JAMIA Open最新文献

{"title":"Impact of possible errors in natural language processing-derived data on downstream epidemiologic analysis.","authors":"Zhou Lan, Alexander Turchin","doi":"10.1093/jamiaopen/ooad111","DOIUrl":"10.1093/jamiaopen/ooad111","url":null,"abstract":"<p><strong>Objective: </strong>To assess the impact of potential errors in natural language processing (NLP) on the results of epidemiologic studies.</p><p><strong>Materials and methods: </strong>We utilized data from three outcomes research studies where the primary predictor variable was generated using NLP. For each of these studies, Monte Carlo simulations were applied to generate datasets simulating potential errors in NLP-derived variables. We subsequently fit the original regression models to these partially simulated datasets and compared the distribution of coefficient estimates to the original study results.</p><p><strong>Results: </strong>Among the four models evaluated, the mean change in the point estimate of the relationship between the predictor variable and the outcome ranged from -21.9% to 4.12%. In three of the four models, significance of this relationship was not eliminated in a single of the 500 simulations, and in one model it was eliminated in 12% of simulations. Mean changes in the estimates for confounder variables ranged from 0.27% to 2.27% and significance of the relationship was eliminated between 0% and 9.25% of the time. No variables underwent a shift in the direction of its interpretation.</p><p><strong>Discussion: </strong>Impact of simulated NLP errors on the results of epidemiologic studies was modest, with only small changes in effect estimates and no changes in the interpretation of the findings (direction and significance of association with the outcome) for either the NLP-generated variables or other variables in the models.</p><p><strong>Conclusion: </strong>NLP errors are unlikely to affect the results of studies that use NLP as the source of data.</p>","PeriodicalId":36278,"journal":{"name":"JAMIA Open","volume":"6 4","pages":"ooad111"},"PeriodicalIF":2.1,"publicationDate":"2023-12-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10752385/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139049459","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Feasibility of implementing patient-reported outcome measures into routine breast cancer care delivery using a novel collection and reporting platform.","authors":"Elena Tsangaris, Colby Hyland, George Liang, Joanna O'Gorman, Dany Thorpe Huerta, Ellen Kim, Maria Edelen, Andrea Pusic","doi":"10.1093/jamiaopen/ooad108","DOIUrl":"10.1093/jamiaopen/ooad108","url":null,"abstract":"<p><strong>Objectives: </strong>imPROVE is a new Health Information Technology platform that enables systematic patient-reported outcome measure (PROM) collection through a mobile phone application. The purpose of this study is to describe our initial experience and approach to implementing imPROVE among breast cancer patients treated in breast and plastic surgery clinics.</p><p><strong>Materials and methods: </strong>We describe our initial implementation in 4 phases between June 2021 and February 2022: preimplementation, followed by 3 consecutive implementation periods (P1, P2, P3). The Standards for Reporting Implementation Studies statement guided this study. Iterative Plan-Do-Study-Act (PDSA) cycles supported implementation, and success was evaluated using the Reach, Effectiveness, Adoption, Implementation, and Maintenance framework.</p><p><strong>Results: </strong>Qualitative interviews conducted during the preimplementation phase elicited 4 perceived implementation barriers. Further feedback collected during each phase of implementation resulted in the development of brochures, posters in clinic spaces, and scripts for clinic staff to streamline discussions with patients, and the resolution of technical issues concerning patient login capabilities, such as compatibility with cell phone software and barriers to downloading imPROVE. Feedback also generated ideas for facilitating provider interpretation of PROM results. By the end of P3, 2961 patients were eligible, 1375 (46.4%) downloaded imPROVE, and 1070 (36.1% of those eligible, 78% of those who downloaded) completed at least 1 PROM.</p><p><strong>Discussion and conclusion: </strong>Implementation efforts across 2 surgical departments at 2 academic teaching hospitals enabled collaboration across clinical specialties and longitudinal PROM reporting for patients receiving breast cancer care; the implementation effort also highlighted patient difficulties with mobile app-based PROM collection, particularly around initial engagement.