- Click here for explanation of classification scheme.

• To examine the role of modifiable predictors of functional decline among community-residing older women
• To validate a clinical prediction tool for functional decline based on modifiable predictors.

• Women
• Aged 65 years or older
• Participants in the Study of Osteoporotic Fractures.

• African American women
• Those unable to walk without the help of another person
• Women with bilateral hip replacements.

Recruitment

Participants were recruited from several sources. In two communities women were identified from membership lists for large health maintenance organizations. In one community they were identified from lists of residents that had been produced for the Hypertension Detection and Follow-up and the Systolic Hypertension in the Elderly Studies and from jury lists. In one community participants were identified from 1985 voter registration lists. In another, they were selected from county lists of holders of drivers' licenses and identification cards. Women received a letter and brochure inviting them to participate in the study.

Design

Prospective cohort study 

Blinding used

Not used

Intervention

Not applicable

Statistical Analysis

• To create a simple tool that could be used by clinicians without much calculation, the authors dichotomized the continuous predictor variables
• A prediction rule based on eight modifiable predictors classified women into three different risk groups for decline in basic activities
• To generate a prediction rule score that would be practical in clinical settings, beta-coefficients were examined in each model, multiplied by a common factor, and rounded to the nearest integer
• To estimate the maximum proportion of functional decline that could be attributed to modifiable predictors, models were constructed with the prediction tool variables in their original continuous state and were compared with those of the prediction tool
• Sensitivity analysis was conducted because decisions regarding sample inclusion criteria, model selections, and predictor variable cutpoints may have influenced findings.

 

Timing of Measurements

• Participants were recruited between September 1986 and October 1988. Visit 1 took place between 1986 and 1988, visit 2 between 1988 and 1990, and visit 4 between 1992 and 1994
• At baseline and at two-year intervals, participants underwent evaluations that included performance tests and interviews to assess functional status. Potential predictors of functional status were measured at visit 1 or visit 2, and function was reassessed at visit 4.
• Exercise level and and visual acuity were measured at visit 1; all other modifiable predictors were measured at visit 2.

 Dependent Variables

• Functional decline as measured by ability to carry out 13 daily activities from the 1984 National Health Interview Survey Supplement on Aging 
• Assessment of functional status was obtained by interviewing participants about their ability to carry out 13 daily activities from the National Health Interview Survey Supplement on Aging and a modified Health Assessment Questionnaire. Activities were classified as vigorous or base activities. Vigorous activities included doing household chores like vacuuming, dusting, etc, heavy housework like scrubbing floors or washing windows, shopping, climbing up 10 steps without resting, and walking two to three blocks on level ground. Base activities were things like getting in and out of bed, turning faucets of and off, getting in and out of a car, dressing yourself, washing and drying your body, bending down to pick clothes up off the floor, preparing your own meals, and lifting a full cup or glass to your mouth.

Independent Variables

• Modifiable risk factors were those that a clinician seeing an older patient for the first time could act upon and reasonably expect to change over a four-year period. These modifiable predictors became candidates for the decision tool: Depression, high body mass (BMI), poor visual acuity, low bone mineral density (BMD), benzodiazepine use, weak grip, slow gait, low social functioning, and low exercise level.
• Fixed predictors of functional decline included age, education, medical comorbidity, cognitive function, presence of spine fracture and smoking history
• Benzodiazepine use (measurement was not described, but it is assumed to be based on self-report from interview/survey/questionnaire)
• Depression as measured by the 15-item Geriatric Depression Scale using a cutpoint of 6 or greater
• Low exercise level (measurement was not described but it is assumed to be based on self-report from interview/survey/questionnaire)
• Social functioning (measurement was not described but it is assumed to be based on self-report from interview/ survey/questionnaire)
• Body mass index as measured using standard techniques not described
• Poor visual acuity as defined by binocular vision 20/40 or worse
• Low bone mineral density as measured by dual-energy X-ray absorptiometry
• Slow gait as measured by the time in seconds needed to walk six meters at a rapid pace
• Weak grip as measured by a grip dynamometer in both hands Control Variables.

 

• Initial N: 6,632 women
• Attrition (final N): All participants were accounted for; no attrition
• Age: 65 years or older at baseline, with a mean age of 73±4.9 years
• Ethnicity: Not specified except to note that African Americans were excluded
• Other relevant demographics:
  • 39.2% of the sample had <12 years education
  • 40.3% had 12 years
  • 20.5% had >12 years education
• Anthropometrics: Mean BMI of participants was 26.3±4.6. Of the sample of 6,632:
  • 12.1% reported having CAD (coronary artery disease)
  • 61.7% had arthritis
  • 3.5% had a previous stroke
  • 2.0 % had CHF (congestive heart failure)
• Location: Participants were recruited from:
  • Portland, Oregon
  • Minneapolis, Minnesota
  • Baltimore, Maryland.

