Health Care Management Science最新文献

Healthcare delivery in the United States has been characterized as overly reactive and dependent on emergency department care for safety net coverage, with opportunity for improvement around discharge planning and high readmissions and emergency department bounce-back rates. Community paramedicine is a recent healthcare innovation that enables proactive visitation of patients at home, often shortly after emergency department and hospital discharge. We establish the first optimization-based framework to study efficiencies in the management and operation of a community paramedicine program. The collective innovations of our modeling include i) a novel hierarchical objective function with the goals of fairly increasing patient welfare, lowering hospital costs, and reducing readmissions and emergency department visits, ii) a new constraint set that ensures priority same-day visits for emergent patients, and iii) a further extension of our model to determine the minimum supplemental resources necessary to ensure feasibility in a single optimization formulation. Our medical-need based objective function prioritizes patients based on their clinical features and seeks to select and schedule patient visits and route healthcare providers to maximize overall patient welfare while favoring shorter tours. We use our methods to develop managerial insights via computational experiments on a variety of test instances based on real data from a hospital system in Upstate New York. We are able to identify optimal and nearly optimal tours that efficiently select, route, and schedule patients in reasonable timeframes. Our results lead to insights that can support managerial decisions about establishing (and improving existing) community paramedicine programs.

The utilization of healthcare services serves as a barometer for current and future health outcomes. Even in countries with modern healthcare IT infrastructure, however, fragmentation and interoperability issues hinder the (short-term) monitoring of utilization, forcing policymakers to rely on secondary data sources, such as surveys. This deficiency may be particularly problematic during public health crises, when ensuring proper and timely access to healthcare acquires special importance. We show that, in specific contexts, online pharmacies' digital footprint data may contain a strong signal of healthcare utilization. As such, online pharmacy data may enable utilization surveillance, i.e., the monitoring of short-term changes in utilization levels in the population. Our analysis takes advantage of the scenario created by the first wave of the Covid-19 pandemic in Mainland China, where the virus' spread lead to pervasive and deep reductions of healthcare service utilization. Relying on a large sample of online pharmacy transactions with full national coverage, we first detect variation that is strongly consistent with utilization reductions across geographies and over time. We then validate our claims by contrasting online pharmacy variation against credit-card transactions for medical services. Using machine learning methods, we show that incorporating online pharmacy data into the models significantly improves the accuracy of utilization surveillance estimates.

In this paper, we present the first Assignment-Dial-A-Ride problem motivated by a real-life problem faced by medico-social institutions in France. Every day, disabled people use ride-sharing services to go to an appropriate institution where they receive personal care. These institutions have to manage their staff to meet the demands of the people they receive. They have to solve three interconnected problems: the routing for the ride-sharing services; the assignment of disabled people to institutions; and the staff size in the institutions. We formulate a general Assignment-Dial-A-Ride problem to solve all three at the same time. We first present a matheuristic that iteratively generates routes using a large neighborhood search in which these routes are selected with a mixed integer linear program. After being validated on two special cases in the literature, the matheuristic is applied to real instances in three different areas in France. Several managerial results are derived. In particular, it is found that the amount of cost reduction induced by the people assignment is equivalent to the amount of cost reduction induced by the sharing of vehicles between institutions.

Patient Activation Measure (PAM) measures the activation level of patients with chronic conditions and correlates well with patient adherence behavior, health outcomes, and healthcare costs. PAM is increasingly used in practice to identify patients needing more support from the care team. We define PAM levels 1 and 2 as low PAM and investigate the performance of eight machine learning methods (Logistic Regression, Lasso Regression, Ridge Regression, Random Forest, Gradient Boosted Trees, Support Vector Machines, Decision Trees, Neural Networks) to classify patients. Primary data collected from adult patients (n=431) with Diabetes Mellitus (DM) or Hypertension (HT) attending Family Health Centers in Istanbul, Turkey, is used to test the methods. [Formula: see text] of patients in the dataset have a low PAM level. Classification performance with several feature sets was analyzed to understand the relative importance of different types of information and provide insights. The most important features are found as whether the patient performs self-monitoring, smoking and exercise habits, education, and socio-economic status. The best performance was achieved with the Logistic Regression algorithm, with Area Under the Curve (AUC)=0.72 with the best performing feature set. Alternative feature sets with similar prediction performance are also presented. The prediction performance was inferior with an automated feature selection method, supporting the importance of using domain knowledge in machine learning.

