We validated our 11-item mortality index in an independent sample of community-dwelling US adults aged 65 and older. Our index demonstrates excellent calibration as shown by similar mortality rates in the new validation cohort, as compared to the development and original validation cohorts. Our index also demonstrates strong discrimination, as shown by increasing risk of 5 and 9-year mortality by point score. This study further validates the use of our index in predicting life expectancy when using NHIS or related datasets, such as MEPS, and strengthens the likelihood that the index will accurately predict mortality among patients, although the index still needs to be validated in a clinical setting. The ability to predict patient life expectancy is important for providing high quality care to older adults.
Ideally older adults who have the life expectancy to potentially benefit from medical interventions would receive these services while those with limited life expectancies would be spared needless interventions. However, physicians find predicting patient life expectancy difficult and may feel uncomfortable discussing life expectancy with their patients.12
Conversely, patients may prefer that their physicians discuss their life expectancy when counseling them on medical interventions.13
Our mortality index can be used to help physicians and older patients with their medical decision-making and it can be used by researchers studying the use of clinical services by patient life expectancy.
It is particularly important to note how well the index performed in predicting 9-year mortality. Although the ability to predict 5-year mortality is important for deciding on many clinical interventions such as mammography screening or determining goals for glycemic control among older persons with diabetes,1,3,14
the ability to identify individuals with nearly 10-years life expectancy may even be more helpful in other situations. For example, most experts agree that older men need at least 10 years life expectancy to potentially benefit from prostate cancer screening if there is a benefit.2
Similarly, older adults need close to 10 years life expectancy to potentially benefit from colon cancer screening.1
Moreover, it is thought than older woman need at least five years but potentially 15 years life expectancy to potentially benefit from radiation therapy after breast conserving surgery in early stage breast cancer.15
In addition, estimating 10-year life expectancy may be important to cardiologists when deciding whether to use a bioprosthesis or a mechanical valve in aortic valve replacement among older adults.16
To date, this is the only validated population-based mortality index that has shown the ability to predict both 5-year and 9-year mortality.
As an example of when our index may be useful, consider a 75 year old male in fair health, with a history of diabetes, at least one IADL dependency, and difficulty walking a quarter of a mile. Using our index, his risk score equals 15. Based on our data this man has a 42% chance of mortality in 5 years and a 75% chance of mortality in 9 years. Given these estimates, it would be reasonable to aim for less intense glycemic goals in this patient to avoid complications of hypoglycemia and it would be reasonable for this patient to forgo colon and prostate cancer screening.1-3
However, a 75 year old male in excellent health, who never smoked, and without any functional limitations would have a risk score of 6 and only have an 8% chance of 5-year mortality and a 26% chance of 9 year mortality. It may be reasonable to discuss colon cancer screening and statins for primary prevention with this patient.1,17
Having a better understanding of patient life expectancy beyond patient age may help clinicians decide where best to focus care of individual patients.
We did find a slight decline in 5-year mortality from years 1997-2000 to years 2001-2004 which may reflect temporal trends towards increased longevity, especially among the oldest adults.18
Despite an overall decline in mortality, our index was still useful in defining which older adults were at greater risk of mortality in 5 to 9 years, with those scoring 18 or more points having a 69% risk of 5-year mortality and a 92% risk of 9-year mortality.
The index does have several limitations. As we noted in our initial report, since the index was developed among community-dwelling adults who can provide self-report, it cannot be generalized for nursing home residents or those with dementia.6
In this study, we included respondents whose answers were given by proxy, and found the model performed as well as in the original development and validation cohorts. This finding has important implications in geriatrics because family members often accompany patients to appointments and may help patients complete survey questions.19,20
Although we tested the index in a new cohort, we still used NHIS data, which is a carefully designed survey that has employed the same general sampling and collection procedures since 1997. Ideally, the index will be validated in a clinical setting or even in a new survey setting using different procedures. It would also be important to validate the index in populations outside the US.
We found that the index demonstrated less discrimination at lower point scores since there is little increase in 5-year mortality among adults who score 0-7 points. However, this should not affect the usefulness of the index since older adults who have less than a 10% chance of mortality in 5-years or about a 25% chance or less chance of mortality in 9 years are likely appropriate candidates for most medical interventions. The discrimination of the index at lower risk scores improves when the index is used to predict 9-year mortality rather than 5-year mortality. Furthermore, it may be reasonable based on our findings to group adults who score from 0-7 at low risk of mortality in 5-9 years, those who score 8-13 at medium risk of mortality in 5-9 years, and those who score 14 or more at high risk of mortality in 5-9 years.
It may seem initially surprising that older adults with BMIs of 25 or less have greater mortality than adults with higher BMIs. However, many studies have also found no association between mortality and obesity among the oldest adults.21;22
Even in further review of our data, we were unable to detect an upper BMI cutoff statistically associated with an increased risk of mortality among adults aged 65 and older. It is possible that some fat mass may serve as nutritional reserve for older adults. It is also possible that adults who were susceptible to the adverse effects of obesity succumbed before old age or that adults who had been obese were losing weight due to severe illness. In addition, the effects of obesity on mortality may be mediated through other factors included in our index.21-23
In summary, we have further demonstrated the utility of an index to predict five and nine year mortality among community dwelling US adults aged 65 and older. The predictive abilities of the index remained accurate even when testing the index on a completely new population of older adults. This index may be useful to researchers, clinicians, and patients who would like to understand the impact of life expectancy when deciding on recommending or accepting medical interventions.