DESCRIPTION OF SHARED ACTUARIAL PROJECTS

UCSB offers a two quarter (20 weeks) upper division course in Actuarial Research. Projects are sponsored by insurance companies or other companies with a risk management project.

Selected projects from this course have been documented and reformatted as case studies for use by the SOA. The case studies are stored on a password protected website, however, faculty from other colleges/universities may request access to these case studies for their own student projects. Materials include a description of the project, data, basic analysis steps, and student results in the form of reports, presentations and a poster. (Instructions to request access are explained at the bottom of this page).

Projects:

  • Analysis of the California Office of Self-Insured Plans Medical Reserving Methodology - California has the largest Workers Compensation self-insurance program in the nation. One of the requirements for holding a certificate to self-insure is that future medical case reserves on Permanent Disability claims must be set using a required reserving formula. The purpose of this project is to test the relative adequacy of case reserves under this formula.
  • California Auto Mileage Forecasting - For insurance companies, higher total mileage driven usually leads to higher auto accident frequency. The purpose of this project is to use publicly available economic indices to create a model to predict future changes in California mileage over the next 12 months.
  • Predictive Modeling for Hospital Readmissions - Predictive models for hospital readmission rates are in high demand due to the Centers for Medicare and Medicaid Services (CMS) Hospital Readmission Reduction Program (HRRP). The purpose of this project is to find the best model for predicting which patients are likely to be readmitted to the hospital and the economic impact of such a model.
  • Simulation of Systemic events based on Random Banking Networks - This project is geared towards students interested in financial mathematics and applied probability. It will investigate a mathematical model that connects the structure of banking networks to degree of systemic risk. The main tool is Monte Carlo simulation.
  • Workers Compensation Survival Models to Predict Life Expectancy – California has the largest Workers Compensation self-insurance program in the nation. One of the requirements for holding a certificate to self-insure is that future medical case reserves on Permanent Disability claims must be set using a required reserving formula. The purpose of this project is to test the relative accuracy of using the 2011 US Life Table to predict expected claim duration.
  • Ruin Probability in the Ordered Risk Model – Ruin probability serves as a risk assessment tool that allows the calibration of initial reserves as well as the premium rate of a non-life insurance company. In this project, we compare different simulation-based methods to compute the finite time ruin probability in a risk model driven by a Mixed Poisson process.
  • Diabetes Risk Factors - It is well documented that diabetes and body mass index (BMI) are positively correlated. The purpose of this study is to investigate the relationship between diabetes and health risk factors. Using data from a wellness improvement company, we developed logistic regression and machine learning models to predict the prevalence of diabetes.
  • Workers Compensation Loss Distribution - Workers compensation insurance was created to protect employees who become injured or ill while on the job. In this project, we model actual workers compensation claim experience in order to create a size of loss distribution for claim severity (average claim size).
  • Modeling HECM Liabilities - Home Equity Conversion Mortgages (HECMs) are reverse mortgage loans insured by the US Federal Housing Administration (FHA). The HECM fund, which accumulates Mortgage Insurance Premiums (MIPs) collected from borrowers as both an upfront premium and ongoing premium, was historically deficient. In October 2017, the premium structure was changed to increase the upfront premium and decrease ongoing premiums. In this project, we evaluate whether the 2017 changes in MIP structure are enough to offset previous loss.
  • Urgent Care Demand - Patients requiring urgent care appointments in a physician office take precedence over regular office visits; physician offices need to plan for urgent care demand in order to be able to schedule regular office visits. The objective of this time-series project is to use physician data to find the best model to predict the demand for urgent care appointments.
  • Workers’ Compensation Medical Claims - Workers compensation insurance was created to protect employees who become injured or ill while on the job. Insurance companies have found that medical claim costs continue to increase the longer a claim is open. Therefore, it is imperative to handle and settle claims on a timely basis. In this project, we analyze closed claim data to determine criteria for predicting the successful settlement of workers compensation medical claims.
  • Pre-Diabetes Transition - Diabetes is a disease that is characterized by elevated blood glucose, which leads to numerous complications in other body systems. It is known that pre-diabetes is a reversible condition and changing from pre-diabetes to healthy status has a very positive impact on personal health. In this project, we study a large database of people with pre-diabetes to look for factors that are associated with the reversal of pre-diabetes.
  • Center for Medicare and Medicaid Services Model Error - CMS (Center for Medicare and Medicaid Services) established the Medicare Shared Savings ACO Program to increase quality of care and to reduce Medicare spending. Physicians participating in the program share gains (difference between expected and actual cost of care). In this project, we assess the model error in the current CMS method used to evaluate Accountable Care Organizations (ACOs): the extent to which a gain is generated by the model, even when no actual savings are present. We compare baseline samples to samples comprised of patients with specific diseases such as diabetes, cancer, chronic heart failure and other heart conditions.
  • Hypertension Transition - Hypertension (i.e., high blood pressure) and its related complications prompt more primary-care visits in the United States than any other medical condition. The objective of this project is to model the rate at which patients with a diagnosis of moderate primary hypertension (moderately high blood pressure) transition to a diagnosis of high hypertension.
  • Alternative Predictive Modeling for Medicare Patient Costs – As health care expenditures increase, patient cost mitigation becomes more essential. Cost mitigation programs rely on the ability to accurately predict patient risk, which is notoriously difficult because of highly-skewed data. In this project, we examine Medicare public use data (which includes demographics, costs, and health conditions) to develop models of patient cost.