China
Africa
Explore chapters and articles related to this topic
Screening and Diagnostic Tests
Published in Marcello Pagano, Kimberlee Gauvreau, Heather Mattie, Principles of Biostatistics, 2022
Marcello Pagano, Kimberlee Gauvreau, Heather Mattie
While we have focused on sensitivity in this section, a parallel argument can be made for specificity. The phenomenon described is called spectrum bias. Spectrum bias occurs when the spectrum of users being tested does not reflect the spectrum of future or intended users. To quote the first sentence of an article by Ransohoff and Feinstein published in The New England Journal of Medicine[150], “Clinical investigations of the efficacy of diagnostic tests have often produced misleading results so that tests initially regarded as valuable were later rejected as worthless.”
Meta-Analysis of Diagnostic Tests
Published in Christopher H. Schmid, Theo Stijnen, Ian R. White, Handbook of Meta-Analysis, 2020
Yulun Liu, Xiaoye Ma, Yong Chen, Theo Stijnen, Haitao Chu
The previous bivariate approaches involving only sensitivities and specificities are appropriate if most studies use case-control designs not enabling estimation of the disease prevalence. When cohort study designs are used based on a random sample from the total population, then the disease prevalence is estimable, and we can derive other clinically meaningful indices such as PPV and NPV. In this case, the potential dependence of test performance on the disease prevalence, which is known as “spectrum bias” (Ransohoff and Feinstein, 1978), may be investigated. This is of most concern when the bivariate diagnostic outcome is based on a continuous trait with a threshold. It may lead to a high risk of misclassification, particularly for the subjects with true values around the cutpoint. Furthermore, the misclassification rate may vary between populations; in other words, it depends on the distribution of the true levels of the underlying trait in the population relative to the cutpoint (Brenner and Gefeller, 1997). The diagnostic misclassification and disease prevalence are related because the disease prevalence in the population is determined by this distribution (Brenner and Gefeller, 1997). To account for this potential dependence, we encourage the use of a trivariate GLMM, which is a generalization of the bivariate GLMM, to jointly model the disease prevalence, sensitivity, and specificity (Chu et al., 2009; Ma et al., 2014, 2016b).
Biostatistics: Issues in study design, analysis, and reporting
Published in Stephen W. Gutkin, Writing High-Quality Medical Publications, 2018
In broad terms, there are two forms of bias: selection bias36 (SB; Table 3.4) and information bias (IB; Table 3.5). Major forms of SB include self-selection bias, sampling bias, ascertainment bias, and Berkson’s bias (Berkson’s fallacy).20,37,38 Other forms include: Medical surveillance bias;Neyman’s bias;Survivor treatment selection bias;Duration-biased sampling;Health-care access bias; andSpectrum bias (ascertainment bias applied to assessing diagnostic tests).
Diagnostic accuracy and feasibility of depression screening in spinal cord injury: A systematic review
Published in The Journal of Spinal Cord Medicine, 2019
Rebecca Titman, Jason Liang, B. Catharine Craven
Risk of bias was assessed using the QUADAS-2 tool and was overall low to unclear. One particular source of bias specific to studies of diagnostic accuracy is spectrum bias. This is a form of sampling bias where performance of a diagnostic test may vary in clinical settings due to a different distribution of patients.70 Thombs et al.71 suggest that including individuals with diagnosed depression or who are receiving depression treatment artificially inflates the diagnostic accuracy of screening tools. Only the studies by Tate et al.28 and Radnitz et al.25 commented on exclusion of those with psychiatric diagnoses. Bombardier et al.31 excluded patients with psychosis only if their ability to complete assessments was impaired. None of the studies commented on whether patients involved were currently receiving treatment for depressive symptoms. It is therefore possible that these measures of diagnostic accuracy are inflated, which should be taken into consideration during clinical implementation.
Role of point of care Hb diagnostic devices in getting the right picture of anemia control: Tangi Rural Anemia Diagnostic Accuracy Study
Published in Journal of Drug Assessment, 2018
Preetam B. Mahajan, Somnath Mukherjee
Spectrum bias will result in clinically important differences in post-test probabilities under two circumstances: (1) presence of bias in the ratio of LR + and prevalence of disease is low, and (2) when bias exist in ratio of LR − and prevalence of disease is high [26]. As per our findings (Table 1), such a situation exists among adult males and females (This is if we were to combine the whole sample instead of looking at subgroup specific estimates). Thus, spectrum effect is likely to result into spectrum bias in these subgroups, unless we use subgroup specific accuracy parameters in decision making. This further strengthens the argument that only those spectrum of patients’ needs to be included in diagnostic accuracy studies that will be similar to the population in which the test will be used in practice [23].