Implications for Successful Implementation of Predictive Modelling in a Healthcare Setting

Implementing algorithms in the health care sector shows promising futuristic prospects for both healthcare personnel, patients, health care economy and innovation. The growing data pools, hospital adap-tion of new data-centric health care technology and the patients use of health care innovation to moni-tor their own health creates new ideas for innovative ways to prolong vitality in patients by findingnew patterns in data for diagnosis and precision medicine.The amount of literature where machine learning algorithms are used for theoretical health diagnosisand improving health care outcomes is exponentially growing. However, we find that concrete casesreporting on implementations and use of machine learning predictions are lacking in the literature.We provide a needed and critical perspective to the socio-technical system from where much of healthcare data is created and put into real context of health care personnel, existing technology and thehospitals. Here, implementations of new hospital information systems are affecting the daily lives ofdoctors, nurses, administration and patients in the pursuit and idealization of the data-driven healthcare sector where costs are brought to a minimum and access to care is increased.We use a prospective machine learning implementation case from a Danish Hospital to learn howdata-mining, statistics, participatory design and implementation can be combined to investigate howpeople, data, professionals and technology can come together to create a better health care sector.The authors used Electronic Health Records from a clinical setting in Denmark to predict patient no-shows with a machine learning algorithm. A participatory design approach was used to create moremeaningful data and getting closer to both the people and the technology where the data was createdand extracted for training a modern class-a algorithm with good results in the initial version of thealgorithmic software.While the predictive model has shown to be highly reliable on test data, the potential of using it face anumber of complications if to be implemented in a complex social-technical environment such as ahealth care hospital setting. The paper contributes by identifying three critical implications needed tobe addressed in order for predictive modelling to be successfully implemented and realized in the hos-pital organization.