Automation is Critical in Modern Life. In Some Key Areas, it’s Not Going Well.
Not a Good History
I’ve been reading a hair-raising story in the current edition of Fortune magazine about the nation’s medical Electronic Health Records system (EHR). The article follows an extensive joint investigation by Fortune and by the Kaiser Health Foundation. What they found will keep you awake at night (http://fortune.com/longform/medical-records/).
But that is not the central story we are looking at today. It’s the fact that this is not all that unusual a story.
I have some familiarity with 5 major government automation projects:
- Electronics Health Records (EHR)
- The Federal Aviation Administration (FAA) systems upgrade
- The Veterans Administration (VA) system of systems upgrades
- The Affordable Care Act implementation
- The TRICARE military health care system implementation
Only one of these could reasonably be called a success, the last one (TRICARE). More on that later. For now, what is going on here that four major programs, of life and death importance and billions of dollars in cost, all started out badly?
All brought some improvements, but all also brought a ton of new problems. Here is a good example. According to an independent study, the average ER doctor must complete 4,000 (!) key strokes in a single work shift. Most doctors doing routine medicine have about 11 minutes with a patient. Doctors are spending an amazing amount of that time doing data entry, not talking with patients. This is not good.
The temptation for many of us will be to conclude that this is proof that government cannot do things as well as the private sector. That would be a hasty, and incorrect, conclusion. Every one of these programs run on commercial programs, developed by the private sector to support government needs. Failures of this scale are an indictment of both the government and the private sector.
There is a Pattern Here
In looking at the EHR case, there is evidence of what went wrong among all these programs.
- First, too much was attempted to be accomplished in one fell swoop. Too many accomplishments and integrations were attempted in ways that have never been attempted. It probably would have been better to set out all those accomplishments as goals, but to set up the processes one category at a time. Get the entry process right, then move on to integration as a phase II. Over time, all the functions can be there, and coordinated.
- Second, much of the data entry and collection processes rely on old approaches. People with other things to do find themselves constantly entering key strokes, reacting to pull down menus, cancelling alerts and warnings, etc. The human mechanical processes required to create automated data is archaic and out of balance. One citation in the Fortune article illustrated this nicely: what doctors and other medical professionals are now required to do is like telling a judge that in addition to running a trial, they have to be their own stenographer. Too many steps – period.
- Third, there was no meaningful phase-in for these programs. They all went on line nationally, all at once. Remember how much fun that was for the Affordable Care Act? There was very little offline simulation testing, no implementation by functional area or region. Just jumping into the deep end of the pool. I would not be surprised to find that related to this would be a lack of testing by people in the field who would actually be using the systems once fielded.
- Fourth, there was little in the way of performance accountability for contractors. When systems fail, vendors blame clients for not using the systems correctly. They could be right in some cases. Hard to say since there were few performance standards in place to judge the results.
A Better Way?
TRICARE certainly was not perfect in implementation (and still has challenges), but overall, it went much better than these other systems. It has the same sort of complex requirements as the other programs. Yet it has gone generally better. Why?
Two factors come to mind. One, the system was implemented over several months, one region at a time. As each implementation took place, lessons were learned and applied to the next region. Secondly, reasonably clear and detailed performance standards were put in place, not just for automation but for all components of health care support. Major contract holders have been dropped in recompetes for failing to meet those standards.
The lessons here are that we have not, as a government, laid out an intelligent model for how to design, implement, and adjust our most important government automation programs. We also lack a widely adapted model of how best to partner government and industry to bring about such programs.
This type of problem will not go away. It will, in fact, only get worse and more prevalent. If there is anything worthy of a Manhattan Project scale project of joint study and process design by government and industry, it is this one.
Let’s stop wasting time, money, and putting lives at risk. Get the model in place and use it.
If you find this blog worthy of your time and curiosity, I invite you to do two things:
(1) Join the conversation. Your voice counts here.
(2) Share the word about this blog with friends and colleagues. Share a link in your emails and social media posts. Let’s grow our circle.