Early studies estimate that 5% to 10% of oropharyngeal squamous cell carcinomas overexpress p16 but are unassociated with transcriptionally-active high-risk human papillomavirus (HPV). Patients with discordant HPV testing may experience clinical outcomes that differ from traditional expectations. To document the rate of p16 and HPV mRNA positivity, characterize patients with discordant testing, and identify features that may warrant selective use of HPV-specific testing after p16 IHC, a multi-institutional, retrospective review of oropharyngeal squamous cell carcinoma patients with p16 IHC and HPV mRNA testing by reverse transcriptase polymerase chain reaction was performed. Of the 467 patients, most had T1 or T2 tumors (71%), 82% were p16 positive, and 84% were HPV mRNA positive. Overall, most tumors were nonkeratinizing (378, 81%), which was strongly associated with p16 and HPV positivity (93% and 95%, respectively). Overall, 81% of patients were double positive, 14% double negative, and 4.9% discordant (3.4% p16 negative/HPV mRNA positive and 1.5% p16 positive/HPV mRNA negative). The survival rates of these discordant patient groups fell squarely between the 2 concordant groups, although in multivariate analysis for both disease-free survival and overall survival, discordant patients were not found to have statistically significantly different outcomes. Reclassifying patients by applying HPV mRNA testing when p16 results and morphology do not match, or when p16 results are equivocal, improved prognostication slightly over p16 or HPV mRNA testing alone. Patients with discordant testing demonstrate a borderline significant trend toward survival differences from those with concordant tests. When evaluated independently, patients who were p16 negative but HPV mRNA positive had a prognosis somewhat closer to double-positive patients, while those who were p16 positive, but HPV mRNA negative had a prognosis closer to that of double-negative patients. We suggest an algorithm whereby confirmatory HPV mRNA testing is performed in patients where p16 status is not consistent with tumor morphology. This captures a majority of discordant patients and improves, albeit modestly, the prognostication.