The recent advances in information technologies and biotechnologies is an opportunity to substantially improve healthcare. To exploit the power of data to benefit patients, however, effective clinical decision support tools and novel, individualized interventions must be designed, tested, and implemented. Although there has been progress in the development of statistical/machine learning methods, numerous challenges remain to tailor and translate them into useful clinical decision support tools. Sudden cardiac arrest (SCA) accounts for 15-20% of all adult deaths and is the industrial world?s leading cause of death. Clinical studies of SCA produce repeated measures on risk factors and multiple different kinds of events over time. We refer to these data as survival, longitudinal, and multivariate (SLAM) data. In this project, we will develop novel statistical learning methods for SLAM data and apply them to two distinct aspects of the SCA problem. First, we propose to develop novel statistical learning algorithms that better predict an individual?s multivariate longitudinal data with a focus on the risk of first and subsequent SCA. Second, we propose to develop micro-randomization and just-in-time adaptive intervention trial designs to reduce behavioral risk factors for SCA among persons at high risk. The methods that we propose to develop will be applicable in many areas of medicine. However, they are motivated by and applied to SCA in this project. Our team has expertise in statistics including causal inference, longitudinal data and survival analyses, plus machine learning, epidemiology, cardiology, and behavioral interventions through mobile health (mHealth). This proposed collaboration has the following specific aims:
Aim 1 : Develop and test statistical learning tools for real-time risk prediction of survival, longitudinal, and multivariate (SLAM) outcome data.
Aim 2 : Estimate the risk of SCA and its dependence on dynamic modifiable and non-modifiable factors in population-based and clinical cohorts.
Aim 3 : Plan and conduct a feasibility-usability study of micro-randomization and just-in-time adaptive intervention trial designs for behavioral change to reduce SCA risk. Upon successful completion of these aims, we will have contributed to the progress of healthcare delivery through the application of computational statistics to medicine. ! !
Although sudden cardiac arrest (SCA) is the industrial world?s leading cause of death, SCA prediction and prevention strategies remain limited. In this project, we contribute towards addressing the global burden of SCAs through the development of novel statistical/machine learning methods for SCA risk prediction and the design of mobile health (mHealth) interventions for risk reduction. While developed in the context of SCA, these methods will be applicable in many areas of medicine and public health. !
