Personal Factors Predictive of Health-Related Lifestyles of Community-Dwelling Older Adults

Design

This descriptive study explored the personal factors that can be used to predict the various aspects of health-related lifestyle of community-dwelling older adults. Data were derived from the development of HELP (Hwang, 2010b).

Participants

The study sample consisted of older adults residing in southern California. Each participant met the following criteria: age ≥55 yr, community dwelling, and adequate cognitive and English-language capabilities for responding to questionnaires. Convenience sampling and snowball sampling methods were used to recruit participants from a diversity of community sites, including senior citizen centers, senior residential communities, independent living facilities, adult day health care centers, local senior social and activities groups, and religious groups and organizations. Despite the use of nonprobability sampling methods, we attempted to ensure that the study sample would demonstrate an adequate distribution in various racial categories, age ranges, and residential locations.

Instrumentation

The measurement of personal factors and health-related lifestyle was completed through HELP (Hwang, 2010b). HELP can be administered as a self-report questionnaire or as an interview in which the interviewer reads aloud the questions and records the respondent’s answers. It consists of two sections. The first section assesses three categories of personal factors:

The second section of HELP comprises seven scales measuring different aspects of health-related lifestyle: (1) Exercise, (2) Diet, (3) Productive and Social Activities, (4) Leisure, (5) Activities of Daily Living (ADLs), (6) Stress Management and Spiritual Participation, and (7) Other Health Promotion and Risk Behaviors. Each scale consists of eight items that examine the frequency of the respondent’s engagement in the previously mentioned areas of health-related occupations (see Figure 1 for sample items). Respondents are asked to answer each question according to their typical or routine performance during the past 3-mo period. Response categories generally include never, 1 to 2 days, 3 to 4 days, 5 to 6 days, and 7 days (per week).

OPEN IN VIEWER

Figure 1.

Sample Items of the Health Enhancement Lifestyle Profile Scales

As suggested by participants in the field pretesting study (see Hwang, 2010b), we added an additional category, 1 to 2 days a month. In addition, it was also suggested that monthly frequency (i.e., never, 1 day, 2 days, 3 days, 4 days, 5 days or more per month) be used for a few items assessing activities that may not be performed on a daily or weekly basis (e.g., carpentering, auto/boat/house fixing) and those assessing noticeable health risk behaviors (e.g., “drink three or more servings of alcohol-containing beverages in 1 day”).

For scoring, each response is given a numerical value reflective of the relative frequency. The result is a 0- to 5-point rating scale (0 = never; 5 = 7 days a week). Responses from negatively worded items (see sample items marked with an asterisk in Figure 1) are reversed for scoring. For each HELP scale, a subtotal score ranging from 0 to 40 can be computed; a higher score indicates a higher frequency of health-promoting behaviors. The total HELP score ranges from 0 to 280. The approximate time needed for completing and administering HELP is 20–40 min.

Psychometric properties of HELP have been supported through multiple procedures and studies. Piloting of HELP was completed through focus group and field pretesting to ensure the relevance and clarity of the measure (see Hwang, 2010b). The Rasch measurement model was used to establish the unidimensionality of each HELP scale through the analyses of principal components of residuals, goodness-of-fit statistics, local dependence, and differential item functioning. Items in each scale form a difficulty hierarchy that provides an expected pattern of health lifestyle behaviors in older adults. Acceptable to good person separation and reliability statistics substantiated the clinical applicability and consistency of HELP scores. In addition, analysis of the rating scale structure confirmed the functioning of the response categories (i.e., 0- to 5-point scale) used in the HELP scales (Hwang, 2010b).

