While regions and countries, health care systems and providers, patients and societies may vary, some things are almost the same everywhere: a week consists of seven days, Saturday and Sunday constitute the weekend, and disease and injury may occur on any day of the week. Patients, families and health care professionals could therefore expect identical outcomes in hospitalized patients irrespective of the day of their admission or hospital stay.
Yet this notion was repeatedly questioned over the last decade; so-called “weekend effects” were identified in some studies, while others failed to do so. Most notably, a British study demonstrated increased risk of death in patients admitted to National Health Service (NHS) hospitals at weekends. This publication not only led to intense debates, but was used to justify changes in junior doctors’ working conditions. That, in turn, sparked protest amongst the affected group and others.
There are some patient groups that may be especially vulnerable to variation in the care and treatment they receive. Critically ill patients are certainly amongst them. We therefore investigated, whether “weekend effects” exist in intensive care medicine. To do so, we looked at a vast patient cohort of more than 150,000 Austrian patients and used elaborate statistical methods to analyze these data.
Avoiding bias
Size and level of detail of the underlying database and meticulous statistical analysis are the defining characteristics of our study
Statistical methodology is key in studies like these. At weekends, patients will usually only be admitted to hospitals due to urgencies or emergencies. The very same effects can be observed in intensive care units. It is therefore vital to include all possible influence factors into the analysis to avoid bias. This encompasses an assessment of severity of illness at the time of admission as well as an adjustment for process-related factors that would otherwise skew results.
Size and level of detail of the underlying database and meticulous statistical analysis are the defining characteristics of our study. Some previous studies were smaller in size, some had to use administrative data for adjustment purposes only, and none were able to incorporate effects such as patient discharge from intensive care units into their analyses. Our study benefits greatly from the use of the well-established SAPS3 risk score and the Fine and Gray-model for sub-distribution hazards, a statistical model that was not used in this scenario before.
Nuanced effects
With this model, we were indeed able to demonstrate “weekend effects”. However, these effects were more nuanced than previously believed. Patients admitted to intensive care units at weekends were significantly more likely to die compared to those admitted during the week. However, the actual risk to die on a weekend was lower than during the week. Furthermore, we did not observe any increase in risk of death after weekend admission in patients admitted to intensive care units after emergency surgery.
Patients admitted to intensive care units at weekends were significantly more likely to die compared to those admitted during the week. However, the actual risk to die on a weekend was lower than during the week
So, what’s to make of these results? First and foremost, we found outcomes to be better than calculated by our risk-assessment score at every day of the week. Furthermore, outcomes got better year by year. In short, intensive care units generally provide good care for critically ill patients. But there is an undoubtable signal that patients admitted to intensive care at weekends have some disadvantage.
However, jumping to any conclusions must strictly be avoided. No matter how large the patient cohort is or how elaborate the used statistical model may be, results from a retrospective, national study must not be liberally extrapolated to other health care systems or used to justify rash top-down interventions.
While that may be the core finding of our study, we tried to go a step further and identify reasons for these effects. We found differences in health care provided to patients depending on the day of intensive care admission.
Although patients admitted to intensive care at weekends were sicker, some key interventions in critical care were less often performed at the very day of admission at weekends. This may indicate that there are indeed causes for “weekend effects” that could be tackled by health care professionals and policy makers.
That is the message our study should convey: Good intensive care medicine is provided right now, but there is obvious potential for further improvement. That necessitates further investigative efforts.
Researchers need to be empowered to identify reasons for the observed differences in care provided at weekdays and weekends. Only based upon such studies policy makers could make evidence-based decisions to improve patient outcomes instead of rushed interventions that could potentially worsen things in the end.
Comments