key: cord-0877487-z1mb34tt
authors: Nielsen, Kent J.; Vestergaard, Jesper Medom; Schlünssen, Vivi; Bonde, Jens Peter; Kaspersen, Kathrine Agergård; Biering, Karin; Carstensen, Ole; Greve, Thomas; Hansen, Karoline Kærgaard; Dalbøge, Annett; Flachs, Esben Meulengracht; Jespersen, Sanne; Hansen, Mette Lausten; Mikkelsen, Susan; Thomsen, Marianne Kragh; Redder, Jacob Dvinge; Würtz, Else Toft; Østergaard, Lars; Erikstrup, Christian; Kolstad, Henrik Albert
title: Day by day symptoms following positive and negative PCR tests for SARS-CoV-2 in non-hospitalised health-care workers: a 90-day follow-up study
date: 2021-05-20
journal: Int J Infect Dis
DOI: 10.1016/j.ijid.2021.05.032
sha: 72a5fb36123dc0ca220e049e0734a9e0c61d9a4a
doc_id: 877487
cord_uid: z1mb34tt

Objective We aimed to compare symptoms day by day for non-hospitalised individuals tested positive and negative for severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). Methods We followed 210 test-positive and 630 test-negative health-care workers of the Central Denmark Region up to 90 days after the test, April-June 2020. They daily reported COVID-19 related symptoms that were compared graphically and by logistic regression. Results Thirty % of test-positive and close to zero of test-negative participants reported reduced sense of taste and smell during all 90 days (adjusted odds ratio [aOR] 86.07, 95% CI 22.86-323). Dyspnoea was reported by an initial 20% of test-positive declining to 5% after 30 days without ever reaching the level of the test-negative participants (aOR 6.88, 95% CI 2.41-19.63). Cough, headache, sore throat, muscle pain, and fever were temporarily more prevalent among the test positive participants; after 30 days, no increases were seen. Women and older participants were more susceptible to COVID-19 symptoms. Conclusion Prevalence of long-lasting reduced sense of taste and smell is highly increased in mild COVID-19 patients. This pattern is also seen for dyspnoea at a low level but not for cough, sore throat, headache, muscle pain, or fever.

studies of non-hospitalised COVID-19 patients including a reference group accounting for symptoms not attributable to SARS-CoV-2 are warranted (Yelin et al., 2020) . The few studies comparing symptom courses of test positive with test negative non-hospitalised participants show increased occurrence of reduced sense of taste and smell and several other symptoms that persist for several weeks and months after a positive SARS-CoV-2 test (Cirulli et al., 2020 , Mizrahi et al., 2020 . We aimed to compare day by day symptoms of SARS-CoV-2 PCR test-positive and test-negative non-hospitalised health-care workers up to 90 days after the test.

We carried out a prospective follow up study of health care workers and other occupational groups from all hospitals in the Central Denmark Region from April 24 until June 30, 2020.

All hospital employees were invited by e-mail to report COVID-19 related symptoms day by day. Participants tested by PCR for SARS-CoV-2 from March 11 until June 30, 2020, at any of the regional hospitals or public test centres, were identified in the Central Denmark Region business intelligence system. We included those with at least one daily report on symptoms from the day being tested and onwards. We excluded those hospitalised for COVID-19 for more than 24 hours because our focus was non-hospitalised individuals.

National surveillance in Denmark of SARS-CoV-2 infection assessed by reverse transcription PCR-based detection of viral RNA in nasopharyngeal and oropharyngeal swabs was initiated on March 2, 2020 (Statens Serum Institut, 2020a, 2020b). Until March 11, only symptomatic J o u r n a l P r e -p r o o f individuals returning from high-risk areas and symptomatic contacts could be tested. From

March 12, also individuals with severe symptoms, individuals at-risk because of high age or comorbidity, or with critical functions could be tested. From April 1, further individuals with mild symptoms; and from April 21, close contacts regardless of symptoms had the opportunity to be tested. From May 18, all adults have been offered testing. Since April 21, all patients have been tested before being admitted to the hospital or undergoing high-risk procedures during outpatient visits.

PCR analysis for SARS-CoV-2 RNA was performed at the Clinical Microbiology Department at Aarhus University Hospital with detection of the ORF-1a/b and E-gene (commercial assay) or exclusively the E-gene, and at the national test-facilities at the TestCentre Denmark, Statens Serum Institut, with detection of the E-gene, both in-house PCRs in accordance with the Charité protocol recommended by the WHO (Corman et al., 2020 , Vogels et al., 2020 .

