Frequently Asked Questions

What is the Archimedes Model?

The Archimedes Model is a large scale simulation model of physiology, disease, interventions, and health care systems, written at a high level of detail using object-oriented programming and run on a distributed computing network.

The core of the Model is a set of ordinary and differential equations that represent the physiological pathways pertinent to diseases and their complications. The Model also includes aspects of diseases and health care systems needed to analyze downstream clinical events, utilization, and costs including: signs and symptoms; patient behaviors in seeking care; patient encounters with the health care system (e.g. emergency room visits, office visits, and admissions); protocols and guidelines; tests and treatments; provider behaviors and performance; patient adherence to treatment recommendations; and clinical events that affect logistics, utilization, and financial costs. Protocols, performance, compliance, and costs can be customized to match different care delivery settings.

The Model represents physiological variables and their relationships as faithfully as possible given current science and evidence. For example, variables that are continuous in reality are continuous in the Model (e.g. blood pressure, glucose levels), time is continuous, interactions between variables are represented realistically, symptoms are driven by underlying variables, tests measure underlying variables, treatments affect underlying variables, and outcomes are determined by the progression of the variables.

Costs related to the conditions that are in the Model are calculated by tracking all the pertinent cost-generating events using micro-costing methods. Costs of other conditions that are not currently calculated in the Model such as osteoporosis (“unrelated costs”), can be added separately as a function of variables that are in the Model (e.g. age, sex, weight, disease states). Costs of the last year of life can also be set separately.

The Model uses person-specific data from real populations to create simulated populations that match the real populations, person-by-person. Each individual can be matched on more than 40 variables relating to demographics, risk factors, behaviors, biological variables, symptoms, current and past medical histories, and current treatments. The methods for creating the copies of real people preserve the distributions and correlations of all the important risk factors and biological variables. » More

ˆ top

What diseases are in the Model?

Currently, the following diseases and conditions have been modeled:

• Diabetes
• Coronary artery disease
• Hypertension
• Congestive heart failure
• Stroke
• Dyslipidemia
• Obesity
• Metabolic syndrome
• Asthma
• Colon cancer
• Breast cancer
• Lung cancer

All of these diseases are integrated into a single simulation model. This enables the Archimedes Model to accurately account for co-morbidities, syndromes, multiple medications, and medications with multiple effects. It also enables the Model to make comparisons and set priorities across different diseases using a consistent, validated methodology.

ˆ top

Where do you get data for building the Model?

The Archimedes Model is based on data from a variety of sources ranging from basic science studies to surveys, clinical data, and clinical trials. For example, models of physiology and diseases are based on basic science studies and clinical trials. Information about the effects of behaviors and risk factors for diseases is obtained from epidemiological studies. Information about the demographics of populations comes from large administrative data sets. Information about the effects of treatments is obtained from clinical research. Protocols and guidelines are based on reports issued by national organizations. Information about personnel, facilities, processes of care, performance, and logistics is based on the experiences of existing medical centers. Cost information is drawn from accounting departments.

ˆ top

When building the Model, what do you do when studies show different or conflicting results?

We use standard methods of systematic reviews to select evidence and determine the strengths of different experimental designs. If two pieces of evidence are both high-quality and yet show apparently different results, we examine the details of each to see if the conflicting results can be explained by subtle differences in the populations, treatment protocols, treatment details (e.g. drug dose or administration), definitions of outcomes, follow-up times, compliance rates, or other factors. If so, then we use one study to build the Model, and use the other to validate it.

If there is no clear explanation of the differences or conflicts in results, then we choose the study with the strongest design for the reference case, but note the existence of the conflicting evidence. The importance of the conflicting evidence can then be studied through sensitivity analysis and "what-if" analyses. In this situation, we present results for the conflicting scenarios so decision makers can understand the implications of the uncertainty.

ˆ top

How has the Archimedes Model been validated?

The core equations of the Model - those that describe the progression of diseases and how they respond to treatments - are validated by simulating actual clinical trials. This is a very deep validation that tests the entire chain of events from the fundamental biology and pathology of the disease, to the development of symptoms, patient behavior in seeking care, performance of tests, and delivery of treatments; to the changes in biological and health outcomes.

The first 74 validations, involving eighteen clinical trials, were supervised by an independent committee of experts in diabetes and cardiovascular disease and have been published. The results of 71 of 74 comparisons between the real-world clinical trial and the Archimedes Model's simulated trial were well within the sampling error. The other three either just missed or the description of the trial was incomplete. These validations involved 18 separate trials, 10 of which were not used to build the Model and provide independent validation of the Model’s accuracy. For the exercises not used to build the Model (100% independent), the correlation between the Model and trial was still very high, with r = 0.96.

These results build confidence that the Model will be accurate for new applications. However, they do not guarantee accuracy for all problems. » More

ˆ top

How does the Archimedes Model differ from other models?

The Archimedes Model differs from other models in many important ways. Our Model includes the following:

• It is a structural model built up from the underlying physiology
• It is built at the level of detail at which real clinical, administrative and policy decisions are made.
• It incorporates multiple diseases in a single integrated model
• It includes physician behaviors, performance, and practice variations
• It includes patient behaviors in seeking care and adhering to treatment recommendations
• It is broad, beginning with the underlying physiology but going through to include important aspects of delivery systems such as visits, admissions, protocols, tests, treatments, personnel, utilization, and costs
• It tracks cost-generating events at the same level of detail as real health plans.
• It can create copies of real populations, described either in the aggregate or at the person-specific level
• It can be customized to specific populations, settings and costs
• It has been checked by accurately simulating more than 50 major clinical trials

ˆ top

How does the Model differ from Markov Models?

Whereas state-transition models assume that diseases can be adequately described as transitions between a small number of discrete "states" at discrete time intervals (e.g. "no diabetes", "uncomplicated diabetes", "diabetes with CVD"), the Archimedes Model represents the physiological variables that actually underlie the diseases (e.g. insulin resistance, hepatic glucose production, glucose uptake by fat and muscle, plasma glucose, …). It retains the continuous nature, interactions, and progression of biological variables, and therefore can accurately describe the complexity of diseases.

The defining characteristic of a Markov model is that it is discrete, which is the opposite of continuous. In Markov models, clinical conditions are characterized as consisting of a small number of discrete states, and time is broken into discrete intervals.

A Markov model works by characterizing a person as being in one and only one state at any time. Time is broken into discrete intervals, usually one year. Thus, whatever state a person is in on January 1st, the model will assume the person stays in that state through to December 31st. The progression of a condition is represented by allowing the person to move at midnight on December 31st to a new state. The pace of the progression is determined by "transition probabilities"; for every state there is a probability that on midnight of December 31st, any patient in that state might jump to any of the other states. For example, a person in the "No complication" state might jump to the state "Have CAD".

Before going any further, at least three characteristics of the Markov approach are immediately apparent. The first is that the Markov model is not set up to model care processes or system resources. If you want a model to include any of those, you will either have to choose a different type of model, or you will have to append a different type of model to the Markov structure.

The second observation is that the discrete states are a highly simplified representation of a disease. This is fine if you are only interested in keeping track of the occurrence of these outcomes at annual intervals, (and if you know the transition probabilities), but it is not the right model if your questions are deeper than that.

A third observation has to do with the transition probabilities. The integrity of the entire model rests on them. More specifically, they are being called on to capture and accurately represent everything that might happen to a person during the year, such as changes in the patients anatomy (e.g. progression of an occlusion in a coronary artery), pathophysiology (e.g. gradual increase in fasting plasma glucose), development of signs and symptoms, patient behaviors in seeking care, visits, workups and decisions by physicians, performance and interpretation of tests , treatments, actions by patients, effects of treatments, and so forth.

By themselves, these three issues rule out the Markov model for healthcare modeling. Healthcare modeling includes care processes, logistics, resources and costs. That is where at least half of decision making in healthcare lives: continuous quality improvement, disease management, case management, the "how to do it" parts of guidelines, efficiency, purchasing, budgeting, forecasting, premiums and rates, and on and on.

The Markov approach will not give us what we need on the clinical side either. A Markov model works well if you are only interested in keeping track of the occurrence of the states (e.g. get CAD", "Get ESRD", at annual intervals, and if you are content to let the transition probabilities finesse everything that is happening to a person between the annual jumps. We were not content to do that. Just the opposite: it is precisely the changes in the patients anatomy and pathophysiology, development of signs and symptoms, patient behaviors, visits, workups , tests , treatments, actions by patients, effects of treatments, and so forth, that we were building our Model to address. The questions we wanted to be able to answer are all happening in between the Markov transitions and underneath the Markov states. For example:

• What if we slow the rate of occlusion with a statin?
• Which statin and dose should be the first line choice?
• What should the goal of treatment be?
• How long should we try the first line choice before escalating the dose, switching drugs, or adding another drug?
• What if we use both a statin and Metformin?
• What if we make a patient more alert to the warning signs of a heart attack?
• How much difference would anti-inflammatory make? I a patient is taking aspirin, could we settle for a looser goal for cholesterol treatment?
• What if we began to use ultrafast CT and triaged patients for angiograms on the basis of their calcium scores?
• What effect would any of these have on office visits, lab costs, the pharmacy budget, or the need for stents?

These and hundreds more like them are impossible to address with a Markov model.

Markov models can be quite useful for certain problems; however, there is a very large class of problems faced by a clinicians and administrators that are simply not amenable to this approach. Our objective was to offer a tool to address those problems. It is for that reason that we developed an entirely different type of model.

ˆ top

Why is customization important?

The effects of interventions can depend as much on the current level of care (the "control") as on the new level of care that is the nominal target of the analysis (the "treatment"). For example, the measured effect of population-wide screening for hypertension will depend on whether it is compared to "no testing at all" versus "sporadic testing at routine office visits". The latter, in turn, can be very different in managed care settings, versus fee for service, versus no insurance. Similarly, the effect of a treatment on costs can depend as much on the cost of routine care (e.g. office visits) and downstream events (e.g. treatment of strokes) as on the cost of the hypertension test or treatment. As shown in studies of performance and practice variations, levels of care can vary widely in different settings. The default setting of the Archimedes Model is based on nationally recommended guidelines, average performance levels, and protocols and costs experienced in managed care settings. Thus it is suitable for a wide range of problems that use a "national" or "standard--of-care" perspective. However, for each new analysis it is important to determine whether a different set of assumptions would be more appropriate. To address this possibility, the Archimedes Model can be customized to represent different levels of care such as different guidelines, protocols, performance levels, and costs.

In the same way that different patterns of care and costs affect the results of an analysis, the population to which an intervention is applied will affect the results. For example, the effect of screening for hypertension will depend not only on current levels of testing, but also on the current levels of blood pressure and other risk factors in the population. Even if all else is equal, an analysis done for people in Los Angeles will be different than an analysis done for people in San Diego. The default setting for the Archimedes Model is a random sample of the US population and is suitable for a wide range of "national" or "standard-of-care" analyses. However, it may need to be changed for applications that target other populations or sub-populations. To accommodate this, the Model includes methods for constructing simulated populations that match virtually any real population using either person-specific or aggregated data.

While the mathematics and programming of the Archimedes Model are very flexible and can theoretically be customized to fit virtually any setting or population, the ability to customize the Model will depend on the available data. Archimedes scientists work with clients to determine the need for customization and the best use of the available data.

ˆ top

What are the limitations of the Model?

For the conditions that are currently in the Model, the Model includes the most important biomarkers and interventions needed to analyze the great majority of problems. However there may be other conditions, biomarkers or interventions that are of interest for a particular problem that are not yet in the Model. These can be added to the Model upon request.

The level of physiological detail of the Model is determined by the level of detail of the questions it is designed to address. Typically this is the level of detail at which clinicians think and make their decisions, the level of detail found in clinical charts, and the level of detail at which clinical trials are designed and reported. For many areas of physiology the causes of clinical phenomena are understood at a deeper level of detail by basic scientists and specialists. In the Archimedes Model there factors are represented phenomenological, not causally. For example, the Model represents causally the relationships between heart rate, stroke volume, cardiac output, peripheral resistance and blood pressure. But the Model but does not represent the effects of cardiac filling on sarcomere length and force of myocardial contraction. In the situation in which one of Archimedes’ clients is interested in representing this type of phenomena, the Model can be extended to deeper levels of physiological detail.

