Objective To establish the validity, reliability and fidelity of use of a screening tool for stroke survivors’ communication support needs during mental capacity assessments. Design Cross-sectional psychometric study and multiple methods case series. Setting Participants’ homes and stroke units in north-west England. Participants Psychometrics: 47 community-dwelling stroke survivors; 10 speech and language therapists experienced in stroke. Fidelity: four nurses and one occupational therapist working in stroke; eight inpatients diagnosed with stroke. Main measures Psychometrics: Stroke survivors were tested using the new communication screening tool and subtests from the Western Aphasia Battery Revised and Comprehensive Aphasia Test to measure convergent and discriminant validity, test-retest and interrater reliability. Data were analysed using Kendall's Tau, Goodman-Kruskal's Gamma and Cohen's Kappa. Fidelity The nurses and occupational therapist were videorecorded using the screening tool with inpatients. An observational checklist and semi-structured interviews were used to investigate adherence to predefined administration behaviours. Data were analysed using descriptive statistics and qualitative content analysis. Results Psychometrics: Screening tool subtests demonstrated moderate-to-strong convergent validity ( τ  = 0.41–0.72) but poor discriminant validity ( τ  = 0.27–0.58). Criterion validity was strong ( κ  = 0.77, 95%CI 0.57–0.98). Test-retest reliability ranged from substantial to near-perfect ( κ  = 0.68–0.95). Interrater agreement ranged from substantial to perfect ( κ  = 0.65–1.00). Fidelity Participants adhered to a mean of 85.79% (SD = 10.33) administration behaviours. User errors related to task instruction delivery and scoring. All participants found the tool easy to use and useful. Conclusions This unique screening tool for stroke survivors’ communication support needs during mental capacity assessments is valid and reliable. Minor refinements to administration training will improve fidelity of use.