Figure 3

Tornado Diagrams. Three separate graphs correspond to the three alternative scenarios for COPD severity in our base case analyses. (a., Mild COPD; b., Moderate COPD; c., Severe COPD). Results of one way sensitivity analyses are illustrated as tornado diagrams with the horizontal bars representing the incremental change in QALYs for Full Code compared to DNI advance directive. The widest bars represent the variables that the model is most sensitive to because changes in their parameter estimates result in large changes in QALY. Variables that cross the 0 mark indicate a change in the recommended AD from Full Code to DNI. For the mild COPD scenario (Figure 2a), there is no change in the recommended directive when parameter estimates for the model variables are changed. For the moderate COPD scenario (Figure 2b), DNI becomes the recommended directive when the probability of having a complication from ETT increases. For the severe COPD scenario (Figure 2c), DNI becomes the recommended directive when the probability of having a complication from ETT increases; and when the probability of failing NIMV decreases. ETT = Invasive mechanical ventilation via endotracheal intubation; NIMV = Noninvasive mechanical ventilation; ECF = Extended Care Facility; CMO = Comfort Measures Only; DNI = Do Not Intubate; ICU = Intensive Care Unit.