Here's something I've heard over and over while on my “executive rounds” as part of our patient safety initiative: Automation will save us. All we need is the right computer system in place, and our errors and mistakes will dissolve into the etherspace.
My response surprises people. I tell them that automating a defective process only makes the defective process more effective at producing mistakes. Yet when I say this, I can't help but get the feeling that few of those listening actually believe me. They nod their heads and silently decide Brody is antitechnology. The fact that in the past I picked up a couple of degrees in electrical engineering and computer sciences is viewed as a strange aberration of my true feelings about technology.
Nonetheless, I have continued insisting that we must first fix our defective processes, then use automation to help implement simpler, less error-prone systems. Without fixing the process first, we might actually make error production more efficient.
Unfortunately, I didn't know how right I was. Last month I learned that a patient I had referred to Hopkins Hospital had a complication leading to transient renal failure after a diagnostic procedure. When I investigated, I found out from the radiologist that the mistake occurred through a miscommunication. But what astounded me was when I was told that since the Hospital implemented the provider order-entry system, “we've encountered more inaccuracies in the information that we use in conducting diagnostic exams.”
I spoke about this incident to Peter Pronovost, our medical director for the Center for Innovation in Quality Patient Care. Peter told me that he analyzes data for a consortium of hospital ICUs across the country. Not long ago, he began to see a changing trend in three hospitals that were reporting an increased error rate in their ICUs. Upon investigation, he discovered that all three had recently installed an automated provider order-entry system.
In my graduate student days in electrical engineering and computer science at MIT, we called this phenomenon GIGO: garbage in, garbage out. If you give a computer bad data, it will very efficiently help you reach bad conclusions. GIGO can actually speed up the propagation of errors, leading in the health care setting to potentially disastrous effects on patient care.
Toyota , the pioneer in zero-defect manufacturing, implemented its revolutionary approach to high-quality automobile production using the so-called kan-ban system.
What many don't know is that the Toyota Production System, as it is now called, doesn't rely on any computer-based automation. The kan-ban was a paper card used among various areas of production and parts suppliers, such as body assembly, drive train, or paint and finishes. It was simple, yet highly effective. Kan-ban worked because Toyota redesigned and simplified the processes first—and then came up with a system to implement them.
As I travel the country, I am inundated with helpful advice from CEOs, health care providers and others, all of whom tell me that automating the medical record is the solution to our health care problems. “Implement a universal medical record,” they say, “and health care costs will melt by 50 percent and errors will disappear.”
Don't believe it. We've got to fix the systemic problems of health care delivery that require an integrated care delivery approach. After that—and only after—can we use information technology effectively.
Garbage In—Garbage Out
Here's something I've heard over and over while on my “executive rounds” as part of our patient safety initiative: Automation will save us. All we need is the right computer system in place, and our errors and mistakes will dissolve into the etherspace.
My response surprises people. I tell them that automating a defective process only makes the defective process more effective at producing mistakes. Yet when I say this, I can't help but get the feeling that few of those listening actually believe me. They nod their heads and silently decide Brody is antitechnology. The fact that in the past I picked up a couple of degrees in electrical engineering and computer sciences is viewed as a strange aberration of my true feelings about technology.
Nonetheless, I have continued insisting that we must first fix our defective processes, then use automation to help implement simpler, less error-prone systems. Without fixing the process first, we might actually make error production more efficient.
Unfortunately, I didn't know how right I was. Last month I learned that a patient I had referred to Hopkins Hospital had a complication leading to transient renal failure after a diagnostic procedure. When I investigated, I found out from the radiologist that the mistake occurred through a miscommunication. But what astounded me was when I was told that since the Hospital implemented the provider order-entry system, “we've encountered more inaccuracies in the information that we use in conducting diagnostic exams.”
I spoke about this incident to Peter Pronovost, our medical director for the Center for Innovation in Quality Patient Care. Peter told me that he analyzes data for a consortium of hospital ICUs across the country. Not long ago, he began to see a changing trend in three hospitals that were reporting an increased error rate in their ICUs. Upon investigation, he discovered that all three had recently installed an automated provider order-entry system.
In my graduate student days in electrical engineering and computer science at MIT, we called this phenomenon GIGO: garbage in, garbage out. If you give a computer bad data, it will very efficiently help you reach bad conclusions. GIGO can actually speed up the propagation of errors, leading in the health care setting to potentially disastrous effects on patient care.
Toyota , the pioneer in zero-defect manufacturing, implemented its revolutionary approach to high-quality automobile production using the so-called kan-ban system.
What many don't know is that the Toyota Production System, as it is now called, doesn't rely on any computer-based automation. The kan-ban was a paper card used among various areas of production and parts suppliers, such as body assembly, drive train, or paint and finishes. It was simple, yet highly effective. Kan-ban worked because Toyota redesigned and simplified the processes first—and then came up with a system to implement them.
As I travel the country, I am inundated with helpful advice from CEOs, health care providers and others, all of whom tell me that automating the medical record is the solution to our health care problems. “Implement a universal medical record,” they say, “and health care costs will melt by 50 percent and errors will disappear.”
Don't believe it. We've got to fix the systemic problems of health care delivery that require an integrated care delivery approach. After that—and only after—can we use information technology effectively.
