Risk of hospitalisation with covid among teachers and healthcare workers

Written by British Medical Journal on September 7, 2021. Posted in Current News

To determine the risk of hospital admission with covid-19 and severe covid-19 among teachers and their household members, overall and compared with healthcare workers and adults of working age in the general population.

School closures have formed part of the response to the covid-19 pandemic in most countries. Although the duration and extent of closures have varied, children and young people across many countries and regions have had limited access to schools throughout the pandemic.1 Such limited access has been found to reduce educational opportunities, limit social interactions, and harm physical and mental health, particularly among children from socioeconomically deprived backgrounds.2

Worldwide, governments have been required to weigh the risks and benefits of school closures. Among the complex considerations is whether providing education in person poses an increased risk to teachers, which has also been a concern for teachers’ representatives. Studies of this risk have been limited by small numbers of events, selection biases, a lack of data on potentially important potential confounders, such as the prevalence of underlying conditions, and too narrow a focus on specific settings resulting in findings of uncertain wider applicability.34567

Using the well established informatics infrastructure for covid-19 health records in Scotland,8910 we examined the risk of covid-19 in teachers in Scotland. Scottish schools closed during the first wave of the pandemic but were fully open with in-person teaching from August to December 2020. At that time class sizes were unchanged from those before the pandemic, and physical distancing was not required between primary school pupils but was recommended between staff and pupils and encouraged among secondary school pupils when possible.

Primary school pupils were not required to wear masks, and pupils in secondary schools were initially only required to wear masks in communal, non-classroom areas, and then from November 2020 in classrooms in areas with high case rates, and from March 2021 mask wearing was required at all times in school.11 In October 2020 a second wave of covid-19 occurred, with overall population rates of infection reaching a seven day case rate of around 150 per 100 000 in early November.

Antibody testing at the end of October 2020 indicated that just over 7 percent of the adult population had antibodies to SARS-CoV-2; and similar levels were observed among staff in educational settings.1213 A further wave of infections associated with the alpha variant prompted further school closures in January 2021, with subsequent phased reopening. In May and June 2021, the delta variant spread in Scotland.

During this time, Scotland had some of the highest rates of covid-19 in Europe, and the covid-19 vaccination programme was underway, with many adults of working age having been offered a first dose. At this time, schools were fully open.

This combination of circumstances provided us with an opportunity to estimate the risk of covid-19 among teachers in Scotland throughout the whole academic year, and during two separate periods of full in-person teaching when community transmission of SARS-CoV-2 was substantial. Before obtaining data on exposures, we prespecified hospital admission with covid-19 as the primary outcome.

We chose hospital admission rather than focusing on any case of covid-19 or severe covid-19, as we judged that cases of covid-19 were highly susceptible to ascertainment bias (affected by both individual behaviour for testing and individual access to testing) and number of events of severe covid-19 were likely to be too small in populations of working age. We estimated the relative risks in summer 2020 and winter 2020/2021 while schools were closed, in spring 2021 during a period of phased reopening, and in autumn 2020 and summer 2021 when schools were fully open.

Methods

We linked datasets of all the teachers to an existing case-control dataset that contains information on covid-19 cases in Scotland and matched population controls. The advantage of linking to an existing case-control study was that we could leverage the extensive data processing and cleaning (especially of covariate data) that we had already performed to produce results more rapidly.

The case-control study uses incidence density sampling such that the effect estimates calculated using these data are identical to hazard ratios obtained from an equivalent whole population cohort study analysed using Cox proportional hazards models.14

As a result of previous work, the existing case-control study includes information on whether participants are healthcare workers.10 This enabled us to compare not only rates of covid-19 in teachers with rates in the general population but also teachers compared with a known high risk group (patient facing healthcare workers) and with an occupational group not expected to be at increased risk (non-patient facing healthcare workers).

For patient facing healthcare workers we used our previous definitions.10 For non-patient facing healthcare workers we applied a stricter definition, including only those staff most likely to be working in non-clinical settings (finance, human resources, information technology, and call centre work) or based in National Health Service organisations not directly involved in patient care (eg, Public Health Scotland). When the patient facing status of healthcare workers was uncertain, we removed them from the case-control dataset.

The complete case data from the case-control study was also used alongside denominator data (not linked to covariates), which included all teachers and healthcare workers (and by subtraction from the population mid-year estimates, adults in neither category) to allow us to estimate absolute risks in all three groups.

