THE SCIENCE BEHIND HALO: THE HALO APPROACH [Best Practices]

The Science Behind HALO: The HALO Approach

White Paper

No one can see perfectly into the future, however, HALO uses innovative methodology and powerful analytics based on millions of data points to make personalized longevity, health, and eldercare cost projections possible in the financial industry. This document provides an overview of our methodology and approach.

HALO offers a unique process for educating clients and prospects and identifying planning opportunities by pinpointing where they have “gaps” in their protection, resulting in moving clients towards informed decisions much faster. This enables clients and advisors to make unbiased decisions regarding risk mitigation strategies and provides personalized client recommendations on how to live well longer.

HEALTH MATTERS

Traditional approaches for planning for retirement ignore or marginalize the effect of health on future finances. However, studies show that health is a major factor in determining one’s economic wellbeing, and high health care costs are associated with bankruptcy1 and significant depletion of retirement assets. Health determines many factors, such as out-of-pocket care costs (e.g. hospital bill co-pays), length of life, and time spent in elder care (e.g., nursing homes) that can affect retirement planning. Thus, any financial advisor wishing to give their clients a full picture of their financial future must take health into consideration.

THE MISSING OPPORTUNITY

Unfortunately, there are very few data-driven resources available to educate financial advisors and their clients on the relationship between health and wealth. Furthermore, most tools available are difficult to use and understand, and lack scientific grounding. The HALO Planner, however, uses a heuristic decision-making model to create a solution that is easy-to-use and based on years of scientific medical research.

HALO DATA SOURCES

HALO’s projections are powered by over 100 million scientifically relevant data points from more than 90 carefully vetted and curated data sources including validated data from large studies by the Center for Disease Control and Prevention (CDC), the SEER Cancer database, the Kaiser Family Foundation, and Social Security Administration.

In addition, the Lumiant team has evaluated hundreds of studies in high-quality, peer-reviewed academic journals, such as The Journal of the American Medical Association and The New England Journal of Medicine, to find the best parameters for inclusion in the HALO models. As new clinical studies are published, the team reviews the data and where appropriate, updates the models accordingly

RISK FACTORS CONSIDERED

The HALO predictive model focuses on the most important scientifically backed factors that affect longevity and years of disability. Unlike other risk models that ask the user to complete long and tedious questionnaires, which often include risk factors shown to have a relatively small impact on mortality and morbidity, the HALO approach covers all of the most statistically important risk factors (some of which, ironically, are excluded by some of the longer questionnaires).

The model puts a strong emphasis on family health history, including factors such as: Alzheimer’s/dementia, stroke, diabetes, heart disease, obesity, cancer (bladder cancer, colon cancer, breast cancer, kidney cancer, lung cancer, ovarian cancer, pancreatic cancer, skin cancer and prostate cancer), and the overall longevity of parents and grandparents. In addition, the most important lifestyle factors (smoking, exercise, diet, alcohol consumption, BMI, and social support), as well as demographic factors like age, gender, personal health history and ethnicity, are also considered. Genomic data, if available, may eventually become factors within the model.

HALO GENETIC AGE

HALO’s concept of Genetic Age lays the foundation for our unique analytic approach. Genetic Age is the age at which a person’s risk of disease, based on their family health history, is comparable to an average person of the same gender in the general population.

For example, if a 35-year-old woman has a health history of breast cancer, her risk of breast cancer may be more comparable to a typical 45-year-old woman. The model would suggest that this woman with an elevated risk of breast cancer has a Genetic Age of 45 years with respect to breast cancer.

An overall Genetic Age is calculated for each person individually by taking a weighted average of the disease-specific Genetic Age for each of the fifteen most common disease conditions. HALO focuses on the most critical and common health conditions in the model, like common forms of cancer and diabetes (it is estimated that heart disease, cancer, and diabetes account for 7 of every 10 deaths in the United States).

Most current metrics of risk (e.g., relative risk, odds ratios, etc.,) are difficult to understand by the average person, but the Genetic Age metric translates risk into familiar terms of age. When explained using the well-understood concept of age, the user can intuitively grasp how their personal family health history is affecting their own risk of disease, and, ultimately, their longevity.

LIMITATIONS OF GENETIC TESTS AND ACTUARIAL TABLES

Genetic tests typically cannot be used alone to accurately predict longevity. Genetic tests are extremely useful in certain families with a very strong family health history of disease, especially Mendelian conditions, to understand potential inherited risks. However, for most individuals with families that include multiple common conditions like diabetes, stroke, and heart disease, family health history and lifestyle are better predictors of risk, as compared to genetic tests alone

LIMITATIONS OF GENETIC TESTS AND ACTUARIAL TABLES CTD

Actuarial tables are based on data from broad populations of people and can be useful in some situations, such as determining the average life expectancy for people of certain age and gender. However, actuarial tables are not personalized to each individual and their particular life journey or circumstance. The tables fail to consider detailed family health history information and many important lifestyle factors such as smoking and exercise.

To understand the life expectancy of a specific individual, such as a 35- year old man with a family health history of diabetes, BMI of 24, and a smoking habit, an actuarial table based on a broad swath of individuals will not be sufficient to provide the accuracy and fidelity of a unique, personalized model.

CONCLUSION

In a complex environment where there is a lot of information and uncertainty about the future, a heuristic model for decision making is superior to regression or probabilistic models because of the demonstrated less-is-more effect. The reliance on rules based on previous knowledge (a heuristic) to guide the result is more accurate than a complex series of statistical tests incorporating all the data points to cover all the bases across all possible outcomes.

People’s health, wealth, and family are inseparable and complex. Employing a heuristic model to the relationship between family health history, personal lifestyle and financial planning is the most effective tool for empowering people to make good decisions about their health, wealth, and financial goals.

The algorithm and science behind Halo was developed by Emily Chang, Ph.D. Dr. Chang completed her doctoral work in theoretical, computational chemistry at Stanford University with post-doctoral work in computational genetics at Stanford Medical School. After that, she was a health scientist at the consumer genetics company, 23andMe. Using her experience in statistics, algorithm design, data analysis, and scientific study, she has created a unique research-driven approach for understanding risk of disease, projecting longevity, and using that for financial planning.

A FEW OF DR CHANG'S CONTRIBUTIONS TO SCIENTIFIC RESEARCH

1. Rains, Emily K., and Hans C. Andersen. "A Bayesian method for construction of Markov models to describe dynamics on various timescales." The Journal of Chemical Physics 133.14 (2010): 144113.

2. Rains, Emily K., and Hans C. Andersen. 2009. A Bayesian approach for the construction of Markov models. Abstract for poster presentation for the World Molecular Kinetics Workshop 2009, Berlin, Germany.

3. Rains, Emily K., and Hans C. Andersen. 2009. Constructing Markov Models for Protein Folding Simulation. Abstract for poster presentation for the World Molecular Kinetics Workshop 2008, Berkeley, California.

4. Marusak, Rosemary A., Kate Doan, and Scott D. Cummings. Integrated Approach to Coordination Chemistry: An Inorganic Laboratory Guide. Hoboken, NJ: Wiley-Interscience, 2007. Print.

Consumer-facing research: 1. 23andMe’s Health Content Scientists. 23andMe Blog. June 28, 2013. Web. https://blog.23andme.com/23and...

[1] Dobkin C, Finkelstein A, Kluender R, Notowidigdo MJ. The Economic Consequences of Hospital Admissions. Am Econ Rev. 2018;102(2):308-352.