Issue link: https://industrialengineer.epubxp.com/i/826827
June 2017 | ISE Magazine 53 candidate Xiaolei Fang and professors Nagi Gebraeel and Kamran Payna- bar of the Stewart School of Industrial and Systems Engineering at the Geor- gia Institute of Technology propose a prognostic modeling functional data analytic framework with algorithms that scale with the size of the data gen- erated by today's industrial assets. To provide accurate predictions of an as- set's remaining lifetime, the proposed framework relies heavily on the covari- ance structure of massive multistream sensor data, which dynamically change over time according to the degradation process. Thus, recomputing matrix factorizations from scratch as new data becomes available for such a large prob- lem becomes infeasible. The proposed framework addresses this key computational challenge by leveraging a randomized low-rank approximation algorithm. This inte- gration has a tremendous impact on computational speed and does not compromise the prediction accuracy. In a case study, the computational time for predicting the remaining lifetime of 100 units with 100 sensors per unit was 13 seconds using the convention- al approach and one second using the proposed model. For 1,000 sensors per unit, the conventional approach re- quired 24 minutes compared to 10 sec- onds for the proposed model. The biggest potential impact of the proposed model will be in industries that rely on the continuous availability of capital-intensive assets, like the en- ergy and power generation sector. CONTACT: Nagi Gebraeel; nagi@isye.gatech. edu; 765 Ferst Drive, Groseclose Building, De- partment of Industrial and Systems Engineer- ing, Georgia Institute of Technology, Atlanta, GA 30332 Investigation improves nurse care coordination in healthcare systems Due to the aging of our society, care for hospitalized patients needs to be well- coordinated within the healthcare team to manage the overall health of a patient effectively. Staff nurses, as the patient's "ever-present" healthcare team mem- bers, play a vital role in the care coor- dination. The recently developed Nurse Care Coordination Instrument (NCCI) is the first of its kind that enables quan- titative data to be collected to measure various aspects of nurse care coordina- tion. To advance the understanding of nurse care coordination in the hospital, we need to examine and reveal how multiple aspects of nurses' care coor- dination activities are related to their demographic and workload variables at the individual/nurse level as well as the structure, policy and climate of the nurses' practice environment at the unit level. Doctoral student Bing Si, profes- sor Jing Li and professor Gerri Lamb from Arizona State University, along with professor Madeline Schmitt from the University of Rochester, addressed this problem in their paper "A Multi- Response Multilevel Model with Ap- plication in Nurse Care Coordination." They proposed a novel statistical model for the unique data structure of NCCI. It jointly models the multiple aspects of nurse care coordination to enable transfer learning between mod- els. Each model quantitatively links one aspect of nurse care coordination (e.g., exchanging information, assisting each other's work, mobilizing people and resources) with multilevel predic- tors, including demographic and work- load variables at the individual/nurse level and variables that characterize the nurses' practice environment at the unit level. The proposed model employs a sparse formulation that enables selection of multilevel predictors that have signifi- cant impact on nurse care coordination for "small n large p" data sets. The proposed method was dem- onstrated on a data set collected across four U.S. hospitals in the metro Atlanta area using NCCI. Two individual-level predictors, i.e., the activities performed by nurses in checking and organizing their own work in an efficient manner, were found to have a positive impact on nurses' coordination between each oth- er. At the unit-level, providing assistance with discharge planning in a unit was found to help create a positive practice environment for the nurses in the unit to conduct care coordination. In the long run, their findings can inform interven- tions to improve staff-nurse care coordi- nation within hospital units, which will improve the quality of life for patients and save the healthcare system a great deal of money. CONTACT: Jing Li; jinglz@asu.edu; (480) 965- 0125; Brickyard Building 310, 699 S. Mill Ave., School of Computing, Informatics and Decision Systems Engineering, Arizona State University, Tempe, AZ 85287 Bing Si studies the nursing care data and modeling problem.