Healthcare IT systems have always been a double-edged sword. Since the beginning, many doctors and nurses have looked at healthcare IT (e.g., electronic medical records system, CPOE, PACS) as nothing more than another tool to assist in their routine clinical work. Yet at the same time, many healthcare IT systems do more than just assist clinical work, they subtly and overtly change the nature of clinical work.
Adrian's dissertation was on the design and implementation of an EMR system within an academic hospital campus. His work explored how EMR-initiated change faced significant obstacles across different clinical sites while at the same time bringing about some form of standardization of work. He drew on extensive research on socio-technical system to explore the underlying reasons for these paradoxical observations.
Adrian has continued to work within the healthcare IT system domain after returning to Singapore. Beyond just EMR system, he and his colleagues studied the establishment and ongoing development of a Singapore-based telegeriatric ecosystem. Using both quantitative and qualitative methods, Adrian and colleagues were able to draw significant insights to the impact of telemedicine on work efficiency and ecosystem value creation. Furthermore, they have published a teaching case (Asian Business Case Centre) that explicates the challenges and approach to sustaining this type of innovative healthcare-based ecosystem. The research paper based on this data was published in Information & Organization 2022.
Adrian and colleagues have begun to work with another Singapore healthcare organization that integrates acute care and community health care centers into one campus. They are working with the clinical and medical informatics team to explore the impacts of predictive and prescriptive AI models on the quality and flow of care. Part of this work won the 2017 Social Science Research Thematic Grant. The key idea in this program of research is to understand the tension between autonomy and AI as prescriptive and predictive analytic algorithms are embedded in various aspects of clinical and medical work within the hospital context. One of the studies from this research won Top 10 Abstract in an international conference (see below). In 2019, we have also been awarded $100,000 grant by ARISE-GERI Research to build a socio-economic profile of geriatric patients who have high risk of readmission through the use of NLP algorithms. Leveraging and building on our SSRTG research, we were awarded in 2022 a MOE Tier 2 research grant (over $852,000) to study how various AI systems are developed, designed, and implemented in different in-patient hospital settings.
In 2020, the team is studying the impact of integrated EMR use on specialist referrals from primary care clinics to specialist clinics. This project is supported by the Jurong Healthcare Fund and is a collaboration among NTU, NUHS, NUS Saw Swee Hock School of Public Health, and SUSS. Our conference paper was accepted in the AIS premier conference (ICIS) to be held in Copenhagen, Denamark (Dec. 2022). It was nominated as Best Paper in the track.
Finally, Adrian is also collaborating on a case study to understand and explicate how community care facility was setup in Singapore to meet the challenges of disruptive events (e.g., the COVID19 pandemic). Our focus is on the use of material and technological artifacts in the emergent coordination and practices deployed within such facilities. A case study report was written for this project and my collaborators and I recently published a case in CHI 2021.
In total (since 2008), Adrian and his colleagues have secured over S$1.7 million worth of grants for healthcare IT related research in Singapore.
Below are journal articles and conference presentations that discuss some of the findings they have found.
Adrian's dissertation was on the design and implementation of an EMR system within an academic hospital campus. His work explored how EMR-initiated change faced significant obstacles across different clinical sites while at the same time bringing about some form of standardization of work. He drew on extensive research on socio-technical system to explore the underlying reasons for these paradoxical observations.
Adrian has continued to work within the healthcare IT system domain after returning to Singapore. Beyond just EMR system, he and his colleagues studied the establishment and ongoing development of a Singapore-based telegeriatric ecosystem. Using both quantitative and qualitative methods, Adrian and colleagues were able to draw significant insights to the impact of telemedicine on work efficiency and ecosystem value creation. Furthermore, they have published a teaching case (Asian Business Case Centre) that explicates the challenges and approach to sustaining this type of innovative healthcare-based ecosystem. The research paper based on this data was published in Information & Organization 2022.
Adrian and colleagues have begun to work with another Singapore healthcare organization that integrates acute care and community health care centers into one campus. They are working with the clinical and medical informatics team to explore the impacts of predictive and prescriptive AI models on the quality and flow of care. Part of this work won the 2017 Social Science Research Thematic Grant. The key idea in this program of research is to understand the tension between autonomy and AI as prescriptive and predictive analytic algorithms are embedded in various aspects of clinical and medical work within the hospital context. One of the studies from this research won Top 10 Abstract in an international conference (see below). In 2019, we have also been awarded $100,000 grant by ARISE-GERI Research to build a socio-economic profile of geriatric patients who have high risk of readmission through the use of NLP algorithms. Leveraging and building on our SSRTG research, we were awarded in 2022 a MOE Tier 2 research grant (over $852,000) to study how various AI systems are developed, designed, and implemented in different in-patient hospital settings.
