Physical functioning and mortality in older women:

Journal of Clinical Epidemiology 56 (2003) 807–813

Physical functioning and mortality in older women: An assessment of energy costs and level of difficulty Ira B. Tager*, Thaddeus J. Haight, Milton Hollenberg, William A. Satariano School of Public Health, University of California at Berkley, 140 Warren Hall, Berkley, CA 94720-7360 Accepted 16 April 2003

Abstract Objective: We tested the hypothesis that the types of activity and the energy equivalent assigned to each activity affect the relationship between self-reported physical functioning and mortality. Methods: We analyzed the relationship between physical functioning and cardiovascular and noncardiovascular mortality in 1230 women (median age 70 years) observed for 7.5 years. We evaluated five separate scores of physical functioning that differed in the method of scoring the responses. Cox proportional hazard models included baseline age, self-reported physical functioning, medical morbidity, and assessment of health. Results: For cardiac and noncardiovascular mortality, greater self-reported functioning was associated independently with a decreased hazard of death. The effects of physical functioning were sensitive to the form of the score used; that is, a score based on ordinal responses was associated with a greater reduction in hazard of death difference in survival between high and low function score: Ordinal: ⫺15.2% (95% confidence interval [CI] ⫺25.2–⫺4.0); dichotomous: ⫺11.6% (95%CI ⫺18.9–⫺3.9). Conclusion: There is a consistent relationship between functional limitation and all causes of mortality. The association is sensitive to the form of the score. Future physical function scores should be based on ordinal responses to individual items used in the scores. 쑖 2003 Elsevier Inc. All rights reserved. Keywords: Mortality; Epidemiology; Functionally impaired elderly

1. Introduction Physical functioning is generally defined as the relative ease with which a person can perform activities of everyday living (ADL). These activities can range from simple, generic tasks such as pushing, lifting, and carrying items to more complicated activities associated with household maintenance, employment, and driving an automobile. Standard items, originally developed by Nagi and Rosow and Breslau [1,2], often are used in epidemiologic studies to measure self-reported level of physical functioning in older populations. Respondents typically are asked to report their level of difficulty in the performance of specific tasks that require different combinations of upper- and lower-body strength, mobility, balance, and fine dexterity. Reported limitations in one or more of these tasks form the basis for the classification of disability [3]. One of the most common subsets of items includes questions that deal with pushing and pulling heavy objects, lifting or carrying items over 10 pounds, lifting or carrying items under 10 pounds, walking a city

* Corresponding author. E-mail address: [email protected] (I.B. Tager). 0895-4356/03/$ – see front matter 쑖 2003 Elsevier Inc. All rights reserved. doi: 10.1016/S0895-4356(03)00149-5

block, and walking up and down stairs without assistance. These items serve as the basis for assessments of probability of future disablement and impaired quality of life [4,5] and have been shown to predict subsequent risk of death in older populations [6–8]. Although these are standard items for self-reported functional assessment [1,2,4,9–11], reported levels of difficulty have been measured and summarized in a variety of ways, and there is no generally agreed upon standard summary protocol [10,12,13]. This makes comparisons across studies difficult; furthermore, it is unknown whether differences in the scoring and summarization of levels of difficulty are associated with quantitative differences in the associations between self-reported function and mortality. It also is unknown whether the type of task itself affects the prediction of mortality and specific causes of death. Difficulty in one task (e.g., lifting items over 10 pounds) is treated as the same as difficulty in performance of another task, such as pushing or pulling objects, even though each task may reflect the use of different muscle groups and levels of energy expenditure. Unlike research on physical activity and health [14,15], estimates of energy expenditure have not been used to distinguish among different functional items and their relationships to future disability and mortality.

808

I.B. Tager et al. / Journal of Clinical Epidemiology 56 (2003) 807–813

We undertook an analysis of the relationship between self-reported physical functioning and mortality in older women. Our hypothesis was that the types of activities included and the energy equivalent assigned to each activity would affect the associations between self-reported physical functioning and mortality. Based on the five functional items described previously, we evaluated several different methods of scoring and summarization of the subjects’ responses to create a summary measure of functioning. We included estimates of the energy equivalents associated with each task to determine the effect on the associations with mortality.

