Skip to main content
PMC home page
Springer logoLink to Springer
. 2011 Oct 18;13(2):183–195. doi: 10.1007/s11121-011-0255-0

Clustering of Unhealthy Behaviors in the Aerobics Center Longitudinal Study

Mariane Héroux 1, Ian Janssen 1,✉, Duck-chul Lee 2, Xuemei Sui 2, James R Hebert 3, Steven N Blair 2,3
PMCID: PMC3304050  PMID: 22006293

Abstract

Background

Clustering of unhealthy behaviors has been reported in previous studies; however the link with all-cause mortality and differences between those with and without chronic disease requires further investigation.

Objectives

To observe the clustering effects of unhealthy diet, fitness, smoking, and excessive alcohol consumption in adults with and without chronic disease and to assess all-cause mortality risk according to the clustering of unhealthy behaviors.

Methods

Participants were 13,621 adults (aged 20–84) from the Aerobics Center Longitudinal Study. Four health behaviors were observed (diet, fitness, smoking, and drinking). Baseline characteristics of the study population and bivariate relations between pairs of the health behaviors were evaluated separately for those with and without chronic disease using cross-tabulation and a chi-square test. The odds of partaking in unhealthy behaviors were also calculated. Latent class analysis (LCA) was used to assess clustering. Cox regression was used to assess the relationship between the behaviors and mortality.

Results

The four health behaviors were related to each other. LCA results suggested that two classes existed. Participants in class 1 had a higher probability of partaking in each of the four unhealthy behaviors than participants in class 2. No differences in health behavior clustering were found between participants with and without chronic disease. Mortality risk increased relative to the number of unhealthy behaviors participants engaged in.

Conclusion

Unhealthy behaviors cluster together irrespective of chronic disease status. Such findings suggest that multi-behavioral intervention strategies can be similar in those with and without chronic disease.

Keywords: Health behaviors, Clustering, Chronic disease, All-cause mortality

Introduction

There is much uncertainty and debate about the proportion of mortality in the United States and other developed countries attributable to tobacco smoking, poor diet, physical inactivity, and excessive alcohol consumption; however, there is agreement that these four unhealthy behaviors account for the vast majority of preventable causes of death (Adler and Rehkopf 2008; Lopez et al. 2006; Mathers and Loncar 2006; World Health Organization 2009). There also is evidence that these four unhealthy behaviors often occur simultaneously within an individual (Chiolero et al. 2006; Fine et al. 2004; Kvaavik et al. 2010; Laaksonen et al. 2001; Poortinga 2007; Schuit et al. 2002). For example, a recent study reported that over 70% of adults with one of these four unhealthy behaviors engaged in at least one of the remaining three (Poortinga 2007). Despite this evidence, the majority of public health and clinical interventions focus only on changing a single unhealthy behavior. A more efficient approach may be to focus on interventions that target the several unhealthy behaviors simultaneously. Indeed, based on previous research examining the clustering of several health behaviors, the National Cancer Institute-funded Behavior Change Consortium was predicated on focusing on two or more health-related behaviors (Hebert et al. 2008). Identifying which unhealthy behaviors typically occur simultaneously within the same individual can help inform these types of interventions.

Several studies have examined which unhealthy behaviors are most likely to occur simultaneously or cluster within individuals (Berrigan et al. 2003; Blakely et al. 2004; Chiolero et al. 2006; Donovan et al. 1993; Fine et al. 2004; Ingledew et al. 1995; Laaksonen et al. 2001; Laaksonen et al. 2002; Laska et al. 2009; Metzger et al. 2008; Pate et al. 1996; Poortinga 2007; Schuit et al. 2002). These studies have used a variety of statistical techniques to assess clustering, and this makes it difficult to compare their findings. Although the results are somewhat varied, the strongest associations between unhealthy behaviors appear to be between smoking and the three remaining behaviors (inactivity, diet, alcohol) (Laaksonen et al. 2002; Poortinga 2007). Also, there is evidence suggesting that although the prevalence of unhealthy behaviors varies with gender, race, and age the clustering of these behaviors is similar irrespective of these demographic characteristics (Berrigan et al. 2003; Laaksonen et al. 2001; Poortinga 2007). For example in a study done by Berrigan et al. (2003), non-Hispanic white, non-Hispanic black, and Mexican American participants displayed similar patterns of clustering among unhealthy behaviors (Berrigan et al. 2003). Health behaviors may improve after diagnosis of a chronic disease (Patterson et al. 2003; Pinto et al. 2000; Wallstrom et al. 2005); however, studies have yet to examine the influence of chronic disease status within the context of unhealthy behavior clustering.

To date, most of the studies that have examined the clustering of unhealthy behaviors have used crude and imprecise methods of measuring these health behaviors, particularly diet (Byun et al. 2010; Mitchell et al. 2010) and physical activity (Blair et al. 1985). For the most part, only one or two dietary variables have been obtained through simple questionnaires (e.g., in the Behavior Change Consortium frequency of fruit and vegetable intake) were used as a marker of dietary intake (Hebert et al. 2008; Yaroch et al. 2008). Similarly, self-reported measures from questionnaires were used to assess physical activity (Blakely et al. 2004; Chiolero et al. 2006; Poortinga 2007; Schuit et al. 2002). Examining dietary patterns obtained through comprehensive diet records and using objective measures of physical activity may yield more accurate measures of these behaviors for cluster analysis (Aadahl et al. 2007; Day et al. 2001; Tokudome et al. 2005).

This study aimed to address some of the gaps and limitations in the previous literature examining the clustering of unhealthy behaviors. Our objectives were to: (1) observe the clustering effects of unhealthy diet, physical inactivity, smoking, and excessive alcohol consumption within adults; (2) use a comprehensive dietary assessment and cardiorespiratory fitness exam to assess how a poor diet and physical inactivity relate to each other and other unhealthy behaviors; (3) determine whether the clustering of unhealthy behaviors differs in those with and without chronic disease; and (4) assess all-cause mortality risk according to the individual unhealthy behaviors and a clustering of unhealthy behaviors.

Methods

Participants

The study sample consisted of a subset of 13,621 participants from the Aerobics Center Longitudinal Study (ACLS) who had 3-day diet record data available between 1987 and 1999. The ACLS is a prospective, observational study of men and women who participated in a preventive medical examination at the Cooper Clinic in Dallas, TX, USA. ACLS participants are predominantly (>95%) non-Hispanic white, well-educated (>75% college graduates), and employed in professional or executive occupations (Blair et al. 1989a; Wei et al. 1999). Individuals in the analyses reported here were similar to the overall ACLS cohort.

The standardized data collection procedures employed in the ACLS are described in detail in previous reports (Blair et al. 1989b; Lee et al. 1999). Briefly, data collection for all of the variables included in this study, with the exception of the all-cause mortality outcome, occurred at baseline. These data included a medical history questionnaire, a thorough physical examination, fasting blood chemistry analyses, symptom-limited maximal exercise test for the measurement of fitness, and a dietary assessment. The medical history questionnaire was self-administered and included information regarding demographics, family history and personal health habits. Physical examination procedures were conducted in a standardized manner by trained personnel following a manual of procedures. All participants gave written informed consent. The study was reviewed and approved annually by the Institutional Review Board at the Cooper Institute (Blair et al. 1989b, 1996; Finley et al. 2006).

Health Behaviors

While we recognize that the four health behaviors examined fall along a continuum, for the purpose of this present study they were all classified as dichotomous variables (healthy vs. unhealthy behavior) in order to simplify the presentation of the results, and to assist in translating the findings into meaningful public health messages (Rothman 2002).

