Rush order containment of critical drugs in ICUs.

The recent SARS CoV-02 pandemic has put enormous pressure on intensive care staff, making it imperative to relieve them of repetitive tasks with little added value such as drug replenishment. We propose a decision support system based on a hybrid policy to manage the inventory of critical drugs with low and intermittent demand at an Intensive Care Unit (ICU). Demand forecasting is at the heart of any inventory policy. We claim that in the ICU setting drug demand patterns must be therapy based. Heterogeneous data have been collected during an on site study, and information have been extracted to provide a faithful abstract representation of the ward as a system, as well as the potential evolutions of ICU patients clinical conditions. Together with medical guidelines, this provides the foundation of a therapy based demand forecasting tool. This study integrates schedule optimization and demand forecasting, and exploits simulation for evaluation purpose in the long run. At the beginning of every period, drug orders are optimally scheduled with respect to forecast demand. Then, scheduled orders are deployed day by day and confronted with the real demand in a simulated environment. Potential stock outs trigger rush orders to restore safety stocks. The comparison between the proposed policy and a standard policy mimicking current practice in an ICU ward shows that information on therapy patterns can be successfully incorporated into drug replenishment processes to reduce the number of rush orders, a primary goal in designing an efficient system.

Addressing consistency and demand uncertainty in the Home Care planning problem.

Optimizing Home Care Services is receiving a great attention in Operations Research. We address arrival time consistency, person-oriented consistency and demand uncertainty in Home Care, while jointly optimizing assignment, scheduling and routing decisions over a multiple-day time horizon. Consistent time schedules are very much appreciated by patients who, in this setting, are very sensitive to changes in their daily routines. Also person-oriented consistency positively impacts on service quality, guaranteeing that almost the same set of caregivers take care of a patient in the planning horizon. Demand uncertainty plays a pivotal role, too, since both the set of patients under treatment and their care plan can change over time. To the best of our knowledge, this is the first paper dealing with all these aspects in Home Care via a robust approach. We present a mathematical model to the problem, and a pattern-based algorithmic framework to solve it. The framework is derived from the model via decomposition, i.e. suitably fixing the scheduling decisions through the concept of pattern. We propose alternative policies to generate patterns, taking into account consistency and demand uncertainty; when embedding them in the general framework, alternative pattern based algorithms originate. The results of a rich computational experience show that introducing consistency and demand uncertainty in pattern generation policies is crucial to efficiently compute very good quality solutions, in terms of robustness and balancing of the caregiver workload. In addition, a comparison with a simpler model, where no kind of consistency is imposed, shows the importance of considering consistency in pursuing a valuable patient-centered perspective, with a positive effect also on the efficiency of the solution approach.

Lagrangean-Based Combinatorial Optimization for Large-Scale S3VMs.

The process of manually labeling instances, essential to a supervised classifier, can be expensive and time-consuming. In such a scenario the semisupervised approach, which makes the use of unlabeled patterns when building the decision function, is a more appealing choice. Indeed, large amounts of unlabeled samples often can be easily obtained. Many optimization techniques have been developed in the last decade to include the unlabeled patterns in the support vector machines formulation. Two broad strategies are followed: continuous and combinatorial. The approach presented in this paper belongs to the latter family and is especially suitable when a fair estimation of the proportion of positive and negative samples is available. Our method is very simple and requires a very light parameter selection. Several medium- and large-scale experiments on both artificial and real-world data sets have been carried out proving the effectiveness and the efficiency of the proposed algorithm.