Laboratory blood testing plays a central role in current diagnostics and therapeutical decisions. Thus, errors have a direct impact on care quality and cost. The majority of errors occur in the pre-analytical phase, when samples are collected, stored and prepared. For biochemical analysis, the most common preparation operation is the separation of blood plasma: performing this operation upon sampling has the potential to simplify and render the testing cycle more reliable. Hence, in this work, a microdevice that performs blood separation at the point of collection is presented. The capillary-driven microdevice processes fingerprick blood microsamples without the need for external equipment. The device relies on sedimentation as a simple and spontaneous driving force for the separation of undiluted whole blood. Blood flows in the device at a velocity allowing cells to settle on the bottom of a constant height channel and create a higher viscosity liquid fraction. The supernatant plasma of lower viscosity is pumped at a higher speed than the sediment in the device. Thus, as the device fills, a plasma plug is generated in the downstream section of the channel. In this work, a unidimensional model of combining Kynch sedimentation and Poiseuille flow theories is established to describe this novel separation phenomenon. The impact of design and blood parameters on the separation is studied. The device can be used to separate fresh or anti-coagulated samples obtained through skin puncture. For both sample natures, the cell-free plug appears after a separation delay of 400 s. This delay is necessary to establish a sufficient viscosity contrast through sedimentation. Subsequently, for anti-coagulated samples, cellfree liquid is extracted with a 17 % yield. For fresh samples, coagulation leads to an increase of yield to 67 %. The combination of sedimentation and filtration through the clot are the reasons for this increase. Separated samples are retrieved from the chip through the use of an integrated ejection mechanism. The device is designed to eject a volume of 2ÎŒL of cell-free liquid out of 25ÎŒL of whole blood. The quality of separated samples was established by measuring particle contaminant concentration and proteomic profile. The contaminant concentration is lower and more repeatable than in centrifuged plasma or serum samples. In the protein profile, only 4.5% of quantified proteins show significantly different levels between serum and chip-separated samples - thus, showing that the separated samples and serum could be used interchangeably. The microdevice was combined with standard clinical automated analyzers to perform blood panels on 12 obese patients. Of the 8 blood markers analyzed, 7 markers showed significant correlation between the chip-separated and standard plasma samples. This shows that the microdevice can be used in combination with standard bench top analytical tools. The novel microdevice presented in this work paves the way to a family of microsystems that perform purely pre-analytical operations. The performance of the device, quality of retrieved samples and combinability with bench-top techniques indicate that the microdevice could impact on current testing cycles: the device could reduce the pre-analytical sources of errors by performing blood microsample separation at the point of collection.