Case-control study

Public Health Scotland maintains a nested case-control study sampled from population-wide healthcare utilisation databases held by the organisation. This study, described in detail elsewhere,9 includes all people in Scotland who are classified as cases of covid-19, and for each case 10 controls randomly selected from the Scottish population who are of the same age (in single years) and sex, and are registered at the same general practice as the case, but who did not (on or before that date) meet the case definition.

The case-control study is regularly updated, with the most recent update on 28 June 2021. Controls were ascertained using the Community Health Index database, which contains the unique healthcare identifier, other identifiers, age, sex, and general practice for people in the total population of Scotland. For the entire analysis, only adults of working age (21 to 65 years) were included.

The case-control dataset is linked to recent hospital admission and prescribing data to identify underlying diseases, and to contemporaneous hospital admission and intensive care data to characterise the severity of each case. Ten controls were matched to each case.15

Outcomes

All events from the onset of the pandemic until 30 June 2021 were included in the analysis. We included a further 28 days of follow-up to determine whether events on 30 June resulted in hospital admission, admission to intensive care, or death.

As in previous analyses, we defined cases of covid-19 as people with a positive reverse transcription polymerase chain reaction (RT-PCR) test result for SARS-CoV-2, or a hospital discharge with a diagnosis of covid-19 regardless of testing positive, or any death with covid-19 included as a cause (regardless of whether it was recorded as the underlying cause and regardless of any previous test result).

The primary outcome was prespecified as hospital admission with covid-19, defined as anyone with a positive test result for SARS-CoV-2 while in hospital, being admitted to hospital within 28 days of a positive test result, or a diagnosis of covid-19 noted on a hospital discharge letter.

The number of teachers and healthcare workers with severe covid-19 was anticipated to be low because this outcome is rare among adults of working age and the number of teachers and healthcare workers is relatively small. As such, despite being more robust to clinical decision making than hospital admission, severe covid-19 was selected as a secondary outcome. Severe covid-19 was defined as covid-19 resulting in death or admission to intensive care within 28 days of a positive test result.

Occupational status

The General Teaching Council for Scotland (GTCS) holds data on every teacher in Scotland, including name, sex, date of birth, home postcode, work sector (nursery, primary school, or secondary school), last known employer, and registration number. Teachers are prompted to update their registration details annually. Teachers were defined as those of working age registered with the GTCS and currently working, or believed to be currently working, in a Scottish school.

The GTCS indicated those who were teaching in February 2020 or November 2020, or both (see supplementary methods for additional details). These data were linked to the case-control study using name, sex, date of birth, and home postcode. Healthcare workers were identified using the General Practitioner Contractor Database and Scottish Workforce Information Standard System databases, as described previously.9

We compared outcomes in teachers with outcomes in patient facing healthcare workers, non-patient facing healthcare workers, and adults of working age who were neither teachers nor healthcare workers (the general population comparator).

Schools in Scotland reopened on or shortly after 12 August 2020. A five day lag period was included to allow for the time between exposure to SARS-CoV-2 and a positive test result. We present the results for two periods when schools were closed in spring/summer 2020 (1 March to 24 August 2020) and winter 2021 (24 December to 26 February 2021); one period of phased reopening in winter/spring 2021; and two periods when schools were fully open (25 August to 23rd December and 24 April to 30 June 2021).

Household members of healthcare workers and teachers were identified through the unique property reference number, which was added to the national general practice registration database register in 2020.

Covariates

We obtained data on age, sex, and Scottish index of multiple deprivation (an area based measure of socioeconomic deprivation) from the national general practice registration database, race/ethnicity through self-report from a range of healthcare utilisation databases (Scottish Morbidity Records (SMR) 01, 02, and 04), and comorbidity from previous hospital admission (SMR01) and drug dispensing (Prescribing Information System) data using the same definitions developed previously.910

Additionally, we used the unique property reference number to obtain the numbers of adults (≥18 years) in the households of cases and controls. Vaccine status was obtained from the Scottish vaccine database.

Statistical analysis

Summary statistics for personal, socioeconomic, and clinical characteristics were calculated for teachers, healthcare workers, and the remaining population of adults of working age. The control arm of the case-control study is effectively a stratified random sample from the entire Scottish population, where the strata are defined by the age and sex of individuals and the general practice within which they are registered.

As such, if the probability of inclusion is known, the control arm can be used to obtain valid summary statistics for the whole population; this is analogous to the reweighting used when analysing survey data. To estimate the inclusion probabilities, we obtained counts of the Scottish population stratified by age (in single years), sex, Scottish index of multiple deprivation, and health board area,16 and for the same strata produced counts for the control arm of the case-control study.

