# Development and validation of nomograms for predicting depression and suicidal ideation in stroke survivors: a community-based study

## TL;DR

This study creates and validates tools to predict depression and suicidal thoughts in stroke survivors using community-based data for better mental health care.

## Contribution

Develops and validates community-based nomograms for depression and suicidal ideation in stroke survivors using real-world data.

## Key findings

- Nomograms showed strong discrimination (AUC: 0.76 for depression, 0.74 for suicidal ideation).
- Key predictors included age, marital status, smoking, and health conditions like cancer and hypertension.
- Models demonstrated clinical utility and reproducibility through internal and external validation.

## Abstract

Stroke is a major global health burden, with post-stroke depression and suicidal ideation prevalent yet often underdiagnosed complications. Existing prediction models rely on acute-phase, hospital-based data, limiting their applicability in community settings. This study aimed to develop community-applicable nomograms for predicting risk of depression and suicidal ideation in stroke survivors.

Using data from the National Health and Nutrition Examination Survey (NHANES, 2005–2018), predictors included sociodemographic, lifestyle, medical history, functional status, and blood test variables. Lasso regression and the Boruta algorithm were used for predictor selection. Logistic regression models were developed, and nomograms were constructed. Prediction model performance was evaluated through discrimination (area under curve [AUC]), calibration (calibration plot and Brier score), and clinical utility (decision curve analysis). Internal validation used bootstrapping, and external validation was performed on temporally distinct NHANES datasets.

The prevalence of depression among stroke survivors was 34.8%, with one-third of cases going unrecognized or untreated. The prevalence of suicidal ideation among stroke survivors was 8.2%. The nomograms demonstrated strong discriminative ability (AUC: 0.76 for depression, 0.74 for suicidal ideation) and calibration accuracy (Brier score: 0.11 for depression, 0.07 for suicidal ideation). Key predictors for depression included age, marital status, current smoking, cancer, healthcare utilization, sleep duration, and inability to work. For suicidal ideation, marital status, hypertension, arthritis, antidepressant use, and inability to work were significant. Decision curve analysis demonstrated clinical utility within probability thresholds of 6%–40% for depression and 3%–17% for suicidal ideation. Internal and external validation supported generalizability and reproducibility of the nomograms.

This study provides validated, community-applicable nomograms for predicting depression and suicidal ideation in stroke survivors, enabling primary care providers to identify high-risk individuals and facilitate timely mental health interventions.

Not applicable.

The online version contains supplementary material available at 10.1186/s12888-025-07767-3.