Relationship Between Physical Activity and Health-Related Utility Among Knee Osteoarthritis Patients

Abstract

Objective

To estimate the relationship between physical activity and health-related utility for people with knee OA and implications for designing cost effective interventions.

Methods

Use GEE regression analysis to estimate partial association of accelerometer-measured physical activity levels with health-related utility after controlling for demographics, health status, knee OA severity level, pain and functioning.

Results

Moving from lowest to middle tertile of physical activity levels is associated with .071 (p<.01) increase in health-related utility after controlling for demographics and .036 (p<.05) increase in utility after controlling for demographics, health status, knee OA severity level, weight, pain, and functional impairments.

Conclusion

Intervention programs that move individuals out of the lowest tertile of physical activity have the potential to be cost effective.

Knee Osteoarthritis (OA) is a major health problem, which affects 6% of adults (1). Exercise programs for persons with arthritis can increase strength and functional status and decrease pain, depressive symptoms, fatigue, and quality of life without adversely affecting joint status (2,3,4). There is evidence that these physical activity-associated benefits can also be realized by interventions that encourage increased physical activity without utilizing formal exercise programs, which can be difficult to sustain long-term (5). The need for increased activity in this population is highlighted by the findings showing that 44% of adults reporting physician-diagnosed arthritis were inactive (6) and that 41.1% of males and 56.5% of females with osteoarthritis of the knee were inactive (7).

When evaluating the relative value of programs directed at improving overall physical activity of people with knee OA, cost effectiveness calculations are increasingly being used to evaluate if the net gain to society is worth the additional costs of the intervention. A generally accepted measure of effectiveness is the effect of an intervention on health-related utility, which is then used to construct changes in Quality-Adjusted-Life-Years (QALYs) (8). A QALY weights one’s remaining lifetime or, alternatively, the period evaluated following an intervention, by the value of health-related utility observed during that time period, where utility is normed to range from death (utility equals zero) to perfect health (utility equals 1). Interventions shown to improve the number of Quality Life Years (QALYs) of an individual have been used to justify a given intervention cost (8). Use of QALYs as an outcome has been formalized in evaluating drug coverage decisions in countries such as Britain, Australia, and Canada (11).

This paper has two objectives: 1) to assess the extent to which increased physical activity is positively associated with health-related utility and 2) to then discuss whether this association is large enough to potentially justify as cost effective interventions that are successful in increasing physical activity. To assess the magnitude of changes in physical activity levels necessary to achieve utility gains, we estimated the association between physical activity levels and health-related utility in a sample of adults with radiographically confirmed symptomatic knee OA (Kellgren-Lawrence grade 2 or greater) recruited for a counseling intervention to improve level of physical activity. While the association between health-related utility measures and physical activity levels among randomized controlled trial participants may differ from the relationship in the general community, it can suggest the extent to which an intervention might change the level of QALYs if it can substantially affect physical activity levels. This information, in turn, suggests limits on the cost of interventions aimed at improving physical activity, if such interventions are to be cost effective.

METHODS

Sample

The sample of knee OA participants for this study comes from an RCT of adults with knee OA conducted to assess the efficacy of a tailored health promotion intervention to increase physical activity. The 155 knee OA participants, were recruited from clinical practices (8%); research registries (40%) and the community (52%). Participants were excluded from the study if they had 1) planned total joint replacement in the subsequent 12 months, 2) contraindication to physical activity due to comorbid condition including a history of peripheral vascular disease, spinal stenosis, residual lower extremity neuromuscular effects of stroke, major signs or symptoms suggestive of pulmonary and cardiovascular disease, 3) were unable to perform basic self-care activities, 4) plans to relocate away from the Chicago area within 24 months, or 5) were missing baseline data for any of the measures described below.

Measures Collected for Variables of Interest

Physical activity was monitored in all study participants using a GT1M Actigraph accelerometer, which measures vertical acceleration and deceleration (12). Accelerometer data were collected at 1-minute intervals and transformed to activity counts per day. An activity count is the weighted sum of the number of accelerations measured each minute, where the weights are proportional to the magnitude of measured acceleration. Participants were instructed to wear the accelerometer upon arising in the morning, and wear continuously (except for water activities) until going to bed at night for seven consecutive days. Skipped days, reported on a daily log, were excluded from the analysis. Consistent with Troiano et al (6), days with less than 10 hours of wear time were excluded from the analysis, where the transformation algorithm converting raw accelerometer readings to wear time was determined using methods validated for knee OA populations (13).

The amount of pre-intervention physical activity was estimated from mean daily activity counts, which represented the summed activity counts for all wear hours divided by the number of valid days (days having > 10 hours of wear time) monitored during the 7-day period. We then divided the sample into tertiles – three groups representing the lowest, middle, and highest physical activity levels at baseline.

