The following five reasons were most prevalent: • Different interpretation of the chart by the reabstractor and the hospital coder; • Documentation in the chart was incomplete at the time of original abstraction; • Conflicting documentation in the chart led the reabstractor and hospital coder to select different diagnosis codes; • Information on the chart was missed by the hospital coder; and • Th [...] This was a particularly surprising result given the rigour involved in the selection process for the study reabstractors, the intensive one-week training and the support and the resources available to the reabstractors during data collection. [...] The few exceptions to this were the following: • The selection of code for the MRDx had a higher agreement rate in the inter-rater data set than in the main data set by an estimated 12% in FY 2002–2003 and 18% in FY 2003–2004; • The assignment of significance to a condition had a higher agreement rate in the inter-rater data set by an estimated 7% in FY 2003–2004; • Improvement between fiscal year [...] The impact of the observed discrepancies in the coding of diagnoses, interventions and assignment of significance to diagnoses affected the output variables from CIHI’s grouping methodology in the following ways: • About 14% of the discharges in FY 2002–2003 and 16% in FY 2003– 2004 changed assignment of Case Mix Group (or CMG ). [...] The following recommendations were designed to address the identified issues and to improve the quality of the clinical data in the Discharge Abstract Database: 1. Review the current concept of diagnosis typing with a view to improving the consistency of implementation.