Dataset Information

Risk of bias in studies on prediction models developed using supervised machine learning techniques: systematic review.

ABSTRACT: To assess the methodological quality of studies on prediction models developed using machine learning techniques across all medical specialties. Systematic review. PubMed from 1 January 2018 to 31 December 2019. Articles reporting on the development, with or without external validation, of a multivariable prediction model (diagnostic or prognostic) developed using supervised machine learning for individualised predictions. No restrictions applied for study design, data source, or predicted patient related health outcomes. Methodological quality of the studies was determined and risk of bias evaluated using the prediction risk of bias assessment tool (PROBAST). This tool contains 21 signalling questions tailored to identify potential biases in four domains. Risk of bias was measured for each domain (participants, predictors, outcome, and analysis) and each study (overall). 152 studies were included: 58 (38%) included a diagnostic prediction model and 94 (62%) a prognostic prediction model. PROBAST was applied to 152 developed models and 19 external validations. Of these 171 analyses, 148 (87%, 95% confidence interval 81% to 91%) were rated at high risk of bias. The analysis domain was most frequently rated at high risk of bias. Of the 152 models, 85 (56%, 48% to 64%) were developed with an inadequate number of events per candidate predictor, 62 handled missing data inadequately (41%, 33% to 49%), and 59 assessed overfitting improperly (39%, 31% to 47%). Most models used appropriate data sources to develop (73%, 66% to 79%) and externally validate the machine learning based prediction models (74%, 51% to 88%). Information about blinding of outcome and blinding of predictors was, however, absent in 60 (40%, 32% to 47%) and 79 (52%, 44% to 60%) of the developed models, respectively. Most studies on machine learning based prediction models show poor methodological quality and are at high risk of bias. Factors contributing to risk of bias include small study size, poor handling of missing data, and failure to deal with overfitting. Efforts to improve the design, conduct, reporting, and validation of such studies are necessary to boost the application of machine learning based prediction models in clinical practice. PROSPERO CRD42019161764.

SUBMITTER: Andaur Navarro CL

PROVIDER: S-EPMC8527348 | biostudies-literature |

REPOSITORIES: biostudies-literature

ACCESS DATA

Similar Datasets

Project description:BackgroundWhile many studies have consistently found incomplete reporting of regression-based prediction model studies, evidence is lacking for machine learning-based prediction model studies. We aim to systematically review the adherence of Machine Learning (ML)-based prediction model studies to the Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) Statement.MethodsWe included articles reporting on development or external validation of a multivariable prediction model (either diagnostic or prognostic) developed using supervised ML for individualized predictions across all medical fields. We searched PubMed from 1 January 2018 to 31 December 2019. Data extraction was performed using the 22-item checklist for reporting of prediction model studies ( www.TRIPOD-statement.org ). We measured the overall adherence per article and per TRIPOD item.ResultsOur search identified 24,814 articles, of which 152 articles were included: 94 (61.8%) prognostic and 58 (38.2%) diagnostic prediction model studies. Overall, articles adhered to a median of 38.7% (IQR 31.0-46.4%) of TRIPOD items. No article fully adhered to complete reporting of the abstract and very few reported the flow of participants (3.9%, 95% CI 1.8 to 8.3), appropriate title (4.6%, 95% CI 2.2 to 9.2), blinding of predictors (4.6%, 95% CI 2.2 to 9.2), model specification (5.2%, 95% CI 2.4 to 10.8), and model's predictive performance (5.9%, 95% CI 3.1 to 10.9). There was often complete reporting of source of data (98.0%, 95% CI 94.4 to 99.3) and interpretation of the results (94.7%, 95% CI 90.0 to 97.3).ConclusionSimilar to prediction model studies developed using conventional regression-based techniques, the completeness of reporting is poor. Essential information to decide to use the model (i.e. model specification and its performance) is rarely reported. However, some items and sub-items of TRIPOD might be less suitable for ML-based prediction model studies and thus, TRIPOD requires extensions. Overall, there is an urgent need to improve the reporting quality and usability of research to avoid research waste.Systematic review registrationPROSPERO, CRD42019161764.