2. Methods 2.1. Subjects The 1246 female participants of the Study of Physical Performance and Age-Related Changes in Sonomans (SPPARCS) project were included in this analysis. SPPARCS is a longitudinal study of physical activity and fitness in people ⭓55 years of age who live in Sonoma, California. The methods of recruitment and the representativeness of the sample to its target population have been described [16]. Protocols were approved by the Committee for the Protection of Human Subjects at the University of California, Berkeley and the Human Subjects Committee at the University of California, San Francisco. This analysis was based on the home interview from the first three evaluations of the cohort (May 1993 through December 1994, September 1995 through November 1996, and June 1998 through October 1999). A standardized questionnaire was administered that contained detailed questions about medical conditions, leisure-time physical activity, cognitive functioning, social factors, and alcohol and tobacco use.

and NRB3. NRB1 is based on the most detailed information and NRB3 on the least. 2.2.1. NRB1 NRB1 includes information on the average energy equivalent associated with a task and multi-level scoring of selfreported difficulty. Each of the five functional items was assigned a metabolic equivalent (MET) value (approximately equal to an oxygen consumption of 3.5 mL/kg min⫺1) based on estimates provided by Ainsworth et al. [17] (Appendix A). When several MET values were possible for a given task, the average of these values was used. The total MET value for the sum of the five items was 25. An energy value (wi) was assigned to each functional item based on its fraction of or contribution to the total 25 METS. A separate value for level of difficulty was assigned (Appendix). These values ranged from 0.0 for a lot/unable to 1.0 for no difficulty. A subject’s final score for an individual functional item was an energy-weighted average of the reported level of difficulty. Scores ranged from 0 to 1. 2.2.2. NRB2 For NRB2, differences in levels of energy equivalents were not included. The difficulty score for each item was calculated as it was for NRB1, and the final NRB2 score was based on the sum of the total difficulty scores. The final score was re-scaled to range from 0 to 1. 2.2.3. NRB3 NRB3 is the simplest summary measure of physical functioning. As with NRB 2, the item/energy values were held constant. Unlike NRB1 and NRB2, the difficulty items were dichotomized: 1 ⫽ no/little difficulty, and 0 ⫽ all other levels of difficulty/nonperformance. The final score ranged from 0 to 5 and was re-scaled to a 0 to 1 scale to facilitate comparison with the other two scores.

2.2. Classification of physical functioning

2.3. Ascertainment of vital status

The assessment of physical functioning was based on five functional measures developed originally by Nagi [1] and Rosow and Breslau [2]. We asked the following questions: In the past month, what level of difficulty did you have in (1) pushing or pulling objects like a living room chair; (2) lifting or carrying items under 10 lb, like a bag of potatoes; (3) lifting or carrying items over 10 lb, like a bag of groceries; (4) walking alone up and down a flight of stairs; and (5) walking two to three neighborhood blocks. Seven responses were allowed: no difficulty, a little difficulty, some difficulty, a lot of difficulty; unable to do activity, do not do activity; under doctor’s orders; never do the activity. Three summary measures were developed. Each measure was based on a different scoring configuration of energy equivalent and level of difficulty. Because the functional items were derived from Nagi [1] and Rosow and Breslau [2], we designate these summary measures as NRB1, NRB2,

Surveillance of mortality involved the checking of obituaries in local newspapers and quarterly receipt of listing of all deaths in Sonoma County, CA. When subjects could not be located at the time that they were scheduled for follow-up evaluation, we (1) contacted next of kin/friend, (2) contacted physicians/health care providers, and (3) searched the Social Security Death Index. At the conclusion of the third evaluation (October 1999), the vital status of 4.9% (60/1230) could not be ascertained due to loss to follow-up (12 could not be contacted, and 48 had refused further participation in the study). The final censoring date for this analysis was 30 October 1999. Death certificates were obtained for 92.3% (192/208) of the reported deaths (16 had not reached the study office). All death certificates were reviewed by one of the investigators (IBT). In cases where the “immediate cause” of death could be ascribed unequivocally to an underlying cause of death

