Background: Mucosal pressure injuries (PIs) are usually caused by pressure from essential medical devices. There is no universally accepted criterion for assessment, monitoring, or reporting mucosal PI. Reliable descriptors are vital to benchmark the frequency and severity of this hospital-acquired complication.Objectives: The objective of this study was to determine whether modified Reaper Oral Mucosa Pressure Injury Scale (ROMPIS) descriptors improved the reliability of mucosal PI assessment. Secondary aims were to explore nurses' knowledge of and attitudes toward mucosal PI. Methods: A prospective cross-sectional survey was distributed to nurses from two tertiary affiliated intensive care units via REDCap ® to capture demographic data, knowledge, attitudes, and inter-rater reliability (IRR) measures. Nurses were randomised at a 1:1 ratio to original or modified ROMPIS descriptors and classified 12 images of mucosal PI. IRR was assessed using percentage agreement, Fleiss' kappa, and intraclass correlation coefficients. Results: The survey response rate was 20.9% (n ¼ 98/468), with 73.5% (n ¼ 72/98) completing IRR measures. Agreement was higher with modified (75%) than original ROMPIS descriptors (69.4%). IRR was fair for the original (k ¼ 0.30, 95% confidence interval [CI] [0.28, 0.33], z 26.5, p < 0.001) and modified ROMPIS (k ¼ 0.29, 95% CI [0.26, 0.31], z 25.0, p < 0.001). Intraclass correlation coefficient findings indicated ratings were inconsistent for the original (0.33, 95% CI [0.18, 0.59], F 18.8 (11 df), p < 0.001) and modified ROMPIS (0.31, 95% CI [0.17, 0.57], F 17.6 (11 df), p < 0.001). PI-specific education and risk factor recognition were common. Conclusion: Modified descriptors had marginally better agreement. Participants understand management and prevention but need to strengthen their perceived capacity for mucosal PI risk assessment. This work provides a foundation for future benchmarking and a platform from which further research to refine and test descriptors specific to mucosal PI can be generated.