Supporting the Billing Process in Outpatient Medical Care: Automated Medical Coding Through Machine Learning

2022 | European Conference On Information Systems | Citations: 0

Authors: Oberste, Luis; Finze, Nikola; Hoffmann, Philipp; Heinzl, Armin

Abstract: Reimbursement in medical care implies significant administrative effort for med ... Expand

Abstract: Reimbursement in medical care implies significant administrative effort for medical staff. To bill the treatments or services provided, diagnosis and treatment codes must be assigned to patient records using standardized healthcare classification systems, which is a time-consuming and error-prone task. In contrast to ICD diagnosis codes used in most countries for inpatient care reimbursement, outpatient medical care often involves different reimbursement schemes. Following the Action Design Research methodology, we developed an NLP-based machine learning artifact in close collaboration with a general practitioner's office in Germany, leveraging a dataset of over 5,600 patients with more than 63,000 billing codes. For the code prediction of most problematic treatments as well as a complete code prediction task, we achieved F1-scores of 93.60 % and 78.22 %, respectively. Throughout three iterations, we derived five meta requirements leading to three design principles for an automated coding system to support the reimbursement of outpatient medical care. Collapse

Topics: health information system open source innovation management knowledge creation organizational context

Methods: design artifact design principle design requirement action design research natural language processing

Augmenting Humans in the Loop: Towards an Augmented Reality Object Labeling Application for Crowdsourcing Communities

2021 | International Conference on Business Informatics | Citations: 0

Authors: Schuir, Julian; Brinkhege, René; Anton, Eduard; Oesterreich, Thuy Duong; Meier, Pascal; Teuteberg, Frank

Abstract: Convolutional neural networks (CNNs) offer great potential for business applicat ... Expand

Abstract: Convolutional neural networks (CNNs) offer great potential for business applications because they enable real-time object recognition. However, their training requires structured data. Crowdsourcing constitutes a popular approach to obtain large databases of manually-labeled images. Yet, the process of labeling objects is a time-consuming and cost-intensive task. In this context, augmented reality provides promising solutions by allowing an end-to-end process of capturing objects, directly labeling them and immediately embedding the data in training processes. Consequently, this paper deals with the development of an object labeling application for crowdsourcing communities following the design science research paradigm. Based on seven issues and twelve corresponding meta-requirements, we developed an AR-based prototype and evaluated it in two evaluation cycles. The evaluation results reveal that the prototype facilitates the process of object detection, labeling and training of CNNs even for inexperienced participants. Thus, our prototype can help crowdsourcing communities to render labeling tasks more efficient. Collapse

Topics: crowdsourcing mobile application usability mobile system user experience

Methods: design artifact survey literature study artificial neural network design principle

Theories: socio technical theory

Yield Prognosis for the Agrarian Management of Vineyards using Deep Learning for Object Counting

2019 | International Conference on Business Informatics | Citations: 0

Authors: Heinrich, Kai; Roth, Andreas; Breithaupt, Lukas; Möller, Björn; Maresch, Johannes

Abstract: In various applications, the counting of objects based on image data plays a piv ... Expand

Abstract: In various applications, the counting of objects based on image data plays a pivotal role. In this paper we first conducted a literature review to display the state of the art in counting objects and summarized the results by extracting several important concepts that describe the counting problem as well as the solution. In a second step we applied this knowledge to yield prognosis in vineyards, where we used Deep Learning models to detect the objects. While these methods used in the detection step are state of the art and perform very well, several problems are usually introduced by the constraint of only counting an object once in the counting step. We provide a solution for this common problem by identifying unique objects and tracking them throughout a sequence of images in order to avoid counting objects more than once, resulting in an automated yield prognosis model for vineyards. Collapse

Topics: cloud computing smart wearable device database system

Methods: literature study computational algorithm artificial neural network deep learning literature sample