range from 1030 cm(-1) to 1230 cm(-1), which is optimal for glucose and lactate but rather unusual for protein analysis, it was possible to selectively determine the albumin and total protein concentrations with sufficient accuracy for POC application.