Dataset Information

Predictive analytics in health care: how can we know it works?

ABSTRACT: There is increasing awareness that the methodology and findings of research should be transparent. This includes studies using artificial intelligence to develop predictive algorithms that make individualized diagnostic or prognostic risk predictions. We argue that it is paramount to make the algorithm behind any prediction publicly available. This allows independent external validation, assessment of performance heterogeneity across settings and over time, and algorithm refinement or updating. Online calculators and apps may aid uptake if accompanied with sufficient information. For algorithms based on "black box" machine learning methods, software for algorithm implementation is a must. Hiding algorithms for commercial exploitation is unethical, because there is no possibility to assess whether algorithms work as advertised or to monitor when and how algorithms are updated. Journals and funders should demand maximal transparency for publications on predictive algorithms, and clinical guidelines should only recommend publicly available algorithms.

SUBMITTER: Van Calster B

PROVIDER: S-EPMC6857503 | biostudies-literature | 2019 Dec

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Predictive analytics in health care: how can we know it works?

Van Calster Ben B Wynants Laure L Timmerman Dirk D Steyerberg Ewout W EW Collins Gary S GS

Journal of the American Medical Informatics Association : JAMIA 20191201 12

There is increasing awareness that the methodology and findings of research should be transparent. This includes studies using artificial intelligence to develop predictive algorithms that make individualized diagnostic or prognostic risk predictions. We argue that it is paramount to make the algorithm behind any prediction publicly available. This allows independent external validation, assessment of performance heterogeneity across settings and over time, and algorithm refinement or updating. ...[more]

PMID: 31373357

Similar Datasets

Project description:BackgroundThe solution to the growing problem of rural residents lacking health care access may be found in the use of telemedicine and mobile health (mHealth). Using mHealth or telemedicine allows patients from rural or remote areas to have better access to health care.ObjectiveThe objective of this study was to understand factors influencing the choice of communication medium for receiving care, through the analysis of mHealth versus telemedicine encounters with a virtual urgent clinic.MethodsWe conducted a postdeployment evaluation of a new virtual health care service, Virtual Urgent Clinic, which uses mHealth and telemedicine modalities to provide patient care. We used a multinomial logistic model to test the significance and predictive power of a set of features in determining patients' preferred method of telecare encounters-a nominal outcome variable of two levels (mHealth and telemedicine).ResultsPostdeployment, 1403 encounters were recorded, of which 1228 (87.53%) were completed with mHealth and 175 (12.47%) were telemedicine encounters. Patients' sex (P=.004) and setting (P<.001) were the most predictive determinants of their preferred method of telecare delivery, with significantly small P values of less than .01. Pearson chi-square test returned a strong indication of dependency between chief concern and encounter mediums, with an extremely small P<.001. Of the 169 mHealth patients who responded to the survey, 154 (91.1%) were satisfied by their encounter, compared with 31 of 35 (89%) telemedicine patients.ConclusionsWe studied factors influencing patients' choice of communication medium, either mHealth or telemedicine, for a virtual care clinic. Sex and geographic location, as well as their chief concern, were strong predictors of patients' choice of communication medium for their urgent care needs. This study suggests providing the option of mHealth or telemedicine to patients, and suggesting which medium would be a better fit for the patient based on their characteristics.

Project description:BACKGROUND:While the application of learning analytics in tertiary education has received increasing attention in recent years, a much smaller number have explored its use in health care-related educational studies. OBJECTIVE:This systematic review aims to examine the use of e-learning analytics data in health care studies with regards to how the analytics is reported and if there is a relationship between e-learning analytics and learning outcomes. METHODS:We performed comprehensive searches of papers from 4 electronic databases (MEDLINE, EBSCOhost, Web of Science, and ERIC) to identify relevant papers. Qualitative studies were excluded from this review. Papers were screened by 2 independent reviewers. We selected qualified studies for further investigation. RESULTS:A total of 537 papers were screened, and 19 papers were identified. With regards to analytics undertaken, 11 studies reported the number of connections and time spent on e-learning. Learning outcome measures were defined by summative final assessment marks or grades. In addition, significant statistical results of the relationships between e-learning usage and learning outcomes were reported in 12 of the identified papers. In general, students who engaged more in e-learning resources would get better academic attainments. However, 2 papers reported otherwise with better performing students consuming less e-learning videos. A total of 14 papers utilized satisfaction questionnaires for students, and all were positive in their attitude toward e-learning. Furthermore, 6 of 19 papers reported descriptive statistics only, with no statistical analysis. CONCLUSIONS:The nature of e-learning activities reported in this review was varied and not detailed well. In addition, there appeared to be inadequate reporting of learning analytics data observed in over half of the selected papers with regards to definitions and lack of detailed information of what the analytic was recording. Although learning analytics data capture is popular, a lack of detail is apparent with regards to the capturing of meaningful and comparable data. In particular, most analytics record access to a management system or particular e-learning materials, which may not necessarily detail meaningful learning time or interaction. Hence, learning analytics data should be designed to record the time spent on learning and focus on key learning activities. Finally, recommendations are made for future studies.

Dataset Information

Predictive analytics in health care: how can we know it works?

Publications

Predictive analytics in health care: how can we know it works?

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure