Natural Language Techniques for Decision Support Based on Patient Complaints

Literaturnachweis - Detailanzeige

Trefferliste

Autor/in	ElMessiry, Adel Magdi
Titel	Natural Language Techniques for Decision Support Based on Patient Complaints
Quelle	(2016), (82 Seiten) PDF als Volltext Verfügbarkeit Ph.D. Dissertation, North Carolina State University
Sprache	englisch
Dokumenttyp	gedruckt; online; Monographie
ISBN	978-1-3696-2067-2
Schlagwörter	Hochschulschrift; Dissertation; Patients; Health Services; Medical Services; Hospitals; Risk; Death; Prevention; Correlation; Costs; Expenditures; Legal Problems; Physician Patient Relationship; Feedback (Response); Intervention; Classification; Early Intervention; Computer Software; Computational Linguistics; Grammar; Prediction; Advocacy; Validity; Natural Language Processing; Accuracy; Decision Making; Statistical Analysis + Suchen Sie Ihr Suchwort? Thesis; Dissertations; Academic thesis; Patient; Health service; Gesundheitsdienst; Gesundheitswesen; Arzt; Krankengymnast; Krankenhaus; Risiko; Sterbefall; Tod; Todesfall; Prävention; Vorbeugung; Korrelation; Cost; Kosten; Ausgaben; Arzt-Patient-Beziehung; Classification system; Klassifikation; Klassifikationssystem; Linguistics; Computerlinguistik; Grammatik; Vorhersage; Sozialanwaltschaft; Gültigkeit; Natürliche Sprache; Decision-making; Entscheidungsfindung; Statistische Analyse
Abstract	Complaining is a fundamental human characteristic that has prevailed throughout the ages. We normally complain about something that went wrong. Patient complaints are no exception; they focus on problems that occurred during the episode of care. The Institute of Medicine estimated that each year thousands of patients die due to medical errors. The number of patient deaths associated with medical errors was 98,000 in 1984. It would be reasonable to assume that this figure should have been considerably reduced given the tremendous scientific progress achieved over the past three decades. The facts are quite contrary; an epidemic of patient harm in hospitals has developed. In 2013, a staggering 210, 000 deaths per year were associated with preventable harm in hospitals, while the number of premature deaths associated with preventable harm was estimated at more than 400,000 per year [James, 2013]. Healthcare in the United States is not monolithic. Rather, it is a broad and ever-increasing range of options. This rapid growth can quickly overwhelm care providers and results in significant medical errors. Although human life is the ultimate casualty, the monetary impact is another aspect of the underlying problem. The cost stemming from countless wasted hours is staggering. Healthcare expenditure constitutes a considerable percentage of the global gross domestic product. The United States comes in at the top of the range with 17.5%, which has more than tripled from 5% in 1960. Healthcare malpractice payouts are a $3.6 billion component of $55.8 billion healthcare risk cost and the overall $2,799 billion healthcare expenditure. An early precursor of malpractice is a patient complaint that can result in an adverse action report, which is an action taken against a practitioner's clinical privileges or medical staff membership in a healthcare entity. Adverse actions have been consistently on the rise over the past ten years. The current mitigation strategy depends on human coders who analyze patient feedback to classify the complaints to help a more advanced team build an intervention plan. Early intervention is critical to prevention and the systematic improvement of the healthcare system. Without the ability to scale the intervention, the tragic loss of human life and the astronomical financial cost will only continue to rise. Nevertheless, due to the growth in the generated patient feedback, scaling the current approach requires automating the initial triage process. However, due to the complexity and diversity of the linguistic representations, building a computational tool for this task is challenging. This dissertation focuses on the problem of developing an understanding of patient complaints to enable a robust approach requiring minimal human supervision to build a patient complaint analysis tool that can be used across healthcare providers. To systematically study this problem, we have identified three important tasks that are critical for building such a tool: (1) Establishing the urgency concept by (a) defining patient complaint urgency and (b) building a framework to predict urgency. (2) Automating the current systems by (a) automatically mapping patient complaints to a sentiment-based model and (b) accurately inferring the class for each complaint. (3) Exploring grammatical analysis by (a) defining a method to build a domain-specific dependency based rules for feature extraction and reduction and (b) applying those rules on the ground truth dataset to predict patient complaint classification. Accordingly, for the first task, we implement a set of classifiers to model urgency and predict subsequent complaints. In the second task, we propose a novel mapping method that maps complaint terms to the linguistic inquiry and word count domains (LIWC) and implement a classification tool that employs supervised learning. In the final task, we build on the grammatical dependency models and develop a set of domain-based rules enabling the extraction of more meaningful features. For all of these tasks, we evaluate the effectiveness of our approaches on a real-life dataset collected by the Vanderbilt Center for Patient and Professional Advocacy and validated against trained human coders.Results show that our approach is quite useful in identifying patient complaint's urgency. Furthermore, our LIWC-based classifiers yield greater overall accuracy than traditional methods in classifying a limited set of categories. The results we obtained using domain-specific grammatically extracted features are promising. Compared to basic unigram features, our method produced superior results across all categories, showing a weighted F-Measure gain from 0.41 to 0.82. Overall, this dissertation indicates the potential benefits of applying natural language techniques in analyzing patient complaints to support decision making in healthcare. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://www.proquest.com/en-US/products/dissertations/individuals.shtml.] (As Provided).
Anmerkungen	ProQuest LLC. 789 East Eisenhower Parkway, P.O. Box 1346, Ann Arbor, MI 48106. Tel: 800-521-0600; Web site: http://www.proquest.com/en-US/products/dissertations/individuals.shtml
Erfasst von	ERIC (Education Resources Information Center), Washington, DC
Update	2020/1/01