Related Experiment Videos
Using patient-reportable clinical history factors to predict myocardial infarction
S J Wang1, L Ohno-Machado, H S Fraser
1Clinical Information Systems Research & Development, Partners HealthCare System, Boston, MA, USA. sjwang@partners.org
Insights
Predicting myocardial infarction (MI) is possible using only patient history. Both logistic regression and neural network models demonstrated high accuracy in predicting MI likelihood, offering potential for early risk assessment outside clinical settings.
Area of Science:
- Cardiology
- Medical Informatics
- Machine Learning
Background:
- Myocardial infarction (MI) risk assessment often relies on objective diagnostic tests.
- Patient-reportable clinical history factors are valuable but underutilized for predictive modeling.
- Developing accessible tools for early MI risk identification is crucial.
Purpose of the Study:
- To develop and validate predictive models for myocardial infarction (MI) likelihood.
- To assess the performance of logistic regression and neural network models using only patient history.
- To explore the utility of these models in non-clinical settings.
Main Methods:
- Logistic regression and neural network models were constructed using a derivation dataset (n=1253).
- Models utilized patient-reportable clinical history factors exclusively.
- Model performance was evaluated on an independent validation dataset (n=500) using C-indices.
Main Results:
- The best logistic regression model achieved a C-index of 0.8444.
- The best neural network model achieved a C-index of 0.8503.
- Both model types demonstrated strong predictive capability for myocardial infarction.
Conclusions:
- Machine learning models, including logistic regression and neural networks, can effectively predict myocardial infarction probability from patient history alone.
- These models show promise for risk stratification in settings lacking immediate diagnostic capabilities.
- Patient-reported data offers a viable alternative for preliminary MI risk assessment.
Abstract:
Using a derivation data set of 1253 patients, we built several logistic regression and neural network models to estimate the likelihood of myocardial infarction based upon patient-reportable clinical history factors only. The best performing logistic regression model and neural network model had C-indices of 0.8444 and 0.8503, respectively, when validated on an independent data set of 500 patients. We conclude that both logistic regression and neural network models can be built that successfully predict the probability of myocardial infarction based on patient-reportable history factors alone. These models could have important utility in applications outside of a hospital setting when objective diagnostic test information is not yet be available.