I.B. Tager et al. / Journal of Clinical Epidemiology 56 (2003) 807–813

(e.g., ventricular fibrillation due to acute myocardial infarction), the death was recorded as the underlying cause. In six cases, cause of death could not be determined unequivocally from the death certificates. These subjects were included only in the analysis of noncardiovascular deaths. 2.4. Analysis of data Proportional hazards models were used to assess the effects of known predictors of mortality (PROC PHREG of SAS 6.12). The following baseline factors were considered as predictors: (1) age, (2) health status (self-reported cardiovascular disease; diabetes mellitus; cancer, except nonmelanoma skin cancers; asthma; chronic obstructive lung diseases; kidney and liver disease), (3) self-reported perception of health and appetite, and (4) smoking history (never, current, former). Health status was classified as (1) pre-existing cardiac disease (e.g., myocardial infarction), (2) pre-existing cardiovascular condition (e.g., any cardiac disease, stroke, and aortic aneurysm), and (3) other pre-existing health conditions. Cardiac deaths (55/208, 26.4%) were those included in International Classification of Disease 9th Revision (ICD9) codes 394 through 429 and 785. Cardiovascular deaths (75/ 208, 35.6%) included cardiac deaths plus deaths caused by diseases included in ICD9 codes 430 through 448. Cancer deaths (55/208, 26.4%) included diagnoses in ICD9 codes 140 through 239. The remaining deaths, which were included in the “all cause” mortality analysis, included pneumonias (ICD9 480–487) (12/208, 5.8%), chronic obstructive lung diseases (ICD9 490–496) (13/208, 6.3%), and other frequent causes (32/208, 15.4%). NRB1, NRB2, and NRB3 were assessed in separate models. All models were fit as follows: first, age was evaluated, and then the NRB score was added. In all cases, age and each NRB score were significantly related (P ⬍ .05) to each measure of mortality. The remaining covariates were entered as a group. The final step involved the evaluation of interactions of age with the NRB scores. The final models were selected based on the Akaike Information Criterion (AIC). All persons who were lost to follow-up or who died from a cause of death other than the one of interest in a specific model were censored at the time of loss or death. Goodness-of-fit procedures were used to test the fits of the models to the data. Models that excluded the most influential observations were re-examined to assess the consistency of the parameter estimates. The proportional hazard assumption was evaluated by plots of the crude survival curves against time and by evaluation of the plots of the estimated log {⫺log S(t)} against log(t) for the various models. To test whether entry times into the study were associated with survival, a variable that signified entry time was included in the models. 3. Results Subjects were followed for up to 7.5 years. Subjects who died were older at baseline and were more likely to report

809

underlying cardiovascular and other health conditions (Table 1). Decedents more frequently reported fair or poor health and appetite, reported lower levels of self-reported leisuretime physical activity, and had a lower percent of predicted peak expiratory flow rate than did survivors. There was little difference in the baseline smoking histories between survivors and decedents. Decedents had lower levels for all NRB scores. Those who died showed greater variability in their function scores; all of the scores showed a similar degree of variability over the 25th and 75th percentile of their distributions. Cardiac diseases and cancers were the two

Table 1 Characteristics of 1230 female subjects for the SPPARCS project Characteristic Age (median, range) At baseline At last follow-up Duration of follow-up in years (median, range) Baseline smoking history (%) Never Current Ex Baseline underlying medical conditionsc(%) Cardiac Caridovascular (includes cardiac) Other health conditions None of the above Baseline reported perception of health (%) Excellent good Fair/poor Baseline reported perception of appetite (%) Excellent/good Fair/poor Baseline physical function scores NRB1 NRB2 NRB3 Baseline leisure-time physical activity (average MET equivalents/week in past 12 mo) Baseline peak expiratory flow rate (% predicted)

