Clinical and microbiological features of maternal sepsis: a retrospective study

Identifying pregnant women with sepsis is challenging because diagnostic clinical and laboratory criteria overlap with normal pregnant physiologic indices. Our primary study aim was to describe clinical and laboratory characteristics of women diagnosed with sepsis, severe sepsis and septic shock. Our secondary aim was to determine positive predictive values for International Classification of Disease (ICD)-9 billing codes for sepsis, severe sepsis, and septic shock.

After gaining Institutional Review Board approval, we identified women with ICD-9 codes for sepsis, severe sepsis and septic shock who were admitted to a single tertiary obstetric center from 2007-2013. Diagnoses were confirmed using criteria from the International Sepsis Definitions Conference report. Demographic, obstetric, vital signs and laboratory data were abstracted by medical chart review.

We identified 190 women with sepsis-related ICD-9 codes: of these, 35 (18%) women met the criteria for a clinical diagnosis of sepsis, severe sepsis or septic shock. Twenty (57%) women had a sepsis-related diagnosis after cesarean delivery. Twenty-one (60%) women had one or more pre-existing medical conditions and 19 (54%) women had one or more obstetric-related conditions. The genital tract was the most common site of infection. We observed considerable heterogeneity in maternal vital signs and laboratory indices for women with ICD-9 codes for sepsis, severe sepsis, and septic shock. The positive predictive value for each sepsis-related ICD-9 code was low: 16% (95% CI 10 to 24%) for sepsis, 10% (95% CI 3 to 25%) for severe sepsis and 24% (95% CI 10 to 46%) for septic shock.

We identified marked heterogeneity in patient characteristics, clinical features, laboratory indices and microbiological findings among cohorts of women diagnosed with maternal sepsis, severe sepsis or septic shock. Based on our findings, the incidence of maternal sepsis using ICD-9 codes may be significantly overestimated.