</p>","PeriodicalId":36278,"journal":{"name":"JAMIA Open","volume":"6 4","pages":"ooad108"},"PeriodicalIF":2.1,"publicationDate":"2023-12-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10750814/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139040597","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"The Arbovirus Mapping and Prediction (ArboMAP) System for West Nile Virus Forecasting","authors":"Dawn M. Nekorchuk, Anita Bharadwaja, Sean Simonson, Emma Ortega, Caio M B França, Emily Dinh, Rebecca Reik, Rachel Burkholder, Michael C Wimberly","doi":"10.1093/jamiaopen/ooad110","DOIUrl":"https://doi.org/10.1093/jamiaopen/ooad110","url":null,"abstract":"\u0000 \u0000 \u0000 West Nile virus (WNV) is the most common mosquito-borne disease in the United States. Predicting the location and timing of outbreaks would allow targeting of disease prevention and mosquito control activities. Our objective was to develop software (ArboMAP) for routine WNV forecasting using public health surveillance data and meteorological observations.\u0000 \u0000 \u0000 \u0000 ArboMAP was implemented using an R markdown script for data processing, modelling, and report generation. A Google Earth Engine application was developed to summarize and download weather data. Generalized additive models were used to make county-level predictions of WNV cases.\u0000 \u0000 \u0000 \u0000 ArboMAP minimized the number of manual steps required to make weekly forecasts, generated information that was useful for decision makers, and has been tested and implemented in multiple public health institutions.\u0000 \u0000 \u0000 \u0000 Routine predictions of mosquito-borne disease risk are feasible and can be implemented by public health departments using ArboMAP.\u0000 \u0000 \u0000 \u0000 West Nile virus (WNV) is the most common mosquito-borne disease in the United States. To reduce the risk of WNV, public health agencies distribute information about how to avoid mosquito bites and use insecticides to reduce the abundances of disease-transmitting mosquitoes. Information about when and where the risk of getting WNV is highest would help these agencies to target their activities and use limited resources more efficiently. To support this goal, we developed the ArboMAP software system for predicting the risk of WNV disease in humans. ArboMAP uses information about recent weather combined with data obtained from trapping mosquitoes and testing them for presence of WNV to predict how many human cases that will occur in future weeks. Predictions extend throughout the current WNV season (typically May-September) and are made for each county within a state. The system is implemented as a set of free software tools that can be used by epidemiologists in state and municipal departments of health. Feedback from public health agencies in South Dakota, Louisiana, Oklahoma, and Michigan has been incorporated to enhance the usability of the system and design visualizations that summarize the forecasts.\u0000","PeriodicalId":36278,"journal":{"name":"JAMIA Open","volume":"136 3","pages":""},"PeriodicalIF":2.1,"publicationDate":"2023-12-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138953349","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Automated informatics may increase the detection rate of suspicious cases of human trafficking-a preliminary study.","authors":"David O Duke, Derin Allard, Suzanne Dysart, Keenan O Hogan, Suzanne Phelan, Luke Rawlings, Hanni Stoklosa","doi":"10.1093/jamiaopen/ooad097","DOIUrl":"https://doi.org/10.1093/jamiaopen/ooad097","url":null,"abstract":"<p><strong>Objectives: </strong>Worldwide, there is an estimated 40.3 million victims trapped in modern day slavery, including 24.9 million in forced labor and 15.4 million in forced marriage. A majority of labor and sex trafficking survivors report at least one healthcare encounter during their victimization. An approach to an informatics technology solution for identifying trafficked persons in real time, in the hospital / emergency department settings is the primary focus of this paper.</p><p><strong>Materials and methods: </strong>Octavia, a software application implemented in 3 California hospitals, scanned all patient encounters for social and clinical determinants that are consistent predictors of HT. Any encounter that matched these criteria was forwarded to a specially trained High-Risk Navigator who screened the data and when able, made direct contact in an effort to build rapport and possibly provide victim assistance.</p><p><strong>Results: </strong>During the observation period, the automated scanning of hospital patient encounters resulted in a notable increase in the detection of persons who had a likelihood of being trafficked when compared to a pre-project baseline.</p><p><strong>Discussion: </strong>Our experience demonstrated that automated technology is useful to assist healthcare providers in identification of potentially trafficked persons, improving the likelihood of care provision.