Optimizing outpatient appointment system using machine learning algorithms and scheduling rules: A prescriptive analytics framework
| Autor: | A. Ravi Ravindran, Sharan Srinivas |
|---|---|
| Rok vydání: | 2018 |
| Předmět: |
021103 operations research
Computer science business.industry 0211 other engineering and technologies General Engineering Scheduling (production processes) Overtime 02 engineering and technology Benchmarking Machine learning computer.software_genre Computer Science Applications Patient satisfaction Artificial Intelligence 0202 electrical engineering electronic engineering information engineering Outpatient clinic 020201 artificial intelligence & image processing Artificial intelligence business computer Resource utilization |
| Zdroj: | Expert Systems with Applications. 102:245-261 |
| ISSN: | 0957-4174 |
| DOI: | 10.1016/j.eswa.2018.02.022 |
| Popis: | Developed prescriptive analytics framework to schedule patients in real-time.Used machine learning to classify patients based on their no-show risk.Proposed scheduling rules by simultaneously considering multiple design decisions.Results provide insights on patient sequencing and overbooking decisions.Helps clinicians to move towards customized, patient-centered care. In the US, the demand for outpatient services is expected to increase, while the supply of physicians to provide the care is projected to decrease. Besides, inefficiencies in the appointment system (AS) and patient no-shows (patients who do not arrive for scheduled appointments) reduce provider productivity, timely access to care, and cost the U.S. healthcare system more than $150 billion a year. To handle increasing demand and compensate for patient no-shows, outpatient clinics tend to overbook appointments. The current scheduling practice at most clinics and majority of the scheduling rules proposed in the literature assume all patients are equally likely to miss an appointment. Further, most scheduling rules in the literature do not make use of the available data, such as electronic health records, when scheduling patients. This paper proposes a prescriptive analytics framework to improve the performance of an AS with respect to patient satisfaction (measured using average patient waiting time and number of patients unable to get an appointment for the day under consideration) and resource utilization (measured using average resource idle time, overflow time and overtime). In the proposed framework, patient-related data from various sources are used to develop predictive models that identify the risk of a patient no-show. Different scheduling rules, that leverage the patient-specific no-show risk is then proposed. A case study, with real data from a Family Medicine Clinic in Pennsylvania, is used to show the feasibility of the proposed framework. The effectiveness of the proposed scheduling rules is evaluated by benchmarking it with three rules adapted from the literature. The results indicate that the proposed scheduling rules consistently outperform the benchmark rules for all the clinic settings tested. Further, the proposed framework is generic and can be adopted by any outpatient clinic characterized by occurrences of no-shows and appointment-based customer arrivals. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
Full Text from ScienceDirect