An autocorrelation-corrected nonparametric control chart technique for health care quality applications.

Recently, Alemi proposed a nonparametric control chart technique (Tukey's control chart) for quality management applications when few data points are available and when data do not conform to the assumptions of traditional control chart techniques. Borckardt et al then published an empirical evaluation of the technique and concluded that the presence of autocorrelation in control-chart data negatively impacted the technique's ability to help managers make accurate decisions about the presence of special-cause variation in their data. Thus, there is still a need for control chart techniques that appropriately handle short data streams that do not necessarily conform to the assumptions of traditional control chart techniques but are not negatively impacted by autocorrelation in the data. In this article, the authors empirically evaluate a modified version of the technique presented by Alemi that is designed to account for autocorrelation. Empirical analyses indicate that the modified technique demonstrates superior false-positive performance with very little degradation of power compared with the original technique proposed by Alemi.

A new approach to tracking the harmfulness of medical errors in health care systems.

A variety of approaches are available to monitor and quantify the frequency of medical errors and the harm associated with them. The University HealthSystem Consortium Patient Safety Net provides a categorical, rank-ordered approach for designating harm associated with medical errors but does not lend itself to routine monitoring of change over time. A brief questionnaire was developed and given to hospital administrators to collect ratings about the harmfulness of events occurring in each of the 10 University HealthSystem Consortium harm-score categories. Results revealed a high degree of agreement among raters for harmfulness ratings for each category. A cubic model fit the data best and provided weights for each of the harm-score categories. This preliminary study proposes a methodology for quantifying harm scores.