Table 3.

Multivariable Cox model for hospital admission or death among non-complicated COVID-19 patients treated with early antiviral treatment

	HRadj (95% CI)a	p-value
Vaccination
No	Reference
Yes	0.31 (0.14–0.70)	0.005
Covid-19 severity
Mild	Reference
Moderate	12.23 (4.65–32.18)	< 0.001
Time from symptoms onset to treatment (days) b (with time-varying HR) c
0–1	Reference
2	0.22 (0.05–1.11)	0.068
3	0.51 (0.04-6.00)	0.593
More than 4	0.09 (0.01–0.66)	0.019

^aSurvey-weighted Cox regression model with inverse-probability weighting and design-based standard errors. HR_adj: Hazard Ratio adjusted for sex

^bCategories are based on quartiles

^cTime-varying HR (HR_tv) was calculated using: HR_tv=exp [β₁X + β₂Xlog(t)] = HR(X)*exp[β₂Xlog(t)]. For each category of time from symtpoms onset to treatment, we considered:

-category “2 days”: β₁= -1.50, β₂ = 0.70, HR = 0.22

HR_tv=exp[-1.50 + 0.70*log(t)] = 0.22*exp[0.70*log(t)]

-category “3 days”: β₁= -0.67, β₂ = 0.54, HR = 0.51

HR_tv=exp[-0.67 + 0.54*log(t)] = 0.51*exp[0.54*log(t)]

-category “More than 4 days”: β₁= -1.18, β₂ = 0.08, HR = 0.09

HR_tv=exp[-1.18 + 0.08*log(t)] = 0.09*exp[0.08*log(t)]

The plot of these HR_tv is reported in Fig. 2