OBJECTIVES: as little has been published about the particular challenges of researching labour and birth events using health information records (HIRs), this study aimed to describe the methodological and quality assurance (QA) issues encountered in reviewing such records for a study of health and recovery after operative birth, and to report on how these issues were tackled. DESIGN AND SETTING: retrospective review of HIRs, recording details of the labour and birth events of 394 women for 630 confinements at 32 hospitals, chiefly situated in the State of Victoria, Australia. PARTICIPANTS: three hundred and ninety-four women, a subset of a cohort of 534 women participating in the Health and Recovery after Operative Birth Project, who gave consent to review of their HIRs. METHODS: a data abstraction form (DAF) and comprehensive accompanying study manual were designed, with the form's layout based on the HIR used by the hospital where the majority of confinements occurred. Amendments were made following piloting of the DAF. Three data abstractors were carefully trained for the task of reviewing records, and issues were dealt with as they arose at fortnightly meetings. Double data abstraction (cross-coding) exercises were conducted three times during the project and reported on for QA purposes. FINDINGS: a number of limitations were found with labour and birth event data recorded in the HIRs. For example, maternal position in active second stage labour could not be established for 65% of births, documentation of onset of second stage was not accompanied by definitive evidence for second stage onset in 23.9% of cases, and maternal position at birth was missing in 26.4% of cases. Other relevant variables also proved problematic. For example, infant head circumference was not documented for 24.4% of births, and 52.9% of records did not document maternal height. Time and effort put into data abstractor training, and careful design and piloting of the DAF enabled both the form and data abstraction methods to be adapted following recognition of some of the limitations identified with the records, and also allowed subsequent analytic possibilities to be maximised. Cross-coding exercises also increased data abstractor reliability. KEY CONCLUSIONS AND IMPLICATIONS FOR PRACTICE: the likelihood of obtaining reliable data from HIRs is increased by a keen awareness of the challenges presented by the records themselves, careful training of staff, careful DAF design which allows for documentation of problems, conducting QA checks such as cross-coding exercises, and cautious reporting of findings with appropriate recognition of inherent limitations of the method.