Identifying Sedentary Subgroups: The NCI Health Information National Trends Survey

Abstract

Background

Developing effective interventions for the 24%–28% of U.S. adults who are sedentary requires a better understanding of the factors related to sedentary lifestyles as well as the communication channels to reach various subgroups. This study identified key sociodemographic and health communication characteristics of various subgroups with high rates of inactivity using signal detection methodology (SDM).

Methods

The sample from the nationally representative Health Information National Trends Survey 2003 (HINTS; N=6369) was randomly split into two samples. Exploratory analyses (conducted 2004–2005) were employed on the first sample to identify various subgroups, and the stability of inactivity rates in those subgroups was examined in the second sample.

Results

Eight subgroups with varying levels of inactivity were identified. Three subgroups had inactivity levels ≥40%, while the lowest subgroup had a level of <15%. The highest inactivity subgroup consisted of individuals with at least some college education who were in fair/poor health and who watched 4+ hours of television/day. The second highest inactivity subgroup was composed of those without a college education who tended not to utilize nor attend to many communication channels. The third highest inactive subgroup consisted of those without a college education who read the newspaper and were obese. Levels of subgroup inactivity in the second independent sample were not significantly different from those found in the exploratory sample.

Conclusions

This study identified empirically-based, physically inactive subgroups that differed on sociodemographic and health communication characteristics. This information should be useful in creating future evidence-based, targeted, and tailored intervention strategies.

INTRODUCTION

Inactivity has been linked to increased risk of many chronic illnesses,^{1, 2} including increased cancer risk and carcinogenesis for many cancers.^3–5 Consequently, one of the major goals of the Institute of Medicine 2003 report—Fulfilling the Potential of Cancer Prevention and Early Detection—is the development of prevention strategies to reduce sedentary behavior and obesity.⁶ The importance of these prevention strategies is highlighted by national surveys indicating that 24%–28% of adults in the United States are completely sedentary, and that levels of obesity in the U.S. are growing to epidemic proportions.^2,7 A better understanding of the characteristics of sedentary populations is required to develop effective intervention strategies.

Those who are physically inactive (i.e., sedentary) in all likelihood do not represent just one group, but may consist of several subgroups. Yet, the identification of specific inactivity subgroups has received relatively little attention, and attempts to characterize inactive groups have typically taken a unidimensional perspective by classifying subgroups along one domain (e.g., age, gender, ethnicity) or examining multiple correlates without evaluating possible interactions among the variables.⁸ The characteristics of inactive groups are likely to be complex and an aggregate of various factors. Examining multiple characteristics of sedentary groups may aid in more effective health promotion intervention programs.

Identification of critical communication channels is key in developing effective programs for reaching sedentary subgroups. In addition to traditional face-to-face approaches (e.g., exercise classes), prior physical activity research has focused on either mass media campaigns to promote physical activity or smaller-scale interventions using the telephone and/or print media to provide physical activity advice and information.⁹ Yet, few studies have simultaneously examined the spectrum of mediated communication channels (e.g., television, newspaper, Internet) available to reach the inactive populations, and how these channels can be used to disseminate effective interventions.

The purposes of this study were to (1) identify distinct subgroups with high proportions of inactivity in a nationally derived sample using signal detection methodology (SDM), and (2) examine the stability of the inactivity patterns using an independent sample.

METHODS

Data Source

Data came from the Health Information National Trends Survey (HINTS) 2003. This survey was developed to collect nationally representative data about the American public’s need for, access to, and use of cancer-relevant information (see http://cancercontrol.cancer.gov/hints/index.jsp).

Data Collection

Data were collected from October 2002 through April 2003. Trained interviewers used a list-assisted random-digit-dialing method from all working “banks” of telephone numbers within the U.S. One adult (aged 18+ years) within each household was selected for the extended interview during a household screener; final household response rate for the household-level screening interview was 55%. Black/African American and Hispanic residents were oversampled at a 1.8 times higher rate than expected under proportional allocation to increase the number of respondents in these typically underserved groups. Base weights were generated for each sampled identification number in the final data release based on the complex sampling strategy for the survey. The final set of replicate weights included calibrations against comparable population data for gender, age, race, and education publicly available from the Current Population Survey. Extended interviews were completed with 6369 adults. The response rate for the extended interview was 62.8%, a rate comparable to other recent telephone surveys of adults in the general population.¹⁰ Details regarding the HINTS 2003 sampling design are published elsewhere.¹¹

MEASURES

Outcome variable

Inactivity

The following item was used to create binary outcome of inactivity: “During the past month, did you participate in any physical activities such as running, calisthenics, golf, gardening, or walking for exercise?” This item was taken from the Behavioral Risk Factor Surveillance Survey (BRFSS) 2000 questionnaire.¹² Individuals answering “No” to this question were categorized as sedentary/inactive, and those answering “Yes” were categorized as “active” (i.e., not sedentary/inactive). Cognitive testing was conducted on this as well as other questions in HINTS 2003.