Nondeceased (n ⫽ 1038) 69 (53–95) 75 (57–102) 6.8 (1.1–7.5)

Deceased (n ⫽ 192)a 79 (58–97) 83 (61–99) 3.6 (0.25–7.1)b

50.3 8.5 41.2

43.2 10.9 45.8

6.5 10.1

16.1 30.2

30.0 59.1

37.0 29.7

86.0 13.5

59.4 39.6

90.9 8.7

70.3 29.2

0.90 0.90 1 36

0.64 (0.24–0.88) (0.74–1)d (0.80–1) 0.64 (0.30–0.90) (0.80–1) 0.60 (0.20–0.80) 20 (0–38, 0–178) (19–62, 0–221)e

98 (16–146)e

89 (25–154)

Abbreviations: SPPARC, Study of Physical Performance and Age-Related Changes in Sonomans; NRB, summary measure based on Nagi [1] and Rosow and Breslau [2]; MET, metabolic. a Total deaths ⫽ 208; 16 death certificates not available. b Represents median survival from time of entry for deceased subjects. c See Methods for classification. For this and other percentages, % ⬍100% due to missing information on the variable. d Median, 25th through 75th percentiles, range 0–1 for all groups. e Median, range.

810

I.B. Tager et al. / Journal of Clinical Epidemiology 56 (2003) 807–813

most frequent causes of deaths (Table 2). Subjects who died from cardiovascular disease had lower physical functioning scores than subjects who died of cancer or other causes of death (Table 2). Sixty subjects were censored before the closing date of the analysis period. Compared with the remaining 1170 subjects, these subjects were more likely to report no underlying health condition (63.3% versus 54.8%) and to rate their health as fair/poor (25.4% versus 17.3%). They were similar with regard to age (median years: 71.5 versus 70.0), selfreported leisure-time physical activity (median METs per week: 31 versus 33.5), percent with fair/poor appetite (11.7 versus 11.0), percent of ever smokers (48.3 versus 49.2), and all NRB scores. For cardiac and cardiovascular mortality, the best fit proportional hazards models included the NRB score, age, selfrated health, and presence of underlying health conditions (Table 3; data for combined cardiovascular deaths not shown because the results were not different). Among the three NRB scores, there was little difference in the fit of the models based on AIC. The largest effect estimates were for NRB1 and NRB2, the two measures with the more detailed scoring algorithm (relative risk of cardiac death for a 0.1 unit change: NRB1 and NRB2, 0.80 [95%CI 0.68–0.93]; NRB 3, 0.83 [95%CI 0.73–0.95]). An interaction with age was identified for the three NRB scores (Table 3) such that the mortalitysparing effect of higher levels of physical functioning diminished with age.

The best-fit model for noncardiovascular mortality also included self-rated appetite (Table 4). There was little difference in the fit of models, and, overall, the effect of the NRB scores on mortality was less than that observed for cardiac mortality. Unlike the results for cardiac and cardiovascular disease, there was no age interaction with the NRB scores. Inclusion of a variable to indicate the exact date of the baseline interview did not affect any of the parameter estimates in any of the models. To compare further the various models, we computed the estimated probability of survival beyond 6 years for women with different observed function scores and who had the following other characteristics: (1) an underlying cardiovascular condition but no noncardiovascular condition or other chronic health condition, (2) age 80 years, and (3) self-rated health of “good.” Comparisons were restricted to NRB2 and NRB3 (i.e., those without energy weights) (Table 5). Overall, the predicted survivals were similar for the ordinal (NRB2) and dichotomous scores, although the gradient of decrease in probability of survival with decreasing function was somewhat greater for the ordinal scores.