Additional evidence of reliability and validity of HELP was provided through classical test theory: The α coefficients ranged from .75 to .92 among the HELP scales, indicating acceptable to good internal reliability (Hwang, 2010c). Construct validity was supported by the method of hypothesis testing that examined (1) the patterns of interrelationship and interdependence among the seven HELP scales (Pearson rs = .37–.62, p < .0001) and (2) the correlations between the HELP results and the global health status (Pearson rs = .42–.71, p < .0001; Hwang, 2010c). Finally, the stepwise multiple regression models revealed that five of the seven HELP scales (i.e., Exercise, Diet, ADLs, Stress Management and Spiritual Participation, and Other Health Promotion and Risk Behaviors) served as significant predictors for health-related quality of life of older adults (Hwang, 2010a).

Procedures

Approval from the institutional review board at California State University, Dominguez Hills, was received before data collection. HELP was completed by each participant either (1) in a direct interview, in which a graduate student or a staff member read aloud the questions and recorded responses for the participant or (2) by means of paper and pen, administered onsite to an individual or a group of participants. In a few cases, the instruments were delivered to potential participants by postal or electronic mail with instructions enclosed.

Data Analysis

Data were analyzed separately for each of the seven HELP scales as well as the total HELP (the total of the seven scales) to yield the distinct sets of personal factors that can serve as predictors of different aspects of the lifestyle measure. First, a series of univariate correlational and comparative analyses, including Pearson correlation coefficient (r), independent t test, and one-way analysis of variance (ANOVA), was conducted to screen factors that may have an effect (prediction potential) on the dependent variables (i.e., seven HELP scales and the total). A probability level (p) of .25 was adopted as the selection criterion for the entry of the predictor variables (factors) into the subsequent regression analyses. The use of the p range of ≤.25 was to avoid the exclusion of potentially important factors from the analyses (Hosmer & Lemeshow, 1989; Tabachnick & Fidell, 2007).

Before running the regression analysis, factors that were of a categorical variable were dummy coded. These factors included gender (reference group: female), race (reference group: White), marital status (reference group: married), and employment status (reference group: unemployed/retired). In addition, the possibility of muiticollinearity between predictor variables was investigated through collinearity diagnostics. The resultant tolerance values ranged from 0.82 to 0.99 (i.e., >0.40), and variance inflation factors ranged from 1.12 to 1.64 (i.e., <10), indicating an absence of muiticollinearity (Tabachnick & Fidell, 2007).

Finally, given the exploratory nature of the analysis, the stepwise option of variable entry was used, with a probability level of p < .05 to enter and p ≥ .10 to remove. An optimal regression model solution was developed for each HELP scale and the total HELP to identify significant predictor variables. With the eight potential predictor variables in this study, to achieve power of .80 and a medium effect size (effect size = .15, α = .05), a sample size of 108 was required to detect a significant model (Tabachnick & Fidell, 2007). Our sample size of 253 was thus considered adequate. All the previously mentioned statistical analyses were performed using SPSS/PC+ Version 17.0 software (SPSS Inc., Chicago).

Results

A total of 253 older adults (148 women and 105 men) were recruited for this study. Participants’ ages ranged from 55 to 92 yr (mean [M] = 71.4, standard deviation [SD] = 10.5). Table 1 provides a summary of participants’ demographics and health-related information. Most participants were unemployed or retired. The most commonly reported health problems or impairments among the participants included hypertension, arthritis, visual impairment, back or neck pain, and diabetes. The total number of chronic diseases and impairments reported by each participant ranged from 0 to 10 (M = 2.77; SD = 2.03). Slightly more than 70% of the participants considered their health status to be “good” or “excellent,” and only approximately 5% believed that they were in poor health.

Table 2 demonstrates the personal factors selected by the preliminary analyses (i.e., Pearson r, t test, ANOVA) for the entry of predictor variables to the regression model of each HELP scale and the total HELP. All eight personal factors were entered at least once into the regression analyses. As seen, the number of diseases or impairments and self-rated health was constantly included in the data entry of the regression analyses of the seven HELP scales as well as the total HELP, whereas marital status was selected only for the Exercise scale.

