What is the ‘real’ admission rate of acute pancreatitis in a regional Australian population?

Objective. Capture-recapture analysis was used to more accurately quantify the admission rate for acute pancreatitis in a regional hospital setting, in comparison to the usual method of case ascertainment. Reasons for differences in capture for the various methods were also sought. Methods. Admissions for acute pancreatitis were enumerated over a 40-month period using three data sources: hospital classification of admission diagnoses, prospective case identification, and receipt of diagnosis-specific pathology specimens. Capture-recapture analysis was applied with log-linear modelling to account for likely dependency between data sources. Covariates were noted to explain capture probability by the various data sources and for eventual stratification in the analysis process. Results. For the census period, there were 304 admissions after merging of data sources, giving a crude admission rate of 7.6 per month. Crude ascertainment rates for discharge records and prospective identification were 44% and 52% respectively. Following log-linear modelling, total admissions more than doubled to 644 (adjusted admission rate 16.1 per month). Of the covariates considered, admissions of less than three days’ duration and those occurring in December and January were significantly associated with increased capture by the hospital discharge records data source. Conclusions. In this clinical setting, admissions for acute pancreatitis are grossly underestimated by the standard case ascertainment method. The reasons for this are not clear. Hospital discharge records are nevertheless more effective than prospective case ascertainment for certain cases, such as brief admissions and those in holiday periods. What is known about the topic? Capture–recapture analysis was originally developed in animal ecology, but has since been used to estimate both prevalent and incident cases of human disease. What does this paper add? This study exposes possible deficiencies in the single-source case ascertainment methods used by most hospitals to enumerate incident cases. It is the first time that capture–recapture techniques have been used to estimate acute pancreatitis admissions. What are the implications for practitioners? To obtain accurate admissions estimates for diseases such as acute pancreatitis, capture–recapture analysis with multiple data sources is advisable. One possible solution may be to conduct intermittent prospective censuses to complement existing retrospective ascertainment methods. On a more general level, clinical staff should be better trained to provide more accurate and detailed information in case records.