Enhancing Large Language Models for Clinical Decision Support by Incorporating Clinical Practice Guidelines (preprint)

Background Large Language Models (LLMs), enhanced with Clinical Practice Guidelines (CPGs), can significantly improve Clinical Decision Support (CDS). However, methods for incorporating CPGs into LLMs are not well studied. Methods We develop three distinct methods for incorporating CPGs into LLMs: Binary Decision Tree (BDT), Program-Aided Graph Construction (PAGC), and Chain-of-Thought-Few-Shot Prompting (CoT-FSP). To evaluate the effectiveness of the proposed methods, we create a set of synthetic patient descriptions and conduct both automatic and human evaluation of the responses generated by four LLMs: GPT-4, GPT-3.5 Turbo, LLaMA, and PaLM 2. Zero-Shot Prompting (ZSP) was used as the baseline method. We focus on CDS for COVID-19 outpatient treatment as the case study. Results All four LLMs exhibit improved performance when enhanced with CPGs compared to the baseline ZSP. BDT outperformed both CoT-FSP and PAGC in automatic evaluation. All of the proposed methods demonstrated high performance in human evaluation. Conclusion LLMs enhanced with CPGs demonstrate superior performance, as compared to plain LLMs with ZSP, in providing accurate recommendations for COVID-19 outpatient treatment, which also highlights the potential for broader applications beyond the case study.

Relapse of Monoclonal Gammopathy of Renal Significance after mRNA Covid-19 vaccination: a case report (preprint)

Introduction This case report represents, to our knowledge, the first suspected case of Light Chain Deposition Disease (LCDD) relapse associated with mRNA COVID-19 vaccination. It must be made clear that timing is the only link between the vaccination and the relapse of LCDD, and since millions of individuals received these vaccines, an event like this may be due to chance. At the same time, this case report may be an entry point into further explorations of the pathogenesis of LCDD, given the mechanism of action of the vaccine and the pathophysiology of the disease, as it is currently understood. Case presentation The 75-year-old female patient of Greek ethnicity was admitted into our clinic for the investigation of worsening renal function that was detected on routine lab examinations two weeks after she received the second dose of the moderna COVID vaccine (mRNA-1273). Rapidly progressive glomerulonephritis and anemia were the most notable findings upon admission. She had a history of Light Chain Deposition Disease (LCDD) which had remained stable under management for four years. Serum protein electrophoresis was performed and showed monoclonal kappa zones, while bone marrow biopsy revealed 5% plasma cell infiltration. All the above, along with other investigations, established the diagnosis of LCDD recurrence. She was started on chemotherapy, which improved her immunological profile but not her renal function. The patient has remained on hemodialysis since. Conclusions The association between mRNA vaccinations and LCDD relapse may be ground for investigations into the pathophysiology of MGRS, given the patient’s previous long-term remission. In this context, we want to stress once again the importance of following standardized routine mandates on COVID vaccination, especially in immunocompromised patients. This case report is not intended to directly inform changes in clinical practice.

Systemic Risk in Supply Chains: A Vector Autoregressive Measurement Approach Based on the Example of Automotive and Semiconductor Supply Chains (preprint)

Supply chain failures and supply shortages have always been a matter of high risk. Especially when considering the scope and velocity of modern supply chains, small disturbances can cause immense damage. However, a framework for quantifying supply chain systemic risk is still missing.To address this, we use the principles of the Diebold and Yilmaz connectedness approach, which is based on assessing the decomposition of the forecast error variance of a vector autoregressive (VAR) model, and adjust it to supply chains. By doing so, we seek to establish a systemic risk measurement of individual supply chains on different aggregation levels. In detail, we examine the automotive and semiconductor supply chains. Looking at specific firms, we identify vulnerable nodes and hubs of these supply chains and, thus, can measure the risk exposure originated by a certain region or supply chain level.Our results show that for both supply chains, risk spillovers were at their highest levels during the COVID-19 pandemic, and firms facing U.S. trade restrictions experienced particularly strong effects during our sample period. In general, our approach provides convincing results, since companies identified as particularly risky are in line with specific company news that indicate risky spillovers during the study period.

Realizing Hyperagility under Time Pressure: Ad Hoc Supply Chains in the COVID-19 Pandemic (preprint)

Early in the COVID-19 pandemic in 2020, the medical product industry faced an unprecedented demand shock for personal protective equipment (PPE) such as face masks, face shields, disinfectants, and gowns. In response, many companies from different industries built ad hoc supply chains for these potentially life-saving products. When building ad hoc supply chains, companies showed supply chain hyperagility that allowed them to respond at an extreme speed. In this paper we investigate how companies realize supply chain hyperagility in the context of ad hoc supply chains built in the COVID-19 pandemic. Following a grounded theory-building approach, we conducted and analyzed 52 interviews with representatives of 37 German companies to develop a theoretical model that explains the relationships of four constructs to supply chain hyperagility: Internal and external (dynamic) capabilities have a direct effect on supply chain hyperagility and are moderated by an entrepreneurial and a temporary orientation. Our theoretical model contributes to the literature stream of supply chain agility by developing a new agility construct that addresses the extreme condition of time pressure. Supply chain managers will be able to develop their capabilities for future external shocks such as natural disasters and economic crises in line with our theoretical model.