An HIT Moment with … is a quick interview with someone we find interesting. Eyal Ephrat, MD is founder and CEO of MedCPU of New York, NY.
What are the shortcomings of clinical decision-support modules of EHRs?
Decision support technology was designed with the best intentions, but accuracy remains a huge problem. Prompting the clinical staff with inaccurate or redundant prompts rapidly leads to frustration, alert fatigue, and loss of reliance on this feature. In most instances I’ve seen, decision-support prompts are ignored or turned off by a busy clinical staff, often because inaccuracy makes them unreliable and therefore unusable.
Roughly 70 percent of the patient’s clinical information exists today in free-form format such as dictations, follow-up notes and discharge summaries. As physicians, we just cannot communicate the clinical picture and plan of care through simple point-and-click pull-down menus and structured fields, so we opt for free-form notes. However, the computer cannot read free text, so the decision-support modules don’t see the 70 percent or 80 percent of critical information that exists exclusively in the free-form formats.
The clinical reasoning and thought process cannot be captured through simplistic “If-Then” rules. If the patient’s hemoglobin is 8gm/dL, it’s wrong to fire a simple prompt that alerts the physician to do something with it. There could be many reasons for such a low hemoglobin, ranging form chronic hereditary conditions that warrant no action to acute conditions that require emergency response.
How do you get the necessary data, including free-text information, to perform decision support?
The industry’s current technologies used for data sharing between systems – HL7 via interface engines and Web services – are not enough. They don’t provide all the data required, in real-time, for the accurate performance of the decision support modules. To resolve this critical barrier in information availability, MedCPU developed a unique Reader technology to collect all the data entered into the organization’s EMR via an API with the operating system (Citrix server, etc.) on which the EMR runs, without touching the EMR itself, without consuming computational resources, and without requiring integration to the EMR or the hospital’s IT infrastructure.
This allows us to see, for the first time in healthcare I believe, all the data entered in real-time. Combined with a limited use of HL7 feeds for getting information entered in the ancillary systems, such as dictations, radiology, and discharge summaries, MedCPU is achieving a complete picture about the patient, in real-time, from history until the present encounter.
What results have users seen?
I’ll give you a couple of examples. One hospital that was an early adopter of our VTE prophylaxis module has seen a significant improvement in compliance with the CMS’s VTE prophylaxis guidelines (above 90 percent from about 50 percent prior to the deployment of MedCPU) over a period of a couple of months. Another health system using our radiology module has seen a significant decrease in the amount of inappropriate imaging performed based on the ACR appropriateness criteria while generating higher revenues because of better appropriate documentation.
But we’re most proud of the daily events we see where the system actually prevents clinical errors. Seeing in the logs how the physician or nurse made a certain decision, got a prompt that the decision may be wrong, and as a result cancelled this decision and reverted back to the appropriate care path makes our huge efforts worthwhile.
What effort, expense, and expertise is required to deploy MedCPU?
The effort, expense, and expertise required is extremely low compared with the typical IT deployments we all know and have traditionally experienced. Using our Reader API, we request very little IT involvement on the part of the hospital, approximately 50 hours. The overall one-time deployment of the MedCPU platform in the organization takes about three to four months, during which time we also work with the organization’s clinical leaders in reviewing the best practice protocols contained in our decision support modules. The ability to deliver low-resourced deployment is critical when dealing with the often-overloaded IT departments.
What is the direction of the product and company going forward?
We want to become the high-precision decision support layer each organization critically needs on top of their existing EMR/IT infrastructure. We’re also really excited about our new initiative with the Health Management Academy. We’re launching a multi-health-system initiative that will foster collaboration in finding and testing advanced solutions in order to bring major improvements to their point-of-care clinical, operational, and financial performance.