Objective: To evaluate the reliability of data registration in calculating the hospital standardised mortality ratio (HSMR).
Design: Retrospective, descriptive.
Method: Data were collected from a research database on all patients who had undergone a partial pancreatoduodenectomy for pancreatic cancer in 2009 and 2010 at our hospital. These data were compared with information about these same patients recorded in the Dutch National Medical Registry (LMR), obtained from the medical administration department of our hospital. The differences between these 2 databases were evaluated on the basis of 3 variables: mortality, main diagnosis and secondary diagnoses (differentiated into complications and co-morbidities). Using the Charlson index, the co-morbidity score from both registries was calculated per patient.
Results: A total of 118 patients had been registered in the research database. Of these patients, 103 appeared in the LMR data; 15 had not been registered in this database. There were no differences in patient characteristics or mortality (2.5%) between the registries. In the LMR, the main diagnosis of 5 patients had been incorrectly recorded. This database contained information on 136 complications and 51 co-morbidities, of which 35 comorbities had been correctly recorded. The research database contained information on 188 complications and 99 comorbidities on these same patients. In the research database, comorbidity comprised 34% of all secondary diagnoses; in the LMR, 19% (p < 0.001). The median score on the Charlson index was 0 for all patients in the LMR and 3 in the research database (p < 0.001).
Conclusion: Comorbidities in patients with pancreatic carcinoma who undergo a resection are being inadequately recorded in the LMR. This results in insufficient correction in the case mix and a low score on the Charlson index, which could result in an incorrect HSMR.