Figure 2.

Simulated power analysis for impact of RVD on days alive and at home at 30 days. Assuming for 1% of the patients DAH₃₀=0, for the remainder DAH₃₀ follows a negative binomial distribution with parameters chosen such that the median DAH₃₀ is 24/25 in one group and 27 in the other group and the shape of the distribution is similar to that seen in the validation cohorts. The simulated DAH₃₀ was then compared between groups using negative binomial regression (repeated) in 1000 samples. The figures show the resulting estimated power for incidences of postoperative RVD from 15% to 50%^#, and for a clinical effect size of 2 (A) or 3 (B) days difference in DAH₃₀*. ^#In our previous work the incidence of postoperative RVD was 50% in thoracic surgical patients but may be significantly less in, for example, an orthopaedic population. *In Chou et al’s study preoperative RV dysfunction prolonged hospital length of stay by over 50%, but this cohort was a very high-risk vascular surgical population.⁷ DAH₃₀, days alive and at home at 30 days postoperatively; RVD, right ventricular dysfunction.