An a priori resource-based classification methodology for specialty/secondary ambulatory patients

This study proposes an approach to predict the health care resource requirements of speciality ambulatory patients at a micro (clinic) level. It employs cluster analysis and learning tools to develop a generalized methodology based on a health provider's patient discharge data to spawn a patient classification system which, on the basis of information available prior to a patient receiving health care service, predicts the clinic resources that a patient may use on the appointment date.;To evaluate its robustness, the methodology has been field-tested at seven secondary/tertiary low vision ambulatory clinics in North America and Sub-Saharan Africa. A minimum of 25% of all available data was collected from each site. After collection, the data were analyzed (by clinic) using the methodology by first employing cluster analysis to develop iso-resource groups, then applying a variety of techniques (decision trees, non-parametric discriminant analysis, nearest neighbour, and neural networks) on data that are available at appointment time. Additionally, the study attempted to determine the generalizable iso-resource variables or groupings which are systemic across clinics/centres in the speciality ambulatory setting of low vision and, therefore, which could, along the lines of length of stay (in acute and long-term health care settings), form the basis for a standard set of measures for resource planning and scheduling in speciality ambulatory low vision settings.;Estimates of apparent and true errors were used in gauging the predictive performance of each learning technique at the sites. Chance criterion served as the benchmark in this evaluation. No learning technique emerged as the universally superior one (and hence the method of choice), however, they typically outperformed the benchmark's predictive ability (frequently doubling or tripling it). This suggests that their usage would make significant contributions to the decision making process.;This research broadens previous work done in this area into a variety of low vision clinical settings to determine (1) the robustness of the proposed methodology, (2) potential additional complexity issues that the proposed methodology must attend, and (3) the generalizable and systemic iso-resource variables across low vision settings that may form the basis for a standard set of measures for ambulatory resource planning and scheduling in speciality low vision settings. It also discovered that an a priori classification can indeed be successfully achieved in this speciality setting. (Abstract shortened by UMI.).