COVID-19 contact tracing at work in Belgium – how tracers tweak … – BMC Public Health

Compliance with guidelines

We observed that, regardless of their profiles, contact tracers did not always rigorously follow contact tracing guidelines (positive agreement of 0.53 and negative agreement of 0.70). If the guidelines had been followed, we would have expected a PA and NA close to 1. There are several possible explanations for this lack of agreement:

Tracers may have deliberately not followed the guidelines after making their own assessment of the risk of infection. Indeed, it was observed during the pandemic that doctors might have been reluctant to apply externally imposed requirements concerning the provision and prioritization of patient care [9].

They may not have known the guideline criteria.

Participants may have erroneously reported the decision of the contact tracer and/or the guideline-based risk classification. We cannot verify this as self-reporting was the only data source used for this study.

Interviewed contacts may have deliberately offered selective information to the contact tracer as a result of mistrust or to obtain benefits associated with a particular contact classification. Mistrust is one of the obstacles to contact tracing described by Megnin-Viggars et al. [10] In their study, several obstacles to efficient contact tracing are identified at the level of the person being traced, not at the level of the contact tracer making their assessment, which is where the current study provides novel insights.

Respondents may have erroneously filled out the questionnaire, e.g. due to lack of attention or recollection bias.

Addressing these obstacles may improve the effectiveness and efficiency of contact tracing by making it more comprehensive and better targeted at individuals at highest risk of infection. Our study was observational, we did not find much literature on the actual implementation of contact tracing strategies. We therefore believe that our results can help to address the gap in knowledge on this topic.

Comparing the PA and NA results of the different contact tracers profiles reveals that they do not vary significantly from one profile to another. Our study does not, therefore, suggest that any of the profiles included applied the guideline criteria differently to the other profiles.

The secondary attack rate among contacts classified as HRCs by the contact tracers was 29%, which is higher than what we found in the literature [11,12,13] but comparable to the results of 27% found by Proesmans et al. who studied the performance of contact tracing in Belgium [14]. Our methodology made it possible to verify whether contact classification as high risk and low risk was useful, by comparing the infection risk in both groups. A Poisson regression found that the infection risk for HRCs was significantly higher than for LRCs. The RR between both groups was, at 3.1 for classifications by the contact tracer, which was significantly greater than for guideline-based classifications, for which the RR was 2.2. Contact tracers may thus have applied criteria that are not included in the definition of a HRC to assess the contacts risk of infection. They also may have had access to additional information for processing their case (existence of a cluster in the company, access to the quantitative results of the index cases PCR test, which may have suggested high viral shedding [15], etc.). We did not come across any research that used the same methodology as ours to calculate the RR of infection between HRCs and LRCs. Several studies, however, have applied a similar methodology to compare positivity rates among HRCs and LRCs. Sahoo et al. [16], Velhal D et al. [17] and Sharma et al. [18] collected information about 3411, 1486 and 1430 health care workers respectively. They were classified as HRCs and LRCs. Sahoo et al. obtained positivity rates of 3.8% for HRCs and 1.9% for LRCs. Velhal D et al. obtained positivity rates of 9.01% for HRCs and 2.72% for LRCs. Sharma et al. obtained positivity rates of 19.5% for HRCs and 0.6% for LRCs. We observe that their contact classification was similar and based on the CDC criteria. Risk stratification in contact tracing was found to be effective, however as it was in our study, even though the risk ratio was not calculated in these studies. Their positivity rates are lower than ours which could be attributed to the timing of the data collection (less contagious variants) and the limitations of our study. While they focused on populations of healthcare workers, we opted to include other professional fields.

The third variable, potential immunization, was included because we suspected that it would reduce transmission risk, as has been described [19, 20], which could have biased our results. Surprisingly, the Poisson regression showed the opposite, with an RR of 2.0 for potentially immune participants compared to non-immune participants. Numerous factors could explain this inconsistency. Our criteria for potential immunization were broad and probably resulted in the inclusion of non-immune participants. Furthermore, we cannot be certain that participants considered to be non-immune were indeed not immune, as case under-ascertainment was common during the first waves of COVID-19 infections [21]. Potentially immune healthcare workers may have been assigned to COVID-19 units, thus being at higher risk of infection but also were more often vaccinated and more intensely screened than other respondents were. Potentially immunized workers may have taken more risk in relation to exposure to others as a consequence of feeling protected from infection and severe disease [22]. The emergence of the Delta variant may have partially neutralized the protection offered by the vaccine against infection [23].

We specifically evaluated the risk classification of contacts and found it to be useful for identifying individuals at high risk of infection. We also demonstrated that targeting of testing is improved when a knowledgeable tracer performs the risk assessment.

Our study has several limitations. It was retrospective and based on an anonymous online survey, both factors which may have reduced the accuracy of the data. Participants could stop filling out the questionnaire at any time, which may also have reduced accuracy and completeness. Although participants received information describing the subject of the study in their email, this information may have led to a selection bias.

Healthcare workers were overrepresented in the study population. This is unsurprising because the organization through which participants were recruited (CESI) is particularly active in this sector, and healthcare workers were probably more exposed to COVID-19, leading to more COVID-19 contacts at work [24]. Stratification by job category would have been useful as risk differs significantly from one field to another. It was unfortunately not feasible due to the sample size.

The questionnaire was created during the first half of 2021, shortly after the start of vaccination and before the administration of booster doses. For this reason, we defined vaccination status rather broadly.

Furthermore, in the section concerning compliance, results for individuals traced by someone else or no one should be interpreted with caution, as many individuals in this category answered I dont know to the question about their contact classification. This latter answer was an exclusion criterion. There was therefore a selection bias that may have influenced the PA and NA.

Finally, the respondents knew the results of their test following contact, which may have had an impact on their questionnaire answers. For example, participants who tested positive may have seen the risk as greater when they replied to the questionnaire than when they responded to the contact tracer. This may have had an impact on both the calculation of the PA/NA and the comparison of the infection risk of HRC and LRC, as per contact tracer and guideline-based classification.

Original post:

COVID-19 contact tracing at work in Belgium - how tracers tweak ... - BMC Public Health