In 2020, the team is studying the impact of integrated EMR use on specialist referrals from primary care clinics to specialist clinics. This project is supported by the Jurong Healthcare Fund and is a collaboration among NTU, NUHS, NUS Saw Swee Hock School of Public Health, and SUSS. Our conference paper was accepted in the AIS premier conference (ICIS) to be held in Copenhagen, Denamark (Dec. 2022). It was nominated as Best Paper in the track.
Finally, Adrian is also collaborating on a case study to understand and explicate how community care facility was setup in Singapore to meet the challenges of disruptive events (e.g., the COVID19 pandemic). Our focus is on the use of material and technological artifacts in the emergent coordination and practices deployed within such facilities. A case study report was written for this project and my collaborators and I recently published a case in CHI 2021.
In total (since 2008), Adrian and his colleagues have secured over S$1.7 million worth of grants for healthcare IT related research in Singapore.
Below are journal articles and conference presentations that discuss some of the findings they have found.
Articles & Cases
A Paradox Theory Perspective on HIT’s Impact on Continuity of Care. by Yeow, A. & Soh, C. (2022)
43rd International Conference on Information Systems (ICIS), Dublin, Ireland **Best Paper Award in IS in Healthcare Track** Regulators of healthcare systems continue to investigate ways to improve continuity of care (COC) for patients given its inherent fragmented nature. Integrated healthcare information technology (HIT) system is touted as one of the ways to improve COC. Yet, studies show that there are still challenges in achieving effective COC even when supported by integrated HIT. These persistent challenges are likely due to deep-seated tensions among the various parts of the healthcare system that are involved in providing care. Drawing on HIT impact literature and paradox theory, we study the implementation of an integrated electronic medical record (EMR) system aimed at improving COC for the specialist referrals process in a hospital cluster. We found that while the system had positive impacts on some aspects of the COC, we uncover two types of paradoxical tensions occurring in this healthcare context that interfered with those positive impacts and contributed to ongoing COC challenges. Social influence is the main driver of emergency physicians’ intention to use an antibiotic clinical decision support mobile application. by Chow, A., Huang, Z.L., Yeow, A., & Lee, J. (2022) Journal of Hospital Infections We conducted an anonymous survey on physicians working in the ED at Sengkang General Hospital, a 1,000-bed acute-care hospital in Singapore, December 2021 through March 2022, on their intention to use an antibiotic CDS mobile application(app) for routine clinical care. Clinical decision support tools based on local epidemiology are preferred by ED physicians and have the potential to reduce unnecessary antibiotic use for URTI patients. This study provided important insights into the behavioral constructs associated with the intention to use an antibiotic CDS app, developed from locally-derived data. While the usefulness of the app is important to the physician user, our study found that having a strong social influence is the strongest determinant of the intention of use. |
How to Smoothen AI Implementation in Healthcare by Yeow, A. and Foong P.S. (2022) Asian Management Insights v. 9 iss. 1. https://cmp.smu.edu.sg/ami/article/20220529/how-smoothen-ai-implementation-healthcare
In this article, we describe these challenges and draw on current research, as well as our collective experience in developing AI-enhanced decision aid tools for HIT, to offer pointers on how to manage the challenges of implementation. We propose three domain-focused prescriptions, which can be used as practical springboards, as healthcare organisations embark on their implementation journey. We believe that by carefully following these prescriptions, healthcare organisations can successfully navigate known AI implementation pitfalls and challenges and be able to repeatedly implement AI tools in an effective manner. |
Managing paradoxical tensions in the development of a telemedicine system by Agarwal, N., Soh, C., and Yeow, A. (2022) v. 32 iss. 1. https://doi.org/10.1016/j.infoandorg.2022.100393
