The AI ethics of digital COVID-19 diagnosis and their legal, medical, technological, and operational managerial implications

The COVID-19 pandemic has given rise to a broad range of research from fields alongside and beyond the core concerns of infectiology, epidemiology, and immunology. One significant subset of this work centers on machine learning-based approaches to supporting medical decision-making around COVID-19 diagnosis. To date, various challenges, including IT issues, have meant that, notwithstanding this strand of research on digital diagnosis of COVID-19, the actual use of these methods in medical facilities remains incipient at best, despite their potential to relieve pressure on scarce medical resources, prevent instances of infection, and help manage the difficulties and unpredictabilities surrounding the emergence of new mutations. The reasons behind this research-application gap are manifold and may imply an interdisciplinary dimension. We argue that the discipline of AI ethics can provide a framework for interdisciplinary discussion and create a roadmap for the application of digital COVID-19 diagnosis, taking into account all disciplinary stakeholders involved. This article proposes such an ethical framework for the practical use of digital COVID-19 diagnosis, considering legal, medical, operational managerial, and technological aspects of the issue in accordance with our diverse research backgrounds and noting the potential of the approach we set out here to guide future research.

Simulation of the mortality after different ex ante (secondary) and ex post (tertiary) triage methods in people with disabilities and pre-existing diseases

The significant increase in patients during the COVID-19 pandemic presented the healthcare system with a variety of challenges. The intensive care unit is one of the areas particularly affected in this context. Only through extensive infection control measures as well as an enormous logistical effort was it possible to treat all patients requiring intensive care in Germany even during peak phases of the pandemic, and to prevent triage even in regions with high patient pressure and simultaneously low capacities. Regarding pandemic preparedness, the German Parliament passed a law on triage that explicitly prohibits ex post (tertiary) triage. In ex post triage, patients who are already being treated are included in the triage decision and treatment capacities are allocated according to the individual likelihood of success. Legal, ethical, and social considerations for triage in pandemics can be found in the literature, but there is no quantitative assessment with respect to different patient groups in the intensive care unit. This study addressed this gap and applied a simulation-based evaluation of ex ante (primary) and ex post triage policies in consideration of survival probabilities, impairments, and pre-existing conditions. The results show that application of ex post triage based on survival probabilities leads to a reduction in mortality in the intensive care unit for all patient groups. In the scenario close to a real-world situation, considering different impaired and prediseased patient groups, a reduction in mortality of approximately 15% was already achieved by applying ex post triage on the first day. This mortality-reducing effect of ex post triage is further enhanced as the number of patients requiring intensive care increases.

Covid-19 triage in the emergency department 2.0: how analytics and AI transform a human-made algorithm for the prediction of clinical pathways

The Covid-19 pandemic has pushed many hospitals to their capacity limits. Therefore, a triage of patients has been discussed controversially primarily through an ethical perspective. The term triage contains many aspects such as urgency of treatment, severity of the disease and pre-existing conditions, access to critical care, or the classification of patients regarding subsequent clinical pathways starting from the emergency department. The determination of the pathways is important not only for patient care, but also for capacity planning in hospitals. We examine the performance of a human-made triage algorithm for clinical pathways which is considered a guideline for emergency departments in Germany based on a large multicenter dataset with over 4,000 European Covid-19 patients from the LEOSS registry. We find an accuracy of 28 percent and approximately 15 percent sensitivity for the ward class. The results serve as a benchmark for our extensions including an additional category of palliative care as a new label, analytics, AI, XAI, and interactive techniques. We find significant potential of analytics and AI in Covid-19 triage regarding accuracy, sensitivity, and other performance metrics whilst our interactive human-AI algorithm shows superior performance with approximately 73 percent accuracy and up to 76 percent sensitivity. The results are independent of the data preparation process regarding the imputation of missing values or grouping of comorbidities. In addition, we find that the consideration of an additional label palliative care does not improve the results.

