In an article in last month’s Modern Healthcare, noted cardiologist and medical futurist Eric Topol lamented that the current set of electronic health record (EHR) vendors do “not properly deliver all the data clinicians need to practice more accurate medicine.” He went on to say that there “[is] so much homeless data out there,” referring to “genomics, readings from wearable sensors, environmental factors, and other items that do not fit neatly into traditional EHR platforms.”
Dr. Topol brings up a valid point about how we (all of us) are generating Big Data (I try to avoid buzz words, but if the shoe fits...) A recent estimate of how much data we create every day worldwide sets the total at 1.145 trillion megabytes per day. I’m no mathematician, but I’m confident that’s a big number. Even if this guess is wrong by a few zeroes, it’s still a big number. This includes non-healthcare information like phone location, tweets, emails, photos, thermostat readings, and the like.
In healthcare, the data seem endless. Genomic data generally include information about the individual chemical bases (As, Cs, Gs, and Ts for the science-focused among us) that compose our genes and chromosomes. The NIH estimates that the storage of a single human genome sequence takes up about 200 gigabytes. Continuous glucose monitors can easily generate a glucose reading every minute. Think about the digital readings that an implanted cardiac pacemaker might create. I remember when physicians were worried that they would be overwhelmed with patients uploading their daily weight every day via the EHR patient portal. Today, that concern seems quaint!
Still, there is other potentially important information that is relevant to the patient and their care, but is not generated by the patient themselves. We know that social determinants of health (SDOH) play a huge role in patient outcomes, hence the need to include related items such as weather factors, the availability of fresh and affordable food, crime statistics, and pollution information for instance. Let’s not forget about health specifics of close (or not so close) family members. Some of their test results and genomic details can be essential in the outcomes of the patient in front of us.
To address Dr. Topol’s point about homeless data, let me put forth two presumptions: first, storage of large amounts of data is becoming easier and cheaper every year. It’s not free, to be certain, but it is becoming less of a concern in the big picture. Second, and more importantly in my opinion, the problem is moving from presenting physicians with reams of data to presenting them with knowledge.
Doctors are drowning in information today. They’ve been inundated with data points for well over a decade. With increasing interoperability, as basic as it is even now, many clinicians currently have access to progress notes, lab results, imaging assessments, and other discrete and narrative information. This includes information from the patient, from their current healthcare organization, and from other clinics and hospitals across the country and even the world. Yet, finding the needed data points can be like looking for a needle in a haystack.
In the past, physicians have used the EHR to ask questions like “Has this patient ever been on this medicine?” or “How has a certain lab result been trending?” With new sources of data come new kinds of questions: “Based on this patient’s genetic profile, medical history, and family history, what is the best class of cholesterol-lowering medication to start this patient on?” or “Can I get a summary of the workup this patient has had over the last five years to address their jaundice?” Physicians need our technology to present them with knowledge, not just information.
I’m not worried about where homeless data will live. I think we’ll figure that out easily enough. I do worry about how we’ll transform data (both with and without an obvious “home”) into knowledge and present it to clinicians at the right time and via the correct workflows to help them care for their patients. This is the difficult work ahead of us.
We are making progress in converting data points into actionable knowledge in the EHR. A recent article in JAMIA highlights how this can be done with an unlikely source of homeless data: a physician’s own ordering patterns and those of their peers. By combining patient-specific data with machine-learning algorithms that examine how physicians have treated similar patients, drug prescription alerts were markedly reduced while the utility of the alerts (as measured by changes in therapy) increased significantly. Cyber assistants such as these are welcomed by doctors at a time when many rule-based algorithms produce more noise than help.
At Nordic, we’ve been saying that better data mean better insights, or rather: healthier data, healthier people. The information that we collect about and from patients constitutes the base for tools such as the EHR. The computer science truism “garbage in, garbage out” rings true with respect to the knowledgebase in the medical record; an incomplete collection of facts, findings, and evidence will not allow the clinical decision support (CDS) algorithms to work as well as possible. Whether the data have been incorporated into the EHR for decades, or they just came into existence a few years ago, we must ensure that information continues to be refined and available for the technology to help us move forward.