The global pandemic has escalated the demand for telemedicine systems across the world, particularly for vulnerable populations such as the elderly in nursing homes. However, challenges in implementation and high failure rates continue to affect the sustainability and capability of telemedicine systems. This study therefore addresses the question of how to sustain and develop telemedicine systems, and offers a conceptual model developed from longitudinal study data and paradox theory. We found that in the inter-organizational context of telemedicine systems, paradoxical tensions arise from conflict between demands and interests of the telemedicine system versus those of its members. We also identified when the specific tensions of belonging, learning, organizing, and performing are likely to occur. These tensions are addressed through responses, initiated by the hub, that address both system and member level demands and interests through creating collaborative governance, uplifting member capabilities, and targeted resourcing. We further demonstrate the temporal dynamics of how the hub's responses create inter-organizational norms and structures that in turn influence responses to tensions in subsequent phases. We examined variations in members' reactions to the responses, and found that they were influenced by member-specific resource factors, suggesting that while the hub does sustain the development of the telemedicine system through addressing common member demands, there are limits with regard to aspects that are more member-specific. Finally, we show how technology can be both enabler-trigger and enabler-response due to its inherent attributes of malleability and reconfigurability. |
Habit and Automaticity in Medical Alert Override: Cohort Study by Le, W., Goh, KH., Yeow, A., Ding, YY., Au, L., Poh, H., Li, K., Yeow, J., and Tan, G, (2022) v. 24 iss. 2. https://www.jmir.org/2022/2/e23355
We conducted a retrospective analysis using the log data comprising 66,049 alerts generated from hospitalized patients in a hospital from March 2017 to December 2018. We analyzed 1152 physicians exposed to a specific clinical support alert triggered in a hospital’s electronic medical record system to estimate the extent to which the physicians’ habit strength, which had been developed from habitual learning, impacted their propensity toward alert dismissal. We further examined the association between a physician’s habit strength and their subsequent incidences of alert dismissal. Additionally, we recorded the time taken by the physician to respond to the alert and collected data on other clinical and environmental factors related to the alerts as covariates for the analysis. We found that a physician’s prior dismissal of alerts leads to their increased habit strength to dismiss alerts. Furthermore, a physician’s habit strength to dismiss alerts was found to be positively associated with incidences of subsequent alert dismissals after their initial alert dismissal. Alert dismissal due to habitual learning was also found to be pervasive across all physician ranks, from junior interns to senior attending specialists. Further, the dismissal of alerts had been observed to typically occur after a very short processing time. Our study found that 72.5% of alerts were dismissed in under 3 seconds after the alert appeared, and 13.2% of all alerts were dismissed in under 1 second after the alert appeared. We found empirical support that habitual dismissal is one of the key factors associated with alert dismissal. We also found that habitual dismissal of alerts is self-reinforcing, which suggests significant challenges in disrupting or changing alert dismissal habits once they are formed. |
Mining of Clinical Notes for Readmission Prediction in Geriatric Patients by Goh, KH., Le, W., Yeow, A., Ding, YY., Au, L., Poh, H., Li, K., Yeow, J., and Tan, G, (2021) v. 23 iss. 10. https://www.jmir.org/2021/10/e26486
Propose a text-mining approach to analyze electronic medical records to identify older adults with key psychosocial factors that predict adverse health care utilization outcomes – measured by 30-day readmission. The psychological factors are appended to the LACE Index for Readmission to improve the prediction of readmission risk. We conduct a retrospective analysis using EMR notes of 43,216 hospitalization encounters in a hospital from 01 Jan 2017 to 28 Feb 2019. We employed text-mining techniques to extract psychosocial topics which are representative of these patients and test the utility of these topics in predicting 30-day hospital readmission beyond the predictive value of the LACE Index for Readmission. We observed that the added text-mined factors improved the AUROC of the readmission prediction by 8.46% for geriatric patients, 6.99% for the general hospital population, and 6.64% for frequent admitters. Medical social workers and case managers capture more psychosocial text topics compared to physicians. |
A Case Study of User Experience Design in a Disrupted Context: Design and Development of a Vital Signs Self-monitoring System (by Lim, CS., Foong, PS., Yeow, A., Koh, G., (2021)
ACM CHI Conference on Human Factors in Computing Systems (doi.org/10.1145/3411763.3443453) We present a case study where we developed an interface for remote vital signs self-monitoring at a large-scale isolation facility for COVID positive patients, under disrupted conditions. These conditions were: a lack of time, lack of access to end users, changing requirements, high risk of infection and supply chain limitations. We first describe the background of the development of the facility and the vital signs self-monitoring system. We use the 5 commonly prescribed activities of user experience design - Empathise, Define, Ideate, Prototype and Test- to describe how our work as user experience designers was affected by these disrupted conditions. Finally, we recommend a focus on Empathy, Prototyping and Communication for user experience practitioners and educators, whose training may be needed in similarly mission-critical, time-constrained circumstances. |
