To identify different phenotypes and screen the prognostic phenotypes by cluster analysis of pulse-indicated continuous cardiac output monitoring technique (PiCCO) parameters in septic patients combined with low cardiac function.

Seventy-eight septic patients with low cardiac function index and PiCCO recordings were screened in the US Intensive Care Database (MIMIC Ⅳ 2.0) (2008-2019). Based on the PiCCO parameters (Cardiac Function Index CI, Whole Heart End-Diastolic Volume Index GEDI, Systemic Vascular Resistance Index SVRI, Extravascular Lung Water Index (ELWI) K-mean clustering characterized patients into different phenotypes. The inter-phenotypic parameters were compared in different phenotypes, such as CI, GEDI, SVRI, ELWI, heart rate (HR), mean arterial pressure (MAP), age, gender, body mass index (BMI), sequential organ failure score (SOFA), history of previous illness, in-hospital mortality for the primary clinical outcome, and incidence of acute kidney injury (AKI grade 3) for the secondary clinical outcome, difference in the length of hospital and ICU stay. Univariate and multivariate logistic regression models were established.

Four different phenotypes were identified in this study. Phenotype 1: hypervolemic, high vascular resistance, very high extravascular lung water. Phenotype 2: normal blood volume, normal vascular resistance, normal extravascular lung water. Phenotype 3: normal blood volume, high vascular resistance, high extravascular lung water. Phenotype 4: high blood volume, normal vascular resistance, very high extravascular lung water. Phenotype 1 had the worst prognosis and the highest in-hospital mortality rate (66.7%). The difference of in-hospital mortality among the four phenotypes was statistically significant different (χ²=7.8, P=0.045). Multifactorial logistic regression showed, compared to phenotype 1, phenotype 2, phenotype 3, and phenotype 4 had an OR and 95%CI of 0.095 (0.017-0.540), 0.087 (0.013-0.580) and 0.067 (0.006-0.719), with significant differences (P<0.05).

Cluster analysis based on PiCCO parameters confirmed different phenotypes of haemodynamic status in patients with sepsis combined with low cardiac function index and identified critically ill patients based on the phenotypes, thus predicting the prognosis of patients.