Bayesian Correction of Misclassification of Pertussis in Vaccine Effectiveness Studies: How Much Does Underreporting Matter?

Abstract

Diagnosis of pertussis remains a challenge, and consequently research on the risk of disease might be biased because of misclassification. We quantified this misclassification and corrected for it in a case-control study of children in Philadelphia, Pennsylvania, who were 3 months to 6 years of age and diagnosed with pertussis between 2011 and 2013. Vaccine effectiveness (VE; calculated as (1 - odds ratio) × 100) was used to describe the average reduction in reported pertussis incidence resulting from persons being up to date on pertussis-antigen containing vaccines. Bayesian techniques were used to correct for purported nondifferential misclassification by reclassifying the cases per the 2014 Council of State and Territorial Epidemiologists pertussis case definition. Naïve VE was 50% (95% confidence interval: 16%, 69%). After correcting for misclassification, VE ranged from 57% (95% credible interval: 30, 73) to 82% (95% credible interval: 43, 95), depending on the amount of underreporting of pertussis that was assumed to have occurred in the study period. Meaningful misclassification was observed in terms of false negatives detected after the incorporation of infant apnea to the 2014 case definition. Although specificity was nearly perfect, sensitivity of the case definition varied from 90% to 20%, depending on the assumption about missed cases. Knowing the degree of the underreporting is essential to the accurate evaluation of VE.

Bias due to differential and non-differential disease- and exposure misclassification in studies of vaccine effectiveness

“…We provided the correction equations for VE estimates in case of non-differential single source (either exposure or disease) misclassification (Section 2.2). Other correction methods include amongst others probabilistic bias analyses [ 24 , 25 ], Bayesian bias analyses [ 26 – 28 ], modified maximum likelihood methods [ 29 ] and imputation-like methods [ 30 – 33 ]. All these methods require assumptions on or estimates of the disease- and exposure misclassification parameters, which—if deemed required—can be obtained using internal or external validation studies.…”

Association of Diphtheria-Tetanus–Acellular Pertussis Vaccine Timeliness and Number of Doses With Age-Specific Pertussis Risk in Infants and Young Children

“…Suspected, probable, and confirmed pertussis cases were included. 19 Pertussis remains highly underreported, 20 , 21 and only cases of patients who were symptomatic and whose caregivers sought medical help through PHSKC were included. Information on age, sex, and home address for patients was available.…”

Data Quality in Electronic Health Record Research: An Approach for Validation and Quantitative Bias Analysis for Imperfectly Ascertained Health Outcomes Via Diagnostic Codes

“…In fact, quantitative bias analyses extend beyond the bounds of data accuracy: similar approaches can handle cases of selection bias and residual confounding (again, see Lash et al, 2009 , and Smith et al, 2021 ). Beyond the case study demonstrated in this article, we refer readers to these other examples of quantitative bias analysis applied to misclassified outcome data ( Bodnar et al, 2010 ; Burstyn et al, 2020 ; Goldstein et al, 2016 ; Goldstein et al, 2021 ; Jurek & Maldonado, 2016 ; Srugo et al, 2021 ; Wesselink et al, 2018 ).…”