All-cause mortality after first ERCP in England: clinically guided analysis of hospital episode statistics with linkage to registry of death.

BACKGROUND

All-cause death within 30 days of ERCP is a candidate indicator of care, but institutional-level statistics require careful interpretation. National-scale, population-based outcome studies of unselected patients undergoing ERCP are needed to define expected levels of real-world mortality risk and the case-mix factors that predict poor outcome.

OBJECTIVE

To develop methods for analyzing administrative data for English hospitals with linkage to death registration to study all-cause mortality after first ERCPs and explore predictors of death and institutional variation.

DESIGN

Hospital episode statistics for 2006 to 2007 and 2007 to 2008 were linked to the statutory death register. First ERCP episodes were extracted and analyzed for demographic characteristics, admission method, diagnoses, and comorbidities. Additional linkages identified the last-coded diagnosis before death. Factors associated with 30-day death were identified by univariate and multiple logistic analyses. Pilot data and a survey were sent to clinicians at each institution. Crude and case-mix adjusted mortality were analyzed at the institutional level.

MAIN OUTCOME MEASUREMENTS

Death within 30 days of the first ERCP procedure.

RESULTS

We analyzed 20,246 first ERCPs from 2006 to 2007 and 20,422 from 2007 to 2008. Diagnostic profile: gallstone related 57.3%; cancer 12.6%; gallstone and cancer 2%; others 28.1%. All-cause 30-day death was 5.3% (2.4% in non-cancer cases). Predictors of 30-day death (adjusted odds ratio [OR]) were as follows: age (OR 6.2, for ≥85 years vs <55 years), male sex (OR 1.2 vs female), emergency admission (OR 2.0 vs elective), cancer (OR 8.6 vs no cancer), and non-cancer comorbidity (OR 1.5 vs none). A mortality risk estimator (look-up table) based on pooled data for >40,000 first ERCPs is provided. Specific procedural complication codes were identified in 1.2% of deaths (0.06% of ERCPs). At the institutional level, analysis of mortality rates was within expected statistical funnel limits, and we found no correlation with ERCP volume (Pearson r = -0.05; P > .05).

LIMITATIONS

The completeness and accuracy of coding may vary between different hospitals. Routine coding does not capture information about procedural complexity or severity of illness.

CONCLUSION

Linkage analysis of hospital episode statistics data for England provides a powerful tool for studying mortality risk after ERCP on an unselected and truly nationwide scale. Institutional-level statistics suggest that the mortality risk for patients requiring ERCP was comparable across English hospitals.