Unhealthy Diet

The dietary assessment consisted of a 3-day diet record that required respondents to keep detailed records of everything they ate over 2 pre-assigned weekdays and 1 weekend day. Participants were provided with written instructions on how to accurately describe foods and estimate portion sizes. Participants kept an ongoing, real-time written record of foods consumed during and between meals, including assessing portion sizes using common household measures. Registered dieticians at the Cooper Clinic coded and analyzed the diet records using the Cooper Clinic Nutrition and Exercise Evaluation system (Finley et al. 2006). This provided detailed dietary information on the overall diet such as the number of foods consumed from specific food groups, total caloric content, the volume of micronutrients (vitamins and minerals) and the volume and caloric content of macronutrients (fat, protein, carbohydrate and alcohol).

In a previous study we developed an Unhealthy Eating Index in the ACLS cohort based on the 3-day diet records (Héroux et al. 2010). The Unhealthy Eating Index was generated using reduced rank regression, and is characterized by a dietary pattern containing foods that predict clinical risk factors for chronic disease (HDL-cholesterol, total cholesterol, blood pressure, triglycerides, white blood cell count, uric acid, body mass index, and fasting glucose). Participants with high scores on the Unhealthy Eating Index were those who had diets high in added fats, non-whole grains, processed meat, red meat, white potato products, and low in non-citrus fruits. For this report, the Unhealthy Eating Index was used as a composite measure of the dietary behavior of the participants. Based on previous analyses examining the relationship between the Unhealthy Eating Index and mortality risk (Héroux et al. 2010), Unhealthy Eating Index scores were first divided into quintiles and were then grouped into two categories: “unhealthy diet” (quintile 5) and “healthy diet” (quintiles 1–4). Groups were divided in this fashion as there was little difference in risk between the 1st and 4th quintiles but a significant difference in risk between the 4th and 5th quintiles (Héroux et al. 2010).

Physically Inactive and Unfit

Physical activity was measured through a self-report questionnaire, which assessed physical activity patterns over the previous 3 months. If the participant reported partaking in an activity, (s)he was asked to provide additional information about the frequency and distance or time spent in the activity. This information was used to divide participants into three mutually exclusive categories: (i) “physically inactive” (no activity); (ii) moderately active (sporting or leisure-time physical activity other than walking or jogging; or walk or jog up to 16 km/week); and (iii) highly active (walk or jog >16 km/week) (Kampert et al. 1996). For the purposes of this study, those in the moderately and highly active groups were combined into a single “physically active” group.

Cardiorespiratory fitness (hereafter referred to as ‘fitness’) was used as an additional marker of physical activity in this study. In fact, because of the measurement errors (e.g., recall and social desirability bias) associated with self-reported physical activity measures (Adams et al. 2005; Sallis and Saelens 2000), the primary analyses were based on measures of fitness rather than physical activity. Although fitness has a genetic component (Bouchard et al. 1997), to a large extent, it is a reflection of an individual’s physical activity habits over recent months and weeks (Aadahl et al. 2007; Finley et al. 2006). Fitness was assessed by a symptom-limited maximal exercise treadmill test using a modified Balke protocol (Balke and Ware 1959). Participants began walking on the treadmill with no elevation at 5.3 km/h. At the end of the first minute the elevation was increased to 2%. Thereafter, the elevation was increased 1% per minute until the 25th minute. For the few participants who were able to continue for >25 min, elevation remained constant and speed was increased each minute by 0.32 km/h. Participants continued the test to the limits of volitional fatigue. Time to completion on the treadmill was used to estimate VO2max in metabolic equivalents (METs) (one MET = 3.5 ml O2 · kg−1 · min−1) from the Balke protocol using the following formula: VO2max = [1.44 × (minutes on treadmill) + 14.99]/3.5 (Katzmarzyk et al. 2005; Pollock et al. 1976). Participants were classified as ‘unfit’ or ‘fit’ based on the lowest 20% and upper 80%, respectively, of the age- and sex-specific distribution of METs in the overall ACLS population. These cut-points are from previous reports on the relation between fitness and all-cause mortality in the ACLS, which indicate that there is little difference in risk across the 2nd through 5th fitness quintiles, but a big difference in risk between the 1st and 2nd fitness quintiles (Blair et al. 1996). Thus, in addition to those who were in the top 20% (quintile 5), participants who scored in the middle 60% (quintiles 2 to 4) were classified as fit. Because there is no widely accepted categorization of fitness, and because we wanted to maintain consistency in our methods with other ACLS studies, we continue to use this approach.

Smoking

The medical history questionnaire asked participants if they currently smoked cigarettes. If they responded yes to this question they were categorized as current smokers. All others were categorized as non-smokers.

Heavy Drinking

Alcohol intake was captured in the medical history questionnaire. Participants were asked how many drinks per week they consumed of beer, wine, and/or hard liquor. One unit of alcohol was defined as 12 oz (3.41 dl) of beer, 5 oz (1.42 dl) of wine, or 1.5 oz (0.43 dl) of liquor. Participants were divided into two categories based on the sex-specific weekly upper limits as recommended by the National Institute on Alcohol Abuse and Alcoholism (National Institute on Alcohol Abuse and Alcoholism 1995). “Heavy drinkers” were those women who consumed ≥8 drinks and those men who consumed ≥15 drinks within the past week. All other participants were considered “non/light drinkers”. Of the heavy drinkers in the present study, males drank an average of 23 drinks per week and females drank an average of 15 drinks per week. There were no differences in the descriptive characteristics or mortality risk estimates in the non-drinkers and light drinkers in the current study (data not shown).

Chronic Disease Status

Participants were divided into two groups: (1) those with no history of chronic diseases, and (2) those who reported having one or more of the following conditions: diabetes, cardiovascular disease (stroke or myocardial infarction), hypertension, and cancer. These chronic diseases were chosen because it is well established that diagnosis of one or more of these diseases can result in changes in lifestyle behaviors including the four health behaviors examined in this study (Neutel and Campbell 2008; World Health Organization 2005). Disease status was obtained at baseline through a standardized questionnaire that inquired specifically about past and current diabetes, stroke, myocardial infarction, hypertension, and cancer.

Mortality Follow-Up

After their baseline exam, participants were followed until they died, or until December 31, 2003, whichever came first. Possible decedents were identified from the National Death Index and verified with official death certificates from their home states. The National Death Index has a sensitivity of 96% and specificity of 100% in cohort studies (Centers for Disease Control and Prevention 2003).

Statistical Analyses

All analyses were performed using SAS version 9.1 (Release 9.1 Cary, NC: SAS Institute Inc.). Baseline characteristics of the study population and bivariate relations between pairs of the health behaviors were evaluated separately for those with and without chronic disease using cross-tabulation and a chi-square test. The odds of partaking in unhealthy behaviors were also calculated.

