Developed nations across the world are facing significant pressures in managing healthcare due to aging population and rising healthcare manpower costs. The United Nations Population Division predicts that in 2050, 22% of the world’s population will be in their 60’s as opposed to just 11%. All else equal, an aging population places negative pressures on the nation’s economy due to labor shortages, and possibly declining productivity. The use of information technology is one commonly cited solution to boost productivity in developed nations to manage these healthcare pressures. The use of information technology to digitize the delivery of healthcare services opens up opportunities for applying data analytics in the management of healthcare. Below are the description of two studies that rely on health analytics. One is a completed study and the other is a research study in progress.

Below are two showcase projects to illustrate our ongoing research efforts.

• Project 1. Healthcare Information Technologies
  
  
  
  The healthcare industry continues to face chronic challenges of rising costs and increased workload for healthcare workers (Kohli & Kettinger, 2004; Porter & Teisberg, 2006). Healthcare information technologies (“HIT”) are often touted as one of the solutions to these problems. While some studies on healthcare and HIT have found that IT investments and use, in general, have led to lower medical errors, mortality rates, and increased financial performance (Amarasingham, Plantinga, Diener-West, Gaskin, & Powe, 2009; S. Devaraj & Kohli, 2000, 2003; Kohli & Kettinger, 2004; Porter & Teisberg, 2006), these positive HIT impact findings are not consistently true as other studies have highlighted cases of HIT issues and failures. Part of this confusion and equivocality of HIT impact findings is due to the fact that many HIT studies were based on cross-sectional data and were ambiguous regarding the type of HIT they were studying (Agarwal, Gao, DesRoches, & Jha, 2010). Others point to the complications in measuring the benefits of HIT as healthcare work is highly complex (Cuellar & Gertler, 2005; Davidson & Chiasson, 2005; Leviss, 2010). Put together, even as HIT impact research continues to evolve, more research is needed to explain how HIT use could help manage healthcare’s rising costs and improve productivity. Early research on IT impacts has identified how IT (in general) could directly change the level of output, usually at the aggregate level, so as to bring about improved organization performance (Hitt & Brynjolfsson, 1996; Hitt, Wu, & Zhou, 2002). Following from this stream of research, most HIT impact studies have focused on the direct productivity impacts of HIT. However, this direct approach is problematic as studies of HIT use have found that many aspects of healthcare work are hard to enhance and automate since this type of work requires ongoing human interactions (Berg, 1998; Davidson & Chismar, 2007). Furthermore, it is common for healthcare medical personnel to put in long hours at work, thus any possible gains in productivity may be less likely to be derived from working harder.
  
  Recent research, especially in the use of telemedicine, has shown that HIT use may indirectly impact work processes and improve healthcare processes (Hui, Woo, Hjelm, Zhang, & Tsui, 2001; Singh, Mathiassen, Stachura, & Astapova, 2011). Building on a small but growing stream of IS research that provides an enhanced and holistic understanding of HIT value (Sarv Devaraj & Kohli, 2002; S. Devaraj, Ow, & Kohli, 2013; Nirup M Menon, Yaylacicegi, & Cezar, 2009), our research study focuses on how specific HIT—telemedicine—impacts healthcare processes and how that, in turn, leads to improved organizational outcomes. In this study, we used the concept of input allocative efficiency and the Theory of Swift and Even Flow (“TSEF”) perspective to explicate how HIT may affect relevant healthcare outcomes.
  
  Our study analyzed the impact of telemedicine use on patient, physician, and healthcare process outputs in a geriatric department of an acute-care hospital. We evaluated the effect of telemedicine on the input allocative efficiency of healthcare process through the re-allocation of organizational resources and assessed whether gains in allocative efficiency resulted in improvements in organizational outcomes. Input allocative efficiency (or in short allocative efficiency) refers to the choice of inputs (resources) mix to produce the outputs while minimizing production cost (Kumbhakar & Lovell, 2000; N.M. Menon, Lee, & Eldenburg, 2000). The allocative efficiency approach allows us to understand how HIT use could improve the assignment of resources to different tasks for efficiency gains (Leibenstein, 1966; N.M. Menon & Lee, 2000; N.M. Menon et al., 2000). In our study, we chose to focus on the impacts of applying a telemedicine system to the geriatric care process. We conducted a longitudinal field study that combined interview, archival, observation and survey data to measure the performance before and after the implementation and use of a telemedicine system in the geriatric specialist clinic.
  
  Our study found that the use of telemedicine and the process changes that accompanied the system had overall positive impact on allocative efficiency for some processes. We observed that applying telemedicine with business process redesign enabled greater visibility of the patient information resulting in patients (tasks) being better assigned to the appropriate physicians (resources). Further, using TSEF principles, we show that the improved allocative efficiency achieved through the new telemedicine process or clinical pathway reduced variance of patient wait-time in the specialist clinic and provided better care to nursing home patients. By tracing the process and mechanisms through which HIT indirectly impact on organizational outcomes via the reallocation of resources and tasks, our study potentially “enhances our understanding of the various positive manifestations of IT” by providing a more holistic perspective of HIT value (Kohli & Grover, 2008 p. 33).
• Project 2. Electronic Medical Records
  
  ​​
  
  In this research program, we partner another hospital located in Singapore to study how the use of Electronic Medical Records (EMR) in Singapore hospitals improves clinical outcomes. More specifically, our over-arching research question seeks to unpack the inter-related processes, agents and issues that influence the way EMR impacts clinical outcomes. As part of the research program, we plan to study the impact of a text-mining based predictive model on the coordination of workflow processes across medical teams. Predictive analytics relies on EMR data to predict health-related event by integrating clinical-based and data-mining based models. Further, these analytics rely on these predictive models to identify at-risk patients and recommend an optimal clinical pathway or intervention.
  
  The use of text-mining tools to analyses the doctors’ notes recorded during the ward “rounds” is one way to leverage upon the capabilities of predictive models. To operationalize this concept, we consider a particular, major complication a patient may suffer during their stay in a hospital – sepsis. Sepsis is a life-threatening condition in which the body is fighting a severe infection that has spread via the bloodstream. Sepsis may be hard to detect in its early stages and may present various ambiguous tell-tale signs and is one major concern for doctors while managing inpatients within hospitals. In this study, we propose to develop a predictive model that will help predict the risk of sepsis prior to the onset of sepsis. The predictive model will utilize textual topics text-mined from doctor’s notes recorded during the ward “rounds” and other codified medical symptoms such as suspicion of infection, systemic inflammatory response syndrome, organ hypoperfusion and dysfunction.
  
  This project is both practically as well as theoretically driven. From a practical perspective, the medical informatics and clinical staff at NTFGH has discussed with us the need to develop robust predictive models to enable handover process of medical teams across shifts to be more efficient and thereby reduce potential gaps in clinical care for patients that are suspected to have sepsis. The development of a sepsis detection predictive model will help to facilitate this operationally as such a predictive model can either provide clinical guideline recommendations or act as an alert function. This predictive model will be designed to be self-learning and could be refined over time as more textual data are made available from the EMR system.
  
  ​​