The inclusion probability was then calculated as the number of controls in each stratum divided by the total population in that stratum. We then produced statistics for the whole of Scotland (and plots of vaccination over time) for all teachers, healthcare workers, and the remaining population of adults of working age using the TableOne package in R, which allows the estimation of summary statistics in the presence of stratified sampling (including counts, proportions, means, and standard deviations) through inverse probability weighting.17

We produced cumulative incidence (risk) plots for hospital admission with covid-19 for all four groups (patient facing or non-patient facing healthcare workers, teachers, and adults of working age in the general population), stratified by age and sex. All events for Scotland were obtained through the case arm of the case-control study, with the denominators obtained directly for teachers and healthcare workers, and through subtraction from the mid-year estimates for the remaining population of adults of working age.

For all covid-19 outcomes (any covid-19, hospital admission, and severe covid-19), we fitted unadjusted conditional logistic regression models. These effect estimates can be interpreted as rate ratios. Unadjusted models were conditional on the matching variables (age, sex, and general practice). In the adjusted models, we additionally included terms for potential confounders such as Scottish index of multiple deprivation, race/ethnicity, number of comorbidities, and whether the individual shared a household with a healthcare worker.

Because the dataset was large, to reduce computational time we restricted the conditional logistic regression models to strata including at least one teacher or one healthcare worker or one member of their household, as strata without such individuals will not contribute to effect estimates for those variables. In the main analysis we included individuals who had received a vaccine dose as this allowed us to examine the risk of hospital admission with covid-19 in teachers in relation to both their access to and their uptake of vaccination.

In exploratory analyses, we censored events (and person time) occurring after vaccination. For the purposes of this analysis, we considered individuals to be post first dose from 14 days after the date of their injection. The standard error for the difference in rate ratios was calculated as the square root of the sum of the individual standard errors for each rate ratio squared. The supplementary file shows the prespecified statistical analysis plan. The analysis code is available at https://github.com/dmcalli2/tchr.

Patient and public involvement

The constraints on time and resources of responding to the covid-19 pandemic for both public health and teaching workforces meant that a formal process of public involvement was not feasible within the timescales of this research. However, this work has been informed from inception by dialogue with representatives of teaching professionals and those working in education policy.

This research was prompted by concern expressed by teaching professionals in Scotland about their potential occupational risk, in discussions before the return to in-person teaching in August 2020. This was communicated through the engagement of professional associations in policy forums, in particular the COVID Education Recovery Group (CERG, www.gov.scot/groups/covid-19-education-recovery-group/). This study design was proposed as part of a programme of enhanced surveillance for education discussed in July 2020 with the range of education partners on CERG.

A communication was prepared in conjunction with GTCS to notify all registered teachers of the proposed sharing of registration data. This included details of the rationale for the research and provided a period during which registrants could raise objections to data sharing.

Results

By the end of June 2021, the case-control study included 132 420 cases, and 1 306 566 controls matched on age, sex, and general practice. Of 66 710 individuals in the teacher dataset 25 687 were selected (as cases or controls) into the case-control study. Of 87 273 patient facing healthcare workers, 38 993 were selected into the case-control study. Of 8501 non-patient facing healthcare workers, 2731 were selected into the case-control study.

Table 1 shows the characteristics of the teachers and healthcare workers compared with the general population using reweighted data from controls. Compared with the general population, teachers and healthcare workers were similar for age and race/ethnicity but were more likely to be women and to have fewer comorbidities. Both teachers and healthcare workers were less likely to live in the most deprived fifth of areas than the general population, with a larger difference for teachers.

Teachers were predominantly women; even in secondary schools, where the proportion of men was higher than that in other schools, two thirds of teachers were women. By the end of follow-up, the proportion who had received a first dose of covid-19 vaccine was similar between teachers and healthcare workers and higher than in adults of working age in the general population. Many more healthcare workers had received second doses than had the other groups.

Conclusion

Compared with adults of working age who are otherwise similar, teachers and their household members were NOT found to be at increased risk of hospital admission with covid-19 and were found to be at lower risk of severe covid-19. These findings should reassure those who are engaged in face-to-face teaching.

Read the rest here: bmj.com

Header image: medicine.net

Please Donate Below To Support Our Ongoing Work To Defend The Scientific Method

PRINCIPIA SCIENTIFIC INTERNATIONAL, legally registered in the UK as a company incorporated for charitable purposes. Head Office: 27 Old Gloucester Street, London WC1N 3AX. 

Trackback from your site.