Traditional cluster and factor analyses were designed for continuous variables. Because the four unhealthy behaviors considered here are dichotomous variables, latent class analysis (LCA) was used to assess clustering as LCA is suitable for dichotomous variables (Lanza et al. 2007; Vermunt and Magidson 2002, 2005). LCA is a statistical method for finding subtypes of related cases (latent classes) from multivariate categorical data. LCA is used in a manner analogous to cluster and factor analysis (Vermunt and Magidson 2002, 2005). That is, given a sample of participants measured on several variables, one can determine if there are a small number of groups into which the participants fall (Laska et al. 2009; Vermunt and Magidson 2002, 2005). The PROC LCA command procedure was used to estimate model parameters (Lanza et al. 2007). A simple two-class model consisting of only the four health behaviors was first assessed. In order to select the appropriate number of classes and maximize model fit, the two-class model was compared with successively fit models which specified an increasing number of latent classes (up to three classes as the 4-class model would not converge). In selecting the final model (e.g., 2- or 3-class model) the BIC and AIC criteria were examined across models as well as model interpretability (Lanza et al. 2007). Based on these criteria, a 2-class model was selected. Once this model was selected, the grouping option was used within the LCA analysis to determine whether or not the probability of class membership differed in those with no history of chronic disease and those with current/former chronic disease. By comparing the G2 difference of the basic model and the grouped model to a chi-square distribution with the degrees of freedom equal to the difference in degrees of freedom between the models, the analysis indicated that measurement invariance should not be rejected (p > 0.05) and therefore that class membership was not significantly different for the two groups. A 2-class LCA model was then run including the covariates option to control for age and gender. Age and gender were not stratified in the model as exploratory analyses suggested that class membership was similar in all age groups and in men and women (data not shown). The Outpram option was also used to output the probability of class membership for each participant. Participants were then divided into class 1 or 2 based on their highest probability score (Goodman 2007).

Cox proportional hazard regression analyses were used to predict the hazard ratio (HR) of mortality over the follow-up period according to engagement in unhealthy behaviors. Age, gender, year of examination, and health status were included as covariates in the regression models. Based on the proportion of the cohort engaged in each unhealthy behavior, and the HR of mortality associated with each unhealthy behavior, the population attributable risk percentage was calculated for each health behavior using the following formula: PAR% (Cole and MacMahon 1971; Walter 1978) = [Pe (HR-1)]/[Pe (HR-1) + 1]*100 where Pe is the prevalence of the health behavior within the population, and HR is the relative risk of all-cause mortality for that unhealthy behavior (using model 2). The population attributable risk is a calculation of the percentage of all-cause mortality (that would occur in a population with similar attributes to the cohort) attributable to each of the unhealthy behaviors.

Results

Participant Characteristics

Baseline characteristics of participants are presented in Table 1. Of the total sample, 4,630 (34%) had at least one of the chronic diseases. The majority (65.6%) of participants were between the ages of 40–59 years and 52.6% were overweight or obese. Approximately 8.9% of the cohort were unfit, 10.9% were smokers, and 3.1% were heavy drinkers. Over half (66.8%) of the cohort engaged in none of the four unhealthy behaviors. The prevalence and distribution of the four unhealthy behaviors was not different in those with and without chronic disease (Table 1).

Table 1.

Aerobics Center Longitudinal Study participant baseline characteristics

No history of chronic disease Current or former chronic disease
N = 8,991 N = 4,630
Age
20–39 y 24.6% 12.9%
40–59 y 65.6% 67.3%
≥60 y 9.8% 19.8%
Females 28.1% 17.1%
Body Mass Index (kg/m2)
Normal weight (<25) 54.2% 34.2%
Overweight (25–30) 37.3% 44.2%
Obese (≥30) 8.5% 21.6%
Diet
Healthy diet 81.3% 77.4%
Unhealthy diet 18.7% 22.6%
Fitness
Fit 93.7% 85.8%
Unfit 6.3% 14.2%
Physical Activity
Physically active 75.3% 71.8%
Inactive 24.7% 28.2%
Smoking Status
Non-smokers 89.1% 89.2%
Smokers 10.9% 10.8%
Alcohol Intake
Non/light drinkers 97.0% 96.8%
Heavy drinkers 3.0% 3.2%
# Unhealthy Behaviors
0 69.2% 62.2%
1 24.1% 27.0%
2 5.7% 9.0%
3 1.0% 1.7%
4 0.1% 0.1%
Open in a new tab

Relationship Between Health Behaviors

Table 2 shows the relationship between the health behaviors according to chronic disease status. The patterns of relationships between health behaviors were similar for the groups with and without chronic disease. Independent of chronic disease status, all four health behaviors were significantly (p < 0.05) related to each other, with the exception of fitness vs. alcohol, and within those with chronic disease, diet vs. alcohol. Within the participants with no history of chronic disease, those with an unhealthy diet were 2.45 (95% confidence interval: 2.10–2.98) times more likely to be unfit, 2.02 (1.74–2.35) times more likely to be a smoker, and 1.61 (1.21–2.15) times more likely to be a heavy drinker by comparison to those with a healthy diet. The strongest relationships were observed between the smoking and alcohol intake behaviors; smokers were 2.69 (2.00–3.62) times more likely to be heavy drinkers in comparison to non-smokers. Gender differences (p < 0.05) were present for all behaviors, with the exception of cardiorespiratory fitness in men and women without chronic disease. For this reason, all further analyses were controlled for gender.

Table 2.

Relationship between health behaviors

Prevalence or Odds Ratio (95% CI)a for Engagement in Unhealthy Behavior
Unhealthy Diet Unfit Smokers Heavy Drinkers
Participants with no history of chronic disease
Diet
Healthy diet,% – 5.2% 8.9% 2.3%
Unhealthy diet,% 11.3% 19.4% 4.5%
Odds ratio (95% CI) 2.45 (2.10–2.98) 2.02 (1.74–2.35) 1.61 (1.21–2.15)
Fitness
Fit,% 17.7% – 10.1% 2.6%
Unfit,% 33.4% 22.4% 3.3%
Odds ratio (95% CI) 2.47 (2.03–3.00) 2.53 (2.05–3.13) 1.22 (0.76–1.97)
Smoking
Non-smokers,% 16.9% 5.5% – 2.2%
Smokers,% 33.2% 13.0% 6.54
Odds ratio (95% CI) 1.99 (1.71–2.32) 2.53 (2.04–3.13) 2.69 (2.00–3.62)
Alcohol Intake
Non/Light drinkers,% 18.3% 6.3% 10.5% –
Heavy drinkers,% 31.4% 7.8% 26.4%
Odds ratio (95% CI) 1.62 (1.22–2.17) 1.23 (0.76–1.98) 2.72 (2.02–3.66)
Participants with chronic disease
Diet
Healthy diet,% – 10.8% 8.9% 2.3%
Unhealthy diet,% 25.8% 17.3% 4.3%
Odds ratio (95% CI) 2.50 (2.07–3.00) 1.66 (1.36–2.05) 1.26 (0.86–1.84)
Fitness
Fit,% 19.5% – 9.6% 2.7%
Unfit,% 41.2% 17.8% 3.0%
Odds ratio (95% CI) 2.50 (2.07–3.00) 1.79 (1.42–2.25) 0.90 (0.55–1.46)
Smoking
Non-smokers,% 20.9% 13.0% – 2.3%
Smokers,% 36.2% 23.4% 6.8%
Odds ratio (95% CI) 1.65 (1.34–2.03) 1.78 (1.41–2.24) 2.45 (1.63–3.69)
Alcohol Intake
Non/light drinkers,% 22.2% 14.1% 10.3% –
Heavy drinkers,% 34.9% 15.5% 26.4%
Odds ratio (95% CI) 1.32 (0.90–1.94) 0.91 (0.55–1.48) 2.46 (1.63–3.71)
Open in a new tab

All relationships were significant (p ≤ 0.05) with the exception of alcohol intake vs. fitness in those with and without chronic disease and diet vs. alcohol intake in those with chronic disease

aOdds ratio (95% confidence interval) for engagement in the corresponding unhealthy behavior. The healthy behavior was used as the referent category. Odds ratios were controlled for age and gender.

