Latent class analysis for identification of sub-phenotypes predicting prognosis in hospitalized out-of-hospital cardiac arrest.

AIM OF THE STUDY

To determine which out-of-hospital cardiac arrest (OHCA) patients should receive advanced treatment is extremely challenging. The objective was to identify sub-phenotypes predicting the prognoses of adult OHCA patients by latent class analysis (LCA) using data up to just after admission.

MATERIAL AND METHODS

We conducted a retrospective observational study using multicentre OHCA registry from 95 Japanese hospitals including adult non-traumatic hospitalized OHCA. The primary outcome was 30-day favourable neurological outcome. Our LCA used clinically relevant variables up to just after admission and the optimal class number was determined from clinical importance and Bayesian information criterion. The associations between subphenotypes and outcomes were analysed using univariate logistic regression analysis with odds ratios (ORs) and 95% confidence intervals (CIs).

RESULTS

Our LCA included 2,162 patients and identified four sub-phenotypes. The base excess on hospital arrival had the highest discriminative power. Thirty-day favourable neurological outcomes were observed in 526 patients (24.3%), including 284 (53.8%) in Group 1, 179 (21.2%) in Group 2, 26 (11.4%) in Group 3, and 37 (6.6%) in Group 4. Prehospital return of spontaneous circulation (ROSC) was achieved in 1,009 patients (46.7%), including 379 (81.8%) in Group 1, 340 (40.3%) in Group 2, 115 (50.4%) in Group 3, and 175 (31.1%) in Group 4. Univariate logistic regression analysis for primary outcome using Group 4 as reference revealed ORs (95% CI) of 16.5 (11.4-24.1) in Group 1, 3.83 (2.64-5.56) in Group 2, and 1.83 (1.08-3.10) in Group 3.

CONCLUSIONS

Our LCA classified OHCA into four sub-phenotypes showing significant differences for prognosis. In cases who achieved prehospital ROSC, it might be meaningful to continue advanced therapeutic interventions.