- The document discusses the issue of missing data values in electronic health records (EHRs), which poses a challenge for developing clinical decision support systems (CDSS) using predictive analytics.
- It introduces a new framework called "Missing Care" to address the high levels of missing values in many EHR variables (up to 70-90% missing). Missing Care aims to select the most important variables with acceptable levels of missingness.
- The document applies Missing Care to analyze a large EHR dataset to develop a CDSS for detecting Parkinson's disease, which currently affects over 1 million Americans but is often undiagnosed or misdiagnosed.