Results from the LCA analysis, which was controlled for age and gender, suggested that two classes were present (Table 3). Because class membership did not differ between those with and without chronic disease, the two groups were merged and results are presented for the whole cohort. The latent class probabilities suggested that 38% of the cohort were in class 1 and 62% were in class 2 (log likelihood of model = −16338). Class 1 membership was characterized by a higher probability of partaking in each of the four unhealthy behaviors, particularly for unhealthy diets (51.2% vs. 1.2%). The number of unhealthy behaviors each participant engaged in was calculated according to class membership (Table 3). A much higher percentage of participants in class 1 engaged in two or more unhealthy behaviors by comparison to participants in class 2 (27.1% vs. 0.2%).

Table 3.

Results of the latent class analysis used to identify classes of participants based on health behaviors, after controlling for age and gender

Unhealthy Behavior Class 1 Class 2
(N = 3,950) (N = 9,671)
Diet
Healthy diet 48.8% 98.8%
Unhealthy diet 51.2% 1.2%
Fitness
Fit 84.6% 94.9%
Unfit 15.4% 5.1%
Smoking
Non-smokers 79.4% 95.0%
Smokers 20.6% 5.0%
Alcohol Intake
Non/light drinkers 93.8% 98.9%
Heavy drinkers 6.2% 1.2%
# Unhealthy Behaviorsa
0 5.9% 91.7%
1 66.7% 8.1%
2 23.0% 0.2%
3 4.1% 0.0%
4 0.2% 0.0%
Open in a new tab

Values represent the probability of class membership

aIncluding an unhealthy diet, low fitness, smoking, and heavy drinking

Relationship Between Health Behaviors and Mortality

During the follow-up period, 445 deaths occurred over the 597,009 person-years of observation. Table 4 shows the risks for all-cause mortality according to engagement in the unhealthy behaviors. The risk estimates for model 1 were controlled for age, gender, year of examination, and history of chronic diseases. The risk estimates for model 2 were controlled for each of the remaining three health behaviors in addition to the aforementioned covariates. After controlling for the other health behaviors, unfit individuals had a 2.16 (95% confidence interval: 1.64–2.85) times higher risk of mortality than fit individuals and smokers had a 1.35 (1.01–1.81) times higher risk of mortality than non-smokers. The risk estimates for mortality were not significantly different between those with a healthy diet and those with an unhealthy diet (p = 0.32) and between non/light drinkers and heavy drinkers (p = 0.52). The population attributable risk calculations, which were a function of the prevalence of the unhealthy behaviors and their independent associations with mortality, suggested that 9.4% of the deaths in the cohort resulted from having low fitness, 2.9% of the deaths resulted from an unhealthy diet, 3.7% from smoking, and 0.6% from heavy drinking. Collectively, these four unhealthy behaviors accounted for 16.6% of the deaths in the cohort.

Table 4.

All-cause mortality risk according to engagement in different unhealthy behaviors

Unhealthy Behaviors HR (95% CI) HR (95% CI) Population Attributable Risk (PAR)
Model 1 Model 2
Diet
Healthy diet 1.0 1.0 2.9%
Unhealthy diet 1.29 (1.00–1.68) 1.15 (0.89–1.50)
Fitness
Fit 1.0 1.0 9.4%
Unfit 2.28 (1.74–3.00) 2.16 (1.64–2.85)
Smoking
Non-smokers 1.0 1.0 3.7%
Smokers 1.50 (1.12–2.01) 1.35 (1.01–1.81)
Alcohol Intake
Non/light drinkers 1.00 1.0 0.6%
Heavy drinkers 1.24 (0.68–2.26) 1.22 (0.67–2.23)
# Unhealthy Behaviors
0 1.0 – –
1 1.46 (1.20–1.83) – –
2 2.37 (1.69–3.34) – –
3 to 4 1.98 (0.81–4.82) – –
Class 1 1.55 (1.22–1.97) – –
Class 2 1.0
Open in a new tab

HR (95% CI) = hazard ratio (95% confidence interval)

Model 1: controlled for age, gender, year of examination, history of chronic disease.

Model 2: controlled for age, gender, year of examination, history of chronic disease, and the remaining 3 health behaviors

PAR% = [Pe (HR-1)]/[Pe (HR-1) + 1]*100 where Pe is the prevalence of the health behavior within the population, and HR is the relative risk of all-cause mortality for that unhealthy behavior (using model 2)

The number of unhealthy behaviors in which individuals partook and health behavior clustering were significant predictors of all-cause mortality (Table 4). Compared to those who partook in no unhealthy behaviors, mortality risk was increased by 1.46 (1.20–1.83) times in those who partook in one unhealthy behavior, by 2.37 (1.69–3.34) times in those who partook in 2 unhealthy behaviors, and by 1.98 (0.81–4.82) times in those who partook in 3 or 4 unhealthy behaviors. There was a small sample size (N = 171) in the group with three or four unhealthy behaviors, which may explain the lack of statistical significance (p = 0.14) for all-cause mortality risk in that group. Finally, by comparison to participants in class 2, participants in class 1 had a hazard ratio of mortality of 1.55 (1.22–1.97).

Findings for Physical Inactivity

All of the analyses shown in Tables 2 to 4 were repeated using self-reported physical activity in place of fitness. The results for physical activity, which are provided in Tables 1 to 3 in the online-only Appendix, were very similar to those reported for fitness in Tables 2 to 4. The major difference in the results presented for physical activity in the online-only appendix versus those presented for fitness were for the hazard ratio for mortality and population attributable risk calculations, which were attenuated for the self-reported physical activity measure by comparison to the objective measure of fitness. Specifically, the hazard ratio for all-cause mortality was 1.20 (0.97–1.50) in the physically inactive participants relative to the physically active participants (Appendix Table 7), and the corresponding population attributable risk calculation for physical inactivity was 4.9% (Appendix Table 7). By comparison, the hazard ratio and population attributable risk estimates for low fitness were 2.16 (1.64–2.85) and 9.4%, respectively (Table 4).

Table 7.

All-cause mortality risk according to engagement in different unhealthy behaviors

Health Behaviors HR (95% CI) HR (95% CI) Population Attributable Risk (PAR)
Model 1 Model 2
Diet
 Healthy diet 1.0 1.0 4.0%
 Unhealthy diet 1.29 (1.00–1.68) 1.21 (0.93–1.58)
Physical Activity
 Physically active 1.0 1.0 4.9%
 Inactive 1.27 (1.02–1.57) 1.20 (0.97–1.50)
Smoking
 Non-smokers 1.0 1.0 4.4%
 Smokers 1.50 (1.12–2.01) 1.42 (1.06–1.90)
Alcohol Intake
 Non/light drinkers 1.0 1.0 0.4%
 Heavy drinkers 1.24 (0.68–2.36) 1.16 (0.64–2.12)
# Unhealthy Behaviors
 0 1.00 – –
 1 1.13 (0.91–1.41) – –
 2 1.67 (1.22–2.28) – –
 3 to 4 2.07 (1.15–3.73) – –
Class 1 1.52 (1.19–1.93) – –
Class 2 1.0
Open in a new tab

HR (95% CI) = hazard ratio (95% confidence interval)

Model 1: controlled for age, gender, year of examination, history of chronic disease.

Model 2: controlled for age, gender, year of examination, history of chronic disease, and the remaining 3 health behaviors

PAR% = [Pe (HR-1)]/[Pe (HR-1) + 1]*100 where Pe is the prevalence of the health behavior within the population, and HR is the relative risk of all-cause mortality for that unhealthy behavior (using model 2)

Discussion

The present study sought to examine the clustering of four unhealthy lifestyle behaviors (unhealthy diet, inactivity/fitness, smoking, and heavy drinking). All four behaviors were related to each other with the exception of fitness and alcohol. The strongest relationship was observed between smoking and alcohol consumption. Interestingly, no differences in health behavior clustering were found between participants with and without chronic disease. As expected, mortality risk increased relative to the number of unhealthy behaviors participants engaged in.

It is difficult to make direct comparisons between the behavioral relationships found in this and other studies as various measurement and analytical approaches have been used. For example, previous studies have included different population groups, used different statistical techniques (e.g., correlation, factor analysis, cluster analysis), and used different tools to measure the health behaviors. What is noteworthy is that, irrespective of these differences, the majority of these studies have concluded that health behaviors conform to a multidimensional model of health behavior patterns (Berrigan et al. 2003; Williams and Wechsler 1972). In fact, several studies have found that the clustering of behaviors is fairly constant across various age and ethnic groups as well as in males and females (Berrigan et al. 2003; Laaksonen et al. 2001; Poortinga 2007). Similar to what was found here, the lack of association between fitness and alcohol has been reported in other studies (Ingledew et al. 1995; Laaksonen et al. 2002; Schuit et al. 2002) as has the strong relationship between drinking and smoking (Laaksonen et al. 2002; Poortinga 2007).

The current study was unique in that it assessed whether or not differences in health behavior clustering patterns existed between those with and without chronic disease. Although previous studies report that the lifestyle behaviors often improve after a chronic disease diagnosis (Patterson et al. 2003; Pinto et al. 2000; Wallstrom et al. 2005), with the exception of fitness, the proportion of those with the unhealthy behavior was similar in the groups with and without chronic disease. Furthermore, the clustering of unhealthy behaviors did not vary according to chronic disease status.

Previous studies examining the relation between the number of unhealthy behaviors and all-cause mortality have consistently reported increasing risks with increasing number of unhealthy behaviors (Byun et al. 2010; Mitchell et al. 2010; Tsubono et al. 2004). For example, in a study done by Tsubono et al. (2004), the relative risks of all-cause mortality when comparing men who partook in 1, 2, 3, or 4 versus 0 unhealthy behaviors were 1.20 (95% confidence interval: 0.83–1.74), 1.66 (1.17–2.35), 1.94 (1.36–2.78), and 3.96 (2.26–6.95), respectively (Tsubono et al. 2004). Similarly, in the present study mortality risk was increased by ~50% in the presence of a single unhealthy behavior and by ~100% in the presence of 2 or more unhealthy behaviors.

Similar to other studies examining the impact of lifestyle factors on mortality (Laaksonen et al. 2001; Laaksonen et al. 2002; Schuit et al. 2002), the pattern of findings in our paper suggest that all-cause mortality risk increases with the number of unhealthy behaviors that a person partakes in. Furthermore, in this cohort 16.6% of deaths were attributed to partaking in the 4 unhealthy behaviors. A study by Chiuve et al. (2006) reported that by leading a healthy lifestyle characterized by not smoking, exercising regularly, eating a healthy diet, consuming moderate alcohol, and maintaining a healthy weight, participants had the potential of avoiding 62% of cardiovascular events (Chiuve et al. 2006). When compared to the present study, the higher attributable risk reported most likely results from the differences in health behaviors assessed, the different outcomes assessed (i.e., cardiovascular events vs. all-cause mortality), and the higher prevalence of the unhealthy behaviors within the study population (i.e., 71% low exercise, 60% poor diet).

The findings of this study are important for several reasons. Firstly, the results support the notion that interventions may be more efficient if they focus on multiple unhealthy behaviors rather than a single behavior. For example, 73.6% of the heavy drinkers in the ACLS were also smokers, suggesting that there is a strong tendency for heavy drinkers to smoke. Thus, interventions aimed at decreasing heavy drinking may benefit from a component aimed at decreasing smoking. Secondly, our results suggest that those with and without chronic disease could benefit from the same interventions. Combining interventions for those with and without chronic disease would be economical, while also bridging the stigmatization gap that often occurs in those with chronic disease (Joachim and Acorn 2000). Thirdly, these findings can be used to influence general prevention policies by demonstrating the need to address multiple health behaviors in political efforts aimed at the general population. Finally, the results add further support to existing literature stressing the importance of living a healthy lifestyle.

As with any study, this one has several strengths and weaknesses that need to be mentioned. The major strengths include the use of 3-day dietary records, laboratory-determined fitness, and the large number of participants who were followed over several years. In terms of limitations, it must be noted that this study used self-reported dietary data which are prone to a variety of measurement errors. One such bias that has become well established is a response set reflecting a tendency to present a diet that adheres more closely to social norms and public health recommendations (Hebert et al. 1995, 1997, 2008). Furthermore, the ACLS participants who completed 3-day diet records did so voluntarily with the knowledge that they would be reviewing their diet records with a dietician. Therefore, additional selection and reporting biases may have been introduced. Additionally, 3 days of dietary intake may not be an accurate representation of habitual intake. Despite the limitations of the methods used to assess dietary intake, 3-day diet records have been shown to be a more accurate representation of actual food intake than food frequency questionnaires (Day et al. 2001; Tokudome et al. 2005), which have been routinely used in similar studies (Laaksonen et al. 2002; Poortinga 2007; Schuit et al. 2002). It also must be noted that both alcohol consumption and smoking status were self-reported. Although alcohol consumption may appear to be difficult to recall accurately, there is good evidence that intake of alcoholic beverages among non-problem drinkers is reported accurately (Hebert et al. 1997b). In addition, because only four health behaviors were included in the LCA analyses, it is unlikely that more than two classes would emerge. Furthermore, the common practice of dichotomizing health behavior variables, and concomitant issues concerning cut-points, may have implications for the findings (MacCallum et al. 2002). It must also be noted that classifying individuals into one discrete class based on probability of being in a class likely resulted in some non-differential misclassification. This measurement error likely biased the relationships between class memberships and all-cause mortality towards the null. The generalizability of the results is restricted given that the ACLS cohort is predominantly white and from middle and upper socio-economic strata. Nevertheless, the homogenous nature of the sample helps to ensure internal control over factors such as ethnicity and socioeconomic status. Furthermore, studies have shown that although the prevalence of partaking in negative behaviors differs with socioeconomic status, ethnicity, and age, the clustering of behaviors is similar in different demographic groups (Berrigan et al. 2003; Laaksonen et al. 2001; Poortinga 2007). Finally, some potentially important demographic variables, such as marital status, were not collected in the ACLS cohort and were therefore not controlled for in the regression models. This may have resulted in some residual confounding.

In conclusion, negative health behaviors cluster together, irrespective of chronic disease status. The risk of all-cause mortality increases with the number of unhealthy behaviors one partakes in. These findings can be used as support for multi-behavioral intervention strategies where several risk behaviors are addressed simultaneously.

Acknowledgements

Supported by National Institutes of Health grants AG06945, HL62508, and R21DK088195, and an unrestricted research grant from the Coca-Cola Company. M. Héroux was supported by a doctoral award from the Canadian Institutes of Health Research. I. Janssen was supported by investigator awards from the Canadian Institutes of Health Research, the Ontario Ministry of Research and Innovation, and the Canada Research Chairs program. J. Hébert was supported by an Established Investigator Award in Cancer Prevention and Control from the Cancer Training Branch of the National Cancer Institute (K05 CA136975). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. We thank the Cooper Clinic physicians and technicians for collecting the baseline data, and staff at the Cooper Institute for data entry and data management.

Open Access

This article is distributed under the terms of the Creative Commons Attribution Noncommercial License which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited.

Appendix

Table 5.

Relationship between health behaviors

Prevalence (%) or Odds Ratio (95% CI) for Engagement in Unhealthy Behavior
Unhealthy Diet Physically Inactive Smokers Heavy Drinkers
Participants with no history of chronic disease
 Diet
  Healthy diet,% – 21.8% 8.9% 2.3%
  Unhealthy diet,% 37.4% 19.4% 4.5%
  Odds ratio (95% CI) 2.12 (1.88–2.39) 2.02 (1.74–2.35) 1.61 (1.21–2.15)
 Physical Activity
  Physically active 15.5% – 8.8% 2.4%
  Inactive 28.2% 17.2% 3.5%
  Odds ratio (95% CI) 2.14 (1.90–2.41) 2.09 (1.81–2.40) 1.36 (1.04–1.80)
 Smoking
  Non-smokers,% 16.9% 23.0% – 2.2%
  Smokers,% 33.2% 39.0% 6.5%
  Odds ratio (95% CI) 1.99 (1.71–2.32) 2.08 (1.81–2.39) 2.69 (2.00–3.62)
 Alcohol Intake
  Non/Light drinkers,% 18.3% 24.5% 10.5% –
  Heavy drinkers,% 31.4% 31.8% 26.4%
  Odds ratio (95% CI) 1.62 (1.22–2.17) 1.34 (1.02–1.77) 2.72 (2.02–3.66)
Participants with chronic disease
 Diet
  Healthy diet,% 23.9% 8.9% 2.3%
  Unhealthy diet,% – 43.0% 17.3% 4.3%
  Odds ratio (95% CI) 2.23 (1.91–2.60) 1.66 (1.36–2.05) 1.26 (0.86–1.84)
 Physical Activity
  Physically active,% 17.9% – 8.7% 2.7%
  Inactive,% 34.4% 16.0% 3.0%
  Odds ratio (95% CI) 2.25 (1.93–2.62) 1.85 (1.52–2.24) 0.96 (0.65–1.41)
 Smoking
  Non-smokers,% 20.9% 26.5% – 2.3%
  Smokers,% 36.2% 41.8% 6.8%
  Odds ratio (95% CI) 1.65 (1.34–2.03) 1.84 (1.51–2.23) 2.45 (1.63–3.69)
 Alcohol Intake
  Non/light drinkers,% 22.2% 28.1% 10.3% –
  Heavy drinkers,% 34.9% 30.2% 26.4%
  Odds Ratio (95% CI) 1.32 (0.90–1.94) 0.92 (0.62–1.35) 2.46 (1.63–3.71)
Open in a new tab

All relationships significant (p ≤ 0.05) with the exception of alcohol intake vs. physical activity and diet vs. alcohol intake in those with chronic disease

* Odds ratio (95% confidence interval) for engagement in the corresponding unhealthy behavior. The healthy behavior was used as the referent category. Odds ratios were controlled for age and gender.

Table 6.

Results of the latent class analysis used to identify classes of participants based on health behaviors, after controlling for age and gender

Health Behaviors Class 1 Class 2
(N = 3,850) (N = 9,771)
Diet
 Healthy diet 47.3% 98.4%
 Unhealthy diet 52.7% 1.6%
Physical Activity
 Active 61.4% 81.2%
 Inactive 38.6% 18.8%
Smoking
 Non-smokers 78.7% 95.0%
 Smokers 21.3% 5.0%
Alcohol Intake
 Non/light drinkers 93.7% 98.8%
 Heavy drinkers 6.3% 1.2%
# Unhealthy Behaviors
 0 1.6%% 77.9%
 1 53.7% 21.2%
 2 36.3% 0.9%
 3 8.0% 0.0%
 4 0.2% 0.0%
Open in a new tab

