Do variations in hospital mortality patterns after weekend admission reflect reduced quality of care or different patient cohorts? A population-based study

Abstract

Background: Proposed causes for increased mortality following weekend admission (the 'weekend effect') include poorer quality of care and sicker patients. The aim of this study was to analyse the 7 days post-admission time patterns of excess mortality following weekend admission to identify whether distinct patterns exist for patients depending upon the relative contribution of poorer quality of care (care effect) or a case selection bias for patients presenting on weekends (patient effect).

Methods: Emergency department admissions to all 501 hospitals in New South Wales, Australia, between 2000 and 2007 were linked to the Death Registry and analysed. There were a total of 3 381 962 admissions for 539 122 patients and 64 789 deaths at 1 week after admission. We computed excess mortality risk curves for weekend over weekday admissions, adjusting for age, sex, comorbidity (Charlson index) and diagnostic group.

Results: Weekends accounted for 27% of all admissions (917 257/3 381 962) and 28% of deaths (18 282/64 789). Sixteen of 430 diagnosis groups had a significantly increased risk of death following weekend admission. They accounted for 40% of all deaths, and demonstrated different temporal excess mortality risk patterns: early care effect (cardiac arrest); care effect washout (eg, pulmonary embolism); patient effect (eg, cancer admissions) and mixed (eg, stroke).

Conclusions: The excess mortality patterns of the weekend effect vary widely for different diagnostic groups. Recognising these different patterns should help identify at-risk diagnoses where quality of care can be improved in order to minimise the excess mortality associated with weekend admission.

Keywords: Health Services Research; Patient Safety; Quality Measurement.

Figure 1

Excess mortality risk associated with weekend admission per diagnosis related group may exhibit different temporal patterns depending on whether the cause is sicker patients (patient effect: H2) or poorer care (care effect: H1).

Figure 2

Excess mortality risks curves for selected diagnosis related groups (DRGs). Excess mortality risk curves for selected DRGs consistent with care effect (top two panels), patient effect (panel 3) and mixed effect (bottom panel).