Article Text

other Versions

Download PDFPDF
When a study doesn’t show what you hoped for
  1. Ellen J Weber
  1. Emergency Medicine, University of California San Francisco, San Francisco, California, USA
  1. Correspondence to Dr Ellen J Weber, Emergency Medicine, University of California San Francisco, San Francisco, CA 94143, USA;{at}

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

One of the arguments often made for eliminating (or at least reducing) the length of time admitted patients spend in the ED waiting for an inpatient bed is that these ‘boarded’ patients are likely, among other undesirable outcomes, to have longer hospital lengths of stay. This argument has been one tactic emergency providers have tried to gain the attention of hospital leadership to do something about boarding. Longer lengths of stay mean fewer patients coming in and out, and this can threaten bottom lines, waiting lists, and targets.

Thus, one may read the article by Lane and colleagues1 with high hopes for additional evidence on the harms of boarding psychiatric patients. These authors look at whether long stays in the ED for psychiatric patients are likely to result in longer hospital stays. The answer is yes, a little. If a patient is boarded in an ED for 24 hours, their length of stay increases…wait for it…0.01 days. That is, 14.4 minutes.

This is a bitter pill for us emergency physicians to swallow, dashing hopes of further convincing our administrations of the need to address patient boarding, as if the discomfort and loss of privacy for patients, the risk of nosocomial disease transmission and the stress on staff of crowded EDs are not reasons in themselves to pay attention.

But as the saying goes, it is what it is. The paper has received careful statistical and content review by our reviewers and editors. Given that we can be reasonably well assured of the validity of the data, it is incumbent on us to publish. Despite our skin in the game, we must avoid publication bias—that is, the tendency (often subconscious) to suppress publication of a study because results were negative. Failing to publish negative studies can lead, and has led, to ineffective treatments being prescribed.2 Moreover, if we do not publish negative studies, why should you believe the positive ones?

An important lesson for both readers and authors is, in fact, the value of a ‘negative’ study. A negative study does not mean there is no learning. A negative study can greatly affect our practice by preventing the use of an ineffective drug. A negative study can also tell us to look elsewhere for the answer, perhaps an answer that is even easier and more effective to implement (consider the ProMISe study). Negative studies that are published after positive ones are also important: In contrast to an earlier publication, a new study might fail to show a connection between a predictor an outcome because prior studies neglected to include important and unrecognised confounders that might mitigate the effect we are studying.

Indeed, the study by Lane et al suggests this. Although psychiatric boarding did not lead to longer inpatient lengths of stay, those who had the longest boarding times were also those that required a high observation inpatient bed, required antipsychotic and sedative/hypnotic medications in the ED, arrived via police or were on Mental Health Act detention. Older and younger (compared with middle-aged) patients had longer boarding times, as did those with conduct or neurodevelopmental psychiatric disorders. This suggests that those most difficult to treat, or to find an appropriate disposition for, had the most important, resource and time-consuming interventions performed in the ED before they were admitted. Conceivably, if these interventions had not occurred in the ED, they would have been necessary in the admitted areas and likely prolong hospital stay. In this case, the confounder is competent ED care.

In general, the literature on boarding suggests that the patients who remain longest in the ED are not the most acutely ill nor are they the least acutely ill. They are generally stable patients who need isolation or monitoring. However, in this ED, at least, the psychiatric patients who boarded were in fact the most acute. So its very likely that these patients were ‘boarded’ because they needed a higher level of care, which the ED was required and able to provide while the inpatient resources were found. Lane et al allude to this in their discussion pointing out that ‘The small absolute effect may be a positive sign of the resilience of healthcare providers finding ways to initiate treatment for patients with acute psychiatric illnesses whey they are boarded in the ED.’ Arguably, this means that the ED simply cannot say ‘the inpatient team will handle it’ when there is an agitated and potentially dangerous or suicidal patient in the department.

This is by no means a defence of boarding. It is simply an explanation of why, in this case, boarding did not prolong length of hospital stay. The analysis does not look at what this effort does to the care of other ED patients or its impact on staff. Although there is much pressure for us to do so, EDs cannot hold patients for extensive periods to pinpoint an exact diagnosis or manage the acute problems so patients arrive on the wards stable and fully ‘worked-up.’ Call me old-fashioned, but then, what is the hospitalisation for?

The study by Lane et al has a few challenging aspects, largely with regard to their use of Bayesian rather than frequentist statistics. In brief, the analysis does not use the concept of ‘statistical significance’—that is whether effect seen is unlikely to have been found by chance. Rather, it estimates how likely is it that there is an association between boarding and hospital length of stay. And they conclude that, for pyschiatric patients, the probability of there being an association, whatever its size, is quite high.

Ethics statements

Patient consent for publication



  • Handling editor Richard Body

  • Contributors The editorial was conceived and written by the author.

  • Funding The author has not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests None declared.

  • Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.

  • Provenance and peer review Commissioned; internally peer reviewed.

Linked Articles