In October of 2010, Jessica Zeppa, five-months pregnant, went four separate times to the emergency room, unable to swallow and fighting a fever. What no one knew was that Jessica had sepsis, a toxic infection that kills 750,000 people a year. Because sepsis resembles common fevers or dehydration, it can be difficult to spot. The signs were there, but it took hours to put the pieces together — too long to save Jessica or her baby.

In this age, one may ask why it took so long to diagnose a well-known, deadly condition. The reason is the tremendous fragmentation built into our health technology.


Today, we deploy dozens of electronic systems to log data, each recording different symptoms. But without a clinician working like Sherlock Holmes to pull the data together and detect patterns, it's all too easy for a case like Jessica's to fall through the cracks.

For all the hype around health information technology (HIT), it's not living up to its potential. Particularly when we look at electronic health records (EHR), most offerings are not intuitive, easy to use or integrated into the daily work of practitioners. And they are woefully inept at integrating data from other systems.

Just a couple of examples illustrate the point. Most HIT systems have built-in "alerts" to notify clinicians of potential problems, like an adverse drug interaction. But nurses at John's Hopkins report "alert fatigue," with bells ringing all the time and a false alarm reported every 90 seconds. In this environment, it's easy for the real emergency to get lost among the white noise.

Moreover, because systems don't work together, we have nurses manually double checking input from these "smart" devices. For instance, an infusion pump reports the level of pain medication being administered to a patient, as does the EHR. But these numbers sometimes don't match and must be double checked by at least two nurses to confirm the dosing.

A recent Health Affairs article noted that today's HIT systems operate less like ATM cards — allowing providers to access patient information anytime, anywhere — and more like frequent flyer club cards designed to preserve brand loyalty. In order to build bridges that connect disparate data sets, care providers must either pay outside vendors thousands of dollars to custom code links so they can "talk" to other assets or do it themselves at an enormous expense.

To unlock the potential of HIT, we have to stop holding data hostage behind proprietary walls. We must demand the use of open application programming interfaces (APIs) that would allow application developers to easily sync different data sets. The classic example of an open API architecture is the smart phone, where Google and Apple make their APIs available to developers, who then create a range of apps that can interact with almost any system. Using these apps, you can seamlessly tell all your friends that you're going out to dinner on Facebook, check in to the restaurant using Four Square and post a review after dessert on Yelp.

Already powerful in the consumer economy, imagine the potential of these apps in health care. Not only would this improve patient safety and make HIT systems more valuable, it would create new opportunities for developers to build cost-effective solutions that providers need to provide better, safer care. In fact, because health care is an untapped market for apps, a recent McKinsey & Company report concluded opening health care data could unleash $300 billion to $450 billion in annual economic potential in the U.S. alone.

Although the "open API" movement is growing and supported by leading experts such as the President's Council of Advisors on Science and Technology, progress has been slow. It's time for the purchasers of HIT — the doctors, hospitals and other large-scale institutions — to push harder and demand open APIs from any vendor selling HIT systems.

It's also time for government policy to usher in new interoperability requirements, making open API part of the mandate for certified technology required for meaningful use.

You wouldn't buy a car if the manufacturer provided no dashboard instruments, telling drivers instead to guess their speed and when it's time to refill the gas tank. So why would we buy such unreliable systems to provide critical care in emergency situations?

As Jessica Zeppa proves, patients are dying because of the HIT systems we are buying. We need to insist on open APIs and smarter devices that work together seamlessly. We owe it to Jessica, and every other patient that has slipped through HIT's cracks, to demand better.

Peter Pronovost is senior vice president for patient safety and quality at Baltimore's Johns Hopkins Medicine; his email is ppronovo@jhmi.edu. Keith Figlioli is senior vice president of health care informatics at Premier, Inc; his email is Keith_Figlioli@PremierInc.com.