Artificial intelligence in sepsis early prediction and diagnosis using unstructured data in healthcare (by Goh, KH., Le, W., Yeow, A., Poh, H., Li, K., Yeow, J., and Tan, G. (2021) v.12 p. 711
Nature Communications (Open Access Article) Sepsis is a leading cause of death in hospitals. Early prediction and diagnosis of sepsis, which is critical in reducing mortality, is challenging as many of its signs and symptoms are similar to other less critical conditions. We develop an artificial intelligence algorithm, SERA algorithm, which uses both structured data and unstructured clinical notes to predict and diagnose sepsis. We test this algorithm with independent, clinical notes and achieve high predictive accuracy 12 hours before the onset of sepsis (AUC 0.94, sensitivity 0.87 and specificity 0.87). We compare the SERA algorithm against physician predictions and show the algorithm’s potential to increase the early detection of sepsis by up to 32% and reduce false positives by up to 17%. Mining unstructured clinical notes is shown to improve the algorithm’s accuracy compared to using only clinical measures for early warning 12 to 48 hours before the onset of sepsis. |
Text Mining of Clinical Progress Notes to Predict Future Onset of Sepsis in Hospitalized Patients (by Goh, KH., Yeow, A., Poh, H., Yeow, J., Le, W., Li, K., and Tan, G (2019)
iSRRS Conference (Top 10 Abstracts) Sepsis is associated with a high mortality rate and presents as a complex clinical problem for clinicians. Identifying and treating sepsis early has been shown to improve outcomes. As such there has been much work done in developing sepsis predictors (sniffers) in order to assist clinicians in identifying patients most at risk. However, there is a wide range of accuracy among these sniffers, which have been built using structured data models (reported ROC AUC ranging from 0.64 to 0.90 (Henry et al. 2015; Mani et al. 2014; Thiel et al. 2010)). Our study attempts to develop a sepsis sniffer that can combine structured data models with clinical information gleaned from clinical progress notes. It is a well-established fact that clinical progress notes are a rich source of unstructured data and many clinical applications have used text-mining as a parser or automated coder to extract symptoms and medication administration from these notes. In this study, we used text mining to extract relevant clinical information from free-text clinical progress notes to improve the accuracy of sepsis prediction. |
Cost Analysis of Implementing a Telegeriatrics Ecosystem with Nursing Homes: Panel Data Analysis (by Low, J. Y. H., Goh, K.H., Agarwal, N., Soh, C., & Yeow, A. 2019)
Health Systems (online publications) Our study analyzed the economic impact of a telegeriatrics programme on care of nursing home residents, from the healthcare system provider’s perspective. This is a retrospective, archival data analysis of multiple data sources in 4 nursing homes of Singapore from 2010 to 2015. Individuals admitted to nursing homes and have undergone telemedicine consultations (N=859) from 2010 to 2015 were recruited. We conducted a cost analysis of the programme by reviewing past hospital admissions’ and specialist outpatient clinic (SOC) visits’ billing records, nurse training records, and key performance indicators’ reports. A significant relationship was observed between teleconsultations and SOC visit cost (β1 = -83.366, p-value<0.01) and between teleconsultations and inpatient cost (β1 =-470.971, p-value <0.05). Remote video consultations could reduce unnecessary SOC visits and hospital admissions, and therefore lead to cost savings. Training of nursing home nurses could translate to cost savings as a result of decreased ED transfers. |
Value Co-creation in Integrated Care Ecosystems: A Member Perspective (by Agarwal, N., Soh, C. and Yeow, A. 2016)
36th International Conference on Information Systems (ICIS), Dublin, Ireland With a growing shift towards a service based economy, organizations are increasingly participating in service ecosystems to jointly co-create value with the keystone organization. Prior research has however taken a cross-sectional view to examine this participation and has assessed limited aspects of value to members. This study hence takes a service dominant view of ecosystems and explicates three value co-creational exchanges that occur between the keystone organization and the members in a service ecosystem—service exchanges, resource exchanges and relational exchanges—and hypothesizes relationship between these exchanges and member value. The proposed model is studied in the context of a telemedicine-enabled ecosystem formed between an acute hospital and several nursing homes. Both economic and other value outcomes are considered —cost effectiveness, clinical outcomes and user satisfaction. With this, the study seeks to provide a more comprehensive analysis of relationships between value co-creational exchanges and value to members in service ecosystems. |