</p>","PeriodicalId":36278,"journal":{"name":"JAMIA Open","volume":"6 4","pages":"ooad097"},"PeriodicalIF":2.1,"publicationDate":"2023-12-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10722470/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138809939","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Predicting pediatric emergence delirium using data-driven machine learning applied to electronic health record dataset at a quaternary care pediatric hospital.","authors":"Han Yu, Allan F Simpao, Victor M Ruiz, Olivia Nelson, Wallis T Muhly, Tori N Sutherland, Julia A Gálvez, Mykhailo B Pushkar, Paul A Stricker, Fuchiang Rich Tsui","doi":"10.1093/jamiaopen/ooad106","DOIUrl":"https://doi.org/10.1093/jamiaopen/ooad106","url":null,"abstract":"<p><strong>Objectives: </strong>Pediatric emergence delirium is an undesirable outcome that is understudied. Development of a predictive model is an initial step toward reducing its occurrence. This study aimed to apply machine learning (ML) methods to a large clinical dataset to develop a predictive model for pediatric emergence delirium.</p><p><strong>Materials and methods: </strong>We performed a single-center retrospective cohort study using electronic health record data from February 2015 to December 2019. We built and evaluated 4 commonly used ML models for predicting emergence delirium: least absolute shrinkage and selection operator, ridge regression, random forest, and extreme gradient boosting. The primary outcome was the occurrence of emergence delirium, defined as a Watcha score of 3 or 4 recorded at any time during recovery.</p><p><strong>Results: </strong>The dataset included 54 776 encounters across 43 830 patients. The 4 ML models performed similarly with performance assessed by the area under the receiver operating characteristic curves ranging from 0.74 to 0.75. Notable variables associated with increased risk included adenoidectomy with or without tonsillectomy, decreasing age, midazolam premedication, and ondansetron administration, while intravenous induction and ketorolac were associated with reduced risk of emergence delirium.</p><p><strong>Conclusions: </strong>Four different ML models demonstrated similar performance in predicting postoperative emergence delirium using a large pediatric dataset. The prediction performance of the models draws attention to our incomplete understanding of this phenomenon based on the studied variables. The results from our modeling could serve as a first step in designing a predictive clinical decision support system, but further optimization and validation are needed.</p><p><strong>Clinical trial number and registry url: </strong>Not applicable.</p>","PeriodicalId":36278,"journal":{"name":"JAMIA Open","volume":"6 4","pages":"ooad106"},"PeriodicalIF":2.1,"publicationDate":"2023-12-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10719078/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138809941","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Selecting patient-reported outcome measures for a patient-facing technology.","authors":"Priyank Raj, Youmin Cho, Yun Jiang, Yang Gong","doi":"10.1093/jamiaopen/ooad104","DOIUrl":"10.1093/jamiaopen/ooad104","url":null,"abstract":"<p><strong>Objective: </strong>This article provides insight into our process and considerations for selecting patient-reported outcome measures (PROMs) designed for self-reporting symptoms and quality-of-life among breast cancer (BCA) patients undergoing oral anticancer agent treatment via a patient-facing technology (PFT) platform.</p><p><strong>Methods: </strong>Following established guidelines, we conducted a thorough assessment of a specific set of PROMs, comparing their content to identify the most suitable options for studying BCA patients.</p><p><strong>Results: </strong>We recommend utilizing the combination of EORTC QLQ-C30 + EORTC QLQ-BR45 as the preferred instrument, especially when developing a dedicated \"breast cancer-only\" application.</p><p><strong>Discussion: </strong>When developing and maintaining a dashboard for a PFT platform that includes multiple cancer types, it is important to consider the feasibility of interface design and workload. To achieve this, we recommend using PRO-CTCAE+PROMIS 10 GH for the PFT. Moreover, it is important to consider adding ad hoc items to complement the chosen PROM(s).</p><p><strong>Conclusion: </strong>This article describes our efforts to identify PROMs for self-reported data while considering patient and developer burdens, providing guidance to PFT developers facing similar challenges in PROM selection.