Thursday, July 12, 2007

Why Don't Electronic Health Systems Lead to Better Quality?

Earlier this week, the Archives of Internal Medicine reported the results of a study that examined the value of electronic record systems with respect to health care (see here and here). Let me stipulate that my career revolves around electronic record systems. I have deep experience in financial, quality, and manufacturing systems, but I don’t have much experience with medical systems.

Being a records system geek, I am predisposed to generally love the idea of automating manual and paper systems. The industry has come a long way from when I first dabbled in this field. We used to only be able to track a very limited number of data points. Today we have electronic record systems that are actual legal instruments (in place of paper documents) and that offer powerful reviewing and analysis capacities that were not even dreamed of when I took my first college computer course.

The aforementioned study focused on “17 ambulatory quality indicators” that researchers felt would determine the comparative quality of electronic health records (EHRs) as opposed to manually maintained health records. Clinics using EHRs “performed slightly better on two indicators: avoiding tranquilizers for patients with depression and avoiding routine urinalysis during general medical examinations,” but performed worse “when it came to prescribing the cholesterol-lowering drugs statins to patients with high cholesterol.” There was no statistical difference on the other 14 indicators.

Researchers concluded that EHRs “make little difference in the quality of medical care, at least when it comes to walk-in doctor visits.” They were surprised by these findings. They expected EHRs to perform significantly better than manual systems.

What is wrong with this picture? Is the whole premise of EHRs for improving medical outcomes flawed, or are the EHR systems themselves flawed? I ask this question because I have seen a lot of bad electronic record systems. The EHR standard specifies only data elements for data exchange. It says nothing about how the data management system must function. It is quite possible to create a lousy system that complies with the EHR data specification.

The study’s lead author, Dr. Jeffrey Linder says that “other studies have shown that most of the electronic health records that have been put in place are not much more than a replacement for the paper chart.” Boston Globe reporter Elizabeth Cooney says (here) that other studies have found that quality increases when systems provide doctors with decision support (i.e. helpful reminders) with respect to “preventive care and chronic disease management.”

Having been involved in electronic record systems for a long time, I would speculate that some or many of the systems used by the clinics studied were poorly designed systems. Design of electronic record systems is extremely difficult. The two main parties involved, users and system creators, come at the issue from very different perspectives. Much give and take is required. A good outcome is achieved only if both sides do their jobs well and both sides break out of narrow thinking.

In my experience, most users have difficulty thinking outside of the box of what they currently have. They hold tenaciously to concepts surrounding how their current paper/manual system functions. These systems are as much about the culture and power structure of the organization as they are about their actual technical uses. It is very difficult for many users to comprehend that some of their notions about what is absolutely essential simply won’t exist or will exist in a very different way in an automated system. This leads to the tendency to simply create an electronic system that does exactly what the current manual system does. Nothing is gained by this, at least initially. Users often soon discover ways that the system could be improved. The trouble is that achieving those improvements costs more money.

Some users have wild ideas about what electronic systems can do. In essence, they fill their requirement lists with science fiction. Computers do many things that people would have thought impossible just a few years ago, but they still can’t do many things that 1970s science fiction imagined. Some desired features can be delivered only at a cost that far outweighs the benefits that would be derived. A good tech designer will help users comprehend what is actually feasible and will help users shift to a new way of thinking about the powerful capacities actually available in an automated system.

Information systems folks have to break through their own barriers if they are going to develop a useful system. It is deucedly difficult for tech geeks to actually get a grasp on what is really important to the users. Because users are so familiar with their tasks and everyone they work with intrinsically understands what is important, they sometimes fail to properly convey what is truly important to them — because everybody should just know it. A good tech designer will learn to draw these factors out of users. System designers need to avoid the tech geek tendency of tacking on features they think are really nifty but that provide little value to the users.

On top of all of this are the issues of culture and budget. It is not necessarily true when it comes to information systems that you get what you pay for. You can spend a lot on a lousy system and you can get a good system for a bargain price if all of the right factors fall into place. But in general, you can’t expect to get a high quality system for cheap. And, as Dr. Linder says, good electronic systems only improve quality if the medical culture that includes “doctors, health systems, patients and payers” changes to focus on quality.

Without saying what kind of system he suggests, Linder asserts, “Part of the solution is to change health-care financing [so] that [it] pays for high quality care, not just more care.” I must say that every electronic record system project on which I have worked has actually ended up delivering pretty much exactly what was actually rewarded. If quality is the chief basis of reward in our health care systems, the supporting electronic systems will likewise end up being tailored to facilitate high quality outcomes. A different chief reward will necessarily lead to using electronic systems that support that reward structure, rather than supporting quality health outcomes.

No comments: