Background: Treatment guidelines recommend the tocilizumab use in patients with a CRP of >7.5 mg/dL. We aimed to estimate the causal effect of glucocorticoids + tocilizumab on mortality overall and after stratification for PaO2/FiO2 ratio and CRP levels. Methods: This was an observational cohort study of patients with severe COVID-19 pneumonia. The primary endpoint was day 28 mortality. Survival analysis was conducted to estimate the conditional and average causal effect of glucocorticoids + tocilizumab vs. glucocorticoids alone using Kaplan–Meier curves and Cox regression models with a time-varying variable for the intervention. The hypothesis of the existence of effect measure modification by CRP and PaO2/FiO2 ratio was tested by including an interaction term in the model. Results: In total, 992 patients, median age 69 years, 72.9% males, 597 (60.2%) treated with monotherapy, and 395 (31.8%), adding tocilizumab upon respiratory deterioration, were included. At BL, the two groups differed for median values of CRP (6 vs. 7 mg/dL; p < 0.001) and PaO2/FiO2 ratio (276 vs. 235 mmHg; p < 0.001). In the unadjusted analysis, the mortality was similar in the two groups, but after adjustment for key confounders, a significant effect of glucocorticoids + tocilizumab was observed (adjusted hazard ratio (aHR) = 0.59, 95% CI: 0.38–0.90). Although the study was not powered to detect interactions (p = 0.41), there was a signal for glucocorticoids + tocilizumab to have a larger effect in subsets, especially participants with high levels of CRP at intensification. Conclusions: Our data confirm that glucocorticoids + tocilizumab vs. glucocorticoids alone confers a survival benefit only in patients with a CRP > 7.5 mg/dL prior to treatment initiation and the largest effect for a CRP > 15 mg/dL. Large randomized studies are needed to establish an exact cut-off for clinical use.