</p>","PeriodicalId":36278,"journal":{"name":"JAMIA Open","volume":"6 4","pages":"ooad104"},"PeriodicalIF":2.1,"publicationDate":"2023-12-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10719077/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138809943","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"The use of electronic health record embedded MRC-ICU as a metric for critical care pharmacist workload.","authors":"Andrew J Webb, Bayleigh Carver, Sandra Rowe, Andrea Sikora","doi":"10.1093/jamiaopen/ooad101","DOIUrl":"10.1093/jamiaopen/ooad101","url":null,"abstract":"<p><strong>Objectives: </strong>A lack of pharmacist-specific risk-stratification scores in the electronic health record (EHR) may limit resource optimization. The medication regimen complexity-intensive care unit (MRC-ICU) score was implemented into our center's EHR for use by clinical pharmacists. The purpose of this evaluation was to evaluate MRC-ICU as a predictor of pharmacist workload and to assess its potential as an additional dimension to traditional workload measures.</p><p><strong>Materials and methods: </strong>Data were abstracted from the EHR on adult ICU patients, including MRC-ICU scores and 2 traditional measures of pharmacist workload: numbers of medication orders verified and interventions logged. This was a single-center study of an EHR-integrated MRC-ICU tool. The primary outcome was the association of MRC-ICU with institutional metrics of pharmacist workload. Associations were assessed using the initial 24-h maximum MRC-ICU score's Pearson's correlation with overall admission workload and the day-to-day association using generalized linear mixed-effects modeling.</p><p><strong>Results: </strong>A total of 1205 patients over 5083 patient-days were evaluated. Baseline MRC-ICU was correlated with both cumulative order volume (Spearman's rho 0.41, <i>P</i> < .001) and cumulative interventions placed (Spearman's rho 0.27, <i>P</i> < .001). A 1-point increase in maximum daily MRC-ICU was associated with a 31% increase in order volume (95% CI, 24%-38%) and 4% increase in interventions (95% CI, 2%-5%).</p><p><strong>Discussion and conclusion: </strong>The MRC-ICU is a validated score that has been previously correlated with important patient-centered outcomes. Here, MRC-ICU was modestly associated with 2 traditional objective measures of pharmacist workload, including orders verified and interventions placed, which is an important step for its use as a tool for resource utilization needs.</p>","PeriodicalId":36278,"journal":{"name":"JAMIA Open","volume":"6 4","pages":"ooad101"},"PeriodicalIF":2.1,"publicationDate":"2023-12-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10697785/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138499661","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Transforming Estonian health data to the Observational Medical Outcomes Partnership (OMOP) Common Data Model: lessons learned.","authors":"Marek Oja, Sirli Tamm, Kerli Mooses, Maarja Pajusalu, Harry-Anton Talvik, Anne Ott, Marianna Laht, Maria Malk, Marcus Lõo, Johannes Holm, Markus Haug, Hendrik Šuvalov, Dage Särg, Jaak Vilo, Sven Laur, Raivo Kolde, Sulev Reisberg","doi":"10.1093/jamiaopen/ooad100","DOIUrl":"10.1093/jamiaopen/ooad100","url":null,"abstract":"<p><strong>Objective: </strong>To describe the reusable transformation process of electronic health records (EHR), claims, and prescriptions data into Observational Medical Outcome Partnership (OMOP) Common Data Model (CDM), together with challenges faced and solutions implemented.</p><p><strong>Materials and methods: </strong>We used Estonian national health databases that store almost all residents' claims, prescriptions, and EHR records. To develop and demonstrate the transformation process of Estonian health data to OMOP CDM, we used a 10% random sample of the Estonian population (<i>n</i> = 150 824 patients) from 2012 to 2019 (MAITT dataset). For the sample, complete information from all 3 databases was converted to OMOP CDM version 5.3. The validation was performed using open-source tools.</p><p><strong>Results: </strong>In total, we transformed over 100 million entries to standard concepts using standard OMOP vocabularies with the average mapping rate 95%. For conditions, observations, drugs, and measurements, the mapping rate was over 90%. In most cases, SNOMED Clinical Terms were used as the target vocabulary.</p><p><strong>Discussion: </strong>During the transformation process, we encountered several challenges, which are described in detail with concrete examples and solutions.</p><p><strong>Conclusion: </strong>For a representative 10% random sample, we successfully transferred complete records from 3 national health databases to OMOP CDM and created a reusable transformation process. Our work helps future researchers to transform linked databases into OMOP CDM more efficiently, ultimately leading to better real-world evidence.