2020
DOI: 10.1093/jamiaopen/ooaa058
|View full text |Cite
|
Sign up to set email alerts
|

Modeling physician variability to prioritize relevant medical record information

Abstract: Objective Patient information can be retrieved more efficiently in electronic medical record (EMR) systems by using machine learning models that predict which information a physician will seek in a clinical context. However, information-seeking behavior varies across EMR users. To explicitly account for this variability, we derived hierarchical models and compared their performance to nonhierarchical models in identifying relevant patient information in intensive care unit (ICU) cases. … Show more

Help me understand this report
View preprint versions

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
1
1

Citation Types

0
4
0

Year Published

2020
2020
2023
2023

Publication Types

Select...
5
1

Relationship

2
4

Authors

Journals

citations
Cited by 6 publications
(4 citation statements)
references
References 37 publications
0
4
0
Order By: Relevance
“…Presently, it is not yet possible to recommend an anticipated 75 g OGTT in overweight women older than 35 years [19], although they possess the same risk of developing GDM as younger women with preconception obesity. However, its monocentric design, with all pregnant women being enrolled and performing 75 g OGTT at the same diabetes care center, is a strength that helps to reduce the risk of biases associated with inter-laboratory analytical variations [19], as well as the misclassification of risk factors related to inter-operator variability in the recording of relevant medical information during routine diagnostic work-ups [45].…”
Section: Discussionmentioning
confidence: 99%
“…Presently, it is not yet possible to recommend an anticipated 75 g OGTT in overweight women older than 35 years [19], although they possess the same risk of developing GDM as younger women with preconception obesity. However, its monocentric design, with all pregnant women being enrolled and performing 75 g OGTT at the same diabetes care center, is a strength that helps to reduce the risk of biases associated with inter-laboratory analytical variations [19], as well as the misclassification of risk factors related to inter-operator variability in the recording of relevant medical information during routine diagnostic work-ups [45].…”
Section: Discussionmentioning
confidence: 99%
“…Additionally, the models exhibit calibrated confidence scores, as the HC-1-ECE ranges from 0.86 [0.76, 0.96] to 0.93 [0.88, 0.98]; due to the calibrated results of the models used in this study, the likelihood of the patient’s worsening (our target outcome) is well grounded. Moreover, whenever predictive models produce calibrated probabilities, the more calibrated the probabilities are, the greater the utility expected from the decisions they generate is [ 50 ]. Non-calibrated models can have a detrimental effect on healthcare.…”
Section: Discussionmentioning
confidence: 99%
“…Accurate uncertainty estimation is required to provide adequate support for human decision-making in machine learning models, especially in high-risk settings (e.g., medicine) where an accurate uncertainty estimation is of primary importance [ 49 ]. The definition of calibration error is the absolute difference between the mean of the predicted probabilities and the proportion of positive outcomes [ 50 ]. The Expected calibration error (ECE) is the weighted average of the calibration errors defined as follows [ 49 ]: where is the average confidence score within bin i (i.e., ), is the relative frequency of the positive class in bin i (i.e., ), is the proportion of instances that fall within .…”
Section: Methodsmentioning
confidence: 99%
“… 5 The LEMR system relies on supervised machine learning models that predict which patient data are likely to be relevant in the context of a clinical task. 6 , 7 However, a critical barrier to building machine learning models for a LEMR system is the acquisition of training data regarding which patient data are relevant for a clinical task. Such data are not recorded in sufficient granularity in currently deployed EMR systems.…”
Section: Introductionmentioning
confidence: 99%