[Simulation of mortality after different ex-ante and ex-post-triage methods in people with disabilities and comorbidities]

The significant increase in patients during the COVID-19 pandemic presented the healthcare system with a variety of challenges. The intensive care unit is one of the areas particularly affected in this context. Only through extensive infection control measures as well as an enormous logistical effort was it possible to treat all patients requiring intensive care in Germany even during peak phases of the pandemic, and to prevent triage even in regions with high patient pressure and simultaneously low capacities. Regarding pandemic preparedness, the German Parliament passed a law on triage that explicitly prohibits ex post (tertiary) triage. In ex post triage, patients who are already being treated are included in the triage decision and treatment capacities are allocated according to the individual likelihood of success. Legal, ethical, and social considerations for triage in pandemics can be found in the literature, but there is no quantitative assessment with respect to different patient groups in the intensive care unit. This study addressed this gap and applied a simulation-based evaluation of ex ante (primary) and ex post triage policies in consideration of survival probabilities, impairments, and pre-existing conditions. The results show that application of ex post triage based on survival probabilities leads to a reduction in mortality in the intensive care unit for all patient groups. In the scenario close to a real-world situation, considering different impaired and prediseased patient groups, a reduction in mortality of approximately 15% was already achieved by applying ex post triage on the first day. This mortality-reducing effect of ex post triage is further enhanced as the number of patients requiring intensive care increases.

[German "Triage Act"-Regulation with fatal consequences]

With the coming into force of § 5c of the Infection Protection Act (IfSG), the so-called Triage Act, on 14 December 2022, a protracted discussion has come to a provisional conclusion, the result of which physicians and social associations but also lawyers and ethicists are equally dissatisfied. The explicit exclusion of the discontinuation of treatment that has already begun in favor of new patients with better chances of success (so-called tertiary or ex-post triage) prevents allocation decisions with the aim of enabling as many patients as possible to beneficially participate in medical care under crisis conditions. The result of the new regulation is de facto a first come first served allocation, which is associated with the highest mortality even among individuals with limitations or disabilities and was rejected by a large margin as unfair in a population survey. Mandating allocation decisions based on the likelihood of success but which are not permitted to be consistently implemented and prohibiting, for example the use of age and frailty as prioritization criteria, although both factors most strongly determine the short-term probability of survival according to evident data, shows the contradictory and dogmatic nature of the regulation. The only remaining possibility is the consistent termination of treatment that is no longer indicated or desired by the patient, regardless of the current resource situation; however, if a different decision is made in a crisis situation than in a situation without a lack of resources, this practice would not be justified and would be punishable. Accordingly, the highest efforts must be set on legally compliant documentation, especially in the stage of decompensated crisis care in a region. The goal of enabling as many patients as possible to beneficially participate in medical care under crisis conditions is in any case thwarted by the new German Triage Act.

Airborne pollen grain detection from partially labelled data utilising semi-supervised learning

Airborne pollen monitoring has been conducted for more than a century now, as knowledge of the quantity and periodicity of airborne pollen has diverse use cases, like reconstructing historic climates and tracking current climate change, forensic applications, and up to warning those affected by pollen-induced respiratory allergies. Hence, related work on automation of pollen classification already exists. In contrast, detection of pollen is still conducted manually, and it is the gold standard for accuracy. So, here we used a new-generation, automated, near-real-time pollen monitoring sampler, the BAA500, and we used data consisting of both raw and synthesised microscope images. Apart from the automatically generated, commercially-labelled data of all pollen taxa, we additionally used a manually created partial classification test set of bounding boxes and pollen taxa, so as to more accurately evaluate the real-life performance. For the pollen detection, we employed two-stage deep neural network object detectors. We explored a semi-supervised training scheme to remedy the partial labelling. Using a teacher-student approach, the model can add pseudo-labels to complete the labelling during training. To evaluate the performance of our deep learning algorithms and to compare them to the commercial algorithm of the BAA500, we created a manual test set, in which an expert aerobiologist corrected automatically annotated labels. For the novel manual test set, both the supervised and semi-supervised approaches clearly outperform the commercial algorithm with an F1 score of up to 76.9 % compared to 61.3 %. On an automatically created and partially labelled test dataset, we obtain a maximum mAP of 92.7 %. Additional experiments on raw microscope images show comparable performance for the best models, which potentially justifies reducing the complexity of the image generation process. Our results bring automatic pollen monitoring a step forward, as they close the gap in pollen detection performance between manual and automated procedure.

COVIDAL: A machine learning classifier for digital COVID-19 diagnosis in German hospitals

For the fight against the COVID-19 pandemic, it is particularly important to map the course of infection, in terms of patients who have currently tested SARS-CoV-2 positive, as accurately as possible. In hospitals, this is even more important because resources have become scarce. Although polymerase chain reaction (PCR) and point of care (POC) antigen testing capacities have been massively expanded, they are often very time-consuming and cost-intensive and, in some cases, lack appropriate performance. To meet these challenges, we propose the COVIDAL classifier for AI-based diagnosis of symptomatic COVID-19 subjects in hospitals based on laboratory parameters. We evaluate the algorithm's performance by unique multicenter data with approx. 4,000 patients and an extraordinary high ratio of SARS-CoV-2 positive patients. We analyze the influence of data preparation, flexibility in optimization targets as well as the selection of the test set on the COVIDAL outcome. The algorithm is compared with standard AI, PCR, POC antigen testing and manual classifications of seven physicians by a decision theoretic scoring model including performance metrics, turnaround times and cost. Thereby, we define health care settings in which a certain classifier for COVID-19 diagnosis is to be applied. We find sensitivities, specificities and accuracies of the COVIDAL algorithm of up to 90 percent. Our scoring model suggests using PCR testing for a focus on performance metrics. For turnaround times, POC antigen testing should be used. If balancing performance, turnaround times and cost is of interest, as, for example, in the emergency department, COVIDAL is superior based on the scoring model.

A Simulation-Based Cost-Effectiveness Analysis of Severe Acute Respiratory Syndrome Coronavirus 2 Infection Prevention Strategies for Visitors of Healthcare Institutions

OBJECTIVES

The aim is to quantitatively evaluate different infection prevention strategies in the context of hospital visitor management during pandemics and to provide a decision support system for strategic and operational decisions based on this evaluation.

METHODS

A simulation-based cost-effectiveness analysis is applied to the data of a university hospital in Southern Germany and published COVID-19 research. The performance of different hospital visitor management strategies is evaluated by several decision-theoretic methods with varying objective functions.

RESULTS

Appropriate visitor restrictions and infection prevention measures can reduce additional infections and costs caused by visitors of healthcare institutions by >90%. The risk of transmission of severe acute respiratory syndrome coronavirus 2 by visitors of terminal care (ie, palliative care) patients can be reduced almost to 0 if appropriate infection prevention measures are implemented. Antigen tests do not seem to be beneficial from both a cost and an effectiveness perspective.

CONCLUSIONS

Hospital visitor management is crucial and effectively prevents infections while maintaining cost-effectiveness. For terminal care patients, visitor restrictions can be omitted if appropriate infection prevention measures are taken. Antigen testing plays a subordinate role, except in the case of a pure focus on additional infections caused by visitors of healthcare institutions. We provide decision support to authorities and hospital visitor managers to identify appropriate visitor restriction and infection prevention strategies for specific local conditions, incidence rates, and objectives.

Homogeneity and Best Practice Analyses in Hospital Performance Management: An Analytical Framework

Performance modeling of hospitals using data envelopment analysis (DEA) has received steadily increasing attention in the literature. As part of the traditional DEA framework, hospitals are generally assumed to be functionally similar and therefore homogenous. Accordingly, any identified inefficiency is supposedly due to the inefficient use of inputs to produce outputs. However, the disparities in DEA efficiency scores may be a result of the inherent heterogeneity of hospitals. Additionally, traditional DEA models lack predictive capabilities despite having been frequently used as a benchmarking tool in the literature. To address these concerns, this study proposes a framework for analyzing hospital performance by combining two complementary modeling approaches. Specifically, we employ a self-organizing map artificial neural network (SOM-ANN) to conduct a cluster analysis and a multilayer perceptron ANN (MLP-ANN) to perform a heterogeneity analysis and a best practice analysis. The applicability of the integrated framework is empirically shown by an implementation to a large dataset containing more than 1,100 hospitals in Germany. The framework enables a decision-maker not only to predict the best performance but also to explore whether the differences in relative efficiency scores are ascribable to the heterogeneity of hospitals.

Balancing control and autonomy in master surgery scheduling: Benefits of ICU quotas for recovery units

When scheduling surgeries in the operating theater, not only the resources within the operating theater have to be considered but also those in downstream units, e.g., the intensive care unit and regular bed wards of each medical specialty. We present an extension to the master surgery schedule, where the capacity for surgeries on ICU patients is controlled by introducing downstream-dependent block types – one for both ICU and ward patients and one where surgeries on ICU patients must not be performed. The goal is to provide better control over post-surgery patient flows through the hospital while preserving each medical specialty’s autonomy over its operational surgery scheduling. We propose a mixed-integer program to determine the allocation of the new block types within either a given or a new master surgery schedule to minimize the maximum workload in downstream units. Using a simulation model supported by seven years of data from the University Hospital Augsburg, we show that the maximum workload in the intensive care unit can be reduced by up to 11.22% with our approach while maintaining the existing master surgery schedule. We also show that our approach can achieve up to 79.85% of the maximum workload reduction in the intensive care unit that would result from a fully centralized approach. We analyze various hospital setting instances to show the generalizability of our results. Furthermore, we provide insights and data analysis from the implementation of a quota system at the University Hospital Augsburg.

Towards automatic airborne pollen monitoring: From commercial devices to operational by mitigating class-imbalance in a deep learning approach

Allergic diseases have been the epidemic of the century among chronic diseases. Particularly for pollen allergies, and in the context of climate change, as airborne pollen seasons have been shifting earlier and abundances have been becoming higher, pollen monitoring plays an important role in generating high-risk allergy alerts. However, this task requires labour-intensive and time-consuming manual classification via optical microscopy. Even new-generation, automatic, monitoring devices require manual pollen labelling to increase accuracy and to advance to genuinely operational devices. Deep Learning-based models have the potential to increase the accuracy of automated pollen monitoring systems. In the current research, transfer learning-based convolutional neural networks were employed to classify pollen grains from microscopic images. Given a high imbalance in the dataset, we incorporated class weighted loss, focal loss and weight vector normalisation for class balancing as well as data augmentation and weight penalties for regularisation. Airborne pollen has been routinely recorded by a Bio-Aerosol Analyzer (BAA500, Hund GmbH) located in Augsburg, Germany. Here we utilised a database referring to manually classified airborne pollen images of the whole pollen diversity throughout an annual pollen season. By using the cropped pollen images collected by this device, we achieved an unweighted average F1 score of 93.8% across 15 classes and an unweighted average F1 score of 75.9% across 31 classes. The majority of taxa (9 of 15), being also the most abundant and allergenic, showed a recall of at least 95%, reaching up to a remarkable 100% in pollen from Taxus and Urticaceae. The recent introduction of novel pollen monitoring devices worldwide has pointed to the necessity for real-time, automatic measurements of airborne pollen and fungal spores. Thus, we may improve everyday clinical practice and achieve the most efficient prophylaxis of allergic patients.

Managing admission and discharge processes in intensive care units

The intensive care unit (ICU) is one of the most crucial and expensive resources in a health care system. While high fixed costs usually lead to tight capacities, shortages have severe consequences. Thus, various challenging issues exist: When should an ICU admit or reject arriving patients in general? Should ICUs always be able to admit critical patients or rather focus on high utilization? On an operational level, both admission control of arriving patients and demand-driven early discharge of currently residing patients are decision variables and should be considered simultaneously. This paper discusses the trade-off between medical and monetary goals when managing intensive care units by modeling the problem as a Markov decision process. Intuitive, myopic rule mimicking decision-making in practice is applied as a benchmark. In a numerical study based on real-world data, we demonstrate that the medical results deteriorate dramatically when focusing on monetary goals only, and vice versa. Using our model, we illustrate the trade-off along an efficiency frontier that accounts for all combinations of medical and monetary goals. Coming from a solution that optimizes monetary costs, a significant reduction of expected mortality can be achieved at little additional monetary cost.

Simulation-based evaluation of operating room management policies

Background

Since operating rooms are a major bottleneck resource and an important revenue driver in hospitals, it is important to use these resources efficiently. Studies estimate that between 60 and 70% of hospital admissions are due to surgeries. Furthermore, staffing cannot be changed daily to respond to changing demands. The resulting high complexity in operating room management necessitates perpetual process evaluation and the use of decision support tools. In this study, we evaluate several management policies and their consequences for the operating theater of the University Hospital Augsburg.

Methods

Based on a data set with 12,946 surgeries, we evaluate management policies such as parallel induction of anesthesia with varying levels of staff support, the use of a dedicated emergency room, extending operating room hours reserved as buffer capacity, and different elective patient sequencing policies. We develop a detailed simulation model that serves to capture the process flow in the entire operating theater: scheduling surgeries from a dynamically managed waiting list, handling various types of schedule disruptions, rescheduling and prioritizing postponed and deferred surgeries, and reallocating operating room capacity. The system performance is measured by indicators such as patient waiting time, idle time, staff overtime, and the number of deferred surgeries.

Results

We identify significant trade-offs between expected waiting times for different patient urgency categories when operating rooms are opened longer to serve as end-of-day buffers. The introduction of parallel induction of anesthesia allows for additional patients to be scheduled and operated on during regular hours. However, this comes with a higher number of expected deferrals, which can be partially mitigated by employing additional anesthesia teams. Changes to the sequencing of elective patients according to their expected surgery duration cause expectable outcomes for a multitude of performance indicators.