4. Discussion We observed a consistent relationship between self-reported functional limitation and mortality from all causes. Function scores based on multiple levels of response (NRB1

Table 2 Physical function scores and leisure-time physical activity in relation to causes of death in 1230 female subjects for the SPPARCS project Function scoresd a

Cause of death

Cardiac Acute ischemic heart disease Congestive heart failure All other (arrhythmias, cardiomyopathy, valvular) Cardiovascular Caridac Noncardiac Cerebrovascular Other (aortic aneurysms) Other Cancer (excluding nonmelanoma skin) Chronic lung diseases/respiratory failure Pneumonia All other

Number (%) 55 10 15 30

e

Age at death

b

Survival time

c

NRB2

NRB3

(28.6 ) (18.2f) (27.3) (54.5)

84 (61–99)

3.4 (0.4–6.8)

0.60 (0.26–0.84)

0.60 (0.20–0.80)

74 (38.5e) 55 (74.3f) 19 (25.7) 15 (78.9f) 4 (21.1) 118 (61.5e) 54 (45.8f) 13 (20.4) 12 (10.1) 39 (33.1)

84 (61–99)

3.5 (0.4–7.1)

0.58 (0.25–0.80)

0.50 (0.10–0.80)

82 (62–97) 80 (62–95)

3.8 (0.2–6.9) 3.2 (0.2–6.3)

0.70 (0.38–0.90) 0.84 (0.60–1.00)

0.60 (0.20–1.00) 0.80 (0.40–1.00)

Abbreviations: SPPARCS, Study of Physical Performance and Age-Related Changes in Sonomans; NRB, summary measure based on Nagi [1] and Rosow and Breslau [2]. a Total deaths: 192. b Median (range). c Survival time from date of baseline observation: median years (range). d Median (25th–75th percentilest): NRB1 score is not included because this score and NRB2 had virtually identical distributions and results in the proportional hazards model. e % of 192 deaths. f % of category deaths.

I.B. Tager et al. / Journal of Clinical Epidemiology 56 (2003) 807–813

811

Table 3 Cox proportional hazard models for cardiac mortality in relation to function in females, SPPARCS project Hazard rations (95%CI) for function scoresa

Variables Age (per year)c NRB score (per 1 unit increase) Age × NRB

NRB1b (energy weights ordinal response score)

NRB2 (no energy weights ordinal response score)

NRB3 (no energy weights dichotomous response score)

1.05 (0.97–1.13) 0.10 (0.02–0.46)

1.04 (0.97–1.13) 0.10 (0.02–0.46)

1.07 (1.00–1.14) 0.16 (0.04–0.60)

1.14 (1.01–1.28) P ⫽ .04 AIC ⫽ 601.4

1.13 (1.00–1.13) P ⫽ .04 AIC ⫽ 601.5

1.01 (1.00–1.23) P ⫽ .06 AIC ⫽ 602.2

Abbreviations: SPPARCS, Study of Physical Performance and Age-Related Changes in Sonomans; NRR, summary measures based on Nagi [1] and Rosow and Breslau [2], AIC, Akaike Information Criterion. a Adjusted for self-rated health status and underlying health conditions. Self-rated appetite and smoking not retained in the final model; 1072 subjects with 53 cardiac deaths with data on all variables. b Parameter estimate (SE). c Age centered on median ⫽ 70 years.

and NRB2) seem to contain useful information for the longitudinal study of the disablement process, and it would be useful to see the application of the approaches in other studies. We chose to evaluate self-reported performance of tasks that depend primarily on upper and lower body muscle strength because limitations in these tasks might also be related to the ability or willingness to participate in leisure-time physical activity, which is a major focus of our study. Moreover, limitation in the performance of these tasks is considered an early stage in the “disablement process” [18], a stage that may be reversible or influenced in such a way as to slow the progress toward disability [19]. Because summary measures of self-reported ability to complete specific tasks have not been standardized, particularly in terms of the scoring of responses, we varied the weighting of answers and the content of the scoring. All three summary Table 4 Cox proportional hazard models for non-cardiovascular mortality in relation function scores for women, SPPARCS project

