Authors (including presenting author) :
Guo ZH (1), Zee CY(1), Chee PY(2)
Affiliation :
(1) The Jockey Club School of Public Health and Primary Care, The Chinese University of Hong Kong,(2)Accident & Emergency Department St John Hospital
Introduction :
Human and facility resources are often limited and short in rural hospitals, and for a rural area, aging of population is always the case and thereby more visit to the hospital. In this regard, efficient resource planning plays an important role in providing health care service timely and appropriately. A reliable patient flow prediction model can help managers not only to understand current level and patterns of health care demand of the catchment area, but to predict the need in future.
Objectives :
(1).To construct reliable models for predicting daily patient flow in a rural hospital.
(2).To detect the contributing variables that associate with the hospital's patient flow.
(3).Evaluate the accuracy of the models by making comparisons between different methods.
Methodology :
A prediction modeling study design would be performed. The input variable would be temporal(weekday, holiday, months, seasons, calendar years), aging( proportion of elder people in the catchment area) and disease( discharge diagnosis). The output variable would be the daily patient flow and it would be further categorized by age groups, diseases, residence, triage level and transfer status.
Artificial neural network(ANN) methods would be used to construct prediction model, and other machine learning techniques and multiple linear regression methods are also used to develop the model and to evaluate the accuracy of ANN models. An linear model would be used to interpret the results.
Result & Outcome :
Expected results: The ANN model would capture the temporal, disease and aging patterns of the patent flow of the hospital and accurately predict the demand in future after feeding certain input values. The performance of ANN model would be superior over other methods. The weight and of each input variables can be detected through linear models and ANN methods for interpreting the result and identify major contributing variables. Such a prediction model could be a reference for managers to better allocate limited resources in advance to meet increasingly health care demand in a rural area.
