Does recall bias explain the association of mood disorders with workplace harassment?

Purpose

To determine the contribution of recall bias to the observed excess in mental ill-health in those reporting harassment at work.

Methods

A prospective cohort of 1885 workers in welding and electrical trades was contacted every six months for up to 5 years, asking whether they were currently anxious or depressed and whether this was made worse by work. Only at the end of the study did we ask about any workplace harassment they had experienced at work. We elicited sensitivity and specificity of self-reported bullying from published reliability studies and formulated priors that reflect the possibility of over-reporting of workplace harassment (exposure) by those whose anxiety or depression was reported to be made worse by work (cases). We applied the resulting misclassification models to probabilistic bias analysis (PBA) of relative risks.

Results

We observe that PBA implies that it is unlikely that biased misclassification due to the study subjects' states of mind could have caused the entire observed association. Indeed, the results demonstrated that doubling of risk of anxiety or depression following workplace harassment is plausible, with the unadjusted relative risk attenuated with understated uncertainty.

Conclusions

It seems unlikely that risk of anxiety or depression following workplace harassment can be explained by the form of recall bias that we proposed.

Quantitative bias analysis in practice: review of software for regression with unmeasured confounding

BACKGROUND

Failure to appropriately account for unmeasured confounding may lead to erroneous conclusions. Quantitative bias analysis (QBA) can be used to quantify the potential impact of unmeasured confounding or how much unmeasured confounding would be needed to change a study's conclusions. Currently, QBA methods are not routinely implemented, partly due to a lack of knowledge about accessible software. Also, comparisons of QBA methods have focused on analyses with a binary outcome.

METHODS

We conducted a systematic review of the latest developments in QBA software published between 2011 and 2021. Our inclusion criteria were software that did not require adaption (i.e., code changes) before application, was still available in 2022, and accompanied by documentation. Key properties of each software tool were identified. We provide a detailed description of programs applicable for a linear regression analysis, illustrate their application using two data examples and provide code to assist researchers in future use of these programs.

RESULTS

Our review identified 21 programs with [Formula: see text] created post 2016. All are implementations of a deterministic QBA with [Formula: see text] available in the free software R. There are programs applicable when the analysis of interest is a regression of binary, continuous or survival outcomes, and for matched and mediation analyses. We identified five programs implementing differing QBAs for a continuous outcome: treatSens, causalsens, sensemakr, EValue, and konfound. When applied to one of our illustrative examples, causalsens incorrectly indicated sensitivity to unmeasured confounding whereas the other four programs indicated robustness. sensemakr performs the most detailed QBA and includes a benchmarking feature for multiple unmeasured confounders.

CONCLUSIONS

Software is now available to implement a QBA for a range of different analyses. However, the diversity of methods, even for the same analysis of interest, presents challenges to their widespread uptake. Provision of detailed QBA guidelines would be highly beneficial.

Quantitative bias analysis of prevalence under misclassification: evaluation indicators, calculation method and case analysis

Prevalence estimates are fundamental to epidemiological studies. Although they are highly vulnerable to misclassification bias, the risk of bias assessment of prevalence estimates is often neglected. Quantitative bias analysis (QBA) can effectively estimate misclassification bias in epidemiological studies; however, relatively few applications are identified. One reason for its low usage is the lack of knowledge and tools for these methods among researchers. To expand existing evaluation methods, based on the QBA principles, three indicators are proposed. One is the relative bias that quantifies the bias direction through its signs and the bias magnitude through its quantity. The second is the critical point of positive test proportion in case of a misclassification bias that is equal to zero. The third is the bound of positive test proportion equal to adjusted prevalence at misclassification bias level α. These indicators express the magnitude, direction and uncertainty of the misclassification bias of prevalence estimates, respectively. Using these indicators, it was found that slight oscillations of the positive test proportion within a certain range can lead to substantial increases in the misclassification bias. Hence, researchers should account for misclassification error analytically when interpreting the significance of adjusted prevalence for epidemiological decision making. This highlights the importance of applying QBA to these analyses. In this article, we have used three real-world cases to illustrate the characteristics and calculation methods of presented indicators. To facilitate application, an Excel-based calculation tool is provided.

Mortality following workplace injury: Quantitative bias analysis

PURPOSE

Recent studies have shown increased all-cause mortality among workers following disabling workplace injury. These studies did not account for 2 potentially important confounders, smoking and obesity. We estimated injury-related mortality accounting for these factors.

