Health Status Adjusted Age & Life Expectancy Without Cancer

This calculator is the pathway for people under age 40. It was developed using race, ethnicity and sex specific life tables for death from causes other than oral cancer using U.S. life tables. However, because so few people in the general population die of oral cancer, all cause death is used as an approximation of death from other causes. To compute life expectancy, the estimates for death from other causes had to continue through attained age 97. Attained age is your current age based on your last birthday, in whole years. It is used in this calculator to compute your chance of death from causes other than oral cancer every year after your cancer diagnosis.

From attained ages 20 through 54 U.S. life tables are used, however starting at attained age 55 the estimates of other cause death are estimated using the oral cancer patients in SEER who tend to have a higher chance of death from other causes than the general population (prior to attained age 55 there are not enough non-oral cancer deaths among oral cancer patients to produce stable estimates). For attained ages 55-97 other cause mortality among oral cancer patients is stratified by sex and modeled as a function of race/ethnicity, the socioeconomic status of the census tract of residence (using the Yost index) (1), and stage of disease.

The predicted probability of death from the model was used to compute life expectancy in the absence of oral cancer death. The life expectancy was compared to U.S. all races, sex–specific life tables to obtain HSAA and predicted survival. Since this calculator provides the least amount of customization and it is applied to younger individuals, for whom the predictions extrapolate far into the future, its life expectancy estimates may be less reliable.

This calculator is the pathway for people age 40-86 who have not seen a doctor regularly prior to diagnosis or see a doctor regularly but do not have any of the calculator listed co-existing health conditions. The General Health Self-Assessment Calculator is based on work by Chao et al (2). It generates non-cancer survival predictions based on data from the nationally representative National Health Interview Surveys (NHIS) during the period 1986–2009 and linked to the National Death Index (NDI) to obtain vital status follow-up information as of December 31, 2011 on the cohorts of people included in the analysis. The risk of death before age 40 is not estimated because there were not enough events to provide stable estimates. The NHIS respondents were asked to rate their health status in one of the following categories: Excellent, Very Good, Good, Fair, or Poor. Self–rated health status is a summary measure that represents an individual’s evaluation of his/her own health, captures the full array of illness a person has, and is a well-validated predictor of mortality including other-cause mortality in cancer patients (3-7). While self-reported health status has the disadvantage of being a subjective assessment, it has some advantages in that it is independent of formal interactions with the health care system and includes an overall assessment—including what a person might or might not have told their physician or been formally diagnosed with—and includes individuals who never or rarely visit a physician. While claims-based comorbidity based on Medicare data captures the sicker end of the health spectrum, a large percentage of people at every age have no major coexisting conditions. Self-rated health status in these five categories better captures the healthier end.

A survival model utilizing the survey weights was fit, using age as the time scale, with follow-up starting at the age at which the survey was administered. The model was fit separately by sex, and predictors included self-assessed health status, race/ethnicity, age at survey, calendar year of survey, and smoking status (never, current, former). Smoking status was included because so many oral cancer patients have a history of tobacco use and because it provides additional independent discrimination with respect to the likelihood of other-cause death. Like the Coexisting Condition calculator, age at survey interacted with self-assessed health status to reflect the fact that the impact of self-assessed health status is larger at younger ages.

To use this calculator, the person must make a subjective assessment of his or her health status (Excellent, Very Good, Good, Fair, or Poor). This model developed using the general population is assumed to be applicable in the oral cancer population. While the general population has better other cause of death life expectancy than oral cancer patients, who tend to be tobacco users, it is reasonable to expect that their life expectancies are similar after conditioning on self-reported overall health and smoking status. Since the preponderance of oral cancer patients use tobacco, one should interpret the chance of cancer death with caution for those whose cancer is not thought to be associated with tobacco use.

A smoothed version, extrapolated to age 110, of the predicted probability of death from other causes in the absence of cancer was used to compute the life expectancy. The life expectancy was compared to U.S. all races, sex specific life tables to obtain HSAA and predicted survival. Since this calculator is estimated using the general population, and so few people die of oral cancer, we use all cause of death as an approximation for causes of death other than oral cancer.

This calculator is the pathway for people age 66-86 who have seen a doctor regularly prior to diagnosis and have one or more of the calculator listed co-existing health conditions. Charlson et al. (8) developed a conceptual approach for assessing the impact of coexisting conditions on survival. More recently, Mariotto et al (9) developed a comorbidity (coexisting condition) score based on Charlson’s work. This score enabled researchers and clinicians to obtain improved survival estimates associated with other (non-cancer) causes of death in people with cancer by accounting for the presence of coexisting conditions. Mariotto et al.’s analysis (9) was based on Surveillance, Epidemiology, and End Results (SEER) data linked to Medicare data (10) and included the development of a comorbidity score using Cox regression (11). The comorbidity score was based on data for all people with cancer age 66 and older (in the year before their cancer diagnosis) for people diagnosed in SEER Program areas, enrolled in Parts A and B of Medicare, and not treated in a Health Maintenance Organization. The linkage to Medicare provided information on health conditions used in calculating the comorbidity score. The variables included in the Cox regression were age, race, and sex, along with indicator variables for each of the 14 coexisting conditions (modified from the work of Charlson et al) included in this Coexisting Condition Calculator.

