Keeping up with all new information in medicine is simply not achievable or practical. The healthcare community has been coming to terms with this for some time, but with fifteen thousand biomedical research articles being published per week, we’ve long ago passed the point where any individual clinician can keep up through simple browsing of the literature1. With the rate of information generation increasing, “the gap between what we can learn and what is known is increasing all the time.”2

In response to the challenge of information overload, many services and strategies have begun to be adopted by clinicians and continue to be refined by publishers and software developers.

We have many excellent editorially-run literature update services like McMaster’s EvidenceUpdates and ACP JournalWise which have editorial groups manually review and recommend newly published papers for members of various specialty areas of medicine. To offer a more personalized approach, services such as Read by QxMD algorithmically curate new research to highlight high quality, relevant content based on what it learns from users’ reading behaviour. Publications such as New England Journal of Medicine’s (NEJM) JournalWatch and ACP Journal Club provide an additional level of assistance to the busy clinician, by summarizing and interpreting the results of relevant high-impact clinical trials.

But one of the unspoken limitations of most of these strategies is that it requires consistent usage and the ability to retain the newly reviewed information so that it can be applied when the appropriate clinical encounter arises, often many months or years later. A landmark trial offering a novel treatment for a rare disease doesn’t help the patient suffering from that condition if that article alert was missed during a busy week on call, or was simply forgotten as it was buried among 10 other important publications that were reviewed 6 months previously.

The moment when practice changing research becomes most relevant is when the appropriate clinical scenario arises. Knowing the best treatment for porphyria cutanea tarda really only becomes relevant in the event you encounter a patient with this rare disease. While there is great research which can guide prevention of a second stroke when aspirin fails the first time, this information is never more relevant than at the moment when a patient presents with a stroke despite longstanding adherence to regular preventative aspirin.

One might hope and expect that healthcare providers (HCPs) will look up the best diagnostic and therapeutic options when their memory fails them or they simply are unaware of the literature in the area.

And certainly they do. Services such as UpToDate, and Essential Evidence, as well as high quality review articles published in academic medical journals, are highly utilized for just this purpose.

But with the high rate of medical error in diagnostics and treatment3, clearly a large gap remains between what is unknown to the practitioner and what could be known.

Making the problem even more complex is that there are many things HCPs don’t even know they don’t know. If a patient on azathioprine develops gout, what if they are prescribed allopurinol for gout prevention without their physician knowing it can lead to severe bone marrow suppression and life-threatening infection? For a doctor not familiar with the use of immunosuppressive drugs, this is a serious drug-drug interaction they may simply be unfamiliar with.4

The good news is that more and more, all the data needed to identify knowledge gaps in real time already exists in various electronic systems - diagnostic codes, lab results, drug names and prescription dates are already readily available. At the same time, we’re seeing tools such as IBM’s Watson that can process massive treasure troves of reference material while analyzing the content of clinical databases along with the free-form written notes of a clinical encounter.

What we need, and will undoubtedly see developed, are true decision support solutions that automatically recognize knowledge gaps and present data and suggestions at the appropriate moment in a patient’s journey. These types of interventions need to present themselves at the moment right before a prescription is written, not when the keen-eyed pharmacist (hopefully) notices a potential drug error.

Is this science fiction? Certainly not.

A recent NEJM publication5 suggests that most patients with atrial fibrillation who are undergoing surgery do not need “bridging” blood thinning injections while they are off their long term anticoagulant. Our electronic systems already include surgical booking schedules, a patient’s diagnosis of atrial fibrillation and prescription data that indicates the use of an oral anticoagulant. At the point of booking the surgical date, admission to hospital, or prescription of bridging anticoagulation in hospital, a clinician could be presented with the relevant publication which offers the data showing this intervention is both unnecessary and increases the risk of bleeding, or the UpToDate summary providing a concise recommendation based on this research.

This type of solution could be easily built in a weekend-long Hacking Health project, assuming access to the clinical data is unobstructed. In fact, our greatest challenge will not be identifying the clinical areas amenable to advanced decision support, nor writing the algorithms to ensure prompts have very low false negative and false positive rates. Our greatest challenge as a healthcare community will be forging the partnerships and economic relationships between players who currently often silo this data.

Beyond the need to apply knowledge at the point of clinical relevance, frequently clinicians may not even be aware of the clinical data that would make the landmark research relevant.

In a patient with kidney disease and protein in the urine who is being treated for high blood pressure, the use of ACE inhibitors and Angiotensin II receptor blockers (2 classes of drugs) help prevent kidney failure6. This is well known to most clinicians due to years of well-publicized clinical trials and clinical practice guidelines.

But when a physician is ready to prescribe a blood pressure lowering drug to a hypertensive patient, they may not select the appropriate agent that offers kidney protection in addition to blood pressure lowering if the abnormal kidney function and urine protein values are situated in an electronically siloed database. But systems that can crawl about and find those relevant data points will be able to offer perfectly-timed, evidence-based recommendations by combining a clinical need (blood pressure treatment), clinical data (laboratory test results) and the published evidence in a just-in-time delivery model.

Even within existing integrated data systems, we often bury relevant clinical data. In an anemic patient with enlarged red cells (indicated by an elevated mean corpuscular volume, or MCV), the potential causes, as listed in hundreds of academic review articles or textbooks, include a laundry list of conditions such as liver disease, thyroid disease or B12 deficiency. With today’s electronic health records, a clinician would need to identify the abnormality (a high MCV), go to a reference source to find the potential causes (or rely on the list memorized in their head), and then manually hunt through the existing clinical database to find the results for liver, thyroid, and B12 testing. It would take little imagination to ask for a solution that automatically surfaces a common abnormality, hunts through the existing literature to come up with diagnostic options, then offers up existing test results that may already rule in or rule out potential causes, and finally suggests additional testing that may still be necessary if the mystery persists.

Not only does providing knowledge at the exact point of clinical relevance offer an opportunity to unite best evidence and optimal patient care, it also represents a powerful teachable moment. Teaching in medicine has always been centered on the patient case, whether in case-based learning sessions or clinical rounds. Clinicians learn best in real world scenarios,7 knowing that a real person is impacted by the results of their learning. Knowledge acquired through advanced electronic solutions in the context of a real patient story is more likely to be retained and recalled the next time a similar clinical scenario arises.

Overall, we need to shift from a ‘pull’ to a ‘push’ model of healthcare knowledge dissemination. Rather than expect all clinicians to be tirelessly thirsting for new knowledge and proactive with continuing medical education, we need to recognize that this approach, while necessary, is not sufficient.

Doctors and nurses are people just like everyone else. They get busy and they forget. They make errors of omission and commission. We still have the opportunity to go after incredibly low hanging fruit - use existing data sources, published research and easily constructed decision support to build systems that identify the knowledge that a clinician needs at any point in time and inject it seamlessly into the patient encounter and a healthcare provider’s workflow.

Who the winners will be that build these all-seeing, all-knowing, point-of-care knowledge solutions is unknown. But as they inevitably become ubiquitous, very quickly doctors won’t be able to compete in terms of recall of clinical knowledge. But that’s just fine.

Just as advanced bedside imaging will enhance the physician’s cardiac diagnostics while simultaneously making stethoscopes obsolete, new tools for knowledge recall will enhance the skills of a physician, not replace the need for them. Physicians and other healthcare providers will find themselves freed up to focus on higher order tasks. As knowledge gaps become distant memories of a forgotten era in healthcare, physicians will be able to focus on what they should do best: build relationships, advocate for their patients and community, apply judgment, motivate patients to achieve personal change, and treat their patients holistically with both empathy and compassion.