measures of self-reported physical functioning were associated significantly with increased risk of death from cardiac, cardiovascular, and noncardiovascular diseases. The results with scores based on energy weighting (NRB1) were no different from those for NRB2, which was based only on an ordinal difficulty score. The NRB1 and NRB2 scores had larger magnitudes of association with all categories of mortality than the score based on a dichotomous classification (NRB3). Although the differences were small when viewed in terms of relative risk of mortality (Tables 4 and 5) and probability of survival beyond 6 years, there is a suggestion that the incorporation of more finely graded answers into the standard questions on self-reported disability (NRB1 and NRB2) could allow for more detailed study of

Table 5 Comparison of effect of ordinal versus dichotomous coding of selfreported estimated physical functioning in women on estimated probability of survival beyond 6 years based on model for cardiac deathsa

Parameter estimates (SE) for function scoresa

Variables Agec NRB score (per 1 unit increase) Age × NRB

NRB1b (energy weights ordinal response score)

NRB2 (no energy weights ordinal response score)

NRB3 (no energy weights dichotomous response score)

1.06 (1.00–1.13) 0.26 (0.10–0.70)

1.06 (1.00–1.13) 0.24 (0.09–0.65)

1.08 (1.02–1.13) 0.29 (0.13–0.66)

1.04 (0.96–1.13) P ⫽ .33 AIC ⫽ 1422.3

1.04 (0.96–1.13) P ⫽ .30 AIC ⫽ 1421.9

1.03 (0.96–1.10) P ⫽ .46 AIC ⫽ 1419.9

Abbreviations: SPPARC, Study of Physical Performance and Age-Related Changes in Sonomans; NRB, summary measures based on Nagi [1] and Rosow and Breslau [2]; AIC, Akaike Information Criterion. a Adjusted for self-rated health status, underlying health conditions, smoking, and self-rated appetite. 1130 subjects with 111 deaths with data on all variables. b Parameter estimate (SE). c Age centered on median ⫽ 70 years.

Percentage difference in estimated survival between lower and higher function score (95%CI)

Scores without energy weights

Function scoreb

Estimated probability of survival (95%CI)

Ordinal Responses (NRB2) Dichotomous Responses (NRB3)

0.2 0.8

0.70 (0.55–0.90) 0.83 (0.73–0.93)

⫺15.2% (⫺25.2–⫺4.0)

0.2 0.8

0.73 (0.60–0.90) 0.83 (0.73–0.93)

⫺11.6% (⫺18.9–⫺3.9)

Abbreviations: CI, confidence interval; NRB, summary measure based on Nagi [1] and Rosow Breslau [2]. a Based on models for cardiac disease in Table 3 for women with an underlying cardiac condition, no other co-morbidity, and age 80. b Each score and the estimated survival probability represent actual data from a single subject with the characteristics specified in footnote above. Data are presented only for NRB2 and NRB3 because results for NRB1 and NRB2 were nearly identical.

