Doctors and nurses make many treatment decisions. Only some are guided by gold standard evidence from randomised controlled trials (an RCT). Many are made without evidence and must rely on training and experience. There is often consensus for 'big decisions' but for smaller decisions, since each clinician's training and experience is different, there is frequent arbitrary variation.
For example, most clinicians will agree that prolonged antibiotics for an infection eventually cause more harm than good. The patient will develop side effects, and the microbes become resistant. However, there is no gold standard to say how many hours or days of each antibiotics are enough for each different infection.
Traditional RCTs are too expensive to answer all these questions. We have made some progress in reducing cost by embedding the RCT in an electronic health record (EHR) so that data collection and follow-up are then automated. However, consenting and randomising patients still requires a research nurse to visit each patient for each study.
We propose to also make the consent and randomisation process more efficient. For common treatments with existing arbitrary variation, we would pre-emptively consent patients either at the start of their health care journey, or through an opt-out mechanism supported by excellent communication. And for randomisation, we would deliver an alert in the EHR to the clinician at the point of decision making.
Our innovation is that the randomisation would not be mandatory. In standard RCTs, once a patient consents then both the clinician and the patient are expected to comply with the randomised decision. If this did not happen then the expense of recruiting would be wasted.
Our randomisation would be a 'nudge' to the clinician rather than a directive. This provides safety because without the research nurse to screen, consent and randomise then we protecting the clinician's right to override the randomisation. But because the electronic alert is so efficient then we more than make up for all the times the clinician overrules the nudge.
There are three further advantages.
- We also learn about the situations where clinicians consistently overrule the nudge. We identify scenarios where there is local expert knowledge that we should study and share.
- We better match the randomisation to the individual equipoise of each clinician-patient partnership. With standard RCTs, a single protocol must be followed by all participants. Some teams might disagree with a part of the protocol and so cannot participate. Conversely, other teams agree to participate but may feel unintended pressure to randomise specific patients who technically meet the trial protocol but for whom one treatment arm might seem better than the other. With nudged not mandated randomisation, then more teams can agree because individual exceptions are permitted.
- Because the trial is efficient, we can recruit more widely and examine the effect of the treatment for different groups of patients. Normally, a standard RCT produces a single result which is true on average but not best for all. The more efficient the trial, the better we can personalise our treatment plans.