For many medical conditions (e.g. chlamydia trachomatis infection, tuberculosis in children, pneumonia, Alzheimer’s disease) there is no perfect diagnostic test or measure. This complicates estimation of the prevalence of the condition as well as estimation of the accuracy of diagnostic tests. Latent class models (or finite mixture models) provide a solution for this problem by modeling the observed patterns of test results as if they arise from a mixture of latent, i.e. unobserved, groups with and without the condition of interest.

Here is a link to a presentation on latent class analysis that I gave in 2017 Latent Class Analysis: An Indispensable Method for Diagnostic Accuracy Research. And here is a link to a video of me presenting on Estimating diagnostic test accuracy in the absence of a perfect reference: The importance of quantifying uncertainty

My research program has covered different aspects of latent class analysis motivated by problems in diagnostic test accuracy research. Some of my peer-reviewed publications and associated software are listed below (links to associated programs are provided when available):

Adjustment for conditional dependence in latent class models:

Dendukuri N, Joseph L. Bayesian approaches to modeling the conditional dependence between multiple diagnostic tests. Biometrics. 2001 Mar;57(1):158-67.

WinBUGS program for fixed effects model WinBUGS program for random effects model

Dendukuri N, Hadgu A, Wang L. Modeling conditional dependence between diagnostic tests: a multiple latent variable model. Stat Med. 2009 Feb 1;28(3):441-61.

Wang Z, Dendukuri N, Joseph L. Understanding the effects of conditional dependence in research studies involving imperfect diagnostic tests. Stat Med. 2017 Feb 10;36(3):466-480.

Wang Z, Dendukuri N, Zar HJ, Joseph L. Modeling conditional dependence among multiple diagnostic tests. Stat Med. 2017 Dec 30;36(30):4843-4859.

Below are R Markdown scripts for fixed- and random-effects model.

Bayesian LCA fixed effects

Bayesian LCA random effects

The random effects model takes advantage of the Gauss-Hermite quadrature approximation to greatly improve on computational speed. Both scripts run with minimal input from the user and will produce an output including parameter estimates, convergence diagnostic tools and more. See the READ ME! file for more details. A dataset (Strongyloides_data) to run with the models is provided to familiarize with the scripts.

Applications of latent class models:

Schumacher SG, van Smeden M, Dendukuri N, Joseph L, Nicol MP, Pai M, Zar HJ. Diagnostic Test Accuracy in Childhood Pulmonary Tuberculosis: A Bayesian Latent Class Analysis. Am J Epidemiol. 2016 Nov 1;184(9):690-700.

Pai M, Dendukuri N, Wang L, Joshi R, Kalantri S, Rieder HL. Improving the estimation of tuberculosis infection prevalence using T-cell-based assay and mixture models. Int J Tuberc Lung Dis. 2008 Aug;12(8):895-902.

MacLean EL, Kohli M, Köppel L, Schiller I, Sharma SK, Pai M, Denkinger CM, Dendukuri N. Bayesian latent class analysis produced diagnostic accuracy estimates that were more interpretable than composite reference standards for extrapulmonary tuberculosis tests. Diagn Progn Res. 2022 Jun 16;6(1):11.

Estimating incremental value in latent class models:

Ling DI, Pai M, Schiller I, Dendukuri N. A Bayesian framework for estimating the incremental value of a diagnostic test in the absence of a gold standard. BMC Med Res Methodol. 2014 May 15;14:67.

Adjustment for verification bias in latent class models:

de Groot JA, Dendukuri N, Janssen KJ, Reitsma JB, Bossuyt PM, Moons KG. Adjusting for differential-verification bias in diagnostic-accuracy studies: a Bayesian approach. Epidemiology. 2011

Lu Y, Dendukuri N, Schiller I, Joseph L. A Bayesian approach to simultaneously adjusting for verification and reference standard bias in diagnostic test studies. Stat Med. 2010 Oct 30;29(24):2532-43.

Sample size calculation for planning studies to apply latent class models:

Dendukuri N, Bélisle P, Joseph L. Bayesian sample size for diagnostic test studies in the absence of a gold standard: Comparing identifiable with non-identifiable models. Stat Med. 2010 Nov 20;29(26):2688-97.

Dendukuri N, Rahme E, Bélisle P, Joseph L. Bayesian sample size determination for prevalence and diagnostic test studies in the absence of a gold standard test. Biometrics. 2004 Jun;60(2):388-97.

Wang Z, Dendukuri N, Pai M, Joseph L. Taking Costs and Diagnostic Test Accuracy into Account When Designing Prevalence Studies: An Application to Childhood Tuberculosis Prevalence. Med Decis Making. 2017 Nov;37(8):922-929.

Problems with composite reference standards:

Dendukuri N, Schiller I, de Groot J, Libman M, Moons K, Reitsma J, van Smeden Concerns about composite reference standards in diagnostic research. BMJ. 2018 Jan 18;360:j5779.

Schiller I, van Smeden M, Hadgu A, Libman M, Reitsma JB, Dendukuri N. Bias due to composite reference standards in diagnostic accuracy studies. Stat Med. 2016 Apr 30;35(9):1454-70.

Hadgu A, Dendukuri N, Wang L. Evaluation of screening tests for detecting Chlamydia trachomatis: bias associated with the patient-infected-status algorithm. Epidemiology. 2012 Jan;23(1):72-82.

Hadgu A, Dendukuri N, Hilden J. Evaluation of nucleic acid amplification tests in the absence of a perfect gold-standard test: a review of the statistical and epidemiologic issues. Epidemiology. 2005 Sep;16(5):604-12.