The ability of the Model to address additional conditions, biomarkers, interventions or physiological processes depends on the available data. In general there must be data that relate any new variables to the existing variables in the Model to which they are causally related. Our validation protocols require that there be data from at least two sources: one for building the equations, and at least one other for performing an independent validation. » More

ˆ top

Why is the model needed? Why can't decision makers continue to make decisions as they do now?

It is extremely difficult for decision makers today to anticipate the consequences of the options they face, with any real quantitative or even qualitative accuracy. The reason is that virtually every important decision involves dozens of factors that are intertwined in extraordinarily complex ways. It is barely possible to understand each factor by itself. It is virtually impossible to understand the effects of their interactions.

For example, no one knows precisely how changing performance on some evidence-based guideline for cholesterol will affect, say, the rate of heart attacks or costs. We don’t know this for any particular guideline or performance measure, much less for how the different guidelines or measures compare. One should not assume that “evidence-based” or “prevention” means “cost saving”. Our Model has demonstrated that all evidence-based guidelines or performance measures are generally not equal in their effects on quality and cost. As a consequence, many if not most decisions are based on hope or the best guesses of experts. But the realities of human physiology, diseases, treatments, and behaviors are far too complicated to be solved in our heads.

To address this, ideally decision-makers would try out each option or program they are considering, such as a new guideline, new drug or incentive program, observe how they all do, and choose the one they like. This type of research is ideal, but impossible because:

• There are far too many options to be addressed
• Studies of this type are very expensive and take a long time to set up and conduct
• Patients and physicians often refuse to participate
• The answer will be different in every setting; an evaluation conducted in one setting will not necessarily apply in another setting, because of differences in the populations, practice patterns, costs, and so forth.
• The technology is moving far faster than our ability to conduct the experiments; the answers will be obsolete before the experiments are done.

The Archimedes Model was specifically designed to address decisions of this type. It uses mathematics and computers to do what we would ideally do in our heads if possible. The result is that decision makers will be able to understand the consequences of the choices they face far better than is possible now at a fraction of the cost and time of traditional methods with a high degree of realism and accuracy. » More

ˆ top

What types of projects can the Archimedes Model be used for?

The following are just a few examples of the types of projects the Model can be used for:

• Comparing the effectiveness of two or more interventions
• Identifying optimal populations and indications for tests and treatments
• Forecasting effects of demographic changes on future costs
• Understanding how changes in current care will affect health outcomes, utilization, and costs
• Planning and setting priorities for the allocation of organizational resources
• Evaluating the potential value of a drug or device in selected populations
• Conducting cost-effectiveness and value analysis
• Designing pay for performance programs
• Selecting and optimizing disease management and case management programs
• Evaluating drugs for inclusion in formularies
• Extending clinical trials

ˆ top

How can we tell if the Archimedes Model might have value for us?

You can get value from Archimedes if you work for a Health Plan and currently do or are considering doing any of the following:

• Use risk adjustment tools
• Have a patient-centered medical home initiative
• Use predictive modeling
• Design, evaluate, prioritize or promote evidence-based guidelines
• Set priorities across interventions
• Use disease management
• Use gap analysis
• Use performance measures
• Create incentives around performance
• Want to forecast costs
• Want to calculate cost/effectiveness
• Make formulary decisions

Or if you work for a pharmaceutical company or device manufacturer and do any of the following:

• Determine the potential value of a device or drug in different populations.
• Determine the optimal target populations for achieving specified objectives.
• Understand quantitatively the benefits and risks of different clinical uses of a drug
• Extend clinical trials to longer follow-up times and additional outcomes such as utilization and costs
• Determine the effects on market size, outcomes and costs of a new guideline.
• Compare the effectiveness of two or more interventions for achieving a specified clinical objective
• Identify optimal populations and indications for diagnostic tests
• Plan and set priorities for the allocation of pharmacy budgets
• Conduct cost-effectiveness analysis
• Understand the effects on outcomes and costs of changes in patient adherence and compliance

ˆ top

How do we contact you?

For the fastest response contact us through our online contact form.

Alternatively call +1 415.490.0400, or email us at info@archimedesmodel.com and we will direct your inquiry to the appropriate person.

ˆ top

When we contact you, what should we expect?

When you contact Archimedes you will be contacted via phone or email by a member of our commercial team. They will work with you directly or identify the right member of our technical staff to answer your questions. We look forward to the opportunity to work with you.

ˆ top