The use of microfluidics for biofluid analysis offers a cheaper alternative to conventional techniques in disease diagnosis. However, traditional microfluidics design may be complicated by the need to incorporate separation elements into the system in order to facilitate specific molecular detection. Alternatively, an optical technique known as surface-enhanced Raman spectroscopy (SERS) may be used to enable identification of analyte molecules directly from a complex sample. This will not only simplify design but also reduce overall cost. The concept of SERS-based microfluidics is however not new and has been demonstrated previously by mixing SERS-active metal nanoparticles with a model sample, in situ, within the microchannel. Although the SERS reproducibility of these systems was shown to be acceptable, it is, however, not stable toward variations in the salt content of the sample, as will be shown in this study. We have proposed a microfluidics design whereby periodic SERS-active metal nanostructures are fabricated directly into the microchannel via a simple method of spin coating. Using artificial as well as human urine samples, we show that the current microfluidics is more stable toward variations in the sample's ionic strength.