Working Harder or Working Smarter?: Information Technology and Resource Reassignment in Health-care Processes (by Yeow, A. and Goh KH. 2015)
MIS Quarterly, 39(4), pp. 763-785. **Best Paper Award for SIG Health 2016** In this study, a granular examination of the impact of HIT systems on how resources are allocated to healthcare tasks and processes was undertaken. The effects of telemedicine on the input allocative efficiency of the healthcare process through the reallocation of organizational resources was evaluated and an assessment of whether gains in allocative efficiency resulted in improvements in organizational outcomes, such as lower hospitalization rates and lower uncertainty in patient wait time, was conducted. Applying the theory of swift and even flow, our findings suggest that the gains in allocative efficiency for some processes are associated with improved organizational outcomes. |
GeriCare@North: Building and Sustaining a Tele-Geriatrics Ecosystem (by Soh, C., Agarwal, N. and Yeow, A. 2017)
Asian Business Case Centre (ABCC-2017-006) This case chronicles the evolution, challenges and actions taken by an integrated care telemedicine ecosystem comprising an acute hospital and nursing homes for the elderly. More generally, it describes the challenges faced by the keystone or hub organization in initiating and deploying a novel service that required the contributions and collaboration of diverse partners, many of whom were experiencing resource constraints, in the face of changing industry environment. It details the steps taken to deal with the challenges of getting wary partners with unequal resources and capabilities on board, levelling up skills, and putting in place technology and joint processes necessary for the delivery of a joint service. It demonstrates that the steps resulted in increased partner trust, improved levels of partner competency, increased service quality, and cost savings. The case can be used for courses in: (a) Information Systems, to examine the implementation of systems across multiple organizational partners in order to offer new services. (b) Platforms and ecosystems, particularly those focused on ecosystem evolution. (c) Healthcare, to discuss the development and deployment of integrated care initiatives. |
Technology and Sociomaterial Performation (by Yeow, A. and Faraj, S. 2014)
Information Systems and Global Assemblages: (Re)configuring Actors, Artefacts, Organizations: IFIP WG 8.2 Working Conference, Auckland, New Zealand Organizational researchers have acknowledged that understanding the relationship between technology and organization is crucial to understanding modern organizing and organizational change. There has been a significant amount of debate concerning the theoretical foundation of this relationship. Our research draws and extends Deleuze and DeLanda’s work on assemblages and Callon’s concept of performation to investigate how different sociomaterial practices are changed and stabilized after the implementation of new technology. Our findings from an in-depth study of two ambulatory clinics within a hospital system indicate that “perform-ing” of constituting, counter-performing, calibrating, and stratifying explained the process of sociomaterial change and that this process is governed by an overarching principle of “performative exigency”. Future studies on sociomateriality and change may benefit from a deeper understanding of how sociomaterial assemblages are rendered performative. |
Marrying Work and the Technical Artifact within the Healthcare organization: A Narrative Network Perspective on IT Innovation-Mediated Organizational Change (by Yeow, A. and Faraj, S. 2008) **Nominated for Best Paper Award**
28th International Conference on Information Systems (ICIS), Paris, France. Despite the implicit belief that IT innovations brings beneficial change, medical practitioners and healthcare professionals constantly struggle to realize the innovation potential of electronic medical records (EMR) system in revolutionizing clinical practices. To understand this conundrum, this paper uses an in-depth case study of an EMR implementation to develop a grounded theory of why, when, and how IT-innovation mediated change occur. We propose the Narrative Network Perspective that combines the analysis of the processes of configuration, implementation and use of the system. This combined view allows researchers to understand how "production narrative network", infrastructure and the macrostructure in healthcare environment co-evolve with the idealized production narrative network inscribed in the EMR system within and across the three phases. By tracing and taking into account all these elements time, this perspective provides plausible answers to when and why organizational innovation occur with the introduction of IT innovations. |