Ergin Erdem

Rescheduling of elective patients upon the arrival of emergency patients

In this study, a mixed integer linear programming (MILP) model is developed for rescheduling elective patients upon the arrival of emergency patients by considering two types of clinical units, namely operating rooms and post-anesthesia care units (PACUs). The model considers the overtime cost of the operating rooms and/or the PACUs, the cost of postponing or preponing elective surgeries, and the cost of turning down the emergency patients. The results indicate that a mainstream commercial solver can efficiently find an optimal solution in a particular scenario with light elective surgery load, but becomes very inefficient in searching optimal solutions in all other scenarios. As such, a genetic algorithm is developed to efficiently obtain the approximately optimal solutions in those scenarios that are difficult for the commercial solver. In the genetic algorithm, a novel chromosome structure is proposed and applied to represent the feasible solutions to the MILP model. It is shown that for the scenarios with heavy load of elective surgeries, the genetic algorithm can find approximate optimal solutions significantly faster than the commercial solver. In practice, the two solution methodologies should be used jointly to provide hospitals a solid tool for making sound and timely decisions in admitting emergency patients and rescheduling elective patients.

An Integrated Wind Power Forecasting Methodology: Interval Estimation Of Wind Speed, Operation Probability Of Wind Turbine, And Conditional Expected Wind Power Output Of A Wind Farm

The article presents a novel quantitative methodology for wind farm management. The methodology starts by forecasting the time series mean and volatility of wind speed. The forecasting of wind speed mean and its volatility is built on an autoregressive moving average model with a generalized autoregressive conditional heteroscedasticity process, namely an ARMA-GARCH model. With the prediction of wind speed mean and its volatility, the article establishes the interval estimation of wind speed which makes the prediction of wind speed more accurate and reliable. To facilitate the quantitative management of wind farm, the operation probability (OP) of wind turbine is formulated according to the interval estimation of wind speed. Based on the characteristics power curve of wind turbine, the article develops the conditional expected wind power output equation (CEWPOE). The interval estimation of wind speed, the OP of wind turbine, and the CEWPOE thus comprise an integrated methodology for the quantitative management of wind farm operations.

Investigation of RFID tag readability for pharmaceutical products at item level

Background: The applications of radio frequency identification (RFID) technology carry a tremendous potential for pharmaceutical industry. There is a pressing need to analyze the performance of RFID tags attached to various pharmaceutical dosage forms. Method: The readability of RFID-tagged pharmaceutical products is, for the first time, systematically investigated by experiments. Factors considered include dosage forms, ion concentration in solution, angle of rotation, and distance between the RFID tag and the interrogator. Results: Compared with empty container, the filling of any representative dosage forms causes deteriorated readability for the tag attached to container. Analysis of variance reveals that the effects of dosage forms, angle of rotation, and their interaction are statistically significant. In addition, an increase in distance (equivalent to higher RF attenuation level) and higher ion concentration in solution beyond a certain level have detrimental effect on tag readability. Conclusion: The analysis shows that the RFID tag readability is strongly dependent on the factors that are experimented with. The level of the factors for optimum RFID system performance should be adjusted based on the particular application.

Simulation analysis on patient visit efficiency of a typical VA primary care clinic with complex characteristics

Abstract In this paper, we develop a simulation model to study the performance of clinic operations based on the settings of a typical VA primary care clinic with complex characteristics. The clinic serves three different types of patients, namely, regular appointment patients, walk-in patients, and nurse-only patients, and they each have different patient flow routes. The model captures the patient flow characteristics of the studied clinic, and it is validated by comparing the simulation results with the real key performance measures obtained from in the clinic. The system performance is mainly measured by two metrics: one is how the clinic makes efficient use of its resources, and the other is how long the patients need to wait for services. A scenario analysis is conducted which adopts the experiment design method for multiple factors to study the effect of six parameters on the system performance. The six parameters are walk-in arrival rate, no-show rate, post-triage rate, new patient rate, number of double booking, and nurse-only appointment rate. Based on the results, one major recommendation for the clinic is to reduce or eliminate the use of double booking because it causes the dramatic increase of patient waiting time. Also, based on the projection of the high new patient rate in the near future, it is recommended that the clinic manage the walk-in patient load by adjusting the existing admission policy. In addition, for the clinic with already high provider utilization, interventions for reducing patient no-show rate alone may further deteriorate the congestion of patient visit in the clinic.

Genetic algorithm for instrument placement in smart grid

A genetic algorithm (GA) based approach for reliability placement of phasor measurement units (PMUs) in smart grid is proposed. The algorithm combines two conflicting objectives which are maximization of the reliability of observability and minimization of the number of PMU placements for ensuring full system observability. The multi-objective problem is formulated as a nonlinear optimization problem and genetic algorithm approach is employed for solving the large scale bus systems. The optimization model is solved for IEEE 14, and 30 standard bus systems. The effectiveness of the proposed approach has been demonstrated by comparing results with exact algorithms for smaller problem sizes. The system reliability of observability is improved compared to traditional optimal PMU placement. The proposed approach achieves significant cost savings compared to available reliability based models in literature.

