CASE STUDY 1: UNION GENERAL

• In Northern Georgia, Union General, a modest non-profit healthcare provider, struggled with their handling of 20 TB of data related to patients. This data included digital patient charts and electronic medical records.
• Due to the poor quality of the data stored and the system, hospital staff had to spend time manually using Excel files, which had an impact on the quality of service.
• The introduction of a new advanced analytics system allowed case management workers at Union General to perform a monthly review and analysis of readmissions. This information is used to develop best practices in order to lower rates of readmissions and keep patients away from hospitals.
• Previous to the implementation of the new predictive analytics system, case managers could only manage to perform their review quarterly. 
• It would also allow case management staff to reduce readmissions in real-time rather than retrospectively on the old system.

CASE STUDY 2: JOHN HOPKINS

• John Hopkins Hospital has developed a predictive model in oncology to assist in the customization of chemotherapy treatment plans for people with cancer.
• The help of this technology would allow doctors to minimize the body areas that will be subjected to radiation during treatment. 
• Performing this task manually is a complicated and time-consuming process that is made quicker and simpler by the introduction of predictive analytics.
• The algorithm developed by the team at John Hopkins allows predicting with precision the amount of radiation that will be taken by a vital organ, basing itself on its spatial relationship with the affected cancerous tissue.
• The aim is to limit the damage caused by radiation to vital organs surrounding the infected tissue.

CASE STUDY 3: PENN HEALTH

• Penn Health is developing predictive analytics for the diagnosis of lethal diseases before they occur.
• The project consists in creating learning models from historical data.
• The first type of data used was digital health records as well as laboratory information, and the first models were created to predict heart failure and severe sepsis. 
• Penn Health is also planning on exporting these models to other institutions.
• For the diagnosis of severe sepsis, Penn Health used six vital measurements, and its predictive model used more than 200 variables.  
• This has been a successful effort, allowing the medical institution to detect 80% of cases within 30 hours of the manifestation of the symptoms.
• Its hearth failure algorithm has allowed detecting 20% more patients that were at risk of cardiac arrest.