We study the problem of determining the locations and capacities of COVID-19 specimen collection centers to efficiently improve accessibility to polymerase chain reaction testing during surges in testing demand. We develop a two-echelon multi-period location and capacity allocation model that determines optimal number and locations of pop-up testing centers, capacities of the existing centers as well as assignments of demand regions to these centers, and centers to labs. The objective is to minimize the total number of delayed appointments and specimens subject to budget, capacity, and turnaround time constraints, which will in turn improve the accessibility to testing. We apply our model to a case study for locating COVID-19 testing centers in the Region of Waterloo, Canada using data from the Ontario Ministry of Health, public health databases, and medical literature. We also test the performance of the model under uncertain demand and analyze its outputs under various scenarios. Our analyses provide practical insights to the public health decision-makers on the timing of capacity expansions and the locations for the new pop-up centers. According to our results, the optimal strategy is to dynamically expand the existing specimen collection center capacities and prevent bottlenecks by locating pop-up facilities. The optimal locations of pop-ups are among the densely populated areas that are in proximity to the lab and a subset of those locations are selected with the changes in demand. A comparison with a static approach promises up to 39% cost savings under high demand using the developed multi-period model.

Primary care providers (PCPs) are considered the first-line defenders in preventive care. Patients seeking service from the same PCP constitute that physician's panel, which determines the overall supply and demand of the physician. The process of allocating patients to physician panels is called panel design. This study quantifies patient overflow and builds a mathematical model to evaluate the effect of two implementable panel assignments. In specialized panel assignment, patients are assigned based on their medical needs or visit frequency. In equal panel assignment, patients are distributed uniformly to maintain a similar composition across panels. We utilize majorization theory and numerical examples to evaluate the performance of the two designs. The results show that specialized panel assignment outperforms when (1) patient demands and physician capacity are relatively balanced or (2) patients who require frequent visits incur a higher shortage penalty. In a simulation model with actual patient arrival patterns, we also illustrate the robustness of the results and demonstrate the effect of switching panel policy when the patient pool changes over time.

Using data from cardiovascular surgery patients with long and highly variable post-surgical lengths of stay (LOS), we develop a modeling framework to reduce recovery unit congestion. We estimate the LOS and its probability distribution using machine learning models, schedule procedures on a rolling basis using a variety of optimization models, and estimate performance with simulation. The machine learning models achieved only modest LOS prediction accuracy, despite access to a very rich set of patient characteristics. Compared to the current paper-based system used in the hospital, most optimization models failed to reduce congestion without increasing wait times for surgery. A conservative stochastic optimization with sufficient sampling to capture the long tail of the LOS distribution outperformed the current manual process and other stochastic and robust optimization approaches. These results highlight the perils of using oversimplified distributional models of LOS for scheduling procedures and the importance of using optimization methods well-suited to dealing with long-tailed behavior.

Assigning inpatients to hospital beds impacts patient satisfaction and the workload of nurses and doctors. The assignment is subject to unknown inpatient arrivals, in particular for emergency patients. Hospitals, therefore, need to deal with uncertainty on actual bed requirements and potential shortage situations as bed capacities are limited. This paper develops a model and solution approach for solving the patient bed-assignment problem that is based on a machine learning (ML) approach to forecasting emergency patients. First, it contributes by improving the anticipation of emergency patients using ML approaches, incorporating weather data, time and dates, important local and regional events, as well as current and historical occupancy levels. Drawing on real-life data from a large case hospital, we were able to improve forecasting accuracy for emergency inpatient arrivals. We achieved up to 17% better root mean square error (RMSE) when using ML methods compared to a baseline approach relying on averages for historical arrival rates. We further show that the ML methods outperform time series forecasts. Second, we develop a new hyper-heuristic for solving real-life problem instances based on the pilot method and a specialized greedy look-ahead (GLA) heuristic. When applying the hyper-heuristic in test sets we were able to increase the objective function by up to 5.3% in comparison to the benchmark approach in [40]. A benchmark with a Genetic Algorithm shows also the superiority of the hyper-heuristic. Third, the combination of ML for emergency patient admission forecasting with advanced optimization through the hyper-heuristic allowed us to obtain an improvement of up to 3.3% on a real-life problem.