Background: Gene fusions and alternate RNA splice forms represent clinically actionable driver and resistance-conferring alterations in NSCLC and other cancers. Where tissue samples are inadequate for molecular testing, liquid biopsies could address the gap in detection of clinically relevant gene fusions. Targeted cell-free (cf) DNA-based NGS methods are typically used for non-invasive blood-based testing of gene fusions, but have limited sensitivity if fusion breakpoints involve long intronic regions or repetitive sequences (e.g. NTRKs and NRG1). RNA-based approaches are not affected by introns, can identify expressed fusion genes and can discriminate splicing. We developed an NGS assay optimized for cfRNA analyte for the detection of actionable gene fusions and exon deletion/skipping events in liquid biopsies. Methods: A highly multiplexed molecular-barcoded primer panel was designed for cfRNA-based detection of actionable fusion genes in NSCLC (ALK, BRAF, FGFR2, FGFR3, MET (including exon 14 skipping), NRG1, NTRK1/2/3, RET and ROS1) covering > 90% of reported driver and partner gene exons in COSMIC database. The panel also targets housekeeping genes as endogenous sample controls. We developed a custom bioinformatics pipeline to call fusions and exon skipping based on split and spanning reads. Results: In initial analytical validation using fragmented RNA from pre-characterized reference material (representing 32 unique fusions at known copy numbers), the assay could detect as low as 10 fusion copies with a sensitivity of 97.6% and a specificity of 100%. For clinical testing, cfRNA co-eluted with cfDNA in nucleic acid extracts from plasma was used. In plasma samples (n = 103) from advanced NSCLC, 76% (15/21) of fusions (5 ALK, 3 RET, 2 ROS1, 5 MET ex14 skipping, 1 FGFR3-TACC3) detected in a clinically validated cfDNA assay (LiquidHALLMARK) were also detected in corresponding cfRNA. In a subset of samples that were driver-negative untreated or tyrosine kinase-inhibitor treated, the cfRNA assay yielded 9 gene fusions or breakpoints (1 CD74-NRG1, 4 BRAF, 2 MET, 2 FGFR3-TACC3 fusions) that could not be detected by the cfDNA assay. This represents an increase of 8.7% (9/103) of actionable alterations (driver or resistance) identified using cfRNA. Together cfRNA and cfDNA resulted in 30 fusion events to be detected compared to 21 by the cfDNA assay alone, representing a 42.8% (9/21) increase in fusion-specific detection of the combined assay. Conclusions: This novel cfRNA assay can detect actionable gene fusions and exon skipping events in liquid biopsies with high sensitivity. Combining cfRNA with more routine cfDNA testing can increase the total actionable diagnostic information from non-invasive testing in NSCLC patients where tissue samples are lacking, especially for gene fusions not amenable to detection in cfDNA.