812

I.B. Tager et al. / Journal of Clinical Epidemiology 56 (2003) 807–813

the progression from full functionality to disability in longitudinal studies. The failure to detect clear differences between the scoring algorithms could be the result of a number of factors. Although we tried to match the specific tasks in the NRB scores to tasks in the Ainsworth compendium of task-specific energy equivalents [17] that were as similar as possible, the energy equivalents are averages and do not reflect necessarily the average energy expenditure by a given individual for a given task. Moreover, several of these tasks incorporate elements of balance (e.g., walking stairs, walking city blocks), the effects of which are not captured in the estimates of energy expenditure. This measurement error is likely to be most pronounced in this elderly population for the three tasks that received the greatest weight (walking stairs, pushing heavy objects ⬎10 lb, and pushing heavy objects; see Appendix). The errors most likely are in the direction of over-weighting these tasks in terms of the energy expended in their performance relative to the less strenuous tasks. That this is the case is suggested by the similarity between the NRB1 and NRB2 score (no energy weighting). It is also possible that the distribution of self-reported functional limitation in this population was too restricted to observe differences in the scores. Although our population was generally a high-functioning population in which scores were skewed toward 1, there were considerable differences in the score distributions between those who died and those who survived during the 6 years of follow-up (Table 2). The validity of the ordinal scoring that was used in the original questions (see Methods) also could have been a problem. The responses, “do not do activity,” “on doctor’s orders,” and “never do the activity” do not have an a priori ordinal structure. Therefore, some subjectivity was involved in the ordinal assignment of these categories. Alternative ordinal classifications were not evaluated. Finally, subject reliability (reproducibility) in the reporting of difficulty with these tasks is a source of error. Previously, we have shown that the test-retest reliability of the various NRB components is variable (kappa range 0.60–0.75) [20]. Walking stairs and pushing heavy objects and lifting objects ⬎10 lb had the highest kappas [20]. The scores were developed from the same questions, and the reliability should have affected the scores equally. We did not evaluate associations between the scores and subsequent disability as measured by ADL scores [21] because virtually all of the subjects in our population who developed ADL abnormalities died during the follow-up period, and only 5% had at least one ADL abnormality at baseline evaluation. Finally, we focused our evaluation on tasks that predominantly reflect activities with a major aerobic component because the focus of the larger study relates to relationships between leisure-time physical activity and physical functioning. Inclusion of range of motion and standing up from chairs may have provided greater separation among the scores. Similar to the results of other studies, we observed that the risk of death from cardiovascular and noncardiovascular

diseases in an elderly population of women was associated independently with summary scores based on self-reported ability to carry out specific physical tasks [6–8]. Ninety-five percent of women in our study reported no ADL disabilities at baseline, which makes it unlikely that the relationships between the function scores and mortality are just a reflection of the greater mortality risk associated with more advanced disability at the baseline evaluation. The beneficial effects of reports of higher levels of self-reported performance of selected physical tasks at baseline on subsequent cardiac and cardiovascular mortality were decreased with age for all formulations of the function scores. This finding likely reflects intervening events that reduced physical function for tasks that place stress on the cardiovascular system over the follow-up period. Although this possibility could not be explored with precision because only three observation points are available for the cohort, the emphasis of the items on aerobic tasks and the lack of an age interaction with noncardiovascular deaths makes this a plausible supposition. In summary, based on our data, there seems to be a justification for the development of a standardized method for the items to be included and the method of scoring for self-reported physical functioning. Such standardization should be based on a specific set of self-reported functioning that includes measures of upper and lower body strength and balance and a set of ordinal responses such as those used in this study. Based on our data, energy weighting does not seem to provide additional information relative to unweighted ordinal answers; however, pooling of the actual data from a large number of population-based studies would provide for a more precise determination of whether weighting by energy equivalents provides additional information for the classification of self-reported physical functioning. Appendix: Details of formulation of NRB score

Question itema

MET value assignedb

Walking stairs Walking city blocks Lifting ⬎10 lb Lifting ⬍10 lb Pushing heavy object

8.0 2.5 6.0 3.5 6.0

Total The final score was

25

Ainsworth codes 17130 (up) 17070 (down) 17170 17027 17020 05120

Fraction of 25 METS (wi) 0.32 0.10 0.20 0.14 0.24 1.00

5

wf calculated as follows: 兺 1 i i a Each of the seven possible answers to the questions was assigned a score ( fi): A lot/unable 0.0 Never/MD 0.2 Some/little 0.5 No difficulty 1.0 “Never/MD” is rate higher than “A lot/unable” because “never” includes people who could do the activity but choose not to do so and people who never do it because they cannot. The standard question does not permit distinction between these two groups. b MET-value assigned from Ainworth et al. in 1993–94 [17]. The values are identical to those in the updated version [22].

