Predicting colorectal surgical complications using heterogeneous clinical data and kernel methods

Soguero-Ruíz, Cristina; Hindberg, Kristian; Mora-Jiménez, Inmaculada; Rojo-Álvarez, José Luis; Skrøvseth, Stein Olav; Godtliebsen, Fred; Mortensen, Kim Erlend; Revhaug, Arthur; Lindsetmo, Rolv-Ole; Augestad, Knut Magne; Jenssen, Robert

doi:10.1016/j.jbi.2016.03.008

Cited by 61 publications

(49 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Although the target of a study by Shahn et al does not include readmission classification, but focuses on predicting strokes in patients with prior diagnoses of Atrial Fibrillation, it demonstrates the importance of temporal information to achieve meaningful improvements in predictive performance. Similarily, the use of temporal information from electronic health records (EHRs) for prediction of Anastomosis Leakage was demonstrated in a study by Soguero-Ruiz et al (2016). The predictive performance gain was proven when combining the data from heterogenus data sources (extracted free text, blood samples values, and patient vital signs).…”

Section: Introductionmentioning

confidence: 81%

Contribution of temporal data to predictive performance in 30-day readmission of morbidly obese patients

Bržan

Obradović

Štiglic

2017

PeerJ

View full text Add to dashboard Cite

BackgroundReduction of readmissions after discharge represents an important challenge for many hospitals and has attracted the interest of many researchers in the past few years. Most of the studies in this field focus on building cross-sectional predictive models that aim to predict the occurrence of readmission within 30-days based on information from the current hospitalization. The aim of this study is demonstration of predictive performance gain obtained by inclusion of information from historical hospitalization records among morbidly obese patients.MethodsThe California Statewide inpatient database was used to build regularized logistic regression models for prediction of readmission in morbidly obese patients (n = 18,881). Temporal features were extracted from historical patient hospitalization records in a one-year timeframe. Five different datasets of patients were prepared based on the number of available hospitalizations per patient. Sample size of the five datasets ranged from 4,787 patients with more than five hospitalizations to 20,521 patients with at least two hospitalization records in one year. A 10-fold cross validation was repeted 100 times to assess the variability of the results. Additionally, random forest and extreme gradient boosting were used to confirm the results.ResultsArea under the ROC curve increased significantly when including information from up to three historical records on all datasets. The inclusion of more than three historical records was not efficient. Similar results can be observed for Brier score and PPV value. The number of selected predictors corresponded to the complexity of the dataset ranging from an average of 29.50 selected features on the smallest dataset to 184.96 on the largest dataset based on 100 repetitions of 10-fold cross-validation.DiscussionThe results show positive influence of adding information from historical hospitalization records on predictive performance using all predictive modeling techniques used in this study. We can conclude that it is advantageous to build separate readmission prediction models in subgroups of patients with more hospital admissions by aggregating information from up to three previous hospitalizations.

show abstract

Section: Introductionmentioning

confidence: 81%

Contribution of temporal data to predictive performance in 30-day readmission of morbidly obese patients

Bržan

Obradović

Štiglic

2017

PeerJ

View full text Add to dashboard Cite

show abstract

“…Time series analysis is an important and mature research topic, especially in the context of univariate time series (UTS) prediction [1,2,3,4]. The field tackles real world problems in many different areas such as energy consumption [5], climate studies [6], biology [7], medicine [8,9,10] and finance [11]. However, the need for analysis of multivariate time series (MTS) [12] is growing in modern society as data is increasingly collected simultaneously from multiple sources over time, often plagued by severe missing data problems [13,14].…”

Section: Introductionmentioning

confidence: 99%

Time series cluster kernel for learning similarities between multivariate time series with missing data

Mikalsen

Bianchi

Soguero-Ruíz

et al. 2018

Pattern Recognition

Self Cite

View full text Add to dashboard Cite

Similarity-based approaches represent a promising direction for time series analysis. However, many such methods rely on parameter tuning, and some have shortcomings if the time series are multivariate (MTS), due to dependencies between attributes, or the time series contain missing data. In this paper, we address these challenges within the powerful context of kernel methods by proposing the robust time series cluster kernel (TCK). The approach taken leverages the missing data handling properties of Gaussian mixture models (GMM) augmented with informative prior distributions. An ensemble learning approach is exploited to ensure robustness to parameters by combining the clustering results of many GMM to form the final kernel.We evaluate the TCK on synthetic and real data and compare to other state-of-the-art techniques. The experimental results demonstrate that the TCK is robust to parameter choices, provides competitive results for MTS without missing data and outstanding results for missing data.

show abstract

“…Health care claims (e.g., diagnoses, procedures, drugs, and clinical variables) provide a wealth of diverse information pertaining to a patient’s clinical history [8910]. However the heterogeneity and volume of this data have only limited utility with conventional analyses.…”

Section: Introductionmentioning

confidence: 99%

Predicting Future Elective Colon Resection for Diverticulitis Using Patterns of Health Care Utilization

Thornblade

Flum

Flaxman

2018

eGEMs

View full text Add to dashboard Cite

Background:Recurrent diverticulitis is the most common reason for elective colon surgery and, although professional societies now recommend against early resection, its use continues to rise. Shared decision making decreases use of low-value surgery but identifying which patients are most likely to elect surgery has proven difficult. We hypothesized that Machine Learning algorithms using health care utilization (HCU) data can predict future clinical events including early resection for diverticulitis.Study Design:We developed models for predicting future surgery among patients with new diagnoses of diverticulitis (2009–2012) from the MarketScan® database. Claims data (diagnosis, procedural, and drug codes) were used to train three Machine Learning algorithms to predict surgery occurring between 52 and 104 weeks following diagnosis.Results:Of 82,231 patients with incident diverticulitis (age 51 ± 8 years, 52% female), 1.2% went on to elective colon resection. Using maximal training data (152 consecutive weeks of claims), the Gradient Boosting Machine model predicted elective surgery with an area under the curve (AUC) of 75% (95% uncertainty interval [UI] 71–79%). Models trained on less data resulted in less accurate prediction (AUC: 68% [64–74%] using 128 weeks, 57% [53–63%] using 104 weeks). The majority of resections (85%) were identified as low-value.Conclusion:By applying Machine Learning to HCU data from the time around a diagnosis of diverticulitis, we predicted elective surgery weeks to months in advance, with moderate accuracy. Identifying patients who are most likely to elect surgery for diverticulitis provides an opportunity for effective shared decision making initiatives aimed at reducing the use of costly low-value care.

show abstract

Predicting colorectal surgical complications using heterogeneous clinical data and kernel methods

Cited by 61 publications

References 27 publications

Contribution of temporal data to predictive performance in 30-day readmission of morbidly obese patients

Contribution of temporal data to predictive performance in 30-day readmission of morbidly obese patients

Time series cluster kernel for learning similarities between multivariate time series with missing data

Predicting Future Elective Colon Resection for Diverticulitis Using Patterns of Health Care Utilization

Contact Info

Product

Resources

About