Appendix 2. Sensitivity Analysis

Sensitivity Analysis / Rationale
1. Restricted to patients receiving antimicrobials in the first 24 hours / Patients who receive antimicrobials greater than 24 hours after an episode of hypotension might be different than those receiving antimicrobials within the first 24 hours; for example, having a more subtle clinical presentation, or, have specific chronic illnesses in which hypotension may be overlooked.
2. Restricted to patients admitted via emergency department / Hospitals might employ a variety of vital sign charting techniques for inpatients that could differentially influence the identification of a first episode of hypotension.
3. Restricted to less complex patients (patients with no more than one co-morbidity, and no immunosuppression) / Academic hospitals might care for inherently more complex patients with a higher baseline risk of death.
4. Restricted to patients younger than 65 years old / Some hospitals may systematically admit patients with older age (e.g. if they are located near nursing homes or certain communities with older populations) and these patients may engender systematically different clinical approaches or have different outcomes.
5. Restricted to the most common types of infections (pneumonia and intra-abdominal infections) / Clinical syndromes may differ in the signs and symptoms of sepsis.
6. Restricted to the later years of the cohort (2008 to 2012) / As hospitals entered the cohort at different times, we postulated that secular trends could bias our results.

Fig e1. Main model and sensitivity analysis.

Fig e2. Main model and sensitivity analysis, with hospitals clustered at the country level.