Explore chapters and articles related to this topic
Designing AI Systems for Clinical Practice
Published in Lia Morra, Silvia Delsanto, Loredana Correale, Artificial Intelligence in Medical Imaging, 2019
Lia Morra, Silvia Delsanto, Loredana Correale
Results from observational studies are often criticized for being vulnerable to confounding factors, including confounding by indications, which occurs, for instance, when the clinical indications that led to the use of a certain diagnostic or predictive AI tool also affect the patient outcome, radiologist’s experience, and lack of standardization in medical care that follows an acquisition of diagnostic or predictive information. Relevant statistical procedures such as stratification and multivariable regression modeling can be used to account for confounding factors. Stratification refers to separation of patients into groups or strata according to the level of a possible confounding. This permits an analysis of each group separately, thus removing the confounding effect. Multivariable regression analysis adjusts for potential confounding factors by including them as explanatory variables in the analysis.
Drinking Water Characteristics and Cardiovascular Disease in a Cohort of Wisconsin Farmers
Published in Frederick C. Kopfler, Gunther F. Craun, Environmental Epidemiology, 2019
Elaine A. Zeighami, Gunther F. Craun, Charlotte A. Cottrill
In studies of drinking water associations with cardiovascular disease, accurate information must be obtained not only for each subject’s exposures to the drinking water constituents of interest, but also for other exposures and other risk factors, as the data relating exposure to disease may convey an appearance of association because of confounding bias. Although negative confounding can also occur, the primary concern is that confounding has led to the erroneous observation of an association. Confounding bias is a basic characteristic of any epidemiologic study, and does not necessarily result from any error on the part of the investigator. Information should be collected on known or suspected confounding characteristics. If a characteristic can be demonstrated to have no association with either the exposure or disease being studied, that characteristic cannot be confounding. To prevent confounding, matching is generally employed in the study design. To assess and control confounding during data analysis, stratification or multivariate techniques are employed.
Training Effectiveness Evaluation
Published in Florian Jentsch, Michael Curtis, Eduardo Salas, Simulation in Aviation Training, 2017
Threats to internal validity. A confounding variable is an uncontrolled factor that is allowed to vary along with the variable of interest. Such confounding variables can affect the quality and the interpretability of the results of the evaluation or experiment. When a confounding variable occurs, it can make it impossible to determine whether observed performance changes are the result of the training or whether the confounding variable has contributed to the outcome. In other words, confounding variables reduce the internal validity of the evaluation. Several confounds can occur because the experiment is poorly designed. Cook and Cambell (1975, 1979) and Eberts and others, (1982) discuss several examples of confounds.
Occupational tuberculosis in healthcare workers in sub-Saharan Africa: A systematic review
Published in Archives of Environmental & Occupational Health, 2019
Faith O Alele, Richard C. Franklin, Theophilus I. Emeto, Peter Leggat
The quality of the studies was analyzed by two reviewers (FOA and TIE) using the checklist proposed by Zaza et al.22 Any discrepancies and disagreement were discussed until a consensus was reached. Only items that are applicable to observational studies were included in the quality assessment. Study quality assessment included issues related to study design, sampling, and potential bias and confounding affecting validity. Confounding affects the internal validity of a study. Confounding refers to mixing or muddling of effects that distorts the true relationship between the exposure and outcome.23 Each section was allocated a minimum of 0 and maximum of 5 and the articles were categorised as excellent, very good, satisfactory and unsatisfactory based on the overall score Table 1.
Model justification and stratification for confounding of Chlamydia trachomatis disease
Published in Letters in Biomathematics, 2019
Now, we have considered a hypothetical case-control study investigating the association between gender and Chlamydia for the mentioned dataset. Stratification is the simplest method for controlling the confounding during data analysis. It also represents the preliminary step for invoking the MH formula and standardization. Here we focus on stratification, a statistical technique that allows to control for confounding by creating two or more categories (strata) in which the confounding variable either does not vary or does not too much (Mantel & Haenszel, 1959). The MH analysis allows to calculate an overall, unconfounded, that is adjusted and also consider the combining (pooling) stratum-specific relative risks (RR) or odds ratios (OR). Stratum-specific RRs or ORs are computed within each stratum of the confounding variable. These are also useful for comparing with the corresponding effect estimates in the whole group (that is, with unstratified RR or OR). ‘No confounding’ means that the effect estimates are roughly homogeneous across strata and do not differ from that in the whole group. Also, the ‘presence of confounding’ indicates that the effect estimates are substantially similar across strata but differ from that in the whole group (Maldonado & Greenland, 1993). In this context. it is mentioned that at the time of comparing between stratified and unstratified effect estimates, epidemiologists consider as relevant a RR or OR difference by more than (Maldonado & Greenland, 1993).
Wear conduct of aluminum matrix composites: A parametric strategy using Taguchi based GRA integrated with weight method
Published in Cogent Engineering, 2018
Narinder Kaushik, Sandeep Singhal
Taguchi experimental approach is selected for the robust and sophisticated design. As compared to traditional designs Taguchi method gives an easy, systematic and managed approach by minimizing the experimental time and cost of experiments for the optimization of the design factors (Montgomery, 2001). A factorial design could be used but the main drawback of a fractionated design is that some interactions may be confounded with other effects. It is important to consider carefully the role of potential confounders and aliases. Failure to take account of such confounding effects can result in erroneous conclusions and misunderstandings. While Taguchi’s designs are usually highly fractionated, this makes them very attractive and the interactions effect can be analysed effectively. When using a Taguchi design, we can guess which interactions are most likely to be significant even before any experiment is performed.