Conclusions

Our simulation-based approach allows operating theater managers to test a multitude of potential changes in operating room management without disrupting the ongoing workflow. The close collaboration between management and researchers in the design of the simulation framework and the data analysis has yielded immediate benefits for the scheduling policies and data collection efforts at our practice partner.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12913-021-06234-5.

Determinants of the utilization of allergy management measures among hay fever sufferers: a theory-based cross-sectional study

Background

The quality of life of chronically ill individuals, such as hay fever sufferers, is significantly dependent on their health behavior. This survey aimed to explain the health-related behavior of allergic individuals using the protection motivation theory (PMT) and the transtheoretical model (TTM).

Methods

The influencing variables stated by PMT were operationalized based on data from semistructured pilot interviews and a pretest with 12 individuals from the target population. The final questionnaire inquired perceived seriousness and severity of hay fever, response efficacy, response costs, self-efficacy, and the use of various hay fever management measures in relation to the TTM stages. Multivariate logistic regression was performed to investigate the relationships between the PMT constructs and the examined health behavior.

Results

A total of 569 allergic individuals completed the online questionnaire. Only 33.26% of allergic individuals were in the maintenance stage for treatment under medical supervision, and almost 60% preferred hay fever self-management. A total of 67.56% had a well-established habit of taking anti-allergic medication, but only 25.31% had undergone specific immunotherapy. The likelihood of seeking medical supervision was positively influenced by perceived severity (OR = 1.35, 95% CI: 1.02–1.81), perceived seriousness (OR = 2.12, 95% CI: 1.56–2.89), and self-efficacy (OR = 4.52, 95% CI: 3.11–6.65). The perceived severity of symptoms predicted the practice of hay fever self-management (OR = 1.60, 95% CI: 1.21–2.11), as well as anti-allergic medication intake (OR = 1.65, 95% CI: 1.16–2.35). The latter measure was also positively influenced by self-efficacy (OR = 1.52, 95% CI: 1.01–2.28) and hay fever self-management (OR = 4.76, 95% CI: 2.67–7.49). Undergoing specific immunotherapy was significantly predicted only by medical supervision (OR = 9.80, 95% CI: 8.16–13.80). Allergen avoidance was a strategy used by allergic individuals who preferred hay fever self-management (OR = 2.56, 95% CI: 1.87–3.52) and experienced notable symptom severity (OR = 2.12, 95% CI: 1.60–2.81).

Conclusion

Educational interventions that increase the awareness of health risks associated with inadequate hay fever management and measures to increase self-efficacy might be beneficial for the promotion of appropriate hay fever management among allergic individuals.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12889-020-09959-w.

[Bed capacity management in times of the COVID-19 pandemic : A simulation-based prognosis of normal and intensive care beds using the descriptive data of the University Hospital Augsburg]

BACKGROUND

Following the regional outbreak in China, severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has spread all over the world, presenting the healthcare systems with huge challenges worldwide. In Germany the coronavirus diseases 2019 (COVID-19) pandemic has resulted in a slowly growing demand for health care with a sudden occurrence of regional hotspots. This leads to an unpredictable situation for many hospitals, leaving the question of how many bed resources are needed to cope with the surge of COVID-19 patients.

OBJECTIVE

In this study we created a simulation-based prognostic tool that provides the management of the University Hospital of Augsburg and the civil protection services with the necessary information to plan and guide the disaster response to the ongoing pandemic. Especially the number of beds needed on isolation wards and intensive care units (ICU) are the biggest concerns. The focus should lie not only on the confirmed cases as the patients with suspected COVID-19 are in need of the same resources.

MATERIAL AND METHODS

For the input we used the latest information provided by governmental institutions about the spreading of the disease, with a special focus on the growth rate of the cumulative number of cases. Due to the dynamics of the current situation, these data can be highly variable. To minimize the influence of this variance, we designed distribution functions for the parameters growth rate, length of stay in hospital and the proportion of infected people who need to be hospitalized in our area of responsibility. Using this input, we started a Monte Carlo simulation with 10,000 runs to predict the range of the number of hospital beds needed within the coming days and compared it with the available resources.