The output from this model was used to obtain a comorbidity score for a person, defined as the sum of the regression coefficients for the 14 coexisting conditions multiplied by disease indicators for the presence or absence of each coexisting condition. A score was calculated for all people in the SEER-Medicare linked database. The SEER-Medicare database is updated every two years in order to add recently diagnosed cases to the linked dataset. When the database is updated, the Cox regression analysis (above) is repeated for all cancer cases in the database, and the comorbidity score for each person is recalculated based on new regression coefficients from the modeling and replaced in each person’s record. The comorbidity score obtained from the 2014 SEER-Medicare database update was used in the analysis of people with oral cancer for this project.

A survival model (11) utilizing the SEER-Medicare linked database (10) for oral cancer patients was fit by sex, with an endpoint of non-cancer causes of death as a function of the comorbidity score, age at diagnosis, race/ethnicity, socioeconomic status (SES) of the census tract of residence, and stage. The Howlader algorithm (12) was used to code cause of death to help account for common misclassifications of cause of death. Age was treated as the time parameter in the model, with follow-up starting at the age of diagnosis. An interaction between the age at diagnosis and comorbidity score was included to accommodate the fact that the impact of coexisting conditions on other–cause survival is much larger at younger ages. That is, even if older individuals do not have serious coexisting conditions at any specific age, they tend to be frailer, and are therefore more likely to develop conditions quickly, and these conditions are more likely to lead to death. The SES variable was based on an index developed by Yu et al (1), and it represents the SES of the census tract of residence.

A smoothed version, extrapolated to age 110, of the predicted probability of death from other causes in the absence of cancer was used to compute the life expectancy. The life expectancy was compared to U.S. all races, sex specific life tables to obtain HSAA and predicted survival.

The data are sparse for people with extensive coexisting conditions, which makes the calculation of HSAA less reliable at the extremes. Specifically, in cases where the HSAA is based on coexisting condition scores over 2.0 (which is approximately equivalent to four or more moderate coexisting conditions or three or more severe coexisting conditions).

"Caution: Because there are very few persons in our database with as many and/or as severe coexisting health conditions as you have selected, the calculated health status adjusted age may be less reliable. Therefore the resulting estimates of surviving, dying of other causes, or dying of cancer should be interpreted with caution."

We have found that even after adjusting for coexisting conditions, the HSAA still varies by cancer site and stage of disease. For example, oral cancer patients tend to have higher HSAA than people with other cancers after adjusting for coexisting conditions and demographics, because many have a history of tobacco and alcohol use. These risk factors will have an influence on the chance of death from other causes, even if they have not yet manifested as an overt specific condition. Higher-stage patients have higher HSAAs. These differences reflect unmeasured factors related to people who tend to be diagnosed with different cancers, and at what stage they are diagnosed. For this reason, for any given coexisting condition profile, the HSAA can differ by cancer site and stage.

1. Yu M, Tatalovich Z, Gibson JT, Cronin KA. Using a composite index of socioeconomic status to investigate health disparities while protecting the confidentiality of cancer registry data. Cancer Causes Control. 2014 Jan;25(1):81–92.
2. Cho H, Wang Z, Yabroff KR, et al. Life expectancy adjusted by self-rated health status in the United States: National Health Interview Survey linked to the mortality data. Submitted 2018.
3. McCallum J, Shadbolt B, Wang D. Self-rated health and survival: a 7-year follow-up study of Australian elderly. Am J Public Health. 1994 Jul;84(7):1100–5.
4. Idler EL, Benyamini Y. Self-rated health and mortality: a review of twenty-seven community studies. J Health Soc Behav. 1997 Mar;38(1):21-37.
5. Idler EL, Russell LB, Davis D. Survival, functional limitations, and self-rated health in the NHANES I Epidemiologic Follow-up Study, 1992. First National Health and Nutrition Examination Survey. Am J Epidemiol. 2000 Nov 1;152(9):874-83.
6. DeSalvo KB, Bloser N, Reynolds K, He J, Muntner P. Mortality prediction with a single general self-rated health question. A meta-analysis. J Gen Intern Med. 2006 Mar;21(3):267–75. Epub 2005 Dec 7.
7. Hoffman RM, Koyama T, Albertsen PC, Barry MJ, Daskivich TJ, Goodman M, Hamilton AS, Stanford JL, Stroup AM, Potosky AL, Penson DF. Self-reported health status predicts other-cause mortality in men with localized prostate cancer: results from the Prostate Cancer Outcomes Study. J Gen Intern Med. 2015 Jul;30(7):924–34.
8. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987;40(5):373-83.
9. Mariotto AB, Wang Z, Klabunde CN, Cho H, Das B, Feuer EJ. Life tables adjusted for comorbidity more accurately estimate noncancer survival for recently diagnosed cancer patients.  J Clin Epidemiol. 2013 Dec;66(12):1376–85.
10. National Cancer Institute, Division of Cancer Control and Population Sciences, SEER-Medicare Linked Database.  https://healthcaredelivery.cancer.gov/seermedicare/  
11. Cox DR. Regression Models and Life-Tables. Journal of the Royal Statistical Society. Series B (Methodological). 1972;34(2):187–220.
12. Howlader N, Ries LA, Mariotto AB, Reichman ME, Ruhl J, Cronin KA. Improved estimates of cancer-specific survival rates from population-based data. J Natl Cancer Inst. 2010 Oct 20;102(20):1584–98.