Project description:BackgroundPredicting survival of recipients after liver transplantation is regarded as one of the most important challenges in contemporary medicine. Hence, improving on current prediction models is of great interest.Nowadays, there is a strong discussion in the medical field about machine learning (ML) and whether it has greater potential than traditional regression models when dealing with complex data. Criticism to ML is related to unsuitable performance measures and lack of interpretability which is important for clinicians.MethodsIn this paper, ML techniques such as random forests and neural networks are applied to large data of 62294 patients from the United States with 97 predictors selected on clinical/statistical grounds, over more than 600, to predict survival from transplantation. Of particular interest is also the identification of potential risk factors. A comparison is performed between 3 different Cox models (with all variables, backward selection and LASSO) and 3 machine learning techniques: a random survival forest and 2 partial logistic artificial neural networks (PLANNs). For PLANNs, novel extensions to their original specification are tested. Emphasis is given on the advantages and pitfalls of each method and on the interpretability of the ML techniques.ResultsWell-established predictive measures are employed from the survival field (C-index, Brier score and Integrated Brier Score) and the strongest prognostic factors are identified for each model. Clinical endpoint is overall graft-survival defined as the time between transplantation and the date of graft-failure or death. The random survival forest shows slightly better predictive performance than Cox models based on the C-index. Neural networks show better performance than both Cox models and random survival forest based on the Integrated Brier Score at 10 years.ConclusionIn this work, it is shown that machine learning techniques can be a useful tool for both prediction and interpretation in the survival context. From the ML techniques examined here, PLANN with 1 hidden layer predicts survival probabilities the most accurately, being as calibrated as the Cox model with all variables.Trial registrationRetrospective data were provided by the Scientific Registry of Transplant Recipients under Data Use Agreement number 9477 for analysis of risk factors after liver transplantation.

Project description:BACKGROUND:Diabetes Mellitus is an increasingly prevalent chronic disease characterized by the body's inability to metabolize glucose. The objective of this study was to build an effective predictive model with high sensitivity and selectivity to better identify Canadian patients at risk of having Diabetes Mellitus based on patient demographic data and the laboratory results during their visits to medical facilities. METHODS:Using the most recent records of 13,309 Canadian patients aged between 18 and 90 years, along with their laboratory information (age, sex, fasting blood glucose, body mass index, high-density lipoprotein, triglycerides, blood pressure, and low-density lipoprotein), we built predictive models using Logistic Regression and Gradient Boosting Machine (GBM) techniques. The area under the receiver operating characteristic curve (AROC) was used to evaluate the discriminatory capability of these models. We used the adjusted threshold method and the class weight method to improve sensitivity - the proportion of Diabetes Mellitus patients correctly predicted by the model. We also compared these models to other learning machine techniques such as Decision Tree and Random Forest. RESULTS:The AROC for the proposed GBM model is 84.7% with a sensitivity of 71.6% and the AROC for the proposed Logistic Regression model is 84.0% with a sensitivity of 73.4%. The GBM and Logistic Regression models perform better than the Random Forest and Decision Tree models. CONCLUSIONS:The ability of our model to predict patients with Diabetes using some commonly used lab results is high with satisfactory sensitivity. These models can be built into an online computer program to help physicians in predicting patients with future occurrence of diabetes and providing necessary preventive interventions. The model is developed and validated on the Canadian population which is more specific and powerful to apply on Canadian patients than existing models developed from US or other populations. Fasting blood glucose, body mass index, high-density lipoprotein, and triglycerides were the most important predictors in these models.

Project description:BackgroundThe advent of metagenomic sequencing provides microbial abundance patterns that can be leveraged for sample origin prediction. Supervised machine learning classification approaches have been reported to predict sample origin accurately when the origin has been previously sampled. Using metagenomic datasets provided by the 2019 CAMDA challenge, we evaluated the influence of variable technical, analytical and machine learning approaches for result interpretation and novel source prediction.ResultsComparison between 16S rRNA amplicon and shotgun sequencing approaches as well as metagenomic analytical tools showed differences in normalized microbial abundance, especially for organisms present at low abundance. Shotgun sequence data analyzed using Kraken2 and Bracken, for taxonomic annotation, had higher detection sensitivity. As classification models are limited to labeling pre-trained origins, we took an alternative approach using Lasso-regularized multivariate regression to predict geographic coordinates for comparison. In both models, the prediction errors were much higher in Leave-1-city-out than in 10-fold cross validation, of which the former realistically forecasted the increased difficulty in accurately predicting samples from new origins. This challenge was further confirmed when applying the model to a set of samples obtained from new origins. Overall, the prediction performance of the regression and classification models, as measured by mean squared error, were comparable on mystery samples. Due to higher prediction error rates for samples from new origins, we provided an additional strategy based on prediction ambiguity to infer whether a sample is from a new origin. Lastly, we report increased prediction error when data from different sequencing protocols were included as training data.ConclusionsHerein, we highlight the capacity of predicting sample origin accurately with pre-trained origins and the challenge of predicting new origins through both regression and classification models. Overall, this work provides a summary of the impact of sequencing technique, protocol, taxonomic analytical approaches, and machine learning approaches on the use of metagenomics for prediction of sample origin.

Dataset Information

Risk of bias in studies on prediction models developed using supervised machine learning techniques: systematic review.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets