Baseline thrombocytopenia and 48-hour respiratory-support burden in presumed neonatal late-onset sepsis: an exploratory MIMIC-III NICU cohort study

Baseline thrombocytopenia is often treated as a marker of severe neonatal late-onset sepsis (LOS), but platelet count may also reflect prematurity or baseline illness. We examined whether baseline thrombocytopenia was associated with a predefined 48-hour respiratory-support burden endpoint in MIMIC-III and whether interpretation changed after accounting for baseline respiratory support. We conducted an exploratory retrospective cohort study using the neonatal subset of MIMIC-III. For each infant, we included the first antibiotic-microbiology-derived, blood-specimen-linked presumed LOS episode occurring > 72 h after birth. Baseline thrombocytopenia was platelet count < 150 × 10⁹/L within 6 h before onset. The endpoint was 28-day mortality or persistent mechanical ventilation 48 h after onset. Associations were estimated using complete-case logistic regression, including a prespecified analysis among infants not ventilated at onset and a contextual sensitivity model including ventilation at onset. Among 397 infants, 174 (43.8%) met the endpoint; 172 had persistent ventilation at 48 h and 8 died by 28 days. Platelet count was available in 328 infants; 28 had thrombocytopenia. In the full-cohort complete-case model ( n = 315), thrombocytopenia had an imprecise adjusted association with the endpoint (aOR 1.34, 95% CI 0.40–5.09). Ventilation at onset was strongly associated with the endpoint; in the contextual model, baseline ventilation had a large adjusted association (aOR 30.19, 95% CI 13.48–73.06), while the platelet estimate remained imprecise (aOR 1.17, 95% CI 0.27–5.49). The non-ventilated subgroup gave the clearest clinical contrast but was underpowered ( n = 183; 21 events; aOR 1.42, 95% CI 0.17–8.32). PIC supported variable-availability assessment only, not endpoint replication. In this historical MIMIC-III NICU cohort, baseline thrombocytopenia did not yield a precise independent association with the endpoint. The endpoint mainly reflected persistent baseline ventilation, and the thrombocytopenic subgroup was small. These findings cannot exclude clinically important platelet associations; larger contemporary cohorts using respiratory escalation or ventilator-free days are needed.