This study explored the feasibility of generating reliable information on the frequency, nature and management of breakthrough pain (BTP) in children with life-limiting conditions and life-threatening illnesses (LTIs) from narrative clinical records. In the absence of standardized ways for documenting BTP, we conducted a consensus exercise to develop a glossary of terms that could denote BTP in the records. Thirteen clinicians who contributed to the records reached consensus on 45 terms which could denote BTP, while emphasizing the importance of contextual information. The results of this approach together with guidance for improving the reliability of retrospective reviews informed a data extraction instrument. A pilot test of this instrument showed poor agreement between raters. Given the challenges encountered, we do not recommend a retrospective review of BTP using narrative records. This study highlighted challenges of data extraction for complex symptoms such as BTP from narrative clinical records. For both clinical and research purposes, the recording of complex symptoms such as BTP would benefit from clear criteria for applying definitions, a more structured format and the inclusion of validated assessment tools. This study also showed the value of consensus exercises in improving understanding and interpretation of clinical notes within a service.