Automated RNA extraction was performed at both facilities. Internal negative and positive controls were included in both the RNA extraction step and in the reverse transcription PCR step.

After giving informed consent, participants received a short baseline questionnaire and then a short text message on their mobile phone or by e-mail every day at 3:30 pm linking to a questionnaire regarding the presence (yes, no) of the following symptoms within the previous 24 hours: cough, sore throat, headache, fever, muscle aches and pains, dyspnoea, and reduced or lost sense of taste and smell (available in supplementary data). Participants could respond within 24 hours from receiving the message and could resume reporting if skipping one or more days. Smoking status was collected in the baseline questionnaire.

Information on occupation, sex, and age was provided by the business intelligence institution of the Central Denmark Region.

We followed participants from the date of the first completed questionnaire after the first positive test, else from the first negative test until the date of the last questionnaire, 90 days after being tested, or June 30, whichever came first. No participants had a positive test after a negative test during the follow-up period.

Because the indication for being tested, testing rate, and infection rate in the study population changed rapidly over time (Supplementary Figure S1 ) we for each participant tested positive, randomly selected three controls with replacement among participants tested negative matching on sex and testing date (+/-2 days). The three-fold number of controls was defined by the maximum allowed within the narrowest strata. When selecting controls, we avoided crossing the specific dates where indications for being tested changed as specified above.

For test-positive and test-negative participants, we computed the prevalence of the seven symptoms as well as any of the symptoms for each day of follow-up. We plotted the prevalences and smoothed the curves with local three-degree polynomial kernels. Standard error based 95% confidence intervals (95% CI) were obtained based on 100 bootstrap samples, resampling among the test-positive participants and repeating the matching of testnegative participants and the smoothing procedure.

We estimated odds ratios (OR) of any symptom and the seven specific symptoms by test result (positive, negative) for three time periods (0-30, 31-60, 61-90 days) since the test by J o u r n a l P r e -p r o o f conditional logistic regression matched by sex and testing date as specified above. We assessed if sex modified the symptom prevalence among test-positive relative to test-negative participants by including an interaction term between test result and sex (male, female). We also assessed the possible modifying effect of age (<45, ≥45 years, the median age) and testing date (≤April 7, >April 7, the median testing date) similarly. We assessed selection bias, i.e. if test-positive and test-negative participants' responding on the questionnaire on a given day were modified by the presence of symptoms the previous day, in a model that included test result, any symptom (present, absent), the interaction term between the two, and responding on the questionnaire (yes, no). The conditional logistic regression models were adjusted for age (<30, 30-39, 40-49, 50-59, and ≥60 years) , except analyses of effect modification by age, occupation (nursing staff, medical doctors, biomedical laboratory scientists, medical secretaries, and other), smoking (current, previous, and never), unless else specified. Overall odds ratios for the entire follow up period were furthermore adjusted by time since test (0-30, 31-60, 61-90 days). The covariates were decided on a priori. Confidence intervals were obtained by bootstrapping as described above. Data handling and statistical analyses were performed in Stata 16.1.

Between April 23 and May 5, 32 413 health-care workers and administrative personnel were invited to participate in the day by day symptom reporting, and 12 115 (37·4%) accepted.

Between March 11 and June 30, 215 respondents were tested PCR-positive for SARS-CoV-2, and 3421 were tested PCR-negative. Five of the test-positive and four of the test-negative participants were hospitalised for >24 hours on the suspicion of COVID-19 and were excluded. Among the remaining 3417 test-negative participants, we randomly selected 630 referents matched on sex and testing date and representing 447 individuals. The study population then included 210 test-positive and 630 test-negative participants. Two referents J o u r n a l P r e -p r o o f were selected five times, the maximum number of repeats observed. Data from a mean of 50 test-positive and 164 test-negative participants were included for day 0-30, 128 and 431 for day 31-60, and 87 and 300 for day 61-90 since the test.

The sex and testing date distributions were identical for test-positive and test-negative participants, as expected due to the matched design, and only minor differences in age, smoking habits, and time from test to responding on the first questionnaire were seen (table 1). The mean daily response rate declined from 80.9% and 79.1% during day 0-30 for the testpositive and test-negative participants to 54.6% and 63.1% during day 61-90, respectively.

The nursing staff was relatively more prevalent compared to other occupations (administrative, service, and technical staff, social workers, and other less prevalent occupations) with limited patient contact among the test-positive participants.

