Yi Zhu

I am a fifth-year Ph.D. candidate in the Department of Information and Decision Sciences at Carlson School of Management, University of Minnesota. My research interests lie in health IT, design science, network science and economics, and innovation and entrepreneurship. My recent research centers on employing data-driven empirical methodologies to unravel new IT-induced adverse outcomes in healthcare (e.g., medical device recall, human sedentary behavior) and developing theory-driven algorithmic approaches for their predictions. My work has been published in leading academic journals, including Journal of the American Medical Association (JAMA), Journal of General Internal Medicine, and Medical Care Research and Review. I also have presented my work at top information system conferences, including International Conference on Information Systems (ICIS), Conference on Information Systems and Technology (CIST), Workshop on Information Technologies and Systems (WITS), and Hawaii International Conference on System Sciences (HICSS). I am the recipient of NSF Innovation Corps (I-Corps) program grant for technology commercialization and customer discovery, INFORMS ISS Design Science Award, Carlson Ph.D. Student Teaching Award, University of Minnesota Early Innovation Fund, and University of Minnesota BOLD IDEAS grant.

Journal Publications 

1. Everhart A, Sen S, Stern A, Zhu Y, Karaca-Mandic P (2023) Association Between Regulatory Submission Characteristics and Recalls of Medical Devices Receiving 510(k) Clearance. Journal of the American Medical Association (JAMA) 329(2):144-156.
2. Zhu Y, Carroll C, Vu K, Sen S, Georgiou A, Karaca-Mandic P (2022) COVID-19 Hospitalization Trends in Rural Versus Urban Areas in the United States. Medical Care Research and Review 80(2):236-244.
3. Karaca-Mandic P, Sen S, Georgiou A, Zhu Y, Basu A (2020) Association of COVID-19-Related Hospital Use and Overall COVID-19 Mortality in the USA. Journal of General Internal Medicine. 

Work in Progress (full manuscripts are available upon request)

1. Zhu Y, Sen S, Everhart A, Karaca-Mandic P. Predicting Medical Device Recalls: An Evidence-Informed Deep Learning Approach Leveraging Regulatory Submission Characteristics. In final preparation for submission to Information Systems Research.
Abstract: Medical devices in the US entering the market through the FDA’s predominant 510(k) clearance pathway have garnered significant safety concerns. The medical industry worries that devices approved under this pathway, primarily based on new devices’ equivalency to previously approved devices (predicate devices), may be more likely to encounter recalls. These recalls could impose substantial patient harm and financial strains on the healthcare system. In response, this work proposes a data-driven approach for predicting 510(k) device recalls, aiming to alleviate such safety concerns. Following the design science paradigm and informed by the empirical findings from prior research, our approach leverages the predictive power of the characteristics in the network formed by predicate device citing relationships (predicate network). It incorporates natural language processing and deep learning techniques to tackle three design challenges, including creating the predicate database to construct predicate network, learning the predicate network structure, and capturing the temporal patterns of predicate network features. Rigorous tests based on 45,236 devices approved from 2003 to 2020 show that our approach significantly outperforms standard prediction models and the performance varies by device subgroups and key attributes applied. The improved medical device recall prediction performance and the analysis insights into the performance variations provide actionable implications for preemptive reactions to potential recalls and improving the current 510(k) pathway requirements to reduce device safety issues.

2. Zhu Y, Bi X, Adomavicious G, Curley S. Forecasting Goal-Oriented Activities: A Deep Learning Approach Leveraging Cross-Activity Relationships. Under review at INFORMS Journal on Computing. (received 2023 INFORMS Annual Meeting and Workshop on Data Science Student Scholarship)
Abstract: Many individuals struggle to maintain consistent efforts toward their goals, such as maintaining a desired physique or mastering a second language. Although online platforms have introduced various goal-oriented programs to help enhance motivation and adherence, these programs often suffer from high dropout rates, diminishing their effectiveness. To intervene in early dropouts and recommend programs with suitable goals, we propose an approach to accurately forecast individuals’ program goal completion and their actual performance toward completing the goal. Following the design science paradigm, the proposed approach leverages a combination of convolutional neural networks and gated recurrent neural networks, surpassing traditional methods by capturing the latent interactions among different activity types (e.g., running, cycling) and measures (e.g., distance, duration) recorded in sequential format and temporal dependencies between daily activities. Its applicability extends beyond goal-oriented programs, as it can be utilized in forecasting contexts involving multiple sequences and benefiting from cross-sequence interrelationships. We test the effectiveness of the proposed approach using simulated and real-world datasets, demonstrating its advantages over canonical machine learning models with both datasets. This performance enhancement positions the approach as a promising tool for improving adherence in goal-oriented programs and goal attainment. It also provides valuable insights for goal-oriented program intervention and personalization.

3. Zhu Y, Mani A. Investigating Churn in Online Wellness Programs: Evidence from a US Online Social Network. Under review at Management Science. 
Abstract: Online wellness and fitness platforms are increasingly utilizing social support in their wellness programs to motivate healthy activities and improve user engagement. However, many wellness programs suffer from high churn rates that discount their expected efficacy. This introduces negative social influence that may lead to a churn contagion and amplify churn speed and scale. Hence, a need arises to understand why users churn wellness programs and how social contagion contributes to the churns. Leveraging the exercise challenge setting, exercise field data, and a large social network on a renowned U.S. online fitness platform, we investigate the effect of peers' tendency to churn exercise challenges on ego's churn likelihood. To achieve the research goal, we employ an instrumental variable framework, using the exogenous variation of peers' weather in locations that differ from the ego's location as instruments. The framework untangles the endogeneity of the estimated effect using variations created by peers' weather as a shock to the ego's churn. We measure churn as a decision the ego makes after being inactive for one to two weeks and define peers as the ones the ego follows on the platform. We find that exercise challenge churn is socially contagious and demonstrates a complex contagion effect. Interestingly, our analyses reveal that the social contagion of churn diffuses from the sub-central or peripheral egos who have fewer friends in the social network to central egos who have more friends in the social network. Such churn contagion is mostly confined to low-density network communities with members who are poorly connected with one another. Our findings have important implications for designing intervention plans to stop wellness program churn based on social contagion.

4. Zhu Y, Chan J, Bi X, Guo Y, Wu J. Consumer Responses to Puffery: Empirical Evidence from a Cellular Network Upgrade Advertising. In final preparation for submission to Journal of Marketing.
Abstract: Firms often use puffery to attract consumer attention, but doing so can backfire and negatively impact the consumer experience, resulting in a boomerang effect. Despite the potential risks, there is limited empirical evidence on the boomerang effect of puffery, which has led to extensive discussion among researchers. To fill this gap, we conducted an empirical study to assess and quantify the impact of advertising puffery on consumer response to a cellular network upgrade advertising campaign using a large proprietary field dataset of consumer service calls about the same cellular network. To achieve our research objective, we relied on a continuous difference-in-differences causal identification strategy to estimate the impact of advertising puffery. This strategy leverages the varied advertisement intensities across different locations and the comparison of consumer response during pre- and post-period of the cellular network upgrade. We measured consumer response by analyzing their service call sentiment and found that the use of puffery had a negative effect on the sentiment. Interestingly, the negative sentiment was mainly driven by the advertised attributes related to the upgraded cellular network's most significant and highly-weighted strength. Moreover, the effect varied with location-based socioeconomic and demographic characteristics. Our findings have important implications for managers in terms of the appropriate use of puffery for different product attributes and its consequent response from various consumer subgroups.