METHODS

We followed workers receiving New Mexico workers' compensation benefits (1994-2000) through 2013. Using data from the Panel Study of Income Dynamics, we derived the joint distribution of smoking status and obesity for workers with and without lost-time injuries. We conducted a quantitative bias analysis (QBA) to determine the adjusted relationship of injury and mortality.

RESULTS

We observed hazard ratios after adjusting for smoking and obesity of 1.13 for women (95% simulation interval (SI) 0.97 to 1.31) and 1.12 for men (95% SI 1.00 to 1.27). The estimated fully adjusted excess hazard was about half the estimates not adjusted for these factors.

CONCLUSIONS

Using QBA to adjust for smoking and obesity reduced the estimated mortality hazard from lost-time injuries and widened the simulation interval. The adjusted estimate still showed more than a 10 percent increase for both women and men. The change in estimates reveals the importance of accounting for these confounders. Of course, the results depend on the methods and assumptions used.

Sensitivity analysis for random measurement error using regression calibration and simulation-extrapolation

Objective

Sensitivity analysis for random measurement error can be applied in the absence of validation data by means of regression calibration and simulation-extrapolation. These have not been compared for this purpose.

Study design and setting

A simulation study was conducted comparing the performance of regression calibration and simulation-extrapolation for linear and logistic regression. The performance of the two methods was evaluated in terms of bias, mean squared error (MSE) and confidence interval coverage, for various values of reliability of the error-prone measurement (0.05-0.91), sample size (125-4000), number of replicates (2-10), and R-squared (0.03-0.75). It was assumed that no validation data were available about the error-free measures, while correct information about the measurement error variance was available.

Results

Regression calibration was unbiased while simulation-extrapolation was biased: median bias was 0.8% (interquartile range (IQR): -0.6;1.7%), and -19.0% (IQR: -46.4;-12.4%), respectively. A small gain in efficiency was observed for simulation-extrapolation (median MSE: 0.005, IQR: 0.004;0.006) versus regression calibration (median MSE: 0.006, IQR: 0.005;0.009). Confidence interval coverage was at the nominal level of 95% for regression calibration, and smaller than 95% for simulation-extrapolation (median coverage: 85%, IQR: 73;93%). The application of regression calibration and simulation-extrapolation for a sensitivity analysis was illustrated using an example of blood pressure and kidney function.

Conclusion

Our results support the use of regression calibration over simulation-extrapolation for sensitivity analysis for random measurement error.

Medical record bias in documentation of obstetric and neonatal clinical quality of care indicators in Uganda

OBJECTIVE

To achieve a high quality of care (QoC), accurate measurements are needed. This study evaluated the validity of QoC data from the medical records for childbirth deliveries and assessed whether medical records can be used to evaluate the efficacy of interventions to improve QoC.

STUDY DESIGN AND SETTING

This study was part of a larger study of QoC training program in Uganda. Study data were collected in two phases: (1) validation data from 321 direct observations of deliveries paired with the corresponding medical records; (2) surveillance data from 1,146 medical records of deliveries. Sensitivity, specificity, and predictive values were used to measure the validity of the medical record from the validation data. Quantitative bias analysis was conducted to evaluate QoC program efficacy in the surveillance data using prevalence ratio and odds ratio.

RESULTS

On average, sensitivity (84%) of the medical record was higher than the specificity (34%) across 11 QoC indicators, showing a higher validity in identifying the performed procedure. For 5 out of 11 indicators, bias-corrected odds ratios and prevalence ratios deviated significantly from uncorrected estimates.

CONCLUSION

The medical records demonstrated poor validity in measuring QoC compared with direct observation. Using the medical record to assess QoC program efficacy should be interpreted carefully.

Quantitative bias analysis for study and grant planning

PURPOSE

Epidemiologists often think about the balance between study error and cost-efficiency in terms of study design and strategies to reduce random error. We less often consider cost-efficiencies in terms of dealing with systematic errors that arise within a study, such as in deciding how to measure study variables and misclassification implications.

METHODS

Given the information used to inform a study size calculation, the expected study data can be simulated during study planning, and the impact of anticipated biases can be estimated using quantitative bias analysis. This would allow investigators and stakeholders to identify areas where better data collection through more valid instruments is critical and where additional investment will not yield strong validity benefits. This could promote better use of study resources and help increase investigators' chances of funding by demonstrating they have thought through biases and have a plan for mitigating the impact.

