Early prediction of post-dialysis systolic and diastolic blood pressure using intradialytic data in Senegalese hemodialysis patients: a single-center retrospective cohort study

Q: What is the clinical evidence from this study?

Study design: Cohort. Population: Chronic kidney disease on maintenance hemodialysis (n=124). Intervention: Supervised machine learning (Random Forest model) using intradialytic data up to hour 3 vs. Naive baseline (median predictor) and alternative ML models (Ridge, LightGBM, XGBoost). Primary outcome: Mean absolute error (MAE) across post-dialysis systolic blood pressure (SBP) and diastolic blood pressure (DBP) (MAE 12.67 mmHg (SBP), MAE 8.93 mmHg (DBP), Overall MAE 10.80 mmHg, R² 0.553, p=≥0.063).

Q: What does this research mean for the field?

Supervised tree-based machine learning models using intradialytic data collected up to the third hour can predict post-dialysis systolic and diastolic blood pressure with moderate accuracy and without systematic bias in hemodialysis patients. Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

Mouhamed Lamine Diankha; Modou Ndongo; Ibrahima Lyra Sarr; Baratou Coundoul; Babacar Mbodj; Issa Ndiaye; Elhadj Babacar Fall; Bamba Gaye; Chérif Bachir Deme; Sidy Mohamed Seck

doi:10.3389/fdgth.2026.1919498

November 2, 2026Frontiers in Digital HealthOpen Access

Early prediction of post-dialysis systolic and diastolic blood pressure using intradialytic data in Senegalese hemodialysis patients: a single-center retrospective cohort study

View Full Paper

Ask AI

Bookmark

Share

Key result

Random Forest models predict post-dialysis systolic blood pressure within ~13 mmHg using intradialytic data.

MAE 12.67 mmHg (SBP), MAE 8.93 mmHg (DBP), Overall MAE 10.80 mmHg, R² 0.553
P≥0.063
n=124

Why the study?

Intradialytic blood pressure variability poses a major challenge in chronic hemodialysis, especially in resource-limited settings, contributing to cardiovascular complications and session interruptions. This study aimed to develop and evaluate supervised machine learning models to predict post-dialysis SBP and DBP using only data available up to the third hour of the session.

Do supervised machine learning models using intradialytic data up to the third hour accurately predict post-dialysis systolic and diastolic blood pressure in Senegalese chronic hemodialysis patients?

Key Points

To develop and evaluate supervised machine learning models that predict post-dialysis systolic and diastolic blood pressure using only intradialytic data recorded up to the third hour of a session.
Retrospective, single-center cohort study evaluating 7,600 hemodialysis sessions from 124 chronic patients (52.4% male, mean age 49.84 ± 15.45 years) at Hôpital Militaire Ouakam in Dakar, Senegal, between February 2023 and November 2024.
Extracted intradialytic hemodynamic, machine, and demographic predictors recorded up to hour 3 while excluding end-of-session data to prevent leakage.
Evaluated four tuned models (Random Forest, XGBoost, LightGBM, and Ridge Regression) using patient-level 5-fold cross-validation, with mean absolute error (MAE) across SBP and DBP as the primary performance metric.
Random Forest achieved the best overall performance, with an MAE of 12.67 ± 0.80 mmHg for SBP, 8.93 ± 0.35 mmHg for DBP, an overall root mean square error of 14.68 ± 0.75 mmHg, and an overall R² of 0.553 ± 0.033.
Bland-Altman analysis showed no systematic prediction bias, with a mean bias of 0.015 mmHg for SBP and 0.395 mmHg for DBP.
Predictive accuracy varied across demographic subgroups, performing best in younger underweight males (SBP MAE = 10.46 mmHg) and weakest in middle-aged normal-weight females (SBP MAE = 17.07 mmHg).

Study Design

Type

Cohort (n=124)

Multicenter

No

Structured PICO

Do supervised machine learning models using intradialytic data up to the third hour accurately predict post-dialysis systolic and diastolic blood pressure in Senegalese chronic hemodialysis patients?

P

Population

124 adult patients (mean age 49.84 years, 47.6% female) undergoing chronic hemodialysis across 7,600 sessions at a single center in Senegal over 20 months.

E

Exposure

Supervised machine learning models (Ridge Regression, Support Vector Regression, Random Forest, LightGBM, and XGBoost) predicting post-dialysis systolic and diastolic blood pressure using intradialytic hemodynamics, dialysis machine parameters, and demographic and contextual factors recorded up to hour 3 of the 4-hour session.

C

Comparator

Naive baseline model (median predictor) and cross-comparison among tuned machine learning models evaluated via patient-level 5-fold GroupKFold cross-validation.