During the first days after being tested, about 80% of the test-positive and 75% of the testnegative participants reported at least one of the seven symptoms (figure 1). Ninety days later, these prevalences had gradually declined to about 40% and 10%, respectively. This corresponded with four-fold increased odds ratios for the complete follow-up period (adjusted odds ratio [aOR] 3.79, 95% CI 2.54-5.66) and for each of the three periods since testing (table   2) .

Reduced or lost sense of taste and smell was consistently reported by 30% of the test-positive participants, except for a somewhat higher level during the initial days (figure 1). Almost none of the test-negative participants reported these symptoms. The odds ratio tended to increase by time since testing and the overall estimate was 80-fold increased (aOR 86.07, 95% CI 22.86-323, table 2). Dyspnoea was reported by an initial 20% of test-positive participants and declined gradually to about 5% after 30 days without ever reaching the level J o u r n a l P r e -p r o o f of the test-negative participants (figure 1). During the first 30 days of follow-up, the odds ratio was 11-fold increased (aOR 10.93, 95% CI 2.29-52.10) compared to test-negative participants. This ratio was reduced during subsequent days and an overall adjusted odds ratio of 6.88 (95% CI 2.41-19.63) was observed. Half of the test-positive and 15% of the testnegative participants reported cough during the initial days (figure 1). The adjusted odds ratio for the first 30 days was 2.19 (95% CI 1.10-4.37). After 30 days, no difference between the two test results was observed. At the time of the test, sore throat, muscle aches or pain, and fever were reported by 35%, 30%, and 20% of the test-positive and this was slightly more than among the test-negative participants. No differences were indicated for these symptoms after 30 days of follow-up.

Test-positive participants aged 45 years or older showed an overall five-fold increased odds ratio (aOR 5.37, 95% CI 2.84-10.14) of any symptom compared with same-age test-negative participants (table 3) . The corresponding odds ratio obtained among participants <45 years of age was 2.43, (95% CI 1.42-4.16) and the P-value of the interaction term was 0.07. Similar patterns were seen for day 31-60 and day 61-90, but not for day 0-30. When breaking this analysis down by the seven symptoms, it appeared that this effect modification by age was primarily seen for reduced or lost sense of taste and smell and headache more than 30 days after the test (Supplementary Table S1 ).

Women who tested positive reported any symptom more often than women who tested negative (aOR 4.38, 95% CI 2.90-6.60) while this was not the case for men (aOR 1.44, 95% CI 0.48-4.36, table 3) and the P-value of the interaction term was 0.05. A similar pattern was seen for day 30-60 and day 61-90 but not for day 0-30. After day 30, much higher prevalences of reduced sense of taste and smell were seen for test-positive relative to test-negative women J o u r n a l P r e -p r o o f than for test-positive relative to test-negative men (Supplementary table S2) . A similar pattern was suggested for dyspnoea but at a lower level.

Early vs late testing date (≤April 7 vs >April 7) did not modify the association between a positive test and any symptom (table 3) .

Among study participants reporting any symptom the previous day, those who tested positive did not respond more often on the present-day questionnaire than those tested negative (aOR 0.93, 95% CI 0.75-1.15, table 4). This was also the case among participants reporting no symptoms the previous day (aOR 1.15, 95% CI 0.88-1.51). The P value of the interaction term was 0.19 and indicated that responding to the questionnaire did not depend on the presence of symptoms the previous day and test result.

Women, middle-aged employees, and nursing staff were more prevalent in the study population than in the source population (Supplementary table S3 ).

Nearly one-third of SARS-Cov-2 test-positive and close to zero of test-negative participants reported reduced sense of taste and smell during all 90 days of follow-up. Dyspnoea was reported by an initial 20% of test-positive participants and declined gradually to about 5% after 30 days without ever reaching the level of the test-negative participants. Cough, headache, sore throat, muscle aches, and fever were temporarily higher among the test positive participants, but after 30 days, no increases were seen. Women tended to be more J o u r n a l P r e -p r o o f susceptible to reduced sense of taste and smell and dyspnoea, and participants aged 45 years or older to reduced sense of taste and smell and headache beyond 30 days.

The major limitation is the study participants' awareness of their test results before reporting symptoms, which is expected to have inflated reporting among the test-positive participants.

Such an effect is probably strongest for loss of sense of taste and smell that has contracted public awareness worldwide and nationally (DR (Danish Broadcasting Corporation), 2020, Mahase, 2020) .