Predictor Variables

Sociodemographic Characteristics

Ethnicity and race were assessed according to Office of Management and Budget standards.¹³ Responses to the ethnicity and race questions were combined into the following four categories: (1) non-Hispanic whites, (2) non-Hispanic African Americans, (3) Hispanics (all races), and (4) non-Hispanic others or more than one race. Other demographic characteristics included age (continuous); gender; marital status (married, divorced, widowed, separated, never been married); education level (less than high school, high school graduate or General Educational Development Test [GED], some college or technical school, college graduate); employment status (employed for wages, self-employed, out of work, homemaker, student, retired, unable to work); household income (>$25K, $25K–$34K, $35K–$49K, $50K–$75K,$75K+); and residential setting (metropolitan, suburban, or rural). Respondents were asked about their general health (excellent to poor); whether they had a regular healthcare provider (yes/no); health insurance coverage (yes/no); cancer history (yes/no); and family history of cancer (yes/no). Body mass index (BMI = weight in kg/height in m²) was calculated based on self-reported height and weight.

Health Behaviors

Respondents were asked about whether they believed that exercise reduces cancer risk (yes/no), their fruit and vegetable consumption (continuous), and if they smoked 100+ cigarettes in their lifetime (yes/no).

Communication channels

Respondents were asked whether they ever went online to access the Internet or send/receive e-mails, number of hours TV watched, cable or satellite TV use, number of hours of radio listening on a typical weekday, number of days of newspaper reading during the past 7 days, and number of days of magazine reading during the past 7 days. Respondents were also asked how much attention they paid to health or medical information on the Internet, TV, radio, in newspapers, and in magazines (1=not at all to 4=a lot).

Analysis Plan

Descriptive analyses with the full sample were first conducted, and t-test analyses were used to compare inactive versus active individuals on the various characteristics, which provided points of comparison with the results of the signal detection analyses.

The sample was randomly split from HINTS (N=6369) into two samples. Data on inactivity were missing for 106 and 103 respondents in samples 1 and 2, respectively. The sample sizes for analyses were n=3076 and n=3084 for samples 1 and 2, respectively. SDM (using ROC4 statistical program¹⁴) was applied to the first sample to identify the various subgroups, their levels of inactivity, and the key factors that delineated each subgroup. All sociodemographic, health behavior, and health communication variables were included in the analysis.

Signal detection methodology is used with recursive partitioning (classification trees), an empirically driven iterative nonparametric process, to produce a series of “and/or” (Boolean) rules on the predictor variables that identify subgroups of individuals who are more or less likely to have a particular outcome according to a selected criterion.^{15, 16} The subgroups identified are mutually exclusive and maximally discriminated from each other with respect to a dichotomous outcome. SDM is ideal for hypothesis-generating activities and when higher order interactions among variables are possible. The partitioning (i.e., cutpoint) process was set up in this study to identify sets of predictors that would optimize both sensitivity and specificity in predicting those who were inactive. After selecting the cut point for the first optimally efficient variable (Step 1), the signal detection program repeats the partitioning process for each “branch” or subgroup, using all initial predictor variables (Step 2 to Step n). The stopping rules for the partitioning process included: (1) no evaluation with fewer than 100 participants in a subgroup and (2) p<0.001 significance level at each step; these rules are consistent with prior studies that have employed SDM on large nationally representative samples.^17–19 Unweighted data were used in this exploratory analysis because SDM cannot incorporate sampling weights that adjust for complex sampling designs.

