The aim of this study was to assure the validity and reliability of the Intercollegiate Stroke Audit Package as used in the National Sentinel Audit of Stroke. The Intercollegiate Working Party for Stroke, which included most stakeholders, including patients, devised the audit standards. These were submitted to a formal consensus (modified Delphi) survey before the audit questions were developed and piloted for validity and reliability. Following the pilot, Help Booklets were developed to promote the involvement of all disciplines as auditors in the national sentinel audit of stroke and ensure inter-rater reliability. During the national audit each Trust was asked to double rate the first five cases with auditors of different disciplines working independently. A total of 886 case notes were double-rated in 184 separate sites (median 5, range 1-5 per site). Trusts used auditors from different disciplines in 77% of cases. After excluding the 'No answer' cases the kappa score for items ranged from 0.49 to 0.87 (median 0.70, IQR 0.63-0.78). Very good agreement was found for seven of the 45 items, good agreement for 30 items, and moderate agreement for eight items. This large study, across a range of hospital sites and involving many disciplines, demonstrates that careful piloting of audit tools, with use of clear instructions to auditors, promotes the reliability of data.