</p>","PeriodicalId":36278,"journal":{"name":"JAMIA Open","volume":"6 4","pages":"ooad100"},"PeriodicalIF":2.1,"publicationDate":"2023-12-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10697784/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138499662","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Development of an interactive dashboard for gun violence pattern analysis and intervention design at the local level","authors":"Rashaud Senior, Lisa Pickett, Andrew Stirling, Shwetha Dash, Patti Gorgone, Georgina Durst, Debra Jones, Richard Shannon, N. Bhavsar, Armando Bedoya","doi":"10.1093/jamiaopen/ooad105","DOIUrl":"https://doi.org/10.1093/jamiaopen/ooad105","url":null,"abstract":"Abstract Introduction Gun violence remains a concerning and persistent issue in our country. Novel dashboards may integrate and summarize important clinical and non-clinical data that can inform targeted interventions to address the underlying causes of gun violence. Methods Data from various clinical and non-clinical sources were sourced, cleaned, and integrated into a customizable dashboard that summarizes and provides insight into the underlying factors that impact local gun violence episodes. Results The dashboards contained data from 7786 encounters and 1152 distinct patients from our Emergency Department’s Trauma Registry with various patterns noted by the team. A multidisciplinary executive team, including subject matter experts in community-based interventions, epidemiology, and social sciences, was formed to design targeted interventions based on these observations. Conclusion Targeted interventions to reduce gun violence require a multimodal data sourcing and standardization approach, the inclusion of neighborhood-level data, and a dedicated multidisciplinary team to act on the generated insights.","PeriodicalId":36278,"journal":{"name":"JAMIA Open","volume":" 16","pages":""},"PeriodicalIF":2.1,"publicationDate":"2023-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138611536","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Development and external validation of multimodal postoperative acute kidney injury risk machine learning models","authors":"G. Karway, J. Koyner, John Caskey, Alexandra B Spicer, Kyle A. Carey, Emily R. Gilbert, D. Dligach, A. Mayampurath, Majid Afshar, M. Churpek","doi":"10.1093/jamiaopen/ooad109","DOIUrl":"https://doi.org/10.1093/jamiaopen/ooad109","url":null,"abstract":"Abstract Objectives To develop and externally validate machine learning models using structured and unstructured electronic health record data to predict postoperative acute kidney injury (AKI) across inpatient settings. Materials and Methods Data for adult postoperative admissions to the Loyola University Medical Center (2009-2017) were used for model development and admissions to the University of Wisconsin-Madison (2009-2020) were used for validation. Structured features included demographics, vital signs, laboratory results, and nurse-documented scores. Unstructured text from clinical notes were converted into concept unique identifiers (CUIs) using the clinical Text Analysis and Knowledge Extraction System. The primary outcome was the development of Kidney Disease Improvement Global Outcomes stage 2 AKI within 7 days after leaving the operating room. We derived unimodal extreme gradient boosting machines (XGBoost) and elastic net logistic regression (GLMNET) models using structured-only data and multimodal models combining structured data with CUI features. Model comparison was performed using the receiver operating characteristic curve (AUROC), with Delong’s test for statistical differences. Results The study cohort included 138 389 adult patient admissions (mean [SD] age 58 [16] years; 11 506 [8%] African-American; and 70 826 [51%] female) across the 2 sites. Of those, 2959 (2.1%) developed stage 2 AKI or higher. Across all data types, XGBoost outperformed GLMNET (mean AUROC 0.81 [95% confidence interval (CI), 0.80-0.82] vs 0.78 [95% CI, 0.77-0.79]). The multimodal XGBoost model incorporating CUIs parameterized as term frequency-inverse document frequency (TF-IDF) showed the highest discrimination performance (AUROC 0.82 [95% CI, 0.81-0.83]) over unimodal models (AUROC 0.79 [95% CI, 0.78-0.80]). Discussion A multimodality approach with structured data and TF-IDF weighting of CUIs increased model performance over structured data-only models. Conclusion These findings highlight the predictive power of CUIs when merged with structured data for clinical prediction models, which may improve the detection of postoperative AKI.","PeriodicalId":36278,"journal":{"name":"JAMIA Open","volume":"60 24","pages":""},"PeriodicalIF":2.1,"publicationDate":"2023-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138985926","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}