Development of a practical predictive nomogram for cognitive impairment risk following traumatic brain injury: a retrospective cohort study

Background Accurately predicting long-term cognitive impairment following moderate–severe traumatic brain injury (TBI) remains a significant clinical challenge. Existing prognostic tools have limited predictive accuracy for individual-level risk stratification. Objective To develop a comprehensive prognostic nomogram for predicting cognitive impairment in adults who suffered a moderate-to-severe TBI. Methods We retrospectively included 482 adults with moderate-to-severe TBI admitted to our hospital. All participants completed the standardized 6-month cognitive function assessments at 6 months following TBI using the Neuropsychological Test Battery. Variables were selected from demographic data, clinical factors, laboratory tests and questionnaire scores. A prediction model was developed through multivariable logistic regression analysis. Results Cognitive impairment was observed in 28.8% of individuals. The prediction model showed that previous TBI history (OR = 2.002), Marshall CT classification IV (OR = 4.072,)/V (OR = 4.613)/ VI (OR = 5.008), elevated neurological function markers ≥2 (OR = 2.255), elevated proinflammatory cytokines ≥2 (OR = 2.209), Systemic Inflammation Response Index scores (OR = 3.066) and incidence of lower urinary tract symptoms (OR = 4.169), seizures (OR = 5.243) or post-traumatic stress disorders (OR = 5.245) were independent risk factors of cognitive impairment following TBI (all p < 0.05). Receiver-operating characteristic curve demonstrated a good discrimination with an area under the curve equaling 0.920. A prognostic index value more than 6.425 were categorized as high-risk to develop cognitive impairment after TBI. Conclusion We developed a prognostic nomogram that ‌showed promise for early risk stratification‌ of 6-month cognitive impairment in adults with moderate–severe TBI, utilizing information available during acute hospitalization. ‌While our results were encouraging, external validation in independent, prospective cohorts was required‌ to confirm its generalizability and potential for integration into clinical practice before widespread implementation.