O

Outcome

Mean absolute error (MAE) across post-dialysis systolic blood pressure (SBP) and diastolic blood pressure (DBP) measured at the end of the 4-hour hemodialysis session (T4).surrogate

Main Result

Effect estimate: MAE 12.67 mmHg (SBP), MAE 8.93 mmHg (DBP), Overall MAE 10.80 mmHg, R² 0.553

Absolute Event Rate: 10.8% vs 18.1%

p-value: p=≥0.063

Supervised machine learning models, particularly Random Forest, can moderately predict end-of-session blood pressure using the first three hours of intradialytic data in a Senegalese hemodialysis cohort, potentially supporting proactive clinical monitoring in resource-limited settings.

Limitations

Single-center retrospective design and relatively small patient cohort limit generalizability
Highly skewed number of sessions per patient may weight model learning toward patients with more observations
Wide Bland-Altman limits of agreement (especially for SBP) limit individual-level precision
Reduced predictive performance at extreme blood pressure values and in certain subgroups (e.g., older obese women)
Absence of external validation and relative blood volume monitoring
Single-center retrospective cohort design
Substantial individual-level prediction uncertainty indicated by SBP MAE of 12.67 mmHg and wide limits of agreement
Limited statistical power with five cross-validation folds to detect differences between model pairs
Negative R² for SBP in the Female/Older adult/Obese subgroup (n = 143 sessions), indicating the model did not explain SBP variance better than the global mean in this subgroup
Limited sample representation within some demographic strata
Highly skewed distribution of sessions per patient

Ask about this paper

Answers are grounded in this paper and cite their sources.

Implication

May support early intradialytic BP monitoring in hemodialysis; leaves open prospective validation before clinical use.

Abstract

Background Intradialytic blood pressure variability poses a major challenge in chronic hemodialysis, especially in resource-limited settings, contributing to cardiovascular complications and session interruptions. This study aimed to develop and evaluate supervised machine learning models to predict post-dialysis systolic blood pressure (SBP) and diastolic blood pressure (DBP) using only data available up to the third hour of the session. Methods In this retrospective, single-center cohort study at Hôpital Militaire Ouakam (Dakar, Senegal; February 2023 to November 2024), we analyzed 7,600 hemodialysis sessions from 124 chronic patients after quality-based exclusions. Predictors included intradialytic hemodynamics, dialysis machine parameters, and demographic and contextual factors recorded up to hour 3, while end-of-session data were excluded to prevent leakage. Four models were retained after initial screening of six candidates (Ridge Regression, Random Forest, LightGBM, and XGBoost) and evaluated using patient-level 5-fold cross-validation, with mean absolute error across SBP and DBP as the primary performance metric. Results Tree-based models outperformed linear and kernel-based methods, with the optimized Random Forest model achieving the best performance, with an MAE of 12.67 ± 0.80 mmHg for SBP and 8.93 ± 0.35 mmHg for DBP, a root mean square error of 14.68 ± 0.75 mmHg, and R² of 0.553 ± 0.033. Diastolic blood pressure predictions were more stable than systolic predictions, with no systematic bias observed in Bland-Altman analyses (bias = 0.015 mmHg for SBP and 0.395 mmHg for DBP). Subgroup performance showed moderate heterogeneity related to age, sex, and body mass index, with the best results observed in younger male patients. Conclusion These findings demonstrate the feasibility of supervised machine learning for unbiased prediction of post-dialysis blood pressure in Senegalese chronic hemodialysis patients, supporting the potential development of proactive interventions during hemodialysis sessions. Prospective and multicenter studies are needed to validate and measure the clinical impact of artificial intelligence use in the management of dialysis patients in resource-limited settings.

Discussion

No takes yet. Share an insight, caveat, or question.

Cite This Study

Diankha et al. (2026) conducted a cohort in Chronic kidney disease on maintenance hemodialysis (n=124). Supervised machine learning (Random Forest model) using intradialytic data up to hour 3 vs. Naive baseline (median predictor) and alternative ML models (Ridge, LightGBM, XGBoost) was evaluated on Mean absolute error (MAE) across post-dialysis systolic blood pressure (SBP) and diastolic blood pressure (DBP) (MAE 12.67 mmHg (SBP), MAE 8.93 mmHg (DBP), Overall MAE 10.80 mmHg, R² 0.553, p=≥0.063). Random Forest models using 3-hour intradialytic data predicted post-dialysis blood pressure with an MAE of 12.67 mmHg for SBP and 8.93 mmHg for DBP (overall R² 0.553).

synapsesocial.com/papers/6aea772189a8acbed9c87ac6 https://doi.org/10.3389/fdgth.2026.1919498

View Full Paper

Ask AI

Bookmark

Share

Early prediction of post-dialysis systolic and diastolic blood pressure using intradialytic data in Senegalese hemodialysis patients: a single-center retrospective cohort study

Key Points

Study Design

Structured PICO

Main Result

Limitations

Ask about this paper

Implication

Abstract

Discussion

Cite This Study

Early prediction of post-dialysis systolic and diastolic blood pressure using intradialytic data in Senegalese hemodialysis patients: a single-center retrospective cohort study

Key Points

Study Design

Structured PICO

Main Result

Limitations

Ask about this paper

Implication

Abstract

Discussion

Cite This Study

Keep reading

Keep reading