Study population and setting
This retrospective cohort study explored the symptoms of COVID-19 disease longitudinally. This study used self-reported symptoms (collected from January 3to July 6, 2020) from Israel’s nationwide health survey in conjunction with primary care electronic health record (EHR) data from Maccabi Health Services. The study population included 95,663 adult patients (4,006 SARS-CoV-2 positive; 91,597 negative) and 21,567 pediatric patients (862 positive; 20,705 negative) who had a primary care visit with a positive RT-PCR test for SARS-CoV-2, as well as 11,483 adult survey respondents with SARS-Cov-2 test results (499 positive; 10,984 negative). The authors analyzed symptom data from each source separately. Standard regression adjustment and inverse probability weighting were used to adjust estimates for possible confounding variables including age, gender, presence of chronic medical conditions, and time relative to PCR testing. The authors estimated symptom prevalence among COVID-19 positive patients and identified discriminative symptoms for COVID-19.
Summary of Main Findings
Results between primary care EHR records and self-reported symptoms on survey were frequently discordant. The most prevalent adult symptoms recorded by primary care physicians were emotional disturbance (15.9%), cough (11.6%), fever (10.3%), myalgia (7.7%), and fatigue (5.9%). Prevalent adult self-reported symptoms included cough (21%), fatigue (19%), rhinorrhea/nasal congestion (17%), headache (16%), and myalgia (11%). The most common pediatric symptoms recorded by primary care physicians were fever (7%), cough (5.5%), abdominal pain (2.4%), and fatigue (2.3%). Among recovered patients (n=2,045), mean disease duration was 23.5 +/- 9.9 days. Children experienced shorter disease duration than adults (hazard ratio [HR] for recovery 1.18; 95% CI 1.01-1.39). The following symptoms were present in at least 10 positive cases and more prevalent in the three weeks prior to testing among COVID-19 positive adults than among negative adults: loss of taste or smell (OR 11.18; 95% CI 6.43-19.44 and OR 5.47; 95% CI 3.69-8.09 for self-reported and primary care visit documentation respectively), self-reported headache (OR 2.03; 95% CI 1.29-3.19), self-reported fatigue (OR 1.73; 95% CI 1.08-2.79), as well as syncope (OR 2.09; 95% CI 1.13-3.88), runny nose (OR 2.09; 95% CI 1.47-2.95), and fever (OR 1.62; 95% CI 1.44-1.83) documented by a primary care physician. These results were robust to adjustment for possible confounding variables. Among pediatric cases, no common symptoms experienced in the 3 weeks prior to testing were significantly associated with a positive test.
This large cohort study utilized data from a Health Maintenance Organization that serves a quarter of Israel’s population as well as a nationwide survey, increasing generalizability of these study results across Israel. Utilization of two separate data sources helped compensate for selection and reporting bias in each individual source. EHR data across multiple time points allowed for a longitudinal analysis of multiple symptoms.
This study looked exclusively at patients within the state of Israel, limiting generalizability to other settings. During this study, a major indication for COVID-19 testing was presence of fever; this may have overrepresented patients with fever and associated symptoms. Selection bias is likely to exist in both self-report and physician documentation, as patients with more severe and classic symptoms may be more likely to engage with the healthcare system, to be recommended for COVID-19 testing, and to respond to COVID-19 specific survey questions. Similarly, individuals responding to the nationwide survey may differ systematically from non-responders; for example, they may be more attentive to perceived symptoms. This study is not randomized, and residual confounding is likely to exist. There was poor agreement between symptoms from self-reported data and from EHRs, prompting the authors to analyze the two sources separately; this undermines the potential advantages of linking EHRs with survey data. Individuals who are tested with RT-PCR three weeks after COVID-19 symptom onset are unlikely to test positive; allowing a three-week window between symptoms and testing is thus likely to create considerable misclassification.
This study is among the first to explore longitudinal symptoms of mild to moderate COVID-19 disease in a large nationally representative outpatient cohort.
This review was posted on: 28 January 2021