Skip to main content

Table 2 Multivariate index for prediction of long-term survival

From: Indicators of "Healthy Aging" in older women (65-69 years of age). A data-mining approach based on prediction of long-term survival

Variable

HR (P-Value)

Number of step-ups completed in 10 seconds1

0.832 (< 0.001)

Smoking: indicator with value 1 if subject is a current smoker

2.354 (< 0.001)

Diabetes: indicator with value 1 if a subject is not diabetic

0.443 (< 0.001)

Age at baseline examination (65 - 69 for all subjects)

1.146 (< 0.001)

Response to Question: How is your health compared to others your age? (categories: excellent, good, fair, poor, very poor)

1.205 (< 0.001)

Smoking: indicator with value 1 if subject is a past smoker

1.390 (< 0.001)

Contrast sensitivity score, average of high and low spatial frequencies2

0.879 (< 0.001)

Pulse Lying Down (beats/60 seconds)3

1.131 (< 0.001)

Hypertension: indicator with value 1 if systolic blood pressure exceeds 160, diastolic blood pressure exceeds 90, or if subject used thiazide4

1.324 (< 0.001)

Past thiazide use: indicator variable with value 1 if the subject has previously used thiazide

1.785 (< 0.001)

Height change since the age of 25 (self-reported at baseline exam)

1.137 (< 0.001)

Participant's clinic throughout the study: indicator with value 1 if subject has attended clinic B

1.357 (< 0.001)

Marriage: indicator with value 1 if subject was married at the time of the baseline examination

0.822 (< 0.001)

  1. The table lists 13 variables that comprise a subset of variables that predict long-term survival as components of a Cox PH model (mean C = 0.673 ± 0.001). The 13 variables were identified through forward search (see Results and Figure 6), and are listed in the order they were introduced into the model during the forward selection procedure (i.e., from most to least important). The second column lists the estimated hazard ratio (HR) and p-value, based upon coefficients obtained from the 13-variable Cox PH model. For variables that require physical measurement in a clinical setting, footnotes list potential alternative variables, along with the average concordance index of an index that includes the alternate variable rather than the variable listed in Table 2. The listed variables were used to generate the risk scores evaluated in Figures 8, 9 and 10, where risk scores for test subjects represent the output of a 13-variable Cox model with coefficients corresponding to measures listed in the table below.
  2. 1Alternative variables: walking speed (m/sec) (mean C = 0.671); average step length at usual pace (m) (mean C = 0.671); average of right and left hand grip strength (kg) (mean C = 0.668); 12-variable model with variable omitted (mean C = 0.668).
  3. 2Alternative variables: indicator with value 1 if subject does not wear glasses (mean C = 0.671); Randot test of stereoacuity, 10-level test scored based upon the highest level correct twice (mean C = 0.671); 12-variable model with variable omitted (mean C = 0.671).
  4. 3Alternative variables: pulse while standing (beats/60 seconds) (mean C = 0.672); diastolic blood pressure while standing (mm hg) (mean C = 0.671); systolic blood pressure while standing (mean C = 0.671); 12-variable model with variable omitted (mean C = 0.671).
  5. 4Alternative variables: waist-to-hip ratio (mean C = 0.671); body mass index (mean C = 0.670); 12-variable model with variable omitted (mean C = 0.670).
Back to article page

BMC Geriatrics

ISSN: 1471-2318

Contact us