Broadening symptom criteria improves early case identification in SARS-CoV-2 contacts

Abstract

Background: The success of case isolation and contact tracing for the control of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) transmission depends on the accuracy and speed of case identification. We assessed whether inclusion of additional symptoms alongside three canonical symptoms (CS), i.e. fever, cough and loss or change in smell or taste, could improve case definitions and accelerate case identification in SARS-CoV-2 contacts.

Methods: Two prospective longitudinal London (UK)-based cohorts of community SARS-CoV-2 contacts, recruited within 5 days of exposure, provided independent training and test datasets. Infected and uninfected contacts completed daily symptom diaries from the earliest possible time-points. Diagnostic information gained by adding symptoms to the CS was quantified using likelihood ratios and area under the receiver operating characteristic curve. Improvements in sensitivity and time to detection were compared with penalties in terms of specificity and number needed to test.

Results: Of 529 contacts within two cohorts, 164 (31%) developed PCR-confirmed infection and 365 (69%) remained uninfected. In the training dataset (n=168), 29% of infected contacts did not report the CS. Four symptoms (sore throat, muscle aches, headache and appetite loss) were identified as early-predictors (EP) which added diagnostic value to the CS. The broadened symptom criterion "≥1 of the CS, or ≥2 of the EP" identified PCR-positive contacts in the test dataset on average 2 days earlier after exposure (p=0.07) than "≥1 of the CS", with only modest reduction in specificity (5.7%).

Conclusions: Broadening symptom criteria to include individuals with at least two of muscle aches, headache, appetite loss and sore throat identifies more infections and reduces time to detection, providing greater opportunities to prevent SARS-CoV-2 transmission.

FIGURE 1

Flow diagram for the inclusion and exclusion of INSTINCT and ATACCC study participants for each analysis. Cohort A was used for a time-to-event analysis describing symptom onset in time post-index case symptom onset (ISO). Cohorts B and C were used to create Spiegelhalter Knill-Jones (SKJ) models for each study day. Cohort D was used to evaluate the performance of simple case definitions at different time-points following exposure. 383 participants were recruited within INSTINCT, of which 138 were indexes and were excluded. Six out of 245 contacts were excluded because of missing PCR data or enrolment symptom data and one was excluded as they became PCR-positive after study day 7. 73 out of 238 contacts were PCR-positive and assigned to the “infected” group. 43 out of 165 PCR-negative contacts were seropositive at study day 0 or 7 (possible prior infection or vaccination), and two out of 165 seroconverted at day 27 (possible separate exposure event) and were excluded. None of the 165 PCR-negative participants seroconverted at study day 14. 25 out of 165 PCR-negative participants had no serology data available and were excluded. 21 out of 168 contacts were excluded from cohort A due to missing ISO date and one was excluded as they reported symptoms several days prior to ISO. 411 contacts were recruited in ATACCC. 34 were excluded because of missing PCR or enrolment symptom data and one was excluded as they became PCR-positive after study day 7. 15 were excluded because they were PCR-positive at only one time-point with a high cycle threshold (C _t) value. 91 of the remaining 361 contacts had at least one positive PCR result by study day 7. Six out of 361 were excluded from cohort D because of missing exposure date.

FIGURE 2

Adjusted likelihood ratios (LRs) for individual symptoms within symptom combinations. Symptoms were considered as a series of binary tests based on their occurrence by each study day. Spiegelhalter Knill-Jones (SKJ) models were created using a) three predictors (fever, cough and anosmia) to evaluate the diagnostic performance of the canonical symptoms (CS) and b) four predictors to evaluate the effect of adding one of the nine candidate variables to the CS (fever, cough, anosmia and candidate). Models were created for the day of enrolment and each of the first 7 study days. Positive and negative LRs for the presence or absence of each symptom by each study day were calculated (supplementary material S11) and then adjusted for dependency with the other predictors within the model to measure the independent predictive value of each symptom within the symptom combination. See supplementary material S4 for a full description of the SKJ method and a worked example. Adjusted LRs for study days 0, 2 and 4 are presented in table 1. Adjusted positive LRs are shown on the left and adjusted negative LRs are shown on the right. In each plot the horizontal line drawn is drawn at 1: LRs above the line increase post-test odds and LRs below the line reduce post-test odds. Bootstrap confidence intervals for adjusted LRs could not be calculated because some bootstrap iterations resulted in samples with singularities.

FIGURE 3

Direct comparison of the area under the receiver operating characteristic curve (AUC) in training and test datasets. Symptoms were considered as a series of binary tests based on their occurrence by each study day. A series of Spiegelhalter Knill-Jones (SKJ) models were created, one for each study day, using three predictors (fever, cough and anosmia) to evaluate the canonical symptoms (CS) (black data points). Nine further series of models were created using four predictors (fever, cough, anosmia and candidate) to evaluate the effect of adding one of the nine candidate variables to the CS at each study day (coloured data points). Model predictions were evaluated in training and test datasets by calculating the AUC. AUC_Train: AUC in training dataset (figure 1, cohort B). AUC_Test: AUC in test dataset (figure 1, cohort C). The solid line marks AUC_Test for the CS model. The dashed line marks AUC_Train for the CS model. Models where the addition of a candidate symptom yielded better predictions in the training dataset lie to the right of the dashed line and models where better predictions were yielded in the test dataset lie above the solid line.

FIGURE 4

a) Sensitivity, b) specificity and c) number needed to test (NNT) for the canonical symptoms (CS) and broadened symptom criteria. The early-predictors (EP) (sore throat, headache, muscle aches and appetite loss) were each combined individually with the CS (fever, cough and anosmia) using an “OR” operator and all were added together using “AT LEAST” and “OR” operators (as described in box 1). Sensitivity, specificity and NNT were calculated for each symptom criterion by day post-exposure (index case symptom onset for household contacts) against a serial PCR reference standard. Full results are given in supplementary material S13. NNT is calculated by dividing the number of false-positives by the number of true-positives and adding 1. Rarely, symptoms were reported at enrolment without an onset date. We imputed onset dates for these symptoms by assuming the median number of days pre-enrolment (maximum two participants (0.55%) for rhinitis).

FIGURE 5

Kaplan–Meier plots showing time from exposure until positive identification by the canonical symptoms (CS) and broadened symptom criteria. The early-predictors (EP) (sore throat, headache, muscle aches and appetite loss) were a) each combined individually with the CS using an “OR” operator and b) all were added together using “AT LEAST” and “OR” operators (as described in box 1). The proportion of PCR-positive (left) and PCR-negative (right) participants who were positively identified by each case definition by each day following exposure (index case symptom onset for household contacts) is shown using a Kaplan–Meier plot. The plot for the CS is shown in black. Life-tables are presented in supplementary material S14 and median time to diagnosis in table 2. Rarely, symptoms were reported at enrolment without an onset date. We imputed onset dates for these symptoms by assuming the median number of days pre-enrolment (maximum two participants (0.55%) for rhinitis).