RESULTS

We demonstrate how this would work with a practical example using the relationship between smoking during pregnancy as measured on birth certificates and incident breast cancer.

CONCLUSIONS

We show that although exposure sensitivity would likely be poor, spending more money to get a better smoking measure is unlikely to yield more valid estimates.

Quantitative bias analysis of the association of type 2 diabetes mellitus with 2,2',4,4',5,5'-hexachlorobiphenyl (PCB-153)

An association between serum concentrations of persistent organic pollutants (POPs), such as 2,2',4,4',5,5'-hexachlorobiphenyl (PCB-153), and risk of type 2 diabetes mellitus (T2DM) has been reported. Conditional on body mass index (BMI) and waist circumference (WC), a higher serum PCB-153 concentration may be a marker of T2DM risk because it reflects other aspects of obesity that are related to T2DM risk and to PCB-153 clearance. To estimate the amount of residual confounding by other aspects of obesity, we performed a quantitative bias analysis on the results of a specific study. A physiologically-based pharmacokinetic (PBPK) model was developed to predict serum levels of PCB-153 for a simulated population. T2DM status was assigned to simulated subjects based on age, sex, BMI, WC, and visceral adipose tissue mass. The distributions of age, BMI, WC, and T2DM prevalence of the simulated population were tailored to closely match the target population. Analysis of the simulated data showed that a small part of the observed association appeared to be due to residual confounding. For example, the predicted odds ratio of T2DM that would have been obtained had the results been adjusted for visceral adipose tissue mass, for the ≥90th percentile of PCB-153 serum concentration, was 6.60 (95% CI 2.46-17.74), compared with an observed odds ratio of 7.13 (95% CI 2.65-19.13). Our results predict that the association between PCB-153 and risk of type 2 diabetes mellitus would not be substantially changed by additional adjustment for visceral adipose tissue mass in epidemiologic analyses. Confirmation of these predictions with longitudinal data would be reassuring.

Excretion of Di-2-ethylhexyl phthalate (DEHP) metabolites in urine is related to body mass index because of higher energy intake in the overweight and obese

A higher body mass index (BMI) has been positively associated with the rate of excretion of di-2-ethylhexyl phthalate (DEHP) metabolites in urine in data from the National Health and Nutrition Examination Survey (NHANES), suggesting an association between DEHP exposure and BMI. The association, however, may be due to the association between body mass maintenance and higher energy intake, with higher energy intake being accompanied by a higher intake of DEHP. To examine this hypothesis, we ran a Monte Carlo simulation with a DEHP physiologically-based pharmacokinetic (PBPK) model for adult humans. A realistic exposure sub-model was used, which included the relation of body weight to energy intake and of energy intake to DEHP intake. The model simulation output, when compared with urinary metabolite data from NHANES, supported good model validity. The distribution of BMI in the simulated population closely resembled that in the NHANES population. This indicated that the simulated subjects and DEHP exposure model were closely aligned with the NHANES population of interest. In the simulated population, the ordinary least squares regression coefficient for log(BMI) as a function of log(DEHP nmol/min) was 0.048 (SE 0.001), as compared with the reported value of 0.019 (SE 0.005). In other words, given our model structure, the higher energy intake in the overweight and obese, and the concomitant higher DEHP exposure, describes the reported relationship between BMI and DEHP.

Applying quantitative bias analysis to estimate the plausible effects of selection bias in a cluster randomised controlled trial: secondary analysis of the Primary care Osteoarthritis Screening Trial (POST)

Background

Selection bias is a concern when designing cluster randomised controlled trials (c-RCT). Despite addressing potential issues at the design stage, bias cannot always be eradicated from a trial design. The application of bias analysis presents an important step forward in evaluating whether trial findings are credible. The aim of this paper is to give an example of the technique to quantify potential selection bias in c-RCTs.

Methods

This analysis uses data from the Primary care Osteoarthritis Screening Trial (POST). The primary aim of this trial was to test whether screening for anxiety and depression, and providing appropriate care for patients consulting their GP with osteoarthritis would improve clinical outcomes. Quantitative bias analysis is a seldom-used technique that can quantify types of bias present in studies. Due to lack of information on the selection probability, probabilistic bias analysis with a range of triangular distributions was also used, applied at all three follow-up time points; 3, 6, and 12 months post consultation. A simple bias analysis was also applied to the study.