Health-related utility was measured using the SF-6D utility measure, which is based on Brazier et al’s cross-walk between the SF-36 and the SF-6D measure (9). More specifically, the SF-6D measures the following six health domains: physical functioning, role limitations, social functioning, pain, mental health, and vitality. For example, the pain domain varies from having no pain to having pain that interferes with work (both outside the home and housework) extremely (14). . Because the SF-36 was collected at baseline, 3-month, 6-month and 12-month follow-up interview periods, we had multiple measures of the SF-36 measure for each individual, which were then converted to SF-6D utility scores using the preference weights estimated from a random sample of community dwellers by Brazier et al (9). Each individual had between one and four measures of health-related utility.

Baseline interviews also collected demographic (age, whether female, nonwhite, without any college education) ; clinical health factors (Kellgren-Lawrence knee OA severity level=2, 3, or 4]; number of comorbidities ascertained from baseline medications, overweight [<25 body mass index [BMI]<=30),obese [BMI>30]) ; WOMAC pain; and disability (instrumental activity of daily living (IADL) and activity of daily living (ADL) limitations (14)). Design variables include membership in the intervention or control group and the follow-up interview at which utility was assessed.

Statistical Methods

A repeated measures hierarchical general estimating equations (GEE) regression analysis related health-related utility at baseline, 3-month, 6-month, and 12-month follow-up periods to baseline physical activity tertile levels (using the lowest tertile as reference). This association of physical activity levels with utility level controlled for 1) design factors and demographics, then added 2) clinical health factors 3) pain, and disability.

RESULTS

Of the 155 individuals with knee OA enrolled in the study 142 had complete baseline data. These 142 individuals formed the analysis sample. There was an average of 3.27 SF-6D utility observations per individual. Figure 1 shows the distribution of baseline utility scores for each of the three physical activity tertiles, indicating roughly normal distributions and no significant floor or ceiling effects on the dependent variable. While the utility scale is normed from 0 (death) to 1 (perfect health), Figure 1 illustrates that, for this sample, utility ranged from .41 to 1. It is notable that higher utility scores (>0.8) occurred more frequently among the two upper tertiles. However, the overall relationship of physical activity tertiles with health-related utility is unclear from the figure.

Figure 1.

Baseline SF-6D Utility Distribution For Each Physical Activity ScoreTertile Percent of Sample at Each Utility Score (Tertile 1 is lowest activity tertile)

We defined our physical activity tertiles based on average daily total accelerometer counts during the measurement week. Actual physical activity counts for the middle and highest tertiles, respectively, are approximately double and more than triple those for the lowest physical activity group. The means and range of average daily counts for each tertile are provided in Table 1.

Table 1.

Baseline Characteristics by Physical Activity Tertile, N=142

Variable	Tertile 1- Low Activity Counts	Tertile 2- Medium Activity Counts	Tertile 2 vs. Tertile 1 p-value a	Tertile 3- High Activity Counts	Tertile 3 vs. Tertile 1 p-value a
Sample size	N=47	N=49	-	N=46	-
Mean SF-6D Utility (s.d.)	0.74 (0.11)	0.76 (0.10)	0.352	0.73 (0.13)	0.69
Mean No. PA Counts (s.d.)	113,273 (30,893)	207,398 (28,738)	<.001	346,176 (88,993)	<.001
Counts Range	(36,498 to 163,024)	(163,412 to 250,553)		(255,077 to 607,090)
Mean Age (s.d.)	72.36 (11.14)	61.48 (12.89)	<.001	56.54 (9.49)	<.001
%Female	65.96%	65.31%	0.95	45.65%	0.05
%Non-White	31.91%	48.98%	0.09	54.35%	0.03
%No College	21.28%	26.53%	0.55	23.91%	0.76
BMI			0.91		0.52
%normal
%Normal	12.77%	14.29%		17.39%
%Overweight	29.79%	32.65%		36.96%
%Obese	57.45%	53.06%		45.65%
%Any Comorbidities?	48.94%	32.65%	0.10	41.30%	0.46
Kellgren-Lawrence					0.26
% =Grade 2	44.68%	53.06%	0.63	60.87%
% =Grade 3	29.79%	28.57%		23.91%
%=Grade 4	25.53%	18.37%		15.22%
Mean WOMAC Pain (s.d.)	18.68 (10.62)	15.55 (11.98)	0.18	19.29 (12.57)	0.80
%IADL only	4.26%	2.04%	0.53	6.52%	0.63
%Any ADL	6.38%	4.08%	0.61	4.35%	0.66

^aStatistical difference from Tertile 1 value based on P-value from Chi-square test for categorical variables and t-test for continuous variables.

