Introduction
The long-term effects of COVID-19, often termed 'long-COVID', pose a significant challenge to healthcare systems worldwide. Understanding the long-term trajectory of symptoms following SARS-CoV-2 infection is crucial for effective resource allocation and treatment strategies. Prior research on long-COVID has been hampered by several limitations. Many studies have reported prevalence at a single timepoint, hindering understanding of symptom progression. Others have lacked appropriate comparison groups, making it difficult to differentiate symptoms genuinely attributable to the infection from pre-existing conditions or coincidental illnesses. Furthermore, studies have often focused on specific patient subgroups (e.g., hospitalized patients, older adults) or particular outcomes (e.g., mental health), limiting generalizability. The inconsistencies across previous studies regarding symptom trajectories—showing improvement, no change, or deterioration over time—highlight the need for a robust, large-scale, population-based study to clarify the natural history of long-COVID. The aim of this study is to address these gaps by providing a comprehensive analysis of symptom changes and recovery trajectories in a nationwide cohort of individuals with and without confirmed SARS-CoV-2 infection, using a rigorous methodology that includes a contemporaneous control group.
Literature Review
Existing literature on the natural history of long COVID is characterized by heterogeneity in study design, population selection, and follow-up duration. Many studies report prevalence of long COVID at a single time point, providing a snapshot rather than a dynamic view of symptom progression. Some studies adjusted for pre-existing symptoms, yet the majority lacked a proper comparison group, making it challenging to separate true long-COVID symptoms from pre-existing or coincidental conditions. Studies employing serial exposure measurements have mostly been restricted to selective groups (hospitalized patients, older patients, or veterans with breakthrough infections) or specific outcomes (mental health), further limiting generalizability. Reports on symptom trajectory vary; some hospital cohorts found no change, while others observed improvement or deterioration over time. Even studies with longer follow-up periods (e.g., up to one year) have shown mixed results regarding the proportion of patients achieving full recovery. Meta-analyses, while offering a broader perspective, are often limited by the quality and heterogeneity of included studies. These inconsistencies underscore the need for a large, population-based study with a robust control group to accurately characterize long-COVID's natural history and identify patterns of symptom change.
Methodology
The Long-COVID in Scotland Study (Long-CISS) employed an ambidirectional, general population cohort design. Participants were identified using the National Health Service Scotland's SARS-CoV-2 PCR results notification platform. Every adult (≥16 years) in Scotland with a positive PCR test from April 2020 onwards was invited to participate via automated SMS text messages. A comparison group of individuals with negative PCR tests and no subsequent positive tests were also recruited. Participants completed self-reported questionnaires at 6, 12, and 18 months post-index test (PCR test date), collecting data on pre-existing health conditions and current health status. The questionnaires included a comprehensive list of 26 symptoms frequently associated with long-COVID, along with assessment of overall recovery status (full recovery, partial recovery, no recovery) and quality of life (EQ-5D-5L). Data linkage to various national databases was employed to obtain additional information on demographics, socioeconomic status (Scottish Index of Multiple Deprivation - SIMD), vaccination status, pre-existing conditions, and infection severity (hospitalization). Statistical analysis involved descriptive statistics, χ² tests, Mann–Whitney U tests, binary logistic regression (to identify factors associated with improvement/deterioration in recovery status), McNemar's tests (to compare symptom changes between groups), and Poisson regression (to analyze changes in EQ-5D scores). All analyses were adjusted for potential confounders (age, sex, deprivation, ethnicity, pre-existing conditions, vaccination status, and dominant SARS-CoV-2 variant).
Key Findings
Of 4,409,590 questionnaires sent, 345,673 (9%) were completed. After exclusions (previous unrecorded positive tests, asymptomatic infections, recruitment beyond 6 months), 160,781 eligible participants remained (80,332 with symptomatic SARS-CoV-2 infection, 80,449 never infected). At 6 months post-infection, 49.5% of those previously infected reported incomplete recovery; this remained largely unchanged at 12 and 18 months. However, individual symptom trajectories varied significantly. While altered taste, smell, and confusion decreased over time among the infected group, late-onset dry productive cough and hearing problems increased. These changes persisted after adjusting for confounders and pre-existing conditions. The prevalence of at least one symptom remained similar between 6 and 12 months (72% vs. 71%) in the infected group but increased among the never-infected group (64% to 71%). Furthermore, median EQ-5D scores decreased slightly over time in both the infected and never-infected groups, with the infected group exhibiting a larger fall. Socioeconomic factors and pre-existing mental health conditions were associated with recovery status and symptom changes. Specifically, depression prior to SARS-CoV-2 infection and socioeconomic deprivation were more common among people who reported deterioration in recovery status between 6 and 12 months.
Discussion
This study offers a comprehensive and nuanced understanding of long-COVID's natural history in a large, general population cohort. The stability of overall recovery status at 12 and 18 months, despite fluctuating individual symptoms, suggests a complex interplay of persistent and resolving effects following SARS-CoV-2 infection. The identification of late-onset cough and hearing problems warrants further investigation into potential pathophysiological mechanisms. The comparison group was crucial in distinguishing genuine long-COVID effects from the expected natural course of illness. The study's findings underscore the importance of individualized symptom assessment, rather than relying solely on composite long-COVID outcomes. The association between pre-existing conditions and long-COVID trajectories highlights the need for risk stratification and targeted interventions. These findings have important implications for clinical practice, resource allocation, and public health strategies related to COVID-19.
Conclusion
This large, nationwide study provides compelling evidence regarding the natural history of long-COVID. While many individuals experience relatively stable long-COVID symptoms over time, others show improvement or deterioration. The observed improvements in altered taste, smell, and confusion are encouraging. Conversely, the identification of late-onset cough and hearing problems necessitates further research to understand their underlying causes and develop effective interventions. Future studies should investigate the biological mechanisms behind these symptom trajectories and explore the effectiveness of specific interventions for improving long-COVID outcomes.
Limitations
Despite its strengths, the study has limitations. Differential attrition by exposure status could not be fully assessed due to data sharing restrictions. The possibility of undetected asymptomatic SARS-CoV-2 infections in the comparison group cannot be entirely ruled out. Although efforts were made to mitigate this by excluding participants who reported COVID-19 despite having only negative PCR tests, a small degree of misclassification may remain. As with any observational study, residual confounding from unmeasured factors may influence the findings.
Related Publications
Explore these studies to deepen your understanding of the subject.