Statistical process control as quantitative method to monitor and improve medical quality
AuthorDriesen, Kevin E.
MetadataShow full item record
PublisherThe University of Arizona.
RightsCopyright © is held by the author. Digital access to this material is made possible by the University Libraries, University of Arizona. Further transmission, reproduction or presentation (such as public display or performance) of protected items is prohibited except with permission of the author.
AbstractStatistical Process Control (SPC) methods, developed in industrial settings, are increasingly being generalized to medical service environments. Of special interest is the control chart, a graphic and statistical procedure used to monitor and control variation. This dissertation evaluates the validity of the control chart model to improve medical quality. The research design combines descriptive and causal comparative (ex-post facto) methods to address the principal research question, How is the control chart model related to medical quality? Hospital data were used for patients diagnosed with Community Acquired Pneumonia (CAP). During the initial research phase, five medical quality "events" assumed to affect CAP medical quality indicators were pre-specified by hospital staff. The impact of each event was then evaluated using control charts constructed for CAP quality indicators. Descriptive analysis was undertaken to determine whether data violated the statistical assumptions underlying the control chart model. Then, variable and attribute control charts were constructed to determine whether special cause signals occurred in association with the pre-specified events. Alternative methods were used to calibrate charts to different conditions. Sensitivity was computed as the proportion of event-sensitive signals. The descriptive analysis of CAP indicators uncovered "messy," and somewhat complex, data structure. The CAP indicators were marginally stable showing trend, seasonal cycles, skew, sampling variation and autocorrelation. Study results need to be interpreted with the knowledge that few events were evaluated, and that the effect sizes associated with events were small. The charts applied to the CAP indicators showed limited sensitivity; for three chart-types (i.e. XmR, Xbar, and P-charts), there were more false alarms than event-associated signals. Conforming to expectation, larger sample size increased chart sensitivity. The application of Jaehn Decision Rules led to increases in both sensitivity and false alarm. Increasing subgroup frequency from month, to week samples, increased chart sensitivity, but also increased data instability and autocorrelation. Contrary to expectation, the application of hybrid charting techniques (EWMA and CUSUM) did not increase chart sensitivity. Study findings support the conclusion that control charts provide valuable insight into medical variation. However, design issues, data character, and causal logic provide conditions to the interpretation of control charts.
Degree ProgramGraduate College