Figure 1 | True detection rate (power) of Variable life adjusted (VLAD) plots for different observed and expected outcome rates. An expected population mortality rate is given. Mortality is specified here for clarity, but the outcome could be any binary event. A mortality rate is set for individual clinicians which may be higher or lower than that of the population. Simulations are performed at the population and individual event rate. The variation resulting from chance is demonstrated by the plot. Red and black lines represent simulations at the individual and population event rate respectively.
Control limits (blue lines) can be set on the chart based on the variation seen in the normal population.
This app demonstrates how many procedures require to be performed by an individual with a mortality rate higher than that of the population. For example, using the default population and individual mortality rate of 5% and 10% respectively, 75 procedures are required to be performed for around 50% of individuals to trigger at the 2 standard deviation control limit.
Use the “change in practice” (orange lines) and “cluster of deaths” (green lines) options to explore the true detection rate (power) of these charts with different control limits set.
Figure 2 | Receiver operator curves for Observed-expected CuSUM / Variable life adjusted (VLAD) plots. The cut-off with greatest discrimination is given. That is, the rate which maximises the sensitivity and specificity.