TABLE 3.
Predictors of change in respiratory outcomes and QoL between 3 and 6 months after coronavirus disease 2019 (COVID-19) symptom onset
Model | Outcome | Predictor (at 3 months) | Coefficient | 95% CI | p-value | Prespecified covariates |
1a | DLCO % predicted | Ground glass score | −2.2 | −5.9–1.4 | 0.23 | Sex, age, smoking pack-years |
Time | 0.8 | −0.5–2.0 | 0.21 | |||
Ground glass score*Time | 0.6 | 0.05–1.2 | 0.03 | |||
1b | DLCO % predicted | Reticulation score | −8.7 | −12.1– −5.4 | <0.001 | Sex, age, smoking pack-years |
Time | 1.2 | 0.1–2.2 | 0.03 | |||
Reticulation score*Time | 0.6 | −0.01–1.25 | 0.05 | |||
1c | DLCO % predicted | Total CT score | −5.3 | −8.7– −1.8 | 0.003 | Sex, age, smoking pack-years |
Time | 0.8 | −0.5–2.2 | 0.23 | |||
Total CT score*Time | 0.5 | −0.04–1.05 | 0.07 | |||
2a | UCSD | Ground glass score | 0.8 | −3.4–5.0 | 0.69 | Sex, age, smoking pack-years |
Time | −0.2 | −2.9–2.5 | 0.88 | |||
Ground glass score*Time | −0.1 | −1.4–1.1 | 0.87 | |||
2b | QoL | Ground glass score | −0.003 | −0.05–0.04 | 0.89 | Sex, age, smoking pack-years |
Time | −0.003 | −0.02–0.02 | 0.77 | |||
Ground glass score*Time | 0.01 | −0.002–0.02 | 0.15 | |||
3a | UCSD | Unexplained dyspnoea | 5.16 | −7.95–18.3 | 0.43 | Sex, age, smoking pack-years |
Time | −0.45 | −1.98–1.08 | 0.56 | |||
Unexplained dyspnoea*Time | 1.01 | −2.98–5.01 | 0.61 | |||
3b | QoL | Unexplained dyspnoea | 0.01 | −0.11–0.13 | 0.88 | Sex, age, smoking pack-years |
Time | 0.01 | −0.001–0.02 | 0.07 | |||
Unexplained Dyspnoea*Time | −0.01 | −0.04–0.01 | 0.33 |
Time was categorised as 3 months (reference) and 6 months from symptom onset. Ground glass, reticulation and total CT scores were continuous variables that were log-transformed to make them normally distributed and to meet model assumptions. Unexplained dyspnoea at 3 months (defined as the presence of a UCSD dyspnoea score >10 with DLCO % predicted ≥80%) was categorical (present or absent). Time was included as an interaction term to evaluate whether time modified the effect of the primary predictor on the outcome. The primary predictor variables are denoted in bold. Example of Model 1a interpretation: the coefficient of −2.2 for ground glass score indicates that each 1% increase in ground glass score is associated with a 2.2% decrease in DLCO % predicted. This association is modified by time. At 6 months, for each 1% increase in ground glass, the coefficient will increase by 1.4 (0.8+0.6), which means there will be a 0.8% (−2.2+1.4) decrease in DLCO % predicted at 6 months compared to a 2.2% decrease at 3 months. DLCO : diffusing capacity of the lung for carbon monoxide; UCSD: University of California, San Diego shortness of breath questionnaire; QoL: quality of life. Asterisks (*) signify interaction terms.