Table 1 also shows the baseline levels of the demographic, health, and pain/functional limitation scores for each tertile. Higher levels of physical activity are significantly correlated with age, being male, and non-white. Table 2 shows the hierarchical regression results. Controlling only for demographic characteristics, the middle activity tertile has a utility score .071 higher (p<.01), and the highest activity tertile has utility score .058 higher (p<.05) than the lowest activity tertile. Adding clinical health factors reduces the effects somewhat, and further adding pain, disability, and design variables, reduces the coefficients on middle and high activity tertile to .036 (p<.05) and .027 (n.s.), respectively. Thus, while moving from the lowest to the middle physical activity tertile is associated with a significant increase in utility level, no further increase is associated with moving from the middle to the highest physical activity tertile. There is no significant difference between the middle and highest tertile coefficients. Significant health factors (Model 2) were being overweignt or obese, which have a negative effect on utility. In the final model, the only other significant predictors of utility were age and pain.

Table 2.

GEE Regression Results Predicting SF-6D Utility Levels

Variables: Baseline Measure	Model 1	Model 2	Model 3
Intercept	.705**	.791**	.817**
Phys Act 2	.071**	.057**	.036*
Phys Act 3	.058*	.035	.027
Age (centered on 60)	.003**	.002**	.001*
Female	−.020	−.029	−.012
Non-White	−.005	−.001	.016
No College	−.014	−.006	.008
Overweight		−.060**	−.025
Obese		−.072**	−.029
Any Comorbidities?		−.008	−.012
Kellgren-Lawrence=3		−.032 (p=.06)	−.022
Kellgren-Lawrence=4		−.020	−.013
WOMAC Pain (21 pt scale)			−.004**
IADL only			.026
Any ADL			−.033
3 month followup	.006	.007	.007
6 month followup	−.002	−.003	−.003
12 month followup	.013	.014	.016
Intervention	.005	.008	.004

^*Statistically significant at .05 level;

^**Statistically significant at .01 level. N=464 observations with 142 individuals and 3.27 observations per individual.

DISCUSSION

The evaluation of interventions to improve physical activity is often put to a cost effectiveness test, asking whether outcomes justify their costs. The standard metric to evaluate cost-effectiveness is the change in quality of life years (QALYs), a measure based on health-related utility. One might therefore wonder, can an intervention that moved knee OA individuals from the lowest tertile of physical activity to higher tertiles be justified in terms of improved health-related utility? And at what cost, given current guidelines that rate programs as cost-effective effective if they are below $50,000–$100,000 per QALY (10).

We find that using a standard health-related utility score based on the SF-36 health status questionnaire, individuals in the middle physical activity tertile have health-related utility scores that are .071 (p<.01) higher than those in the lowest tertile after controlling for demographics. Further control for clinical factors, pain and disability reduced this difference to .036 (p=.032). Interestingly, individuals in the highest tertile appeared to have utility levels similar to (actually, insignificantly lower than) those in the middle tertile. Thus, targeting those with the least active lifestyles and, on average, doubling their physical activity counts, would potentially improve their QALYs significantly, as measured by a standard cost effectiveness measure.

Would such an intervention be cost effective? Consider the .036 difference in utility we found after controlling for demographic, health, pain, and disability differences. Suppose an intervention could obtain such an effect after one year. Starting at zero difference and attaining the .036 utility difference at 12 months yields an average annual QALY gain of .018, if utility gains showed a linear trend over the year. If the intervention cost is $450, the related cost effectiveness ratio is $25,000 per QALY, well within the cost effectiveness range. Under a more conservative assumption that a program was only 33% effective in moving individuals into the higher physical activity tertile zone, then this cost effectiveness ratio becomes $75,000 per QALY. Lowering the cost of the intervention to $300 reduces the cost-effectiveness ratio to $50,000 per QALY.

Of course, the associations found here do not necessarily translate to what we would observe if we changed physical activity levels – they are not necessarily causal. Unobserved baseline differences in health status correlated with baseline physical activity levels might explain observed associations between physical activity and utility levels. However, by the same token we have been conservative here by controlling for levels of pain, overweight/obesity levels, and disabilty in estimating expected differences between the lowest and other activity levels. In fact, if a physical activity intervention improved weight, disability, and pain levels, the overall utility levels would have greater improvment.

In conclusion, while we cannot say whether a controlled intervention would reproduce the observational results reported here, there is potential for a physical activity intervention to be cost effective at existing cost-effectiveness standards. Limiting the cost of such an intervention, targeting to individuals with very low levels of physical activity, and substantially increasing physical activity counts would appear to be necessary ingredients for a successful intervention. However, achieving these goals would not appear to require substantial increases in bouts of moderate/vigorous activity, often associated with successful exercise programs and required to meet current DHHS physical activity guidelines (15).

Significance and Innovative Findings.

• A significant increase in health related utility is observed when moving out of the lowest tertile of physical activity.

• Interventions that target the individuals with the lowest physical activity levels to increase their level of non-vigorous activity have the potential to be cost effective.