What question did this study set out to answer?

Evaluate and compare the performance of multiple machine learning models in predicting 10-year risk of developing hypertension using cardiorespiratory fitness and clinical data.

April 18, 2018PLoS ONEOpen Access

Using machine learning on cardiorespiratory fitness data for predicting hypertension: The Henry Ford ExercIse Testing (FIT) Project

Q: What is the clinical evidence from this study?

Study design: Cohort. Population: Hypertension (n=23095). Intervention: Random Tree Forest model using cardiorespiratory fitness data vs. Other machine learning models (ANN, LB, LWB, SVM, BN). Primary outcome: Prediction of incident hypertension (Area Under the Curve) (AUC 0.93).

View Full Paper

Ask AI

Bookmark

Key result

The Random Tree Forest model demonstrated the best performance for predicting the 10-year risk of developing hypertension using cardiorespiratory fitness data, achieving an AUC of 0.93.

Why the study?

Do machine learning models using cardiorespiratory fitness data accurately predict the 10-year risk of developing hypertension in patients undergoing exercise treadmill stress testing?

Population

23,095 patients who underwent clinician-referred exercise treadmill stress testing at Henry Ford Health…

Comparison

Machine learning models applied to… vs Comparison between different machine learning…

Design

Cohort

Follow-up

10 years

Authors

SSSherif SakrMansoura University RERadwa ElshawiPrincess Nourah bint Abdulrahman University AAAmjad AhmedKing Saud bin Abdulaziz University for Health Sciences

Discussion

Member takes

Implication

Supports ML models using CRF data for hypertension prediction; hypothesis-generating and requires prospective validation before practice change.

Key Points

Evaluate and compare the performance of multiple machine learning models in predicting 10-year risk of developing hypertension using cardiorespiratory fitness and clinical data.
Analyzed clinical data, vital signs, laboratory metrics, and treadmill stress test results from N=23,095 patients with complete 10-year follow-up from the Henry Ford Exercise Testing Project.
Evaluated six machine learning architectures: LogitBoost, Bayesian Network, Locally Weighted Naive Bayes, Artificial Neural Network, Support Vector Machine, and Random Tree Forest.
Random Tree Forest achieved the best predictive performance among all tested models, reaching an area under the ROC curve (AUC) of 0.93.
Comparative analysis demonstrated that predictive accuracy varied substantially across different machine learning algorithms and validation methods.

Study Design

Type

Cohort (n=23,095)

Multicenter

Yes

Structured PICO

Do machine learning models using cardiorespiratory fitness data accurately predict the 10-year risk of developing hypertension in patients undergoing exercise treadmill stress testing?

Population

23,095 adults aged 17 to 96 who underwent clinician-referred treadmill stress testing, followed for 10 years to predict the development of hypertension.

Exposure

Machine learning models (Random Tree Forest, Support Vector Machine, Artificial Neural Network, LogitBoost, Locally Weighted Naive Bayes, Bayesian Network) applied to cardiorespiratory fitness data

Comparator

Comparison between different machine learning models

Outcome

Prediction of incident hypertension at 10-year follow-up

Main Result

Effect estimate: AUC 0.93

Machine learning, specifically Random Tree Forest models, can accurately predict the 10-year risk of developing hypertension using cardiorespiratory fitness data from routine treadmill stress testing.

Limitations

Highly imbalanced dataset requiring synthetic over-sampling techniques (SMOTE) to train models effectively
Retrospective cohort design limited to patients from a single health system in metropolitan Detroit

Cite This Study

Sakr et al. (2018) conducted a cohort in Hypertension (n=23,095). Random Tree Forest model using cardiorespiratory fitness data vs. Other machine learning models (ANN, LB, LWB, SVM, BN) was evaluated on Prediction of incident hypertension (Area Under the Curve) (AUC 0.93). The Random Tree Forest model demonstrated the best performance for predicting the 10-year risk of developing hypertension using cardiorespiratory fitness data, achieving an AUC of 0.93.

synapsesocial.com/papers/6a82db4c6416d6fbd123a4f5 https://doi.org/10.1371/journal.pone.0195344

Topics

Hypertension management

View Full Paper

Ask AI

Bookmark

April 18, 2018PLoS ONEOpen Access

Using machine learning on cardiorespiratory fitness data for predicting hypertension: The Henry Ford ExercIse Testing (FIT) Project

View Full Paper

Ask AI

Bookmark

Key result

The Random Tree Forest model demonstrated the best performance for predicting the 10-year risk of developing hypertension using cardiorespiratory fitness data, achieving an AUC of 0.93.

Why the study?

Do machine learning models using cardiorespiratory fitness data accurately predict the 10-year risk of developing hypertension in patients undergoing exercise treadmill stress testing?

Population

23,095 patients who underwent clinician-referred exercise treadmill stress testing at Henry Ford Health…

Comparison

Machine learning models applied to… vs Comparison between different machine learning…

Design

Cohort

Follow-up

10 years

Authors

SSSherif SakrMansoura University RERadwa ElshawiPrincess Nourah bint Abdulrahman University AAAmjad AhmedKing Saud bin Abdulaziz University for Health Sciences

Discussion

Member takes

Implication

Supports ML models using CRF data for hypertension prediction; hypothesis-generating and requires prospective validation before practice change.

Key Points

Evaluate and compare the performance of multiple machine learning models in predicting 10-year risk of developing hypertension using cardiorespiratory fitness and clinical data.
Analyzed clinical data, vital signs, laboratory metrics, and treadmill stress test results from N=23,095 patients with complete 10-year follow-up from the Henry Ford Exercise Testing Project.
Evaluated six machine learning architectures: LogitBoost, Bayesian Network, Locally Weighted Naive Bayes, Artificial Neural Network, Support Vector Machine, and Random Tree Forest.
Random Tree Forest achieved the best predictive performance among all tested models, reaching an area under the ROC curve (AUC) of 0.93.
Comparative analysis demonstrated that predictive accuracy varied substantially across different machine learning algorithms and validation methods.

Study Design

Type

Cohort (n=23,095)

Multicenter

Yes

Structured PICO

Do machine learning models using cardiorespiratory fitness data accurately predict the 10-year risk of developing hypertension in patients undergoing exercise treadmill stress testing?

Population

23,095 adults aged 17 to 96 who underwent clinician-referred treadmill stress testing, followed for 10 years to predict the development of hypertension.

Exposure

Machine learning models (Random Tree Forest, Support Vector Machine, Artificial Neural Network, LogitBoost, Locally Weighted Naive Bayes, Bayesian Network) applied to cardiorespiratory fitness data

Comparator

Comparison between different machine learning models

Outcome

Prediction of incident hypertension at 10-year follow-up

Main Result

Effect estimate: AUC 0.93

Limitations

Highly imbalanced dataset requiring synthetic over-sampling techniques (SMOTE) to train models effectively
Retrospective cohort design limited to patients from a single health system in metropolitan Detroit

Cite This Study

synapsesocial.com/papers/6a82db4c6416d6fbd123a4f5 https://doi.org/10.1371/journal.pone.0195344

Also Consider

Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context:

Topics

Hypertension management

View Full Paper

Ask AI

Bookmark