BACKGROUND AND PURPOSE

We aimed to build a nomogram, based on patients with spontaneous intracerebral hemorrhage (SICH), to predict the probability of mortality and morbidity at 7 days and 90 days, respectively.

METHODS

We performed a retrospective study, with patients at less than 6 h from ictus admitted to the department of neurosurgery in a single institute, from January 2011 to December 2018. A total of 1036 patients with SICH were included, 486 patients (46.9%) were 47-66 years old at diagnosis, and 711 patients (68.6%) were male. The least absolute shrinkage and section operator method was performed to identify the key adverse factors predicting the outcomes in patients with SICH, and multivariate logistic regression analysis was built on these variables, and then the results were visualized by a nomogram. The discrimination of the prognostic models was measured and compared by means of Harrell's concordance index (C-index), calibration curve, area under the curve (AUC), and decision curve analysis (DCA).

RESULTS

Multivariate logistic regression analysis revealed that factors affecting 7-day mortality, including the following: age, therapy, Glasgow Coma Scale (GCS) admission, location, ventricle involved, hematoma volume, white blood cell (WBC), uric acid (UA), and L-lactic dehydrogenase (LDH); and factors affecting 90-day mortality, including temperature, therapy, GCS admission, ventricle involved, WBC, international normalized ratio, UA, LDH, and systolic blood pressure. The C-index for the 7-day mortality and 90-day mortality prediction nomogram was 0.9239 (95% CI = 0.9061-0.9416) and 0.9241 (95% CI = 0.9064-0.9418), respectively. The AUC of 7-day mortality was 92.4, as is true of 90-day mortality. The calibration curve and DCA indicated that nomograms in our study had a good prediction ability. For 90-day morbidity, age, marital status, and GCS at 7-day remained statistically significant in multivariate analysis. The C-index for the prediction nomogram was 0.6898 (95% CI = 0.6511-0.7285), and the calibration curve, AUC as well as DCA curve indicated that the nomogram for the prediction of good outcome demonstrated good agreement in this cohort.

CONCLUSIONS

Nomograms in this study revealed many novel prognostic demographic and laboratory factors, and the individualized quantitative risk estimation by this model would be more practical for treatment management and patient counseling.