In clinical diagnostics, human breath presents an alternative and more convenient sample than biofluids for detecting the ingestion of nonvolatile drugs. Surface-enhanced Raman spectroscopy (SERS) is a powerful vibrational spectroscopy technique with high sensitivity based on molecular fingerprinting. However, the low affinity of traditional SERS substrates for aerosols and the stochastic fluctuation of the SERS signal at low concentrations limit their application in breath aerosol analysis. In this study, we synthesized hydrogel microsphere SERS substrates with highly reversible swelling/shrinking properties that enhance target analyte accumulation in breath aerosols and promote plasmonic nanoparticle aggregation for intense Raman hotspot formation. Furthermore, these hydrogel microsphere SERS substrates function as a three-in-one system, enabling multilevel selectivity based on size, charge, and hydrophilicity for target molecules simultaneously without pretreatment. Notably, by "digitizing" the SERS signal of each individual hydrogel microsphere and calculating the proportion of positive microspheres, the hydrogel microspheres can serve as a digital SERS platform that circumvents the low stability issues resulting from fluctuations in SERS signal intensity. Consequently, the digital SERS platform achieved a detection limit of 0.5 ppm for fentanyl in simulated breath aerosols. This innovative sensing strategy not only demonstrates a promising approach for screening nonvolatile drugs but also simplifies the sampling process, holding great potential for clinical diagnosis of breath aerosols.