← Clinical Tools

Barrier Dynamics

Put a barrier in front of CT pulmonary angiography and ordering drops. Then the system pushes back. This model follows an ED's PE workups week by week for two years after a barrier goes in, through five feedback loops that can blunt it, reverse it, or make it backfire. Switch loops off to see what each one does.

The causal loops

Read an arrow as "more of this pushes that up (+) or down (−)". A loop with an even number of minus signs reinforces itself (R); an odd number balances (B). Hover or tap a loop under "Run it" to trace it.

Scroll the diagram sideways to see all of it.

Lever you controlStock: builds up and drains slowly Everything else|| Delay

Run it

Blunt friction Justification forms, radiology approval calls: the same hassle whatever the patient's risk.

Decision support A Wells/D-dimer hard stop that steers low- and moderate-risk patients to a D-dimer first.

Scanner capacity Relative to today. Adding a scanner removes the wait, itself a barrier.

Two years, week by week

The hidden stocks driving it

Data table

What the model assumes

100 possible-PE workups a week: 60% low risk (PE prevalence 1%), 30% moderate (8%), 10% high (35%). With no barrier, low-risk patients go straight to CT 30% of the time, moderate 65%, high 92%; the rest get a D-dimer (sensitivity 95%; false-positive rate 35%, 50% and 60% by tier) or no test. 250 other CTs a week share the scanner.

Blunt friction deters direct CT orders by up to 60% of the slider in low-risk, 45% moderate and 25% high-risk patients; deterred orders split between a D-dimer and no test.
Decision support re-routes up to 70% of low-risk and 50% of moderate-risk direct CTs to a D-dimer first, and leaves high-risk patients alone.
Queue (B1). Wait = 0.5 h + 0.9 h × ρ/(1−ρ) for scanner utilization ρ, perceived with a 4-week lag. Waits over an hour deter direct CT orders, mostly in low-risk patients.
Alert fatigue (R1). Habituation grows with alerts fired and fades with a 1-year time constant; it scales the alert's effect down.
Workarounds (R2, B3). Logistic spread by word of mouth, faster under more friction, forgotten over about 3 years.
Fear rebound (B2). 30% of missed PEs become visible (a bounce-back, an M&M case) after 6 weeks. Harm above the usual level raises fear, which fades over about 5 months and moves patients from a D-dimer or no test to CT.
D-dimer creep. Once a D-dimer is the expected first step, it spreads to low-risk patients who used to get no test. Unlike the alert itself, the habit doesn't fatigue.

An illustrative model, not calibrated to any ED's data. The shapes (better before worse, drift, overshoot) are the lesson; the exact numbers are not. Every equation is in this page's source.

Sources and further reading

Sterman JD. Business Dynamics: Systems Thinking and Modeling for a Complex World. McGraw-Hill, 2000.
Senge PM. The Fifth Discipline. Doubleday, 1990. The "fixes that fail" and "shifting the burden" archetypes.
Ancker JS, Edwards A, Nosal S, et al. Effects of workload, work complexity, and repeated alerts on alert fatigue in a clinical decision support system. BMC Med Inform Decis Mak 2017;17:36.
Shain M, Roemer MI. Hospital costs relate to the supply of beds. Mod Hosp 1959;92:71–3. (Roemer's law: supply induces its own use.)
Raja AS, Ip IK, Prevedello LM, et al. Effect of computerized clinical decision support on the use and yield of CT pulmonary angiography in the emergency department. Radiology 2012;262:468–74.

The individual decision behind these curves is what Barriers to Ordering measures, case by case.