An advantage of SDM over more traditional parametric methods (e.g., regression) is that missing data are considered for each variable at each cutpoint (rather than requiring list-wise deletion), allowing for the efficient use of all available data for each variable being evaluated. Furthermore, SDM does not assume linearity or normality and uses a decision tree approach to analyze each variable independently. Thus, multicollinearity among the “predictor” variables is not a concern.²⁰ Detailed comparison of SDM versus traditional methods have been discussed by others.^{15, 21}

Using split sample methodology,^{21, 22} a Cochran-Mantel-Haenzel test was conducted comparing the levels of inactivity between the first and second samples stratified by subgroup. SUDAAN, version 9.0.1 (Research Triangle Institute, Research Triangle Park NC, 2005) was used in these analyses to adjust for the complex sampling design (i.e., sampling weights adjust for unequal sampling probabilities). The mutually exclusive subgroups were created in the second sample (i.e., confirmatory sample) based on the critical variable cut points found using SDM in the first sample. Support for split-sample validation would occur if no significant differences in levels of inactivity were noted between Samples 1 and 2 for the paired subgroups identified.

If Samples 1 and 2 were similar in levels of inactivity, descriptive analyses were planned with the full sample to determine whether additional factors were uniquely associated with the various subgroups. These analyses were exploratory in nature and provided a more complete picture of the characteristics of each subgroup. ANOVA procedures with pair-wise comparisons using SUDAAN were used in these additional analyses.

RESULTS

Descriptive characteristics of the full sample (unweighted data) are displayed in Table 1. Overall, about 28% of respondents were sedentary. Bivariate analysis revealed that inactive respondents were older, less educated and of lower income strata; more likely to be women, not married, nonwhite, overweight, and have smoked 100+ cigarettes in their lifetimes; and less likely to live in metropolitan areas, be in poorer general health, have health insurance, believe that exercise lowers cancer risk, and eat 5+ servings/day of fruits and vegetables compared with more active respondents. Regarding communication channels, inactive respondents read newspapers and magazines less, watched TV more, were less likely to have accessed the Internet, and were less likely to attend to health messages on all media channels compared with more active respondents.

Table 1.

Descriptive statistics of full sample (unweighted data)

Variables	Full sample % or mean(SD)	Inactive % or mean(SD)	Active % or mean(SD)	Bivariate significant level
Sedentary/inactive	27.6%	-	-	-
Sociodemographic and health
Age (years)	47.7(17.4)	49.2(17.6)	47.2(17.2)	p < 0.0001
Gender (female)	60.4%	64.2%	58.9%	p < 0.0001
Married	52.8%	49.2%	54.2%	p < 0.001
Race/ethnicity (white)	67.1%	64.1%	71.5%	p < 0.0001
Educationa	2.8(1.0)	2.4(1.0)	2.9(1.0)	p <0.0001
Employed	59.3%	57%	60%	n.s.
Incomeb	5.5(2.1)	5.0(2.2)	5.8(2.1)	p < 0.0001
County in metro area	81.2%	77.5%	82.7%	p < 0.0001
General healthc	2.7(1.1)	3.1(1.1)	2.5(1.0)	p <0.0001
Provider (yes)	67.3%	68.3%	65.9%	n.s.
Health insurance (yes)	87.3%	82.5%	89.2%	p < 0.0001
Ever had cancer (yes)	12.0%	12.0%	12.3%	n.s.
Family member Ever had cancer (yes)	62.9%	61.3%	63.8%	n.s.
Body mass index	26.7(5.7)	27.9(6.5)	26.3(5.2)	p <0.0001
Health behaviors
Believes that exercise lowers cancer risk	78.0%	69.0%	81.4%	p < 0.0001
Eat 5+ fruits & vegetables per day	16.6%	10.0%	19.0%	p < 0.0001
Smoked 100+ cigarettes in your life	47.2%	52.2%	45.0%	p <0.0001
Health communications
Access the Internet	62.6%	50.9%	67.5%	p < 0.0001
TV (hrs/day)	3.1(2.5)	3.7(2.9)	3.0(2.3)	p < 0.0001
Cable/satellite TV	80.7%	80.4%	80.9%	n.s.
Radio (hrs/day)	2.4(2.8)	2.4(2.9)	2.4(2.8)	n.s.
Newspaper (days/wk)	3.5(2.9)	3.0(3.0)	3.7(2.9)	p < 0.0001
Magazine (days/wk)	1.8(2.1)	1.4(2.0)	2.0(2.1)	p <0.0001
Attends to health messages on…d
Internet	1.8(1.1)	1.6(1.1)	1.9(1.2)	p < 0.0001
TV	3.0(0.9)	2.9(1.0)	3.0(0.9)	p < 0.0001
Radio	2.3(1.1)	2.1(1.1)	2.4(1.1)	p < 0.0001
Newspapers	2.6(1.1)	2.3(1.2)	2.7(1.1)	p <0.0001
Magazine	2.6(1.1)	2.3(1.2)	2.6(1.1)	p < 0.0001

