Table 5. Multivariate analysis of factors affecting seropositivity and neutralizing antibody titer of severe acute respiratory syndrome (SARS) patients.

Variables*†	Seropositivity*		Antibody titer†
	OR (95% CI)	p value	Parameter estimates + SE	p value
Age (y) (n + 1 vs. n)	0.97 (0.94-1.00)	0.065	0.0056 + 0.0079	0.478
Women vs. men	1.24 (0.47-3.3)	0.67	–0.417 + 0.235	0.081
Infection source, known vs. unknown	15.6 (5.9-41.4)	<0.0001	0.248 + 0.313	0.431
Duration of illness (d) (n+1 vs. n, n = 1 through 44 d)	1.08 (1.025-1.143)	0.004	0.0638 + 0.0233	0.008
Time of convalescent-phase serum sample (weeks after fever onset) (n + 1 vs. n, n = 3 through 15 wk)	–	–	0.449 + 0.198	0.026
(Duration of illness) ×(Time of convalescent-phase serum sample)	–	–	–0.005 + 0.0024	0.037
(Time of convalescent-phase serum sample)2	–	–	–0.025 + 0.012	0.042

*Logistic model: age, with every additional year of age, the odds of seropositivity is 0.97 (odds ratio, OR) (see Figure 1); sex, the odds for women to be seropositive is 1.24 (OR) when compared with men; infectious source, the odds of patients with known infection source to be seropositive is 15.6 times that of the patients without known source of infection; duration of illness, for every additional day of illness, the odds of seropositivity increases by 1.08.
†Linear mixed model: log₂ (neutralizing antibody titer) = β₀ + β₁ (age) – β₁ (sex) + β₃ (infection source) + β₄ (duration of illness) + β₅ (time of convalescent-phase serum sample) –β₆ (duration of illness ×time of convalescent-phase serum sample) –β₇ (time of convalescent-phase serum sample)². In results above, the model estimates are based on log₂ (titers), to which the time of convalescent-phase serum collection (in weeks postonset of illness, starting from week 3) contributed in 3 terms; the antibody rise follows the first order of weeks postonset, and decay follows the second order of weeks postonset and an interactive term between duration of illness and weeks postonset.