Support vector machine versus logistic regression modeling for prediction of hospital mortality in critically ill patients with haematological malignancies

Verplancke, T; Looy, Stijn Van; Benoît, Dominique; Vansteelandt, Stijn; Depuydt, P; Turck, Filip De; Decruyenaere, Johan

doi:10.1186/1472-6947-8-56

Cited by 101 publications

(78 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…SVM methods have been increasingly used in a wide variety of medical classification problems. In certain instances they can prove superior in terms of classification accuracy to standard methods such as logistic regression, especially in being able to extract key predictors [18][19][20][21] that can then be used in the simplified algorithm.…”

Section: Discussionmentioning

confidence: 99%

See 1 more Smart Citation

A screening algorithm for HIV‐associated neurocognitive disorders

et al. 2010

View full text Add to dashboard Cite

BackgroundHIV physicians have limited time for cognitive screening. Here we developed an extra-brief, clinically based tool for predicting HIV-associated neurocognitive impairment (HAND) in order to determine which HIV-positive individuals require a more comprehensive neurological/ neuropsychological (NP) assessment. MethodsNinety-seven HIV-positive individuals with advanced disease recruited in an HIV out-patient clinic received standard NP testing. A screening algorithm was developed using support vector machines, an optimized prediction procedure for classifying individuals into two groups (here NP-impaired and NP-normal) based on a set of predictors. ResultsThe final algorithm utilized age, current CD4 cell count, past central nervous system HIV-related diseases and current treatment duration and required approximately 3 min to complete, with a good overall prediction accuracy of 78% (against the gold standard; NP-impairment status derived from standard NP testing) and a good specificity of 70%. ConclusionThis noncognitive-based algorithm should prove useful to identify HIV-infected patients with advanced disease at high risk of HAND who require more formal assessment. We propose staged guidelines, using the algorithm, for improved HAND therapeutic management. Future larger, international studies are planned to test the predictive effect of nadir CD4 cell count, hepatitis C virus infection, gender, ethnicity and HIV viral clade. We recommend the use of this first version for HIV-infected Caucasian men with advanced disease.Keywords: antiretroviral therapy, diagnosis algorithm, HIV/AIDS, HIV-associated neurocognitive disorder, mathematical model Accepted 21 January 2010 IntroductionThe clinical management of HIV-positive persons is increasingly complicated in the era of combination antiretroviral therapy (CART). One aspect of management that requires extensive training relates to the early identification of neurocognitive complications of HIV infection. In many countries the general practitioner or the HIV physician is often the primary patient's interlocutor [1]. Without specific screening using procedures that are still relatively lengthy or require specific training, especially for interpretation [2], physicians may sometimes overlook patients in need of further neurological evaluation.In the CART era, the prevalence of neurocognitive impairment remains high (up to 50% [3]) and HIVassociated neurocognitive disorder (HAND) has shifted towards a milder clinical presentation [4]. Such a mild clinical presentation can escape detection without formal neurological assessment and neuropsychological testing [5]. HAND, even in its mild form, is independently predictive of death [6] The development of the screening algorithm was based on support vector machine (SVM) methodology. Because the aim of our study was to provide a simplified algorithm from a complex set of predictors, SVM was the most appropriate procedure [18]. The SVM has been shown to be extremely robust in solving prediction problems while handling la...

show abstract

Section: Discussionmentioning

confidence: 99%

“…The SVM has been shown to be extremely robust in solving prediction problems while handling large sets of predictors [18]. It is an alternative to more standard statistical techniques such as logistic regression and in certain situations has been found to be superior to logistic regression for finding a robust fit with fewer predictors [18][19][20][21]. …”

mentioning

confidence: 99%

A screening algorithm for HIV‐associated neurocognitive disorders

et al. 2010

View full text Add to dashboard Cite

show abstract

“…Moreover, the classes are separated by a gap (margin) that is as wide as possible and which is defined by support vectors. 84 Verplancke et al 85 compared the use of support vector machines with logistic regression for the prediction of hospital mortalities among patients with haematological malignancies, but they did not find the discrimination of support vector machines to be statistically better than that of logistic regression. In contrast, Van Looy et al 86 compared the ability of support vector machines to predict tacrolimus blood concentrations with that of linear regression and found the support vector machines to be significantly better.…”

Section: Support Vector Machinesmentioning

confidence: 99%

Development and validation of a risk model for identification of non-neutropenic, critically ill adult patients at high risk of invasive Candida infection: the Fungal Infection Risk Evaluation (FIRE) Study