^a2 = high school graduate, 3 = some college, 4 = college graduate

^b4 = “$20,000 to <$25,000”, 5 = “$25,000 to <$35,000”, 6 = “$35,000 to <$50,000, 7 = “$50,000 to <$75,000”

^c2 = very good, 3 = good, 4 = fair

^d1 = not at all, 2 = a little, 3 = some, 4 = a lot

n.s., not significant

As displayed in Figure 1, the overall sedentary proportion was 27.4% in the exploratory sample (Sample 1). Signal detection analyses identified 8 distinct subgroups. While descriptive names are provided for each subgroup, the complex sets of characteristics should be used, rather than the names supplied, to understand and interpret the subgroups. The highest sedentary subgroups are described, followed by intermediate and low inactivity subgroups. Subgroup 5 (Illness-Burdened Couch Potatoes) had the highest level of inactivity (50.0%) and consisted of individuals with at least some college education, who were in fair to poor general health, and watched 4+ hours of TV on a typical weekday. Subgroup 8 (Limited Media Users) had the second highest levels of inactivity (47.8%) and consisted of individuals who did not attend college and did not read the newspaper. Individuals in Subgroup 7 (Obese News Readers) had the third highest levels of inactivity (43.7%) and were composed of those who did not attend college, who read the newspaper at least one day a week and had BMIs ≥30.0. Subgroup 6 (Non-Obese News Readers; inactive=29.3%) had characteristics similar to Subgroup 7, but had BMIs<30.0. Those in subgroups 2–4 had intermediate levels of inactivity (25.0% to 27.8%). Subgroup 2 (Overweight Believers; inactive=25.0%) was composed of college educated individuals, who were in good to excellent general health, believed that exercise can lower cancer risk, and had BMIs ≥29. Subgroup 3 (Healthy Non-Believers; inactive=27.8%) consisted of college-educated individuals, who were in good to excellent general health, and did not believe that exercise can lower cancer risk. Subgroup 4 (Illness-Burdened TV Restricters; inactive=26.2%) consisted of college-educated individuals, who were in fair to poor general health, and watched less than 4 hours of TV on a typical weekday. Subgroup 1 (Non-Obese Believers) had the lowest inactivity level (12.9%) and was composed of college educated individuals, who were in good to excellent general health, believed that exercise can lower cancer risk, and had BMIs <29.

Figure 1.

Signal detection analysis for inactivity, Sample 1

CA, cancer

The levels of inactivity (weighted data) between respective subgroups in Sample 1 versus 2 are displayed in Figure 2. Concerning the stability of subgroup inactivity across the two random samples, some shrinkage in the levels of inactivity was noted between Samples 1 and 2 (as one would expect in split-sample validation analyses). However, results indicated no significant differences in physical inactivity levels between the respective subgroups in Samples 1 and 2 (CMH χ²=0.35; p =0.55).

Figure 2.

Stability of inactivity rates by subgroup, Samples 1 vs. 2

Cochran-Mantel-Haenszel Test =0.35; p =0.55

Descriptive analyses results, examining additional unique characteristics among various subgroups, are displayed in Table 2. In addition to the characteristic noted using SDM, the Limited Media Users (Subgroup 8) were more likely to be nonwhite and of lower income strata; not have a regular healthcare provider nor health insurance; and less likely to have a family history of cancer, be married, have accessed the Internet, read magazines, or attended to health information on various communication channels compared with the other subgroups. Further analyses revealed that Limited Media Users were more likely to be Hispanic/Latino (29%) compared with the other subgroups (5%–14% Hispanic/Latino). The Obese News Readers (Subgroup 7) were less likely to live in metropolitan areas and had the second lowest income level of all the other subgroups. Further analyses revealed that the Obese News Readers were more likely to live in suburban areas (27%) compared with the levels of other subgroups (11%–21%). The Non-Obese Believers (Subgroup 1) were more likely to be white and younger compared with the other subgroups. The Illness-Burdened Couch Potatoes (Subgroup 5) were less likely to be employed, but attended more to health messages on TV compared with the other subgroups. The Overweight Believers (Subgroup 2) had significantly more men compared to the other subgroups, with the exception of the Healthy Non-Believers subgroup.

Table 2.

HINTS subgroup characteristics of identified subgroups – full sample with weighted data

	Subgroup 1	Subgroup 2	Subgroup 3	Subgroup 4	Subgroup 5	Subgroup 6	Subgroup 7	Subgroup 8
	Non-obese believers	Overweight believers	“Healthy” non-believers	Illness-burdened TV restricters	Illness burdened couch potatoes	Non-obese news readers	Obese news readers	Limited media users
	(n = 1798)	(n = 455)	(n = 319)	(n = 319)	(n = 218)	(n = 1245)	(n = 475)	(n = 807)
	Educ ≥ some college,	Educ ≥ Some College,	Educ ≥ Some College,	Educ ≥ Some College,	Educ ≥ Some College,	Educ < Some College,	Educ < Some College,	Educ < Some College,
	Health ≥ Good, Believer = Y,	Health ≥ Good, Believer = Y,	Health ≥ Good, Believer = N	Health < Good,	Health < Good,	Newspaper ≥ 1 day/wk,	Newspaper ≥ 1 day/wk,	Newspaper < 1 day/wk
	BMI < 29	BMI ≥ 29	Inactive = 25%	TV < 4 hrs/day	TV ≥ 4 hrs/day	BMI < 30.0	BMI ≥ 30.0	Inactive = 46%
	Inactive = 14%	Inactive = 27%		Inactive = 25%	Inactive = 43%	Inactive = 29%	Inactive = 40%
Sociodemographic and health
Age (years)	41.5	44.1	46.0	45.4	47.6	48.6	48.9	43.1
Gender (female)	53%	42%	47%	57%	62%	52%	52%	51%
Married	65%	67%	63%	63%	56%	57%	62%	48%
Race/ethnicity (white)	82%	77%	75%	69%	65%	73%	67%	53%
Educationa	3.5	3.4	3.5	3.4	3.3	1.7	1.7	1.5
Employed (yes)	68%	72%	66%	63%	40%	53%	55%	56%
Incomeb	6.6	6.5	6.4	5.8	5.2	5.0	4.7	4.2
County in metro area	87%	84%	88%	81%	78%	76%	67%	80%
General healthc	2.1	2.5	2.2	4.2	4.2	2.7	3.2	3.1
Provider (yes)	67%	74%	67%	74%	75%	62%	68%	47%
Health insurance (yes)	92%	92%	92%	90%	87%	84%	82%	70%
Ever had cancer (yes)	9%	7%	10%	15%	19%	14%	12%	9%
Family member ever had cancer (yes)	63%	67%	67%	67%	70%	61%	64%	54%
BMI	23.8	33.1	26.1	27.9	30.2	24.7	34.8	27.1
Health behaviors
Believes that exercise lowers cancer risk	100%	100%	0%	80%	77%	72%	67%	72%
Eat 5+ fruits & vegetables per day	20%	17%	14%	14%	13%	14%	11%	13%
Smoked 100+ cigarettes in your life	39%	41%	49%	53%	62%	53%	52%	54%
Health communications
Access the Internet	88%	85%	80%	78%	67%	46%	43%	35%
TV (hrs/day)	2.4	3.0	2.8	2.0	5.8	3.5	3.9	3.7
Cable/satellite TV	85%	88%	87%	84%	82%	81%	84%	74%
Radio (hrs/day)	2.2	2.6	2.4	2.3	2.5	2.6	2.8	2.6
Newspaper (days/wk)	3.7	3.8	4.2	3.7	3.7	4.4	4.4	0
Magazine (days/wk)	2.1	2.1	2.1	2.1	2.0	1.7	1.6	0.8
Attend Internetd	2.2	2.0	2.0	2.2	2.1	1.5	1.5	1.4
Attend TVd	3.0	3.1	2.9	3.0	3.2	2.9	3.1	2.8
Attend radiod	2.4	2.4	2.3	2.3	2.2	2.2	2.2	2.1
Attend newspaperd	2.8	2.7	2.6	2.8	2.7	2.7	2.8	1.6
Attend magazined	2.5	2.8	2.5	2.6	2.7	2.4	2.4	1.9

^a1 = Less than HS, 2 = High School Graduate, 3 = Some College, 4 = College Graduate

^b4 = “$20,000 to <$25,000”, 5 = “$25,000 to <$35,000”, 6 = “$35,000 to <$50,000, 7 = “$50,000 to <$75,000”

^c2 = very good, 3 = good, 4 = fair

^d1 = not at all, 2 = a little, 3 = some, 4 = a lot

DISCUSSION

Approximately one quarter of the U.S. adult population remains sedentary. This study was conducted to identify empirically distinct subgroups with varying sedentary levels. Overall, results suggest that subgroups with similar levels of inactivity can have quite different defining characteristics. Furthermore, the inactivity patterns found using SDM were found to stable in an independent sample.

