Chapter 8 246 All Bayesian models were fitted using Markov Chain Monte Carlo (MCMC) with PyMC3[16], implementing a no-u-turn sampler. Four MCMC chains were run with at least 4000 burnin steps and 2000 saved posterior samples. This procedure resulted in a posteriori distribution, from which its mode defined the parameter estimate and its 2.5% and 97.5 quantiles defined the 95% credible interval (CrI). If 1 was not included in the 95%CrI, the parameter estimate was considered statistically significant. Convergence for all parameters were verified by checking trace plots, ensuring their R-hat values were <1.05 with sufficient effective sample size (>200). Full formulations of the models used are listed in the Supplementary Materials. Statistical analyses were performed using Stata (v.15.1, StataCorp LLC, College Station, TX, USA) and the Python PyMC3 programme described above. RESULTS Study population Of 349 enrolled participants by 1 November 2021, 316 had at least 3 months of follow-up, of whom 186 (58.9%) developed PASC (Table 1). Those with PASC were older (p<0.001) and more frequently had moderate or severe/critical COVID-19 (p<0.001), higher BMI (p=0.002) and were more likely to have a lower educational level (p<0.001) compared to those who fully recovered from symptoms within 3 months of illness onset (Table 1). Whilst all participants were unvaccinated for COVID-19 prior to enrolment, the majority of participants had been vaccinated against SARS-CoV-2 by 1 November 2021.
RkJQdWJsaXNoZXIy MTk4NDMw