I want to use this post as an opportunity to show how an insurance medicine physcian looks at the literature. Basically, we compare the actual or observed death rates by age and with the expected death rates based on something like the US Vital Statistics.
I was recently retained (by a lawyer) to review the insurability of people infected with HIV. Clearly HAART treatments are making great strides and improving mortality, but what does current mortality actually look like?
I took a recent article Bhaskaram, et al, "Changes n the risk of Death After HIV Seroconversion Compared with Mortality in the General Population," JAMA July 2, 2008, Vol 300 p 51. and applied the actual to expected methodology. I think you may find the results interesting. I look forward to your comments.
Observed to general population mortality results by age and duration
Table 4 also can also be used to show observed to expected mortality results which are surprisingly favorable. In order to best apply this study to life insurance (actuarial) methodology, I translated the study’s “overall probability of deaths in percent” in each age and duration group to deaths per thousand. I used the 2004 to 2006 period to reflect the most current treatment status. I compared the number of deaths in the article in the article to the number of expected deaths by age in the 2004 US Vital Statistics http://www.cdc.gov/nchs/data/nvsr/nvsr56/nvsr56_09.pdf. (Note that I substituted US for European mortality to make these calculations.)
Comparing the most current treatment mortality data, that from 2004 to 2006, with US life table mortality (referenced above), gives the following charts:
Mortality for 2004 to 2006 Period for Those Infected 5 Years (study “overall probability %” translated to deaths per 1000) | ||||
Age group |
From the study, observed deaths expressed per 1000 for years 2004 to 2006 |
From the study, observed deaths per 1000 per year (n/3) |
From US Vital Statistics, expected deaths per 1000 per year, average age from column 1 + 5 years (rounded). |
Ratio of observed deaths in the study to expected deaths from US population as a % (total, not excess deaths) |
15 to 24 |
0 |
0 |
1 |
< 100 % |
25 to 34 |
3 |
1 |
1 |
100 % |
35 to 44 |
8 |
3 |
3 |
100 % |
≥ 45 |
49 |
16 |
6 |
267% |
Mortality for 2004 to 2006 Period for Those Infected 10 Years (study “overall probability %” translated to deaths per 1000) | ||||
Age group |
From the study, observed deaths expressed per 1000 for years 2004 to 2006 |
From the study, observed deaths per 1000 per year (n/3) |
From US Vital Statistics, expected deaths per 1000 per year, average age from column 1 + 10 years (rounded). |
Ratio of observed deaths in the study to expected deaths from US population as a % (total, not excess deaths) |
15 to 24 |
54 |
18 |
1 |
1800 % |
25 to 34 |
20 |
6.7 |
2 |
335 % |
35 to 44 |
37 |
12.3 |
4 |
307 % |
≥ 45 |
122 |
47 |
6 |
783 % |
Mortality for 2004 to 2006 Period for Those Infected 15 Years (study “overall probability %” translated to deaths per 1000) | ||||
Age group |
From the study, observed deaths expressed per 1000 for years 2004 to 2006 |
From the study, observed deaths per 1000 per year (n/3) |
From US Vital Statistics, expected deaths per 1000 per year, males, average age from column 1 + 15 years (rounded). |
Ratio of observed deaths in the study to expected deaths from US population as a % (total, not excess deaths) |
15 to 24 |
81 |
27 |
1 |
2700 % |
25 to 34 |
75 |
25 |
3 |
833 % |
35 to 44 |
72 |
24 |
8 |
300 % |
≥ 45 |
249 |
83 |
16 |
519 % |
I suspect that some of the volatility in observed to actual percents is due to small numbers of patients in the groups.
