In 2001, when my colleagues and I ranked nearly 100 patient safety practices on the strength of their supporting evidence (for an AHRQ report), healthcare IT didn’t make the top 25. We took a lot of heat for, as one prominent patient safety advocate chided me, “slowing down the momentum.” Some called us Luddites.

Although we hated to be skunks at the IT party, we felt that the facts spoke for themselves. While decent computerized provider order entry (CPOE) systems did catch significant numbers of prescribing errors, we found no studies documenting improved hard outcomes (death, morbidity). More concerning, virtually all the research touting the benefits of HIT was conducted on a handful of home-grown systems (most notably, by David Bates’s superb group at Brigham and Women’s Hospital), leaving us concerned about the paucity of evidence that a vendor-developed system airlifted into a hospital would make the world a better place.

Since that time, there have been lots of studies regarding the impact of HIT on safety and, while many of them are positive, many others are not. In fact, beginning about 5 years ago a literature documenting new classes of errors caused by clunky IT systems began to emerge. A study from Pittsburgh Children’s Hospital found a significant increase in mortality after implementation of the Cerner system – a study that was criticized by IT advocates on methodologic grounds, and because “they didn’t implement the system properly.” Studies by Ross Koppel of Penn and Joan Ash of Oregon (such as here and here) chronicled the unintended consequences of IT systems, and urged caution before plunging headfirst into the HIT pool. I raised similar concerns in a 2006 JAMA article, and also recounted the iconic story of Cedars-Sinai’s 2003 IT implementation disaster, where a poorly designed interface, combined with physician resistance to overly intrusive decision support, led the plug to be pulled on the $50 million CPOE system only a few weeks after it was turned on.

A new story line was emerging, and its theme was that implementing an effective HIT system is harder than it looks. And yet one could not deny the political attractiveness of computerization – during the last presidential campaign, the need for HIT was the only thing that McCain and Obama seemed to agree on, and a prominent proponent of a Manhattan Project-like push for HIT has been none other than Newt Gingrich, not exactly a freewheeling spender.

These politics led to $19 billion being included in February’s stimulus package to support HIT implementation in American hospitals and clinics. Once federal HIT Czar David Blumenthal figures out how to divvy up the money (it hinges on coming up with a workable definition of “meaningful use” of HIT, which would unlock the door to the federal vault), HIT will have its big Coming Out Party. Cue the balloons and streamers!

It’s all pretty exciting… if the systems work.

Fast forward to today’s Washington Post, where an article describes a new crusade by Iowa Senator Charles Grassley to confront HIT vendors who sell defective products. Grassley, who has taken on the role of keeping professions honest (including scrutinizing physician relationships with pharma and device companies, attention that is long-past due), has raised a number of concerns about the safety of store-bought IT systems, and has sent a letter of inquiry to one of the companies (Cerner) that looks a lot like a prelude to a Senate subpoena.

Interestingly, many of the issues raised in the Grassley letter mirror an argument advanced by Penn sociologist Ross Koppel in JAMA earlier this year (for which he was vilified by the IT community – both implementers and vendors). Ross (who is a friend) noted that most of today’s IT implementation contracts insist on gag clauses for clinicians who identify errors caused by faulty software, and virtually all contain hold-harmless clauses for the vendors in the event that an IT-related error leads to patient harm. The vendors’ case seems like a version of the “guns don’t kill people…” argument: there is nothing wrong with the software, the errors reflect poor implementation practices or screw-ups by users, yada yada. This is certainly true at times, but Ross and others have documented scores of errors that are absolutely inevitable given clunky software and poor user interfaces. It seems right that the vendors would at least share responsibility if patients were harmed in such circumstances.

This is all acutely interesting to me right now, since my own hospital (UCSF Medical Center) recently entered the not-very-proud fraternity of hospitals who aborted their implementation after an IT system failed to live up to expectations. In our case, nearly a decade ago, we put our IT nickel (actually, more than a billion of them) down on a system built by a vendor named IDX. A few years later, when IDX became wobbly as a company, we were reassured when technology titan General Electric gobbled them up. “They’re GE,” we thought, “they’ll get this right.” We eagerly signed on to be development partners with GE and found ourselves in a world of missed deadlines and inadequate support; in short, neither product nor vendor seemed ready for prime time. After years of negotiations, hair-pulling, and prayer, we recently pulled the plug on our GE relationship. (I’ve been on the committee overseeing our transition, and won’t divulge any confidences, but all of this has now been reported in the media.) Just last week, our CEO Mark Laret publicly announced our intention to pursue an implementation with Epic, the Wisconsin company that appears to be emerging as the best of the breed. I’m hoping that this is the system we’ve been waiting for (the reviews from colleagues who are using Epic elsewhere are generally reassuring), and not, as my friend Jim Reinertsen sometimes quips, simply “The Cream of the Crap.”

Last year, I wrote about the Technology Hype Cycle, a predictable roller coaster in which new technologies are over-hyped (the “Peak of Inflated Expectations”), fail to live up to their expectations (the “Trough of Disillusionment”), and ultimately (if they’re any good) traverse their “Slope of Enlightenment” before reaching a “Plateau of Productivity.” CPOE, and HIT more generally, are clearly on this roller coaster – somewhere between the “Trough…” and the “Slope…” Since we are about to invest 19 billion tax dollars on nationwide implementation of these systems, let’s collectively hope it is the latter.

Obviously, we simply must computerize American healthcare: in 2009, how can we possibly improve our care and coordination when we document our work by writing in chicken scratch on pieces of dead trees? And there are healthcare organizations that have enjoyed successful implementations and are beginning to reap real benefits, in quality, safety, and efficiency.

But it is not a slam-dunk, and there are some crummy systems out there that have the capacity to cause harm. Having folks like Ross Koppel, and maybe even Senator Grassley, push the IT companies to do better and be accountable for their products is critical if we’re going to get this complex but crucial task right.

Bob Wachter is chair, American Board of Internal Medicine and professor of medicine, University of California, San Francisco. He coined the term “hospitalist” and is one of the nation’s leading experts in health care quality and patient safety. He is author of Understanding Patient Safety, Second Edition, and blogs at Wachter’s World, where this post originally appeared.