Dataset Information

Harnessing repeated measurements of predictor variables for clinical risk prediction: a review of existing methods.

ABSTRACT:

Background

Clinical prediction models (CPMs) predict the risk of health outcomes for individual patients. The majority of existing CPMs only harness cross-sectional patient information. Incorporating repeated measurements, such as those stored in electronic health records, into CPMs may provide an opportunity to enhance their performance. However, the number and complexity of methodological approaches available could make it difficult for researchers to explore this opportunity. Our objective was to review the literature and summarise existing approaches for harnessing repeated measurements of predictor variables in CPMs, primarily to make this field more accessible for applied researchers.

Methods

MEDLINE, Embase and Web of Science were searched for articles reporting the development of a multivariable CPM for individual-level prediction of future binary or time-to-event outcomes and modelling repeated measurements of at least one predictor. Information was extracted on the following: the methodology used, its specific aim, reported advantages and limitations, and software available to apply the method.

Results

The search revealed 217 relevant articles. Seven methodological frameworks were identified: time-dependent covariate modelling, generalised estimating equations, landmark analysis, two-stage modelling, joint-modelling, trajectory classification and machine learning. Each of these frameworks satisfies at least one of three aims: to better represent the predictor-outcome relationship over time, to infer a covariate value at a pre-specified time and to account for the effect of covariate change.

Conclusions

The applicability of identified methods depends on the motivation for including longitudinal information and the method's compatibility with the clinical context and available patient data, for both model development and risk estimation in practice.

SUBMITTER: Bull LM

PROVIDER: S-EPMC7346415 | biostudies-literature | 2020

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Harnessing repeated measurements of predictor variables for clinical risk prediction: a review of existing methods.

Bull Lucy M LM Lunt Mark M Martin Glen P GP Hyrich Kimme K Sergeant Jamie C JC

Diagnostic and prognostic research 20200709

<h4>Background</h4>Clinical prediction models (CPMs) predict the risk of health outcomes for individual patients. The majority of existing CPMs only harness cross-sectional patient information. Incorporating repeated measurements, such as those stored in electronic health records, into CPMs may provide an opportunity to enhance their performance. However, the number and complexity of methodological approaches available could make it difficult for researchers to explore this opportunity. Our obje ...[more]

PMID: 32671229

Similar Datasets

Project description:Methane (CH4) emissions from ruminants are of a significant environmental concern, necessitating accurate prediction for emission inventories. Existing models rely solely on dietary and host animal-related data, ignoring the predicting power of rumen microbiota, the source of CH4. To address this limitation, we developed novel CH4 prediction models incorporating rumen microbes as predictors, alongside animal- and feed-related predictors using four statistical/machine learning (ML) methods. These include random forest combined with boosting (RF-B), least absolute shrinkage and selection operator (LASSO), generalized linear mixed model with LASSO (glmmLasso), and smoothly clipped absolute deviation (SCAD) implemented on linear mixed models. With a sheep dataset (218 observations) of both animal data and rumen microbiota data (relative sequence abundance of 330 genera of rumen bacteria, archaea, protozoa, and fungi), we developed linear mixed models to predict CH4 production (g CH4/animal·d, ANIM-B models) and CH4 yield (g CH4/kg of dry matter intake, DMI-B models). We also developed models solely based on animal-related data. Prediction performance was evaluated 200 times with random data splits, while fitting performance was assessed without data splitting. The inclusion of microbial predictors improved the models, as indicated by decreased root mean square prediction error (RMSPE) and mean absolute error (MAE), and increased Lin's concordance correlation coefficient (CCC). Both glmmLasso and SCAD reduced the Akaike information criterion (AIC) and Bayesian information criterion (BIC) for both the ANIM-B and the DMI-B models, while the other two ML methods had mixed outcomes. By balancing prediction performance and fitting performance, we obtained one ANIM-B model (containing 10 genera of bacteria and 3 animal data) fitted using glmmLasso and one DMI-B model (5 genera of bacteria and 1 animal datum) fitted using SCAD. This study highlights the importance of incorporating rumen microbiota data in CH4 prediction models to enhance accuracy and robustness. Additionally, ML methods facilitate the selection of microbial predictors from high-dimensional metataxonomic data of the rumen microbiota without overfitting. Moreover, the identified microbial predictors can serve as biomarkers of CH4 emissions from sheep, providing valuable insights for future research and mitigation strategies.