Values represent the probability of class membership4

References

  1. Aadahl M, Kjaer M, Kristensen JH, Mollerup B, Jorgensen T. Self-reported physical activity compared with maximal oxygen uptake in adults. European Journal of Cardiovascular Prevention & Rehabilitation. 2007;14:422–428. doi: 10.1097/HJR.0b013e3280128d00. [DOI] [PubMed] [Google Scholar]
  2. Adams SA, Matthews CE, Ebbeling CB, Moore CG, Cunningham JE, Fulton J, Hebert JR. The effect of social desirability and social approval on self-reports of physical activity. American Journal of Epidemiology. 2005;161:389–398. doi: 10.1093/aje/kwi054. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Adler NE, Rehkopf DH. U.S. disparities in health: Descriptions, causes, and mechanisms. Annual Review of Public Health. 2008;29:235–252. doi: 10.1146/annurev.publhealth.29.020907.090852. [DOI] [PubMed] [Google Scholar]
  4. Balke B, Ware RW. An experimental study of physical fitness of Air Force personnel. U S Armed Forces Medical Journal. 1959;10:675–688. [PubMed] [Google Scholar]
  5. Berrigan D, Dodd K, Troiano RP, Krebs-Smith SM, Barbash RB. Patterns of health behavior in U.S. adults. Preventive Medicine. 2003;36:615–623. doi: 10.1016/S0091-7435(02)00067-1. [DOI] [PubMed] [Google Scholar]
  6. Blair SN, Jacobs DR, Jr, Powell KE. Relationships between exercise or physical activity and other health behaviors. Public Health Reports. 1985;100:172–180. [PMC free article] [PubMed] [Google Scholar]
  7. Blair SN, Kannel WB, Kohl HW, Goodyear N, Wilson PW. Surrogate measures of physical activity and physical fitness. Evidence for sedentary traits of resting tachycardia, obesity, and low vital capacity. American Journal of Epidemiology. 1989a;129:1145–1156. doi: 10.1093/oxfordjournals.aje.a115236. [DOI] [PubMed] [Google Scholar]
  8. Blair SN, Kohl HW, Paffenbarger RS, Clark DG, Cooper KH, Gibbons LW. Physical fitness and all-cause mortality. A prospective study of healthy men and women. Journal of the American Medical Association. 1989b;262:2395–2401. doi: 10.1001/jama.1989.03430170057028. [DOI] [PubMed] [Google Scholar]
  9. Blair SN, Kampert JB, Kohl HW, Barlow CE, Macera CA, Paffenbarger RS, Gibbons LW. Influences of cardiorespiratory fitness and other precursors on cardiovascular disease and all-cause mortality in men and women. Journal of the American Medical Association. 1996;276:205–210. doi: 10.1001/jama.1996.03540030039029. [DOI] [PubMed] [Google Scholar]
  10. Blakely F, Dunnagan T, Haynes G, Moore S, Pelican S. Moderate physical activity and its relationship to select measures of a healthy diet. Journal of Rural Health. 2004;20:160–165. doi: 10.1111/j.1748-0361.2004.tb00023.x. [DOI] [PubMed] [Google Scholar]
  11. Bouchard C, Malina R, Perusse L. Genetics of fitness and physical performance. 1. Champaign, IL: Human Kinetics; 1997. [Google Scholar]
  12. Byun W, Sieverdes JC, Sui X, Hooker SP, Lee CD, Church TS, Blair SN. Effect of positive health factors and all-cause mortality in men. Medicine & Science in Sports & Exercise. 2010;42:1632–1638. doi: 10.1249/MSS.0b013e3181d43f29. [DOI] [PubMed] [Google Scholar]
  13. Centers for Disease Control and Prevention. (2003). National Death Index. National Center for Health Statistics. Available at http://www.cdc.gov/nchs/ndi.htm
  14. Chiolero A, Wietlisbach V, Ruffieux C, Paccaud F, Cornuz J. Clustering of risk behaviors with cigarette consumption: A population-based survey. Preventive Medicine. 2006;42:348–353. doi: 10.1016/j.ypmed.2006.01.011. [DOI] [PubMed] [Google Scholar]
  15. Chiuve SE, McCullough ML, Sacks FM, Rimm EB. Healthy lifestyle factors in the primary prevention of coronary heart disease among men: Benefits among users and nonusers of lipid-lowering and antihypertensive medications. Circulation. 2006;114:160–167. doi: 10.1161/CIRCULATIONAHA.106.621417. [DOI] [PubMed] [Google Scholar]
  16. Cole P, MacMahon B. Attributable risk percent in case-control studies. British Journal of Preventive & Social Medicine. 1971;25:242–244. doi: 10.1136/jech.25.4.242. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Day N, McKeown N, Wong M, Welch A, Bingham S. Epidemiological assessment of diet: A comparison of a 7-day diary with a food frequency questionnaire using urinary markers of nitrogen, potassium and sodium. International Journal of Epidemiology. 2001;30:309–317. doi: 10.1093/ije/30.2.309. [DOI] [PubMed] [Google Scholar]
  18. Donovan JE, Jessor R, Costa FM. Structure of health-enhancing behavior in adolescence: A latent-variable approach. Journal of Health & Social Behavior. 1993;34:346–362. doi: 10.2307/2137372. [DOI] [PubMed] [Google Scholar]
  19. Fine LJ, Philogene GS, Gramling R, Coups EJ, Sinha S. Prevalence of multiple chronic disease risk factors. 2001 National Health Interview Survey. American Journal of Preventive Medicine. 2004;27:18–24. doi: 10.1016/j.amepre.2004.04.017. [DOI] [PubMed] [Google Scholar]
  20. Finley CE, LaMonte MJ, Waslien CI, Barlow CE, Blair SN, Nichaman MZ. Cardiorespiratory fitness, macronutrient intake, and the metabolic syndrome: The Aerobics Center Longitudinal Study. Journal of the American Dietetic Association. 2006;106:673–679. doi: 10.1016/j.jada.2006.02.012. [DOI] [PubMed] [Google Scholar]
  21. Goodman L. On the assignment of individuals to latent classes. Sociological Methodology. 2007;37:1–22. doi: 10.1111/j.1467-9531.2007.00184.x. [DOI] [Google Scholar]
  22. Hebert JR, Clemow L, Pbert L, Ockene IS, Ockene JK. Social desirability bias in dietary self-report may compromise the validity of dietary intake measures. International Journal of Epidemiology. 1995;24:389–398. doi: 10.1093/ije/24.2.389. [DOI] [PubMed] [Google Scholar]
  23. Hebert JR, Ma Y, Clemow L, Ockene IS, Saperia G, Stanek EJ, Merriam PA, Ockene JK. Gender differences in social desirability and social approval bias in dietary self-report. American Journal of Epidemiology. 1997;146:1046–1055. doi: 10.1093/oxfordjournals.aje.a009233. [DOI] [PubMed] [Google Scholar]
  24. Hebert JR, Ockene IS, Hurley TG, Luippold R, Well AD, Harmatz MG. Development and testing of a seven-day dietary recall. Dietary Assessment Working Group of the Worcester Area Trial for Counseling in Hyperlipidemia (WATCH) Journal of Clinical Epidemiology. 1997b;50:925–937. doi: 10.1016/S0895-4356(97)00098-X. [DOI] [PubMed] [Google Scholar]
  25. Hebert JR, Hurley TG, Peterson KE, Resnicow K, Thompson FE, Yaroch AL, Ehlers M, Midthune D, Williams GC, Greene GW, Nebeling L. Social desirability trait influences on self-reported dietary measures among diverse participants in a multicenter multiple risk factor trial. Journal of Nutrition. 2008;138:226–234. doi: 10.1093/jn/138.1.226S. [DOI] [PubMed] [Google Scholar]
  26. Héroux M, Janssen I, Lam M, Lee DC, Hebert JR, Sui X, Blair SN. Dietary patterns and the risk of mortality: Impact of cardiorespiratory fitness. International Journal of Epidemiology. 2010;39:197–209. doi: 10.1093/ije/dyp191. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Ingledew D, Hardy L, Cooper C. Latent class analysis applied to health behaviours. Personality and Individual Differences. 1995;19:13–20. doi: 10.1016/0191-8869(95)00045-8. [DOI] [Google Scholar]
  28. Joachim G, Acorn S. Stigma of visible and invisible chronic conditions. Journal of Advanced Nursing. 2000;32:243–248. doi: 10.1046/j.1365-2648.2000.01466.x. [DOI] [PubMed] [Google Scholar]
  29. Kampert JB, Blair SN, Barlow CE, Kohl HW. Physical activity, physical fitness, and all-cause and cancer mortality: A prospective study of men and women. Annals of Epidemiology. 1996;6:452–457. doi: 10.1016/S1047-2797(96)00059-2. [DOI] [PubMed] [Google Scholar]
  30. Katzmarzyk PT, Church TS, Janssen I, Ross R, Blair SN. Metabolic syndrome, obesity, and mortality: Impact of cardiorespiratory fitness. Diabetes Care. 2005;28:391–397. doi: 10.2337/diacare.28.2.391. [DOI] [PubMed] [Google Scholar]
  31. Kvaavik E, Batty D, Ursin G, Huxley R, Gale C. Influence of individual and combined health behaviors on total and cause-specific mortality in mean and women. Archives of Internal Medicine. 2010;170:711–718. doi: 10.1001/archinternmed.2010.76. [DOI] [PubMed] [Google Scholar]