Prior studies on inactivity have typically not considered the complex interactions among predictive factors and, therefore, may have missed important subgroups. For example, other studies using correlational and regression techniques with nationally representative samples have found lower educational attainment to be associated with greater inactivity.^8,23 This study had similar findings in bivariate analyses and when evaluating the first partitioning; specifically, those with higher educational attainment had significantly lower inactivity levels (20.2%) compared with lower educational attainment (37.5%). Yet, substantive differences in sedentary levels existed between subgroups within education level. The Illness-Burdened Couch Potatoes (a high-inactivity subgroup) was composed of those with higher (not lower) educational attainment in conjunction with other health and behavioral factors (i.e., in fair to poor general health and greater TV watching). This important sedentary subgroup would likely have been overlooked if the complex interactions were not examined.

Results may facilitate the development of interventions targeted and tailored specifically for the subgroups identified.^15,24 Study results also highlight the role of communication variables as key factors that define inactive subgroups. As noted earlier, watching TV 4+ hours/day was one of the significant characteristics defining a high sedentary subgroup, the Illness-Burdened Couch Potatoes. A possible intervention program targeting this subgroup might focus on substituting TV with more active behaviors,^25–28 taking into consideration the health status of individuals in this subgroup. Given the fair to poor general health status of individuals in this subgroup, the content of the health message may need to emphasize activities aimed as secondary prevention (e.g., improve strength function, activities to manage disability), rather than primary health promotion. Another high-inactive subgroup, the Limited Media Users, was less-educated, did not read the newspaper, and had the lowest scores on most communication variables. This subgroup was also more likely to be Hispanic/Latino, not have a regular healthcare provider or health insurance, and have lower incomes. This may be a difficult subgroup to reach for health promotion purposes, particularly since they attend less to health messages through most media channels. However, this subgroup did complete the HINTS telephone interview and may be amenable to culturally sensitive, health promotion counseling by phone or potentially through community settings that they regularly visit (e.g., community centers, churches). Prior research has found telephone counseling aimed at increasing physical activity among adults, including Latino adults, to be effective.^29–31 As another example, newspaper ads or tailored print may be effective in recruiting those in the Obese News Reader subgroup (particularly in suburban areas) to a health promotion program focused on increasing activity and weight loss, although these materials should be written for an audience with a lower education level. This tailored information could then be incorporated within a theoretically based, behavioral intervention.²⁴

Some limitations of this study include the cross-sectional study design, which limits the ability to establish causality. Furthermore, HINTS 2003 did oversample ethnic/minority groups such as African Americans and Hispanics, but there was limited information on other groups (e.g., Asian Americans, American Indians); hence, results cannot be generalized to these other underrepresented groups. Another limitation concerns the one item measure used to assess inactivity. Although taken from another national survey and cognitively tested prior to being used, this measure may not accurately capture the levels of inactivity of the subgroups identified. However, given that reporting bias in the physical activity arena tends to involve underreporting of inactivity (and overreporting of activity), it is likely that those individuals identifying themselves as inactive were truly sedentary. Moreover, respondents provided self-reports on all the behaviors measured, including BMI and media use (e.g., television, radio, Internet), which could have resulted in overreported or underreported this information (due to poor recall, social desirability, or other reporting biases). Addressing these potential self-reporting biases may require objective measures or measurement triangulation in future research. Although this study focuses on various media communication channels used by respondents, little information is available concerning the specific information/message content that would be most useful or appropriate for decreasing sedentary behavior among the subgroups. Tailoring programs on both the communication channel and message content will likely be needed to create significant and sustained behavior change.

Overall, this study’s inactivity proportion (i.e., 27.6%) was comparable to the national averages. This study adds to the literature in that it is the first to validate the results of a SDM using a nationally representative sample. Researchers have argued that a “one-size-fits-all” approach to health promotion may not be as effective as programs that target specific groups and/or tailor according to individual characteristics,³² and the current results represent a first step in attempting to identify relevant subgroups for subsequent intervention targeting and development.