Sylvester Rohan Devasahay1,2, Sylvia Karpagam3, Nang Laik Ma4. 1. Data Science, School of information Systems, Singapore Management University, Singapore. 2. Information Technology specialized in Analytics, Bangalore, India. 3. Rajiv Gandhi University of Health Sciences, Bangalore, India. 4. School of Business, Senior Lecturer, SIM University, Singapore.
Abstract
BACKGROUND: There is growing attention over the last few years about non-attendance in hospitals and its clinical and economic consequences. There have been several studies documenting the various aspects of non-attendance in hospitals. Project Predicting Appoint Misses (PAM) was started with the intention of being able to predict the type of patients that would not come for appointments after making bookings. METHODS: Historic hospital appointment data merged with "distance from hospital" variable was used to run Logistic Regression, Support Vector Machine and Recursive Partitioning to decide the contributing variables to missed appointments. RESULTS: Variables that are "class", "time", "demographics" related have an effect on the target variable, however, prediction models may not perform effectively due to very subtle influence on the target variable. Previously assumed major contributors like "age", "distance" did not have a major effect on the target variable. CONCLUSIONS: With the given data it will be very difficult to make any moderate/strong prediction of the Appointment misses. That being said with the help of the cut off we are able to capture all of the "appointment misses" in addition to also capturing the actualized appointments.
BACKGROUND: There is growing attention over the last few years about non-attendance in hospitals and its clinical and economic consequences. There have been several studies documenting the various aspects of non-attendance in hospitals. Project Predicting Appoint Misses (PAM) was started with the intention of being able to predict the type of patients that would not come for appointments after making bookings. METHODS: Historic hospital appointment data merged with "distance from hospital" variable was used to run Logistic Regression, Support Vector Machine and Recursive Partitioning to decide the contributing variables to missed appointments. RESULTS: Variables that are "class", "time", "demographics" related have an effect on the target variable, however, prediction models may not perform effectively due to very subtle influence on the target variable. Previously assumed major contributors like "age", "distance" did not have a major effect on the target variable. CONCLUSIONS: With the given data it will be very difficult to make any moderate/strong prediction of the Appointment misses. That being said with the help of the cut off we are able to capture all of the "appointment misses" in addition to also capturing the actualized appointments.
Authors: Richard D Neal; Mahvash Hussain-Gambles; Victoria L Allgar; Debbie A Lawlor; Owen Dempsey Journal: BMC Fam Pract Date: 2005-11-07 Impact factor: 2.497