Building Archimedes – a Q&#38A with ‘Dr. Data’

CIO INSIGHT: How did Archimedes get started?

EDDY: The origins go back a long way. This project is really the culmination of my career. In the late 1960s and early 1970s, I began to build mathematical models of diseases. Each model was focused on a particular disease, like cancer of the breast or cervix, and addressed a particular type of intervention, like screening. I was pleased with that work, but I was troubled by at least four things.

The first was that I kept feeling the need for a much broader model that spanned all the types of diseases and all the types of interventions, so that you could make comparisons across diseases and interventions. That is the type of model you would need if you wanted to help solve problems like setting priorities, or designing strategic goals, or designing performance measures.

The second thing that troubled me was that, with the exception of Archimedes, mathematical models in clinical medicine have been very superficial. They have no understanding of the underlying pathology, and simply skim across the surface of a disease. For example, a very prominent model of the coronary artery disease (CAD) complications of diabetes has a patient in one of two states; the patient either has CAD or does not. Needless to say, this is an extraordinarily simplified interpretation of coronary artery disease in people with diabetes. While this degree of superficiality might be acceptable for some problems – like simply counting the occurrence of a CAD complication – it is far too crude to address the overwhelming number of questions that clinicians and health administrators have.
The third problem was the lack of comprehensiveness of other models. When I analyze a clinical problem, I am certainly interested in the effects of the proposed options on clinical outcomes, like morbidity and mortality. But to be helpful in realistic settings, you also have to think about the effects on care processes, logistics and costs. So I began to want to build a model that would go from soup to nuts.
The fourth issue was validation. As strange as it might seem, very few models in clinical medicine have any independent validation. This has always bothered me a lot. While I had always tried to validate my earlier models, and while I can point to some success, I was limited by the fact that the other models weren’t realistic enough to perform a proper validation. I wanted to build a model that had the necessary biological and clinical realism, so that we could check it against reality.

When you put all those together you get Archimedes. Because it is such a large task, I wasn’t able to pursue it until about 10 years ago when the leaders of Southern California Kaiser Permanente agreed to fund it. They had the vision to appreciate its value, and got the project launched. Shortly after that we hired Len Schlessinger. His training is in particle physics. I am proud of my previous work, but Len has much more mathematical horsepower than I do. Whatever my contributions, I don’t believe this model could have been built without him.

CIO Insight Staff
CIO Insight Staff
CIO Insight offers thought leadership and best practices in the IT security and management industry while providing expert recommendations on software solutions for IT leaders. It is the trusted resource for security professionals who need network monitoring technology and solutions to maintain regulatory compliance for their teams and organizations.

Get the Free Newsletter!

Subscribe to Daily Tech Insider for top news, trends, and analysis.

Latest Articles