Development and internal validation of a nomogram for predicting 90-day poor functional outcome after mechanical thrombectomy in very elderly patients: A retrospective cohort study.

Among very elderly patients (aged ≥80 years) who underwent mechanical thrombectomy for acute ischemic stroke, 37.4% had poor functional outcomes at 90 days.

• Total sample size was 214 patients enrolled retrospectively from a single center between January 1, 2020 and August 1, 2025.
• Poor functional outcome was defined as a modified Rankin Scale (mRS) score of 3 to 6 at 90 days.
• 80 patients (37.4%) had poor outcomes and 134 patients (62.6%) had good outcomes at 90 days.
• All patients were aged 80 years or older with acute ischemic stroke treated with mechanical thrombectomy.

The final nomogram incorporated six independent predictors of 90-day poor functional outcome after mechanical thrombectomy in very elderly patients.

• The six predictors were: National Institutes of Health Stroke Scale (NIHSS) score at admission, diabetes mellitus, puncture-to-recanalization time, postoperative pneumonia, symptomatic intracranial hemorrhage (sICH), and poor collateral status.
• Candidate predictors included baseline, peri-procedural, and early postprocedural variables screened using univariable logistic regression.
• Complete separation was observed for collateral status, so the final model was constructed using Firth penalized logistic regression rather than standard logistic regression.
• The model was designed for in-hospital prognostic assessment after early postprocedural events had become clinically apparent, not for pretreatment decision-making.

The nomogram demonstrated good discrimination with an area under the receiver operating characteristic curve of 0.874.

• The AUC was 0.874 with a bootstrap 95% confidence interval of 0.811 to 0.929.
• Internal validation was performed using bootstrap resampling.
• The bootstrap-corrected confidence interval suggests the model maintains good discriminative ability after accounting for optimism.

Calibration of the nomogram was acceptable across multiple calibration metrics.

• The Brier score was 0.112, indicating good overall predictive accuracy.
• The calibration intercept was 0.057, close to the ideal value of 0, indicating minimal mean calibration error.
• The calibration slope was 1.090, close to the ideal value of 1.0, indicating appropriate spread of predictions.
• These three metrics together suggest the model's predicted probabilities closely match observed outcomes.

Decision curve analysis demonstrated that the nomogram provided meaningful clinical net benefit across a wide range of threshold probabilities.

• Net benefit was demonstrated across threshold probabilities ranging from 0.01 to 0.99.
• This range suggests the model would be clinically useful for virtually all clinically plausible risk thresholds.
• Decision curve analysis was used to assess clinical utility beyond discrimination and calibration metrics alone.

The nomogram was explicitly designed for in-hospital prognostic stratification and not for pretreatment decision-making regarding endovascular therapy.

• The model incorporates early postprocedural variables (postoperative pneumonia and sICH) that are only identifiable after the procedure has been completed.
• The authors state the model 'may assist in-hospital prognostic stratification, risk communication, and discharge planning in already treated patients after early postprocedural complications have been identified or excluded.'
• The authors explicitly noted: 'it was not designed for pretreatment decision-making or for estimating the treatment effect of endovascular therapy.'
• This design choice reflects the intended clinical use case of post-procedural risk communication rather than patient selection for MT.

The study was conducted at a single center with a retrospective cohort design, and internal validation was performed using bootstrap resampling.

• The retrospective single-center design limits generalizability and external validity.
• Internal validation via bootstrap resampling was used to estimate optimism-corrected model performance.
• No external validation cohort was available, representing a key limitation acknowledged in the study design.
• The study period spanned from January 1, 2020 to August 1, 2025, yielding 214 eligible patients over approximately 5.5 years.