On an average shift at St Emlyn’s we admit a high number of elderly patients, many of whom are frail. These patients are often unable to fully care for themselves due to low mobility, cognitive impairment, and chronic pain.
Hospital wards are usually well-equipped to meet these additional needs, but sourcing a ward bed can be difficult. It is not uncommon at this time of year for our elderly patients to wait hours, if not the better part of a day, for one to become available.
In the meantime, they are cared for in ED – which is not designed or staffed for frailty. We do not have soft beds, pillows, or lights that turn off. We do not store many long-term medications. We do not have enough nurses and support workers to provide regular analgesia, repositioning, and personal care. The list goes on.
With these problems in mind, for February’s journal club we are looking at a study published in JAMA Internal Medicine last year investigating whether long waits in A&E lead to adverse outcomes for older patients. Specifically, the researchers wanted to know whether an overnight stay in the department makes an elderly person more likely to die.
The abstract is below – followed by our summary of the paper – but as ever, we recommend that you read the full text and come to your own conclusions.
What type of study is this?
This is a prospective cohort study. Retrospective studies have been previously been published on admission delays but interpreting their results has been challenging.
Retrospective data collection is a methodology with key limitations and it is difficult to imagine that it would have allowed the authors to answer their clinical question. To establish with confidence whether elderly patients are especially vulnerable to prolonged ED stays, confounding factors must be accounted for.
For example, we know that cardiac problems can simultaneously increase the wait for a suitable (i.e. monitored) ward bed and the likelihood of dying. Retrospective data collection from medical notes is an unreliable way to investigate factors like these.
Tell me about the patients
1598 elderly ED patients (>75 years) were recruited consecutively at 97 participating French hospitals over a 48-hour period in December 2022.
Patients were eligible for inclusion if they required hospital admission. They were excluded if they were admitted to a ward at certain time points (more on this later) or required critical care.
The authors used the Groupe Iso-Ressources (GIR) and Charlson Comorbidity Index (CCI) to estimate baseline levels of frailty and comorbidity, which were high. Cardiovascular, renal, and metabolic disease was common, as was dementia. Trauma, respiratory illness, and ‘asthenia’ (fatigue) accounted for the majority of admissions.
What did they study?
Patients were split into two groups: those who spent a night in ED (midnight to 08:00) while waiting for a ward bed, and a “control” group who reached the ward before midnight. Any patients who spent some of the night in ED but went to the ward sooner (i.e. between midnight and 08:00) were excluded from the analysis.
At first glance, this might seem strange. A patient in this study could be waiting for a ward bed for fifteen hours (e.g. 08:00 to 23:00) and still be allocated to the “control” group. But the authors were interested in the effect of an overnight ED stay – not the total length of time spent in the department. In their introduction, they discuss the specific risks of prolonged immobilisation and sleep deprivation in elderly patients.
As it stands, the patients who spent the night on an ED trolley waited on average fifteen hours longer in the department.
Tell me about the outcomes
The primary outcome was in-hospital mortality. Secondary outcomes included adverse events and increased length of hospital stay.
Results were presented as risk ratios with 95% confidence intervals. Adjusted risk ratios were also provided, correcting for clustered factors including (1) age, GIR score, and CCI score, and (2) presenting complaint, time of presentation, and length of ED stay.
The researchers also used propensity score matching (PSM) to explore the relationship between baseline characteristics and outcomes. PSM is a statistical technique used to adjust for a particular form of bias seen in observational studies: factors that predispose patients to receiving the treatment or exposure of interest and experiencing the outcome of interest. To use a blunt example: the success rates of resuscitative thoracotomy are low because the sort of patients who have a ‘propensity’ to receive thoracotomy have catastrophic, often un-survivable injuries.
Because it would unethical to randomise patients to spending a night on an ED trolley, there is a strong possibility that this paper’s analysis was biased by factors that predict the primary outcome and patient ‘propensity’ to be kept in ED. A good example would be baseline physiology: patients who were severely unwell and required stabilisation in ED may have been more likely to die and less likely to be accepted by a ward right away. Baseline physiology was one of the many variables included in the authors’ PSM.
What did they find?
Overall, 210 patients died in hospital: 111 (15.7%) of those who spent the night on an ED trolley, and 99 (11.1%) of those who made it to the ward before midnight. The absolute risk increase was 4.6%, which means the ‘number needed to kill’ for an overnight stay in ED was 21.
Translated into adjusted risk ratios: patients who spent the night in ED were 50% more likely [aRR 1.5] to die in hospital – albeit with quite a wide confidence interval of 9% to 106%. Virtually the same figure [aRR 1.48] was reported after PSM had been performed, also with wide confidence intervals.
Adverse events were 7% less common in patients who were admitted to a ward before midnight. Typical adverse events reported in the study included, nosocomial infections, falls, and pressure ulcers. However, in the adjusted analysis this finding was shown to be non-significant [aRR 1.07; 95% CI 0.84-1.36] as were findings relating to hospital length of stay [aRR 1.05; 95% CI 0.93-1.18].
The bottom-line is this: when our elderly patients spend the night on an ED trolley, they are more likely to die, regardless of their baseline co-morbidity, frailty, and illness severity. However, it is not clear that this increased mortality risk can be linked to a longer hospital stay, or a greater likelihood of in-hospital adverse events.
What should we take away from this?
This study supports our ongoing efforts to improve flow in emergency care. It provides further evidence that getting our elderly patients off the trolleys is, quite literally, a matter of life and death. We reviewed data from our own electronic patient record system in Virchester and found similar figures.
With that said, there are some caveats to keep in mind as we interpret this paper. These were discussed at our journal club in November last year and felt to be the most significant points for critical appraisal.
Firstly, and perhaps most importantly: this was an observational study. It was clearly a well-conducted observational study, but any analysis involving non-randomised patients is vulnerable to bias, and there were multiple potential confounders that the authors were unable to control for. The most glaring were race and ethnicity, which cannot be included in patient records under French law. This is problematic, because we know from prior research that ethnic minority groups are more likely to experience poor outcomes in emergency care.
Another potential confounder missing from the baseline data was ED over-crowding. This is despite recruitment taking place during the early phases of the COVID-19 pandemic!
We know that the ratio of nurses to patients determines quality of care. On a quiet night, or at least a well-staffed night, elderly patients receive a high standard of care in ED – similar to what they would receive on the ward. When the department is overflowing, they do not. Nursing ratios would have been a helpful adjustment to the authors’ risk analysis.
Finally: it is worth thinking about the primary outcome used in this study. The researchers selected in-hospital mortality, which is simple to measure, and of obvious clinical significance. But for research with older patients, is it the most appropriate outcome? We know that many of the elderly people we see in ED have injuries and illnesses that they are not going to survive. Our goals of care for these patients should be comfort and dignity, not prevention of death.
I would have personally been interested to see a sub-group analysis, with an outcome measure centred on the successful delivery of palliative care. I suspect this too is negatively affected by spending the night on an ED trolley.
Should this study change our practice?
I think so. We are not going to see a randomised study on a topic like this, and the authors have published high-quality, prospective research. Their findings should galvanise changes in the way we practice emergency medicine. Here are some ideas.
- ED seniors should regularly comb the list for elderly, frail patients who are likely to require hospitalisation: if they are seen out-of-order, a decision to admit can be made earlier, and an overnight stay avoided.
- ED doctors and nurses should urge bed managers to prioritise elderly patients when ward beds become available, regardless of the impact this has on hospital performance statistics.
- ED staff at all levels should be advocating for novel approaches to de-congesting ED – such as immediate transfer of patients to the ward after decision to admit.
vb
Greg Yates.