  32. Laaksonen M, Prattala R, Karisto A. Patterns of unhealthy behaviour in Finland. European Journal of Public Health. 2001;11:294–300. doi: 10.1093/eurpub/11.3.294. [DOI] [PubMed] [Google Scholar]
  33. Laaksonen M, Luoto R, Helakorpi S, Uutela A. Associations between health-related behaviors: A 7-year follow-up of adults. Preventive Medicine. 2002;34:162–170. doi: 10.1006/pmed.2001.0965. [DOI] [PubMed] [Google Scholar]
  34. Lanza ST, Collins LM, Lemmon DR, Schafer JL. PROC LCA: A SAS procedure for latent class analysis. Structural Equation Modeling. 2007;14:671–694. doi: 10.1080/10705510701575602. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Laska MN, Pasch KE, Lust K, Story M, Ehlinger E. Latent class analysis of lifestyle characteristics and health risk behaviors among college youth. Prevention Science. 2009;10:376–386. doi: 10.1007/s11121-009-0140-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Lee CD, Blair SN, Jackson AS. Cardiorespiratory fitness, body composition, and all-cause and cardiovascular disease mortality in men. American Journal of Clinical Nutrition. 1999;69:373–380. doi: 10.1093/ajcn/69.3.373. [DOI] [PubMed] [Google Scholar]
  37. Lopez AD, Mathers CD, Ezzati M, Jamison DT, Murray CJ. Global and regional burden of disease and risk factors, 2001: Systematic analysis of population health data. Lancet. 2006;367:1747–1757. doi: 10.1016/S0140-6736(06)68770-9. [DOI] [PubMed] [Google Scholar]
  38. MacCallum RC, Zhang S, Preacher KJ, Rucker DD. On the practice of dichotomization of quantitative variables. Psychological Methods. 2002;7:19–40. doi: 10.1037/1082-989X.7.1.19. [DOI] [PubMed] [Google Scholar]
  39. Mathers CD, Loncar D. Projections of global mortality and burden of disease from 2002 to 2030. Public Library of Science Medicine. 2006;3:2011–2030. doi: 10.1371/journal.pmed.0030442. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Metzger JS, Catellier DJ, Evenson KR, Treuth MS, Rosamond WD, Siega-Riz AM. Patterns of objectively measured physical activity in the United States. Medicine & Science in Sports & Exercise. 2008;40:630–638. doi: 10.1249/MSS.0b013e3181620ebc. [DOI] [PubMed] [Google Scholar]
  41. Mitchell JA, Bornstein DB, Sui X, Hooker SP, Church TS, Lee CD, Lee DC, Blair SN. The impact of combined health factors on cardiovascular disease mortality. American Heart Journal. 2010;160:102–108. doi: 10.1016/j.ahj.2010.05.001. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. National Institute on Alcohol Abuse and Alcoholism . The physicians’ guide to helping patients with alcohol problems. Rockville, MD: United States Department of Health and Human Services. NIH publication no; 1995. pp. 95–3769. [Google Scholar]
  43. Neutel CI, Campbell NR. Changes in lifestyle after hypertension diagnosis in Canada. Canadian Journal of Cardiology. 2008;24:199–204. doi: 10.1016/S0828-282X(08)70584-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Pate RR, Heath GW, Dowda M, Trost SG. Associations between physical activity and other health behaviors in a representative sample of US adolescents. American Journal of Public Health. 1996;86:1577–1581. doi: 10.2105/AJPH.86.11.1577. [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Patterson RE, Neuhouser ML, Hedderson MM, Schwartz SM, Standish LJ, Bowen DJ. Changes in diet, physical activity, and supplement use among adults diagnosed with cancer. Journal of the American Dietetic Association. 2003;103:323–328. doi: 10.1053/jada.2003.50045. [DOI] [PubMed] [Google Scholar]
  46. Pinto BM, Eakin E, Maruyama NC. Health behavior changes after a cancer diagnosis: What do we know and where do we go from here? Annals of Behavioral Medicine. 2000;22:38–52. doi: 10.1007/BF02895166. [DOI] [PubMed] [Google Scholar]
  47. Pollock ML, Bohannon RL, Cooper KH, Ayres JJ, Ward A, White SR, Linnerud AC. A comparative analysis of four protocols for maximal treadmill stress testing. American Heart Journal. 1976;92:39–46. doi: 10.1016/S0002-8703(76)80401-2. [DOI] [PubMed] [Google Scholar]
  48. Poortinga W. The prevalence and clustering of four major lifestyle risk factors in an English adult population. Preventive Medicine. 2007;44:124–128. doi: 10.1016/j.ypmed.2006.10.006. [DOI] [PubMed] [Google Scholar]
  49. Rothman KJ. Epidemiology: An introduction. 1. New York: Oxford University Press; 2002. [Google Scholar]
  50. Sallis J, Saelens B. Assessment of physical activity by self-report: Status, limitations, and future directions. Research Quarterly for Exercise & Sport. 2000;71:1–14. doi: 10.1080/02701367.2000.11082780. [DOI] [PubMed] [Google Scholar]
  51. Schuit AJ, van Loon AJ, Tijhuis M, Ocke M. Clustering of lifestyle risk factors in a general adult population. Preventive Medicine. 2002;35:219–224. doi: 10.1006/pmed.2002.1064. [DOI] [PubMed] [Google Scholar]
  52. Tokudome Y, Goto C, Imaeda N, Hasegawa T, Kato R, Hirose K, Tajima K, Tokudome S. Relative validity of a short food frequency questionnaire for assessing nutrient intake versus three-day weighed diet records in middle-aged Japanese. Journal of Epidemiology. 2005;15:135–145. doi: 10.2188/jea.15.135. [DOI] [PMC free article] [PubMed] [Google Scholar]
  53. Tsubono Y, Koizumi Y, Nakaya N, Fujita K, Takahashi H, Hozawa A, Suzuki Y, Kuriyama S, Tsuji I, Fukao A, Hisamichi S. Health practices and mortality in Japan: Combined effects of smoking, drinking, walking and body mass index in the Miyagi Cohort Study. Journal of Epidemiology. 2004;14:39–45. doi: 10.2188/jea.14.S39. [DOI] [PMC free article] [PubMed] [Google Scholar]
  54. Vermunt J, Magidson J. Latent class cluster analysis. In: McCutcheon A, editor. Applied latent class analysis. Cambridge: Cambridge University Press; 2002. pp. 89–106. [Google Scholar]
  55. Vermunt J, Magidson J. Factor analysis with categorical indicators: A comparison between traditional and latent class approaches. In: Van der Ark A, Croon M, Sijtsma K, editors. New developments in categorical data analysis for the social and behavioural sciences. New Jersey: Lawrence Erlbaum Associates; 2005. pp. 41–62. [Google Scholar]
  56. Wallstrom P, Mattisson I, Tyden P, Berglund G, Janzon L. Dietary habits after myocardial infarction—results from a cross-sectional study. Journal of Internal Medicine. 2005;257:329–337. doi: 10.1111/j.1365-2796.2005.01461.x. [DOI] [PubMed] [Google Scholar]
  57. Walter SD. Calculation of attributable risks from epidemiological data. International Journal of Epidemiology. 1978;7:175–182. doi: 10.1093/ije/7.2.175. [DOI] [PubMed] [Google Scholar]
  58. Wei M, Gibbons L, Mitchell T, Kampert J, Lee C, Blaire S. The association between cardiorespiratory fitness and imparied fasting glucose and type 2 diabetes mellitus in men. Annals of Internal Medicine. 1999;130:89–96. doi: 10.7326/0003-4819-130-2-199901190-00002. [DOI] [PubMed] [Google Scholar]
  59. Williams AF, Wechsler H. Interrelationship of preventive actions in health and other areas. Health Services Representative. 1972;87:969–976. doi: 10.2307/4594704. [DOI] [PMC free article] [PubMed] [Google Scholar]
  60. World Health Organization. (2005). Preventing chronic diseases: A vital investment. Available at http://www.who.int/chp/chronic_disease_report/en/
  61. World Health Organization. (2009). World Health Organization: World health statistics. Available at http://www.who.int/whosis/whostat/EN_WHS09_Full.pdf
  62. Yaroch AL, Nebeling L, Thompson FE, Hurley TG, Hebert JR, Toobert DJ, Resnicow K, Greene GW, Williams GC, Elliot DL, Goldman Sher T, Stacewicz-Sapuntzakis M, Salkeld J, Rossi S, Domas A, McGregor H, Defrancesco C, McCarty F, Costello RB, Peterson KE. Baseline design elements and sample characteristics for seven sites participating in the Nutrition Working Group of the Behavior Change Consortium. Journal of Nutrition. 2008;138:185–192. doi: 10.1093/jn/138.1.185S. [DOI] [PubMed] [Google Scholar]

Articles from Prevention Science are provided here courtesy of Springer

RESOURCES