Project description:We present prediction and variable importance (VIM) methods for longitudinal data sets containing continuous and binary exposures subject to missingness. We demonstrate the use of these methods for prognosis of medical outcomes of severe trauma patients, a field in which current medical practice involves rules of thumb and scoring methods that only use a few variables and ignore the dynamic and high-dimensional nature of trauma recovery. Well-principled prediction and VIM methods can provide a tool to make care decisions informed by the high-dimensional patient's physiological and clinical history. Our VIM parameters are analogous to slope coefficients in adjusted regressions, but are not dependent on a specific statistical model, nor require a certain functional form of the prediction regression to be estimated. In addition, they can be causally interpreted under causal and statistical assumptions as the expected outcome under time-specific clinical interventions, related to changes in the mean of the outcome if each individual experiences a specified change in the variable (keeping other variables in the model fixed). Better yet, the targeted MLE used is doubly robust and locally efficient. Because the proposed VIM does not constrain the prediction model fit, we use a very flexible ensemble learner (the SuperLearner), which returns a linear combination of a list of user-given algorithms. Not only is such a prediction algorithm intuitive appealing, it has theoretical justification as being asymptotically equivalent to the oracle selector. The results of the analysis show effects whose size and significance would have been not been found using a parametric approach (such as stepwise regression or LASSO). In addition, the procedure is even more compelling as the predictor on which it is based showed significant improvements in cross-validated fit, for instance area under the curve (AUC) for a receiver-operator curve (ROC). Thus, given that 1) our VIM applies to any model fitting procedure, 2) under assumptions has meaningful clinical (causal) interpretations and 3) has asymptotic (influence-curve) based robust inference, it provides a compelling alternative to existing methods for estimating variable importance in high-dimensional clinical (or other) data.

Project description:BackgroundResearchers should examine existing evidence to determine the need for a new study. It is unknown whether developers evaluate existing evidence to justify new cardiovascular clinical prediction rules (CPRs).ObjectiveWe aimed to assess whether authors of cardiovascular CPRs cited existing CPRs, why some authors did not cite existing CPRs, and why they thought existing CPRs were insufficient.MethodDerivation studies of cardiovascular CPRs from the International Register of Clinical Prediction Rules for Primary Care were evaluated. We reviewed the introduction sections to determine whether existing CPRs were cited. Using thematic content analysis, the stated reasons for determining existing cardiovascular CPRs insufficient were explored. Study authors were surveyed via e-mail and post. We asked whether they were aware of any existing cardiovascular CPRs at the time of derivation, how they searched for existing CPRs, and whether they thought it was important to cite existing CPRs.ResultsOf 85 derivation studies included, 48 (56.5%) cited existing CPRs, 33 (38.8%) did not cite any CPR, and four (4.7%) declared there was none to cite. Content analysis identified five categories of existing CPRs insufficiency related to: (1) derivation (5 studies; 11.4% of 44), (2) construct (31 studies; 70.5%), (3) performance (10 studies; 22.7%), (4) transferability (13 studies; 29.5%), and (5) evidence (8 studies; 18.2%). Authors of 54 derivation studies (71.1% of 76 authors contacted) responded to the survey. Twenty-five authors (46.3%) reported they were aware of existing CPR at the time of derivation. Twenty-nine authors (53.7%) declared they conducted a systematic search to identify existing CPRs. Most authors (90.7%) indicated citing existing CPRs was important.ConclusionCardiovascular CPRs are often developed without citing existing CPRs although most authors agree it is important. Common justifications for new CPRs concerned construct, including choice of predictor variables or relevance of outcomes. Developers should clearly justify why new CPRs are needed with reference to existing CPRs to avoid unnecessary duplication.

Project description:Milling chatter, a form of self-excited vibration, can cause significant damage in machining and manufacturing processes. By selecting appropriate milling parameters, milling chatter can be effectively mitigated without sacrificing milling efficiency. Within the framework of the semi-discretization scheme, this paper introduces the Newton-Simpson-based predictor-corrector methods to compute milling stability lobe diagrams. Firstly, the milling delay differential equation is transformed into the state space form, and then the time-delayed term and the periodic coefficient matrix of the state space equation are treated as an operator. Secondly, the tooth passing period is divided into the free vibration period and the forced vibration period. During the forced vibration period, the time-delayed term and the periodic coefficient matrix are approximated as a holistic operator over two different time intervals using the Newton interpolation polynomials and the Simpson formula, respectively. Finally, the state transition matrix is constructed based on the predictor-corrector scheme, and the stability lobe diagrams are obtained by applying Floquet theory. The convergence rate and calculation accuracy of the proposed methods are compared with those of the existing predictor-corrector methods, semi-discretization, and full-discretization methods. The results show that the proposed Newton-Simpson-based predictor-corrector methods have a faster convergence rate. For the local stability lobe diagrams, the arithmetic mean of relative error (AMRE), mean squared error (MSE), and the sum of absolute error (SAE) of the proposed methods are in the ranges of 0.003 to 0.004, 2.66 × 10-10 to 6.40 × 10-10, and 6.41 × 10-4 to 9.34 × 10-4, respectively, which are much lower than those of the existing methods, indicating that the proposed methods have higher calculation accuracy than the existing methods. The current work has a broad application prospect in the field of milling stability prediction for precision machining and the selection of chatter-free milling parameters.

Dataset Information

Harnessing repeated measurements of predictor variables for clinical risk prediction: a review of existing methods.

Background

Methods

Results

Conclusions

Publications

Harnessing repeated measurements of predictor variables for clinical risk prediction: a review of existing methods.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets