Long Covid Symptom Clusters, Correlates and Predictors in a Highly Vaccinated Australian Population in 2023

Abstract

Background: Limited data exists regarding long Covid burden following Omicron infection in highly vaccinated populations.

Objective: To (1) characterise long Covid prevalence and predictors and (2) identify key symptom clusters and their correlates among long Covid patients, during an Omicron-predominant period in a highly vaccinated population.

Design: Anonymous, online, cross-sectional survey.

Setting: January 2023, Australia.

Participants: Residents aged ≥ 18 years with self-reported history of test-positive Covid-19. The main variables studied were socio-demographic characteristics, Covid-19 risk factors, vaccination, infection history and experiences with long Covid.

Main outcome measures: Long Covid symptoms. Symptom-based clustering was used to identify long Covid symptom clusters and their functional correlates. Predictors of long Covid occurrence and severity were assessed using multivariable logistic regression.

Results: Overall, 240/1205 participants (19.9%) reported long Covid. Long Covid risk was significantly higher for women OR 1.71 (95% CI: 1.17-2.51), people with comorbidities 2.19 (95% CI: 1.56-3.08) and those using steroid inhalers for Covid-19 treatment (2.34 [95% CI: 1.29-4.24]). Long-Covid risk increased with increasing Covid-19 infection severity (moderately severe symptoms: 2.23 [95% CI: 1.50-3.30], extremely severe symptoms: 5.80 [95% CI: 3.48-9.66], presented to ED/hospitalised: 7.22 [95% CI: 3.06-17.03]). We found no significant difference in the likelihood of long Covid between the Omicron and pre-Omicron periods, vaccination status and participant age. We identified two long Covid clusters (pauci-symptomatic, n = 170, vs. polysymptomatic, n = 66). Polysymptomatic cluster membership was associated with worse functioning (impacts on work, moderate activity, emotions and energy). Severity acute infection was strongly predictive of polysymptomatic cluster membership (5.72 [2.04-17.58]). Monoclonal antibody treatment was strongly associated with pauci-symptomatic cluster membership (0.02 [0.00-0.13]).

Discussion: Our study shows that long Covid is an important health burden in Australia, including during the Omicron era, and identifies several risk factors. We found a subgroup of patients characterised by more symptoms and worse functional outcomes. Our findings can inform policies for protecting vulnerable populations and frameworks for long Covid risk assessment and management.

Conclusions: One-in-five people may suffer long Covid after acute Covid-19 infection, with similar risk across age groups. Omicron variants appear not to have a lower risk compared to earlier variants in our study. A cumulative number of symptoms can help triage long Covid patients.

Patient or public contribution: We did not involve patients or the public in the design of the questionnaire. However, after a soft launch, we revised some survey questions by reviewing early responses from patients and the public.

Keywords: Australia; Covid‐19; Omicron; SARS‐CoV‐2; long Covid.

Figure 1

Study population and inclusion and exclusion criteria.

Figure 2

Predictors of long Covid—multivariable regression model (Model 2).

Figure 3

Characteristics of long Covid participants belonging to the pauci‐symptomatic and polysymptomatic clusters at the time of survey. Kruskal–Wallis test was applied to non‐normally distributed continuous variables and ordinal variables. Pearson's χ ² was applied to categorical predictor variables. Categories with cell counts less than 5 were excluded from analysis, which included ‘Gender: Non‐binary’. For a detailed frequency table, see Table S13. *p value < 0.05, **p value < 0.001.

Figure 4

Forest plot showing adjusted odds ratios and 95% confidence interval for predictors of polysymptomatic cluster membership (PAM method). Blue squares indicate the odds ratio, and horizontal lines depict the 95% confidence interval.

Figure 5

(A–D) Functional correlates and post‐COVID outcomes of long COVID participants belonging to the pauci‐symptomatic and polysymptomatic clusters. Kruskal–Wallis test was applied to non‐normally distributed continuous variables and ordinal variables. Pearson's χ ² was applied to categorical predictor variables. Categories with cell counts less than 5 were excluded from analysis, which included ‘Change in health self‐rating post‐COVID: Improved’. For a detailed frequency table, see Table S14. *p value < 0.05, **p value < 0.001.