 

Key Findings

Five modifiable risk characteristics (slow gait, short-acting benzodiazepine use, depression, low exercise level, and BMI ≥29) are associated with functional decline in both vigorous and basic activities. Weak grip predicted functional decline in vigorous activities, whereas long-acting benzodiazepine use and poor visual acuity predicted functional decline in basic activities.

A prediction rule based on these eight modifiable predictors classified women in the derivation set into three risk groups for decline in vigorous activities (12%, 25%, and 39% risk) and two risk groups for decline in basic activities (2% and 10% risk). In the validation set, the probabilities of functional decline were nearly identical.

Significant Multivariable Predictors of Functional Decline in Vigorous Activities*

Risk Factor	β	OR	95% CI	Points°
y Intercept	-3.53
Slow gait (lowest quintile)	0.57	1.76	1.44-2.16	2
Short-acting benzodiazepine use	0.48	1.62	1.21-2.18	2
Depression (6 or greater on 15-item GDS)	0.41	1.51	1.09-2.08	2
Low exercise level	0.31	1.36	1.12-1.67	1
BMI≥29	0.26	1.30	1.07-1.58	1
Weak grip strength (lowest quintile)	0.19	1.21	0.99-1.49	1

* Model adjusted for age, level of education, medical comorbidity, cognitive function, presence of spine fracture, smoking status, and enrollment site

OR=odds ratio   CI=confidence interval

°Calculated by multiplying β coefficients by 4 and rounding to the nearest integer. The point scheme for predicting functional decline is derived from summing the points from the six significant factors.

Significant Multivariable Predictors of Functional decline in Basic Activities

Risk Factor	β	OR	95% CI	Points
y Intercept	-1.66
Slow gait (lowest quintile)	0.83	2.29	1.66-3.17	2
Depression (6 or greater on the 15-item GDS)	0.62	1.87	1.17-2.98	1
Long-acting benzodiazepine use	0.59	1.80	1.10-2.95	1
Visual acuity worse than 20/40	0.51	1.66	1.09-2.55	1
Short-acting benzodiazepine use	0.50	1.66	1.02-2.68	1
Low exercise level (lowest quintile)	0.39	1.47	1.06-2.05	1
BMI ≥29	0.32	1.37	0.98-1.93	1

* Model adjusted for age, level of education, medical comorbidity, cognitive function, presence of spine fracture, smoking status, and enrollment site

OR=odds ratio   CI=confidence interval

°Calculated by multiplying β coefficients by 4 and rounding to the nearest integer. The point scheme for predicting functional decline is derived from summing the points from the six significant factors.

 Performance of the Prediction Rule for Functional Decline

Risk Group	Points	Derivation Set n/N (95% CI*)² RR	Validation Set n/N (95% CI*)² RR
Vigorous Activities Low Moderate High	0-1 2-3 =4	391/3172(11.2-14.7)     1.0 249/1012(22.0-27.3)      2.0 92/238(32.5-44.9)          3.1	220/1614(12.0-15.3)    1.0 126/474(22.6-30.6)        2.0 48/122(307-48.0)           2.9
Basic Activities Low High	0-1 =2	74/3151 (1.8-2.9)          1.0 130/1271(8.6-11.9)       4.4	38/1602(1.6-3.1)          1.0 42/608(4.5-8.9)              2.9

 *95% confidence intervals of participants in risk group experiencing functional decline

 ²Chi-square trend, P<0.001

The authors conclude that a substantial portion of the variation of functional decline can be attributed to risk factors amenable to intervention over the short term. In addition, using eight modifiable predictors that can be identified in a single office visit, clinicians can identify older women at risk for functional decline.

Government:	Public Health Service
University/Hospital:	UCLA School of Medicine, USCF School of Medicine
Not-for-profit	Robert Wood Johnson Foundation Other non-profit:

• It is not clear to this reviewer how information on benzodiazepine use, social functioning, and low exercise level were obtained but it appears to be self-reported. Self-reporting may not be accurate and could have affected the outcome of the study.
• The authors do not appear to have taken confounding factors such as use of medications (aside from benzodiazepines) into account
• This age group is at risk for dementia, which could affect BMI, physical functioning (gait and exercise level), social functioning and depression. Presence or absence of dementia is not accounted for in this study.
• It is not clear to this reviewer if changes in benzodiazepine use and/or dose were measured over time
• Authors note a major limitation is lack of ethnic and gender diversity among participants; whether findings are generalizable to different populations is unknown