Monitoring clinical indicators

A new control chart for monitoring clinical indicator (CI) data based upon the beta-binomial posterior predictive (BBPP) distribution was compared with the more commonly used Bernoulli cumulative sum (Bernoulli CUSUM) chart. Run lengths were simulated for 3894 parameter combinations. For the case where the underlying proportion of cases with an event of interest had to be estimated, the BBPP chart was shown to have the desired smaller out-of-control ARL in 71.6% of the simulations. This effect was greatest in the parameter space having the: -mean proportion across all healthcare providers (HCPs), π, <0.1; -percentage change in the underlying proportion (required for Bernoulli CUSUM chart), Δ, 0.15 < Δ < 0.35; -number of admissions at risk of the event of interest at the ith HCP, Di, <30; -proportion of admissions having the event of interest at the ith HCP, θi, <0.1; across all values for the standard deviation between HCPs, σ.