Objectives: To assess the association between mortality and the day of elective surgical procedure.

Design: Retrospective analysis of national hospital administrative data.

Setting: All acute and specialist English hospitals carrying out elective surgery over three financial years, from 2008-09 to 2010-11.

Participants Patients undergoing elective surgery in English public hospitals.

Main outcome measure: Death in or out of hospital within 30 days of the procedure.

Results There were 27 582 deaths within 30 days after 4 133 346 inpatient admissions for elective operating room procedures (overall crude mortality rate 6.7 per 1000). The number of weekday and weekend procedures decreased over the three years (by 4.5% and 26.8%, respectively). The adjusted odds of death were 44% and 82% higher, respectively, if the procedures were carried out on Friday (odds ratio 1.44, 95% confidence interval 1.39 to 1.50) or a weekend (1.82, 1.71 to 1.94) compared with Monday.

In this retrospective analysis of a large national representative database and including deaths after discharge we found that the adjusted odds of death in the 30 days after elective surgical procedures were 44% and 82% higher if the procedures were carried out on Friday or at the weekend, respectively, compared with Monday. To our knowledge, this is the first study to report a “weekday effect,” in addition to the well known “weekend effect” on hospital mortality, in a large national representative database and including post discharge deaths. The odds of death within two days of the index procedure were similar for Fridays but even larger (2.67) for procedures carried out at the weekend. We found a significant effect in four out of our five selected higher risk procedures. The lack of a significant association between day of procedure and mortality for repair of abdominal aortic aneurysms might be explained by the low number of procedures analysed (particularly at the weekend) and hence a lack of power to detect any association for mortality. In addition, some of the patients who underwent selected high risk procedures might have received a substantial amount of postoperative care in critical care units that are more likely to provide service 24 hours a day, seven days a week than general hospital wards, which could dilute any effect.7 We acknowledge that although we attempted to adjust for some case mix variables—including age, sex, ethnicity, socioeconomic deprivation, comorbidity, previous emergency admissions, risk of procedure, and method of admission—there might be some residual confounding. Daily variation, although small, in those factors that we were able to account for, however, seemed to suggest that patients operated on towards the end of the week and at the weekend actually had a lower risk profile than Monday patients. We found that weekend patients on average had slightly longer waiting times, which could indicate either that their situation had become more severe because they had been waiting longer or was less severe because it was deemed that they could wait longer, but this could not account for the increased mortality from Monday to Friday. We also found that a lower proportion of procedures in the highest fifth of risk were carried out at the weekend. We did not find any evidence for clustering by hospital.