Communication Enhancement and Best Practices for Co-Managing Dual Care Rural Veteran Patients by VA and Non-VA Providers: A Survey Study

Many rural Veteran patients receive healthcare services from both Veterans Affairs (VA) and non-VA providers. Effective management of dual care Veteran patients to ensure the best clinical outcomes is a VA mission. The previous VA studies indicate that coordination between VA and non-VA providers has been lacking for dual care management of Veteran patients. In this study, we propose that VA proactively shares information with non-VA providers to enhance the communication process and identify the best practices to be carried out by both VA and non-VA providers for better coordination. Structured questionnaires are designed and distributed to VA and non-VA providers to obtain their evaluations on the proposed VA proactive information sharing approaches and the best practice items for dual care management. The non-VA provider respondents largely support the proposed proactive sharing items by VA, with the lowest average score being 3.96 out of a 5.0 scale on one item. In terms of the best practice items on co-managing dual care patients, three out of five items are overall rated higher than 4.0 from both sides. A pair-wise comparison between VA and non-VA perspectives further shows that the difference in average ratings of a proposed item could be significant. For such best practice items, the implementations from both sides may not be most effective.

Large-scale assessment of missed opportunity risks in a complex hospital setting

In this research, we apply a large-scale logistic regression analysis to assess the patient missed opportunity risks at a complex VA (US Department of Veterans Affairs) hospital in three categories, namely, no-show alone, no-show combined with late patient cancellation and no-show combined with late patient and clinic cancellations. The analysis includes unique explanatory variables related to VA patients for predicting missed opportunity risks. Furthermore, we develop two aggregated weather indices by combining many weather measures and include them as explanatory variables. The results indicate that most of the explanatory variables considered are significant factors for predicting the missed opportunity risks. Patients with afternoon appointment, higher percentage service connected, and insurance, married patients, shorter lead time and appointments with longer appointment length are consistently related to lower risks of missed opportunity. Furthermore, the VA patient-related factors and the two proposed weather indices are useful predictors for the risks of no-show and patient cancellation. More importantly, this research presents an effective procedure for VA hospitals and clinics to analyze the missed opportunity risks within the complex VA information technology system, and help them to develop proper interventions to mitigate the adverse effects caused by the missed opportunities.

Performance analysis and improvement of a typical telephone response system of VA hospitals

Purpose – Telephone response system is the frontline of hospital operations. The purpose of this paper is to analyze a representative telephone response system of Veterans Affairs (VA) hospitals, address the existing inefficiency issues such as long call waiting time, and improve system resilience to changes. Design/methodology/approach – Resource sharing schemes are proposed to improve the system performance in answering calls related to appointment booking and medication renewal. Discrete event simulation is adopted to model the current system and the resource sharing schemes. Findings – The resource sharing schemes dramatically improve system performance reflected by the decrease of call waiting time and queue, as well as the extreme high utilization of agents in a key unit. Compared with the less desired alternative of hiring additional employees to mitigate the performance issues, the resource sharing schemes perform at par or even better. Sharing more resource during the peak hours can further balan...

Profitability Analysis of Residential Wind Turbines with Battery Energy Storage

Residential wind turbines are often accompanied by an energy storage system for the off-the-grid users, instead of the on-the-grid users, to reduce the risk of black-out. In this paper, we argue that residential wind turbines with battery energy storage could actually be beneficial to the on-the-grid users as well in terms of monetary gain from differential pricing for buying electricity from the grid and the ability to sell electricity back to the grid. We develop a mixed-integer linear programming model to maximize the profit of a residential wind turbine system while meeting the daily household electricity consumption. A case study is designed to investigate the effects of differential pricing schemes and sell-back schemes on the economic output of a 2-kW wind turbine with lithium battery storage. Overall, based on the current settings in California, a residential wind turbine with battery storage carries more economical benefits than the wind turbine alone.

Comparison of Two ARMA-GARCH Approaches for Forecasting the Mean and Volatility of Wind Speed

In this study, we develop two ARMA-GARCH models for predicting the mean and volatility of wind speed. The first model employs the standalone ARMA-GARCH model for modeling the mean wind speed and the variance simultaneously. For the second model, in the first step, the current wind vector is decomposed into lateral and longitudinal components by using the prevailing wind direction. The mean and variance of the two components are then modeled using two separate ARMA-GARCH processes. Thereafter, the two components are combined back to form the resultant single wind vector. A large wind dataset is employed for model building and prediction so that the two approaches can be compared. It shows that the standalone ARMA-GARCH model is more accurate for predicting the wind speed, whereas the component ARMA-GARCH model performs better for predicting the wind variance.