Results

Worse pain outcomes were observed among intervention participants than control participants (crude odds ratio at 3, 6, and 12 months: 1.30 (95% CI 1.01, 1.67), 1.39 (1.07, 1.80), and 1.17 (95% CI 0.90, 1.53), respectively). Probabilistic bias analysis suggested that the observed effect became statistically non-significant if the selection probability ratio was between 1.2 and 1.4. Selection probability ratios of > 1.8 were needed to mask a statistically significant benefit of the intervention.

Conclusions

The use of probabilistic bias analysis in this c-RCT suggested that worse outcomes observed in the intervention arm could plausibly be attributed to selection bias. A very large degree of selection of bias was needed to mask a beneficial effect of intervention making this interpretation less plausible.

Revisiting an old disease? Risk factors for bovine enzootic haematuria in the Kingdom of Bhutan

Bovine enzootic haematuria (BEH) is a debilitating disease of cattle caused by chronic ingestion of bracken fern. Control of BEH is difficult when bracken fern is abundant and fodder resources are limited. To fill a significant knowledge gap on modifiable risk factors for BEH, we conducted a case-control study to identify cattle management practices associated with BEH in the Bhutanese cattle population. A case-control study involving 16 of the 20 districts of Bhutan was carried out between March 2012 and June 2014. In Bhutan sodium acid phosphate and hexamine (SAP&H) is used to treat BEH-affected cattle. All cattle greater than three years of age and treated with SAP&H in 2011 were identified from treatment records held by animal health offices. Households with at least one SAP&H-treated cattle were defined as probable cases. Probable case households were visited and re-classified as confirmed case households if the BEH status of cattle was confirmed following clinical examination and urinalysis. Two control households were selected from the same village as the case household. Households were eligible to be controls if: (1) householders reported that none of their cattle had shown red urine during the previous five years, and (2) haematuria was absent in a randomly selected animal from the herd following clinical examination. Details of cattle management practices were elicited from case and control householders using a questionnaire. A conditional logistic regression model was used to quantify the association between exposures of interest and household BEH status. A total of 183 cases and 345 controls were eligible for analysis. After adjusting for known confounders, the odds of free-grazing for two and three months in the spring were 3.81 (95% CI 1.27-11.7) and 2.28 (95% CI 1.15-4.53) times greater, respectively, in case households compared to controls. The odds of using fresh fern and dry fern as bedding in the warmer months were 2.05 (95% CI 1.03-4.10) and 2.08 (95% CI 0.88-4.90) times greater, respectively, in cases compared to controls. This study identified two husbandry practices that could be modified to reduce the risk of BEH in Bhutanese cattle. Avoiding the use of bracken fern as bedding is desirable, however, if fern is the only available material, it should be harvested during the colder months of the year. Improving access to alternative fodder crops will reduce the need for householders to rely on free-grazing as the main source of metabolisable energy for cattle during the spring.

Quantitative bias analysis for epidemiological associations of perfluoroalkyl substance serum concentrations and early onset of menopause

An association between increased serum concentrations of perfluoroalkyl substances (PFAS) such as perfluorooctane sulfonate (PFOS) and perfluorooctanoate (PFOA) and early menopause has been reported (Knox et al., 2011; Taylor et al., 2014). This association may be explained by the fact that women who underwent menopause no longer excrete PFAS through menstruation. Our objective was to assess how much of the epidemiologic association between PFAS and altered timing of menopause might be explained by reverse causality. We extended a published population life-stage physiologically-based pharmacokinetic (PBPK) model of PFOS and PFOA characterized by realistic distributions of physiological parameters including age at menopause. We then conducted Monte Carlo simulations to replicate the Taylor population (Taylor et al., 2014) and the Knox population (Knox et al., 2011). The analysis of the simulated data overall showed a pattern of results that was comparable to those reported in epidemiological studies. For example, in the simulated Knox population (ages 42-51) the odds ratio (OR) for menopause in the fifth quintile of PFOA compared to those in the first quintile was 1.33 (95% CI 1.26-1.40), whereas the reported OR was 1.4 (95% CI 1.1-1.8). Using our model structure, a substantial portion of the associations reported can be explained by pharmacokinetics.