The following presents the simulation process and results using the thirteen Paths describes on the previous pages. It demonstrates how the Wellness Model results in fewer deaths from errors and omissions and lower total healthcare costs compared to the Non-Wellness Model driving healthcare today

Simulations Assumptions and Procedures

The Wellness Model includes Paths 4, 7, 11, and 12 — Decision Support PoC, Computer Physician Order Entry, Long-term Care, and Wellness, respectively. The Non-Wellness Model, which is currently used in healthcare today, lacks these four paths.
Since the Wellness Model has four extra pathways, logic dictates that while it might reduce errors and omissions, it might also increase overall healthcare costs. To test this hypothesis, we ran a simulation comparing the Wellness Model (Wellness) to the Non-Wellness Model (Non-Wellness) on each of the three starting points (Start A-C). The purpose of this simulation was to determine if Wellness results in better outcomes and lowers costs than today’s Non-Wellness healthcare system.
The simulator was instructed to generate a random number at each of the six decision points to represent the percentage of patients receiving care through each of the two processes at those points in the path. We assumed that Wellness reduces the number of errors and omissions through use of the Decision Support (Path 4) and Computer Physician Order Entry (Path 7) systems, which help avoid adverse events through alerts and recommendations to clinical staff enabling them to take remedial action before the problems cause adverse events. This assumption was expressed by putting a “bias” in process P14-Error and Omission, which increased the error and omission events by 5% and by 10% for Non-Wellness by moving a small percentage of patients from process P14 to P7.
We defined the following initial simulation conditions under which both models operated:
  • 40 errors per day per 300-bed hospital set the baseline for adverse events at 13%
  • R1 through R6 all used the same random number generator to drive simulated traffic, pulling a unique random number for each decision point. The same random number generator values were used to split work at P14 to P7, and P7 to P9 and P8
  • The process cost numbers stated earlier were used, and the Start A, B and C input values were all set to 100.

The simulation was then run for eleven different death rates, ranging from 10% to 0%, in order to investigate what happens to costs as outcomes improve and to discern the winner between Wellness and Non-Wellness given a specific death rate. For example, if the simulation run gave a Wellness death rate value of 5% and the Non-Wellness death rate value for the same run were 8%, Wellness was declared the winner for that run.
Thus, there were 800 scenarios run in order to investigate the:
  • 2 models (Wellness and Non-Wellness),
  • 3 starting points (P1-P3),
  • 3 error and omission biases (0%, 5%, 10%),
  • 11 death rates (10% to 0%), and
  • 763 concurrent side-by-side comparisons to discern a winner, i.e., the model exhibiting the lowest death to discharge ratio, given three separate initial conditions: no-bias, 5% bias and 10% bias.

Upon completion, the simulation results yielded the frequency of activities performed for each of the eighteen processes (P1-18). By assigning a dollar amount to each activity, we were able to calculate the total cost of care for each scenario.


The measure we used to assess clinical outcomes was based on two elements: Discharge (positive outcome) and Death due to errors and omissions (negative outcome). Each simulation run produced a ratio of Deaths to Discharges. After running multiple trials:
  • There was an 84% correlation between PoC review (P1-Decision Support processing) and Diagnosis (P7) indicating that Wellness Path 4 was heavily leveraged in the Wellness scenarios
  • A 77% correlation between CPOE Review (at R2) and Death associated with errors and omissions (P14 on Path 10) indicated that Wellness Path 7 was heavily leveraged in the Wellness scenarios.

With no bias applied, the simulation demonstrated a 2:1 advantage in the number of times Non-Wellness demonstrated a lower death to discharge ratio, with a tendency towards ties. This means that the death to discharge ratio for Non-Wellness was less than the death to discharge ratio for Wellness for one-hundred concurrent simulation runs. With a bias of 5%, meaning, we took a five-percent benefit for Wellness; at the P14-Error and Omission process,

Wellness and Non-Wellness performed the same. The result was a toss up, both Wellness and Non-Wellness demonstrated near equal death to discharge ratio values over one-hundred observations of concurrent simulation runs. When a bias of 10% was used, the Wellness simulation outperformed the Non-Wellness Model by a 2:1 margin. This means Wellness demonstrated a lower death to discharge ratio over one-hundred concurrent simulation runs.

This improvement continues with increased bias values.
When cost was applied to the zero bias scenarios — giving Wellness and Non-Wellness the same death to discharge ratio — Non-Wellness cost was less, as expected. This is because three Wellness paths — Path 4 (Decision Support), 7 (Medication Management) and 12 (Wellness Preventive Care) — were removed from the simulation.

The situation changed, however, with a bias of 5% applied at P7 affecting P14. Wellness and Non-Wellness appeared to have the same cost for Death Rates of 3% or less due to errors and omissions. When the high and low values are removed for each, the dollar value for Wellness and Non-Wellness are just a few percentage points of each other.

As the scenarios approach Death Rates of 0% due to errors and omissions, only Wellness exhibits continuous improvement tendencies. That is, the cost to maintain Wellness decreases, while the cost to sustain the Non-Wellness increases for 5% bias and 10% bias.

The cost difference between each with a linear trend line intersecting at 0% Death Rates is substantial. Wellness at a Death Rate of 0% is $1.8 million, Non-Wellness at 5% bias is $2.6 million, and Non-Wellness at 10% bias is $3 million. A great deal more work and expense is required to move Non-Wellness toward a Death Rate of 0%. Thus, approaching 0% Death Rates may actually lower costs in the Wellness Model due to continually improving efficiencies and HIT tool effectiveness. The Non-Wellness Model, however, shows the opposite trend and reflects a similar situation that is found in manufacturing, i.e., attempting to reach zero defects is cost-prohibitive; the only reasonable thing to do is set a tolerance and attempt through continuous improvement to achieve that tolerance at the least cost.


The simulation indicates that implementing the Wellness Model, unlike today’s Non-Wellness Model, results in continuously improving care quality resulting in fewer deaths from errors and omissions and lower costs!

We encourage people to question and dispute the models’ processes, paths and decision points; to challenge our assumptions and conclusions; and to offer alternatives — so they may continue to evolve.

Next: From the Authors: Steve Beller, PhD