Another limitation is only few observations during the first weeks after the test. Hereby the study primarily addresses the course of symptoms after the initial acute phase of the infection.

The prospective design with daily collection of symptom reports that provides information with high temporal resolution is a major strength and makes us able to depict the courses of symptoms day by day. Another strength is the inclusion of a reference group of test-negative participants recruited within the same population as the test-positive health-care workers and tested with the same kit at the same time. This allows us to take symptoms among the testpositive participants not attributed to SARS-CoV-2 infection into consideration and also to account for rapid changes in indications for testing, infection rate, and testing rate in the population. Matching on sex and adjustment for age, smoking, and occupation is expected to have further reduced potential confounding.

Our access to the results of all SARS-CoV-2 tests conducted by the Health Authorities on all samples obtained in the Central Denmark Region during the study period independently of the participants should ensure inclusion of all tested participants and precludes selection or information bias related to testing status. One-third of the invited employees volunteered for J o u r n a l P r e -p r o o f symptom reporting and among them, one third was PCR tested. Relatively more nursing staff participated in the study compared to other occupations with limited patient contact. This should have increased the proportion of test-positive participants, but not have affected the validity of symptom comparisons between test-positive and test-negative participants (Jespersen et al., 2020) .

Indication for a SARS-CoV-2 PCR-test, testing and infection rates changed during the course of the study, and for that reason, we matched participants individually on testing date. We observed no difference in the association between the test result and any symptom among participants tested early vs late during spring 2020, indicating that matching had fulfilled the purpose. We observed no indications that responding to the questionnaire on a given day depended on test results and symptoms the previous day, and this indicates no differential attrition.

Our finding of a highly and constantly increased prevalence of reduced or lost sense of taste and smell among the SARS-CoV-2 test-positive compared with the test-negative participants is partly in accordance with two recent reports from general population samples in Israel and the US including few or no participants hospitalised for COVID-19 (Cirulli et al., 2020) .

Both studies showed initial prevalences among the test-positive participants comparable with ours, but prevalences declined to about 5% after 20 days and to 14% after 90 days, respectively. In both studies, symptom prevalences of lost sense of taste and smell among test-negative participants were constantly close to zero during follow-up in line with our findings. High initial prevalences of altered sense of smell and taste of 60-90% followed by were also seen in a follow-up study of mainly COVID-19 outpatients examined with olfactory and gustatory psychophysical tests (Vaira et al., 2020) . The first days after the test, 85% had taste and smell dysfunction, which gradually declined to 7% 60 days later.

A five-fold increased prevalence of dyspnoea among test-positive compared with testnegative participants (16% vs 3%) 90 days after the test has been reported and is in line with our findings but at a higher absolute level (Cirulli et al., 2020) . Others have reported a constant level of dyspnoea of 30% among test-positive participants during 14-21 days of follow-up in a study that included no reference group, as well as minor difference between test-positive and test-negative participants during 20 days of follow-up (Mizrahi et al., 2020 , Tenforde et al., 2020 .

Increased prevalences of cough, sore throat, body aches, and fever among test-positive relative to test-negative individuals 90 days after the test (Cirulli et al., 2020) , high prevalences of the same symptoms among test-positive individuals 14-21 days after the test (Tenforde et al., 2020) , as well as no relative symptom increase in test-positive individuals 20 days after the test, have been reported (Statens Serum Institut, 2020b) . The latter finding being in line with ours. It should be stressed that our study accounted for the testing date, and this may explain some of the inconsistencies between earlier findings and ours (Cirulli et al., 2020 , Mizrahi et al., 2020 .

Our data suggest that women and older individuals are more susceptible than men and younger individuals to suffer from COVID-19 related symptoms. There is ample evidence of men being more severely affected by COVID-19 than women, and our contradictory findings may point towards explanations other than SARS-CoV-2 infection per se (Scully et al., 2020) .

Conclusion We observe a highly increased prevalence of long-lasting reduced or lost sense of taste and smell among participants diagnosed with mild COVID-19. This pattern is also seen for dyspnoea at a low level but not for cough, sore throat, headache, muscle ache or pain, or fever. Women and participants aged 45 years or older tend to be more susceptible to SARS-CoV-2 infection.

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. 

The data underlying this article will be shared on reasonable request to the corresponding author. a Any symptom includes reduced or lost sense of taste and smell, dyspnoea, cough, headache, sore throat, muscle aches or pain, and fever. b n represents number of responses stating the presence of any symptom within the last 24 hours and % represents the proportion of all responses. c Adjusted odds ratios with 95% confidence intervals (CI) are obtained from conditional logistic regression models with 1:3 matching of test-positive with test-negative participants on testing date (+/-2 days) and sex (male, female). Models include test result (positive, negative), age (<45 years, ≥45 years), smoking (current, previous, and never), occupation (nursing staff, medical doctors, biomedical laboratory scientists, medical secretaries, and other), and the interaction term between test result and age, test-result and sex, or test result and testing date (≤April 7, >April 7). Adjusted odds ratios for day 0-90 are furthermore adjusted by time since the test (day 0-30, 31-60, and 61-90). The conditional logistic regression models provide instantaneous odds ratios that cannot be estimated from the period cumulative numbers and percentages of the a Any symptom includes reduced or lost sense of taste and smell, dyspnoea, cough, headache, sore throat, muscle aches or pain, and fever. b n represents number of responses stating the presence of any symptom within the last 24 hours and % represents the proportion of all responses. c Odds ratios with 95% confidence intervals (CI) are obtained from conditional logistic regression models with 1:3 matching of test-positive with test-negative participants on testing date (+/-2 days) and sex (male, female). Models include test result (positive, negative), age (<30, 30-39, 40-49, 50-59, and ≥60 years) , smoking (current, previous, and never), occupation (nursing staff, medical doctors, biomedical laboratory scientists, medical secretaries, and other), time since the test (day 0-30, 31-60, and 61-90), and the interaction term between any symptom the previous day and test result. The conditional logistic regression model provides instantaneous odds ratios that cannot be estimated from the cumulative numbers and percentages of the table. Confidence intervals are obtained by bootstrapping. d The P value is the p value of the interaction term between any symptom the previous day and test result.

Evolution of Altered Sense of Smell or Taste in Patients With Mildly Symptomatic COVID-19

Persistent Symptoms in Patients After Acute COVID-19

Long-term COVID-19 symptoms in a large unselected population

Detection of 2019 novel coronavirus (2019-nCoV) by real-time RT-PCR

Long-term Health Consequences of COVID-19

Multi-organ impairment in low-risk individuals with long COVID

Features of 20 133 UK patients in hospital with covid-19 using the ISARIC WHO Clinical Characterisation Protocol: prospective observational cohort study

Nurse with Corona: The worst I've experienced (in Danish)

European Centre for Disease Prevention and Control. COVID-19 situation update worldwide, as of

Clinical spectrum of coronavirus disease 2019 in Iceland: population based cohort study

Clinical Characteristics of Coronavirus Disease 2019 in China

6-month consequences of COVID-19 in patients discharged from hospital: a cohort study

SARS-CoV-2 seroprevalence survey among 17,971 healthcare and administrative personnel at hospitals, pre-hospital services, and specialist practitioners in the Central Denmark Region

Covid-19: What do we know about "long covid

The lasting misery of coronavirus long-haulers

Longitudinal symptom dynamics of COVID-19 infection

Living with COVID19

Olfactory and Gustatory Outcomes in COVID-19: A Prospective Evaluation in Nonhospitalized Subjects

Subjective smell and taste changes during the COVID-19 pandemic: Short term recovery

Considering how biological sex impacts immune responses and COVID-19 outcomes

Instructions for handling COVID-19 in the health care system (in Danish)

Attributes and predictors of Long-COVID: analysis of COVID cases and their symptoms collected by the Covid Symptoms Study App

Symptom Duration and Risk Factors for Delayed Return to Usual Health Among Outpatients with COVID-19 in a Multistate Health Care Systems Network -United States

a n represents number of responses stating the presence of the specified symptom within the last 24 hours and % represents the proportion of all responses. b Adjusted odds ratios with 95% confidence intervals (CI) are obtained from conditional logistic regression models with 1:3 matching of test-positive with test-negative participants on testing date (+/-2 days) and sex (male , female). Models are adjusted for age (<30, 30-39, 40-49, 50-59, and ≥60 years), smoking (current, previous, and never) , and occupation (nursing staff, medical doctors, biomedical laboratory scientists, medical secretaries, and other), except for analyses of reduced or lost sense of taste and smell and fever due to unstable estimates that did not provide valid confidence intervals by bootstrapping. Adjusted odds ratios for day 0-90 were furthermore adjusted by time since the test (day 0-30, 31-60, and 61-90). The conditional logistic regression models provide instantaneous odds ratios that cannot be estimated from the period cumulative numbers and percentages of the table.