RESULTS

Since 2 February 2020 a total of 306 patients were treated with suspected or confirmed COVID-19 at this university hospital. Of these 84 needed treatment on the ICU. With the help of several simulation-based forecasts, the required ICU and normal bed capacity at Augsburg University Hospital and the Augsburg ambulance service in the period from 28 March 2020 to 8 June 2020 could be predicted with a high degree of reliability. Simulations that were run before the impact of the restrictions in daily life showed that we would have run out of ICU bed capacity within approximately 1 month.

CONCLUSION

Our simulation-based prognosis of the health care capacities needed helps the management of the hospital and the civil protection service to make reasonable decisions and adapt the disaster response to the realistic needs. At the same time the forecasts create the possibility to plan the strategic response days and weeks in advance. The tool presented in this study is, as far as we know, the only one accounting not only for confirmed COVID-19 cases but also for suspected COVID-19 patients. Additionally, the few input parameters used are easy to access and can be easily adapted to other healthcare systems.

Automated Classification of Airborne Pollen using Neural Networks

Pollen allergies are considered as a global epidemic nowadays, as they influence more than a quarter of the worldwide population, with this percentage expected to rapidly increase because of ongoing climate change. To date, alerts on high-risk allergenic pollen exposure have been provided only via forecasting models and conventional monitoring methods that are laborious. The aim of this study is to develop and evaluate our own pollen classification model based on deep neural networks. Airborne allergenic pollen have been monitored in Augsburg, Bavaria, Germany, since 2015, using a novel automatic Bio-Aerosol Analyzer (BAA 500, Hund GmbH). The automatic classification system is compared and evaluated against our own, newly developed algorithm. Our model achieves an unweighted average precision of 83.0 % and an unweighted average recall of 77.1 % across 15 classes of pollen taxa. Automatic, real-time information on concentrations of airborne allergenic pollen will significantly contribute to the implementation of timely, personalized management of allergies in the future. It is already clear that new methods and sophisticated models have to be developed so as to successfully switch to novel operational pollen monitoring techniques serving the above need.

Analyzing economies of scale and scope in hospitals by use of case mix planning

This study analyzes the effect of economies of scale and scope on the optimal case mix of a hospital or hospital system. With respect to the ideal volume and patient composition, the goal is to evaluate (i) the impact of changes in the efficiency of resource use with increasing scale, and (ii) to determine the potential effects of spreading fixed costs over a greater number of patients. The problem is formulated as a non-linear mixed integer program. It turns out that this non-linear program is too difficult to be solved with standard software. As an alternative, an iterative procedure using piecewise linear approximations to derive lower and upper bounds is proposed and shown to converge to the optimum. The procedure is applied using a public database on German hospital costs and performance statistics. Results indicate that changes in the efficiency of resource use with increasing scale have a considerable impact if similar services can be consolidated, e.g., among different departments. However, if the scope for decision-making regarding the case mix of a hospital is limited, such changes may be negligible.

Operations research in intensive care unit management: a literature review

The intensive care unit (ICU) is a crucial and expensive resource largely affected by uncertainty and variability. Insufficient ICU capacity causes many negative effects not only in the ICU itself, but also in other connected departments along the patient care path. Operations research/management science (OR/MS) plays an important role in identifying ways to manage ICU capacities efficiently and in ensuring desired levels of service quality. As a consequence, numerous papers on the topic exist. The goal of this paper is to provide the first structured literature review on how OR/MS may support ICU management. We start our review by illustrating the important role the ICU plays in the hospital patient flow. Then we focus on the ICU management problem (single department management problem) and classify the literature from multiple angles, including decision horizons, problem settings, and modeling and solution techniques. Based on the classification logic, research gaps and opportunities are highlighted, e.g., combining bed capacity planning and personnel scheduling, modeling uncertainty with non-homogenous distribution functions, and exploring more efficient solution approaches.

Case mix planning in hospitals: a review and future agenda

The case mix planning problem deals with choosing the ideal composition and volume of patients in a hospital. With many countries having recently changed to systems where hospitals are reimbursed for patients according to their diagnosis, case mix planning has become an important tool in strategic and tactical hospital planning. Selecting patients in such a payment system can have a significant impact on a hospital's revenue. The contribution of this article is to provide the first literature review focusing on the case mix planning problem. We describe the problem, distinguish it from similar planning problems, and evaluate the existing literature with regard to problem structure and managerial impact. Further, we identify gaps in the literature. We hope to foster research in the field of case mix planning, which only lately has received growing attention despite its fundamental economic impact on hospitals.