Harrison

Muskett

Harvey

et al. 2013

Health Technol Assess

View full text Add to dashboard Cite

BackgroundThere is increasing evidence that invasive fungal disease (IFD) is more likely to occur in non-neutropenic patients in critical care units. A number of randomised controlled trials (RCTs) have evaluated antifungal prophylaxis in non-neutropenic, critically ill patients, demonstrating a reduction in the risk of proven IFD and suggesting a reduction in mortality. It is necessary to establish a method to identify and target antifungal prophylaxis at those patients at highest risk of IFD, who stand to benefit most from any antifungal prophylaxis strategy.ObjectivesTo develop and validate risk models to identify non-neutropenic, critically ill adult patients at high risk of invasiveCandidainfection, who would benefit from antifungal prophylaxis, and to assess the cost-effectiveness of targeting antifungal prophylaxis to high-risk patients based on these models.DesignSystematic review, prospective data collection, statistical modelling, economic decision modelling and value of information analysis.SettingNinety-six UK adult general critical care units.ParticipantsConsecutive admissions to participating critical care units.InterventionsNone.Main outcome measuresInvasive fungal disease, defined as a blood culture or sample from a normally sterile site showing yeast/mould cells in a microbiological or histopathological report. For statistical and economic modelling, the primary outcome was invasiveCandidainfection, defined as IFD-positive forCandidaspecies.ResultsSystematic review: Thirteen articles exploring risk factors, risk models or clinical decision rules for IFD in critically ill adult patients were identified. Risk factors reported to be significantly associated with IFD were included in the final data set for the prospective data collection.Data collection: Data were collected on 60,778 admissions between July 2009 and March 2011. Overall, 383 patients (0.6%) were admitted with or developed IFD. The majority of IFD patients (94%) were positive forCandidaspecies. The most common site of infection was blood (55%). The incidence of IFD identified in unit was 4.7 cases per 1000 admissions, and for unit-acquired IFD was 3.2 cases per 1000 admissions.Statistical modelling: Risk models were developed at admission to the critical care unit, 24 hours and the end of calendar day 3. The risk model at admission had fair discrimination (c-index 0.705). Discrimination improved at 24 hours (c-index 0.823) and this was maintained at the end of calendar day 3 (c-index 0.835). There was a drop in model performance in the validation sample.Economic decision model: Irrespective of risk threshold, incremental quality-adjusted life-years of prophylaxis strategies compared with current practice were positive but small. Incremental costs of the prophylaxis strategies compared with current practice were positive for most strategies, although a few strategies were cost saving. Incremental net benefits of each prophylaxis strategy compared with current practice were positive for most, but not all, of the strategies. Cost-effectiveness acceptability curves showed that risk assessment and prophylaxis at the end of calendar day 3 was the strategy most likely to be cost-effective when the risk threshold was 1% or 2%. At a lower risk threshold (0.5%) it was most cost-effective to assess risk at each time point; this led to a relatively high proportion of patients receiving antifungal prophylaxis (30%), which may lead to additional burden from increased resistance. The estimates of cost-effectiveness were highly uncertain and the value of further research for the whole population of interest is high relative to the research costs.ConclusionsThe results of the Fungal Infection Risk Evaluation (FIRE) Study, derived from a highly representative sample of adult general critical care units across the UK, indicated a low incidence of IFD among non-neutropenic, critically ill adult patients. IFD was associated with substantially higher mortality, more intensive organ support and longer length of stay. Risk modelling produced simple risk models that provided acceptable discrimination for identifying patients at ‘high risk’ of invasiveCandidainfection. Results of the economic model suggested that the current most cost-effective treatment strategy among non-neutropenic, critically ill adult patients admitted to NHS adult general critical care units is a strategy of risk assessment and antifungal prophylaxis at the end of calendar day 3, but this finding is highly uncertain and future studies should consider the potential impact of increased resistance.FundingFunding for this study was provided by the Health Technology Assessment programme of the National Institute for Health Research.A previous version of this report was published in February 2013. The report was subsequently modified to reflect a substantial reduction in the unit cost of fluconazole that took place between the original analysis being conducted and the publication of the report.

show abstract

“…The second comparison in Table I is between the transfer learning approaches and logistic regression (commonly used in clinical decision making and denoted by LR) [10], 1-class SVM, 2-class SVM, and 2-class SVM classification with cost-sensitive weighting. Despite variations in the frequencies of outcomes and the numbers of patients in the training set, the combined transfer approach outperformed all the other methods in 7 out of the total 10 endpoints of interests.…”

Section: Resultsmentioning

confidence: 99%

Improving surgical models through one/two class learning

Chia

Karam

Lee

et al. 2012

2012 Annual International Conference of the IEEE Engineering in Medicine and Biology Society

View full text Add to dashboard Cite

Abstract-Only a minority of patients undergoing in-patient surgical procedures experience complications. However, the large number of in-patient surgeries (over 48 million procedures each year in the U.S.) results in substantial overall mortality and morbidity due to these complications. This burden can be decreased through improvements in the ability to evaluate patients by the bedside, and to assess surgical quality and outcomes across hospitals. Unfortunately, the process of developing clinical models for surgical complications is made challenging by the availability of generally small datasets for model training, and by class imbalance due to the diminished prevalence of many important complications. In this paper, we address this issue and explore the idea of jointly leveraging the benefits of both supervised and unsupervised learning to model surgical complications that occur infrequently. In particular, we study an approach where the problems of supervised and unsupervised model development are treated as tasks that can be transferred. Focussing this work on support vector machine (SVM) classification, we describe a transfer learning algorithm that improves performance relative to both supervised (i.e., binary or 2-class SVM) and unsupervised (i.e., 1-class SVM) methods, as well as the use of cost-sensitive weighting techniques, for predicting different surgical complications within the American College of Surgeons National Surgical Quality Improvement Program registry.

show abstract

Support vector machine versus logistic regression modeling for prediction of hospital mortality in critically ill patients with haematological malignancies

Cited by 101 publications

References 20 publications

A screening algorithm for HIV‐associated neurocognitive disorders

A screening algorithm for HIV‐associated neurocognitive disorders

Development and validation of a risk model for identification of non-neutropenic, critically ill adult patients at high risk of invasive Candida infection: the Fungal Infection Risk Evaluation (FIRE) Study

Improving surgical models through one/two class learning

Contact Info

Product

Resources

About