The Centers for Disease Control and Prevention (CDC) regularly issues newsreleases regarding current trends in US population mortality. In June of 2016, articles in The Washington Post and The Wall Street Journal cited new CDC data from 2015 which showed a rise in the US mortality rate. Figure 1 shows the causes of death and directional impacts on population mortality as noted by the CDC1.
Figure 1 - Mortality Trends (CDC 2015)
|Heart Disease||Unintentional injuries|
|Pneumonia||Chronic Liver Disease|
Reviewing the Data
SCOR’s biometric research group was interested in knowing if this trend in general US population mortality translated to a similar trend in our own reinsured population. The first course of action was to download more detailed information from the CDC website. We particularly wanted to have multiple-year data to assess longitudinal trends rather than just a snapshot from 2015.
Figure 2 shows distribution of deaths by cause for ages 20-89 (male and female combined) from 2010 through 2015. The cause of death categories have been mapped from CDC notation to SCOR’s practice.
Figure 2 - Distribution of Deaths by Cause (ages 20-89)
It was no surprise that cardiovascular (circulatory system diseases) and cancers (neoplasms) were at the top of the list. A similar list limited to ages 20-54 showed both cardiovascular and cancer averaging around 20 percent of all deaths and accidental causes (poisoning, external causes and motor vehicle accidents) at around 25 percent. Suicides for this same age group average around 8 percent.
Next, we reviewed SCOR’s data for ages 20-89 (male and female combined). Figure 3 shows our distribution of death counts (not amounts of reinsurance) by cause category. As with the CDC data, we were able to obtain detailed information from 2010 through 2015.
Figure 3 - Distribution of Death Counts
Similar to the general US population, cardiovascular and cancers (neoplasms) were the leading causes for these ages. Limiting the age group to 20-54 showed cancer as the leading cause, averaging around 35 percent; cardiovascular as number two at 20 percent; accidents (poisoning, external causes and motor vehicle accidents) at 25 percent. Suicides for this age group averaged around 10 percent.
Comparing CDC to SCOR
In order to compare CDC and SCOR data, we reviewed changes in cause of death percentages over time. For the CDC data, we compared the absolute change in the percentages from 2011-2013 to 2010 as well as from 2014 to 2011-2013 and 2015 to 2014. Negative changes indicate a decrease in the percentage from the earlier time periods. Figure 4 shows a sample of the pattern of changes for ages 20-89 (male and female combined).
Historically, claims within SCOR’s portfolio are driven mostly from cancers and cardiovascular disease. Combined, these two causes alone consist of nearly 60 percent of our claims. Similarly, suicides, poisonings, chronic lower respiratory disease and influenza combined make up approximately 10 percent of our claims.
We went on to compare CDC data to SCOR by gender and by age groups 20-54, 55-74 and 75-89. This analysis confirmed the increases in poisoning for the younger male age groups in aggregate in the CDC data; however, this same increase was not observed in the SCOR data.
Over the same time period as the CDC data, the SCOR portfolio has shown improvement in deaths due to cancer and only a slight increase in deaths from cardiovascular disease. However, the SCOR portfolio has also shown an increase in claims from Alzheimer’s/ mental/nervous disorders, chronic lower respiratory diseases and influenza/pneumonia. In 2015, a slowdown in improvements due to cardiovascular and cancer deaths was observed in the SCOR experience. We did not observe similar increases as CDC noted in June, overall or for the age/gender specific cohorts.
Differences Between US and Insured Populations
Of late, there has been much press and scrutiny over a decline in expected longevity for the US population relative to other developed nations. While these studies have received noted publicity in the press, it is important to recognize that trends in the general population do not necessarily translate to trends in the insured population, which underlies SCOR’s reinsurance experience. These individuals tend to be in a higher socio-economic class with access to better health care and living conditions and generally make healthier lifestyle choices. For example, the insured population has a lower percentage of tobacco/smoker risks than the general population (less than five percent in the SCOR experience) with a significantly increased cost of insurance for tobacco users.
In 2016, JAMA published an article discussing the correlation between life expectancy and income in the US (Figure 5)2. The authors stated that “between 2001 and 2014, individuals in the top five percent of the income distribution gained around three years of life expectancy, whereas individuals in the bottom five percent experienced no gains. Most of the variation in life expectancy across various geographic areas (and income levels) was related to differences in health behaviors, including smoking, obesity and exercise. Individuals in the lowest income quartile have more healthful behaviors and live longer in areas with more immigrants, higher home prices and more college graduates.”
Figure 5 - Correlation between Life Expectancy and Income
Other recent papers have come to similar conclusions regarding the link between income and mortality3,4. It should be noted, however, that studies of this nature typically focus on the short-term effects of income inequality. In theory, socio-economic impacts on health are not immediate and, therefore, the lifetime effects may be underestimated.
Due to the socio-economic differences discussed above, deaths due to influenza and pneumonia tend to impact the general population more heavily than the insured population. However, this is not always the case. In 2014 and into 2015, several insurance carriers indicated higher than expected mortality claims related to influenza and cardiovascular diseases due to abnormally bad winters in the midwest and northeast regions. A less than effective influenza vaccine in 2014-2015 was also a contributing factor.
Finally, our industry’s extensive underwriting process (detailed health questions plus medical exam and/or fluid collection) tends to select risks that are generally in much better health than the average individual in the US population. Poorer risks are declined insurance or are rated as substandard and pay appropriately higher premiums. This tends to result in a greater proportion of preferred risks and, for impaired lives, a higher proportion in the upper and middle income status who can afford an increased cost of insurance.
As more companies broaden their focus to increase penetration in the middle income market, there will likely by a shift in the drivers of insured mortality towards that of the general population. Therefore, it is important to pay attention to population trends. For example, the increased use of e-cigarettes and vaping, especially in the younger population, is worth further study.
SCOR’s biometric research team continually monitors mortality trends in the US and compares them to our own reinsured experience. We are committed to understanding and analyzing the socio-economic and environmental factors that may be affecting general death rates in the US and to consider their impact, if any, on life portfolios and how those factors may change based on market segment.
1 CDC website http://www.cdc.gov/nchs/data/databriefs.
2 JAMA, April 10, 2016: “The Association Between Income and Life Expectancy in the United States, 2001-2014.” Raj Chetty, PhD; Michael Stepner, BA; Sarah Abraham, BA; Shelby Lin, MPhil; Benjamin Scuderi, BA; Nicholas Turner, PhD; Augustin Bergeron, MA; David Cutler, PhD.
3 Frontiers in Public Health Services and Systems Research, Vol. 5, No. 5 : Art. 5. Beth C. Truesdale, Harvard University, email@example.com; Christopher Jencks, Harvard University, firstname.lastname@example.org.
4 The Lancet, February 21, 2017: “Future life expectancy in 35 industrialised countries: projections with a Bayesian model ensemble.” Vasilis Kontis, PhD; James E Bennett, PhD; Colin D. Mathers, PhD; Guangquan Li, PhD; Kyle Foreman, PhD