I.B. Tager et al. / Journal of Clinical Epidemiology 56 (2003) 807–813

References [1] Nagi SZ. An epidemiology of disability among adults in the United States. Milbank Quarterly 1976;54:439–68. [2] Rosow I, Breslau N. A Guttman health scale for the aged. J Gerontol 1966;21:556–9. [3] Simonsick EM, Lafferty ME, Phillips CL, et al. Risk due to inactivity in physically capable older adults. Am J Pub Health 1993;83:1443–50. [4] Sinomsick EM, Kasper JD, Guralnik JM, et al. Severity of upper and lower extremity functional limitation: scale development and validation with self-report and performance-based measures of physical function. J Gerontol Soc Sci 2001;56B:S10–9. [5] Jette AM, Assman SF, Rooks D, et al. Interrelationships among disablement concepts. J Gerontol Med Sci 1998;53A:M395–404. [6] Manton KG. A longitudinal study of functional change and mortality in the United States. J Gerontol 1988;43:S153–61. [7] Scott WK, Macera CA, Cronman CB. Functional health status as a predictor of mortality in men and women over 65. J Clin Epidemiol 1997;50:291–6. [8] Corti M-C, Guralnik JM, Salive ME, et al. Serum albumin level and physical disability as predictors of mortality in older persons. JAMA 1994;272:1036–42. [9] Fried LP, Ettinger WH, Lind B, et al. Physical disability in older adults: a physiological approach. J Clin Epidemiol 1994;47:747–60. [10] Beckett LA, Brock DB, Lemke JH, et al. Analysis of change in selfreported physical function among older persons in four population studies. Am J Epidemiol 1996;143:766–78. [11] Seeman TE, Charpentier PA, Berkman LF, et al. Predicting changes in physical performance in a high-functioning elderly cohort: MacArthur Studies of Successful Aging. J Gerontol Med Sci 1994;49A:M97– M108. [12] Guralnik JM, LaCroix AZ, Abbott RD, et al. Maintaining mobility in later life I. demographic characteristics and chronic conditions. Am J Epidemiol 1993;137:845–57.

813

[13] Mendes de Leon CF, Beckett LA, Fillenbaum GG, et al. Black-white differences in risk of becoming disabled and recovering from disability in old age: a longitudinal analysis of two EPESE populations. Am J Epidemiol 1997;145:488–97. [14] Paffenbarger RS Jr. Some interrelations of physical activity, physiological fitness, health, and longevity. In: Bouchard C, Shephard RJ, Stephens T, editors. Physical activity, fitness, and health: international proceedings and consensus statement. Champaign, IL: Human Kinetics Publishers; 1994. p. 119–33. [15] Williams PT. High-density lipoprotein cholesterol and other risk factors for coronary heart disease in female runners. N Engl J Med 1996;334:1298–303. [16] Satariano WA, Smith J, Swanson A, et al. A census-based design for the recruitment of a community sample of older residents: efficacy and costs. Ann Epidemiol 1998;8:278–82. [17] Ainsworth BE, Haskell WL, Leon AS, et al. Compendium of physical activities: classification of energy costs of human physical activities. Med Sci Sports Exerc 1993;25:71–80. [18] Lawrence RH, Jette AM. Disentangling the disablement process. J Gerontol Soc Sci 1996;51B:S173–82. [19] Jette AM. Disability trends and transitions. In: Binstock RH, George LK, Marshall VW, Myers GC, Schulz JH, editors. Handbook of aging and the social sciences. New York: Academic Press; 1996. p. 94–116. [20] Hollenberg M, Ngo LH, Turner D, et al. Treadmill exercise testing in an epidemiologic study of aging and physical activity: application of the Cornell Protocol. J Gerontol Biol Sci 1998;53B:B259–67. [21] Katz S, Akpom CA. A measure of primary socio-biological functions. Int J Health Serv 1976;6:493–508. [22] Ainsworth BE, Haskell WL, Whitt MC, et al. Compendium of physical activities: an update of activity codes and MET intensities. Med Sci Sports Exerc 2000;32(Suppl):S498–516.