Table 3 displays the results of regression analysis after stepwise selection of the significant predictor variables to form an optimal regression model solution for each HELP scale. All the regression model solutions were significant at p < .0001, and the values of adjusted R ² ranged from .205 for the ADLs scale to .571 for the total HELP. Each model solution (HELP scale) consisted of two or more significant predictor variables (personal factors). For example, the Exercise scale can be best predicted by three factors: (1) the number of chronic diseases and impairments, (2) self-rated health, and (3) gender. With the exception of marital status, all other factors emerged as significant predictors for one or more HELP scales. Two factors, the number of chronic diseases or impairments and self-rated health, were found to be predictive of all seven HELP scales as well as the total HELP. In particular, both factors exclusively predict Diet, ADLs, and the total HELP.

Compared with the women, the men had a slightly more positive effect on the Exercise scale but a slightly more negative effect on both the Productive and Social Activities and Stress Management and Spiritual Participation scales. Employment (both full time and part time) contributed positively to Productive and Social Activities, whereas full-time employment had a negative impact on Leisure, in contrast with the unemployed/retired status. With regard to race, relative to White, the Asian/Pacific Islander and African-American groups presented less favorable effects on both the Productive and Social Activities and Leisure scales, whereas the Asian/Pacific Islander and Hispanic/Latino groups presented a more favorable effect on the Stress Management and Spiritual Participation scale. Moreover, education was a positive predictor of scores on the Leisure and Other Health Promotion and Risk Behaviors scales. Finally, age had a small positive effect on the Other Health Promotion and Risk Behaviors scale, along with other factors (i.e., the number of diseases and impairments, self-rated health, education).

Discussion

This study was conducted to determine the significance of eight personal factors in predicting health-related lifestyle of a sample of community-dwelling older adults. Regression analyses revealed that different aspects of lifestyle (as measured by the HELP scales) can be significantly predicted by several of the personal factors entered. Personal health conditions, including the number of chronic diseases/impairments and self-perception of health status, were found to be the best predictors of the lifestyle measure. In particular, these two factors exclusively accounted for 57% of the variance in the total HELP score. Demographic characteristics, including gender, employment status, race, education, and age, also demonstrated varied degrees of capability for predicting different aspects of health-related lifestyle in the older adults. Although all the resultant regression model solutions were significant at p < .0001, their adjusted R ² values were mostly <.30. Not surprisingly, because of complexity in the context of health-related lifestyle, we did not expect these personal factors alone to be held accountable for a great deal of variance of lifestyle behaviors among older adults.

The number of chronic diseases and impairments was a strong inverse predictor for all seven HELP scales and the total HELP. The results suggest that older adults who demonstrate more positive lifestyle behaviors may have fewer adverse health conditions, and vice versa. Many studies have shown that chronic conditions and disability can be prevented, postponed, or diminished through positive lifestyle behaviors. For example, changes in exercise and diet, when used as a preventive intervention in older adults, can significantly reduce the risk of cardiovascular disease, osteoporosis, arthritis, diabetes, some forms of cancer, obesity, cognitive decline, and depression (Depp & Jeste, 2006; Koertge et al., 2003; Serra-Majem, Roman, & Estruch, 2006). Studies have also shown the positive correlations between levels of participation in social and productive activity and a range of health outcomes, including physical and cognitive functions, self-efficacy, and quality of life (Everard, Lach, Fisher, & Baum, 2000; Lawton, Winter, Kleban, & Ruckdeschel, 1999). These research findings support the protective effects of healthy lifestyle behaviors against the occurrence of chronic illness or disabling conditions among older adults. Our research findings, however, may only suggest the ability of the disease or impairment count to “mathematically” predict the pattern of health-related lifestyle in older adults. Clinically, all these relationships need not be causal. For example, people who demonstrate prolonged health risk behaviors may consequently suffer from more chronic diseases; conversely, it is possible that those who have chronic diseases, such as diabetes, become more health conscious and incorporate a range of health behaviors into their day-to-day self-management routine.

Self-rated health was also a significant predictor for all aspects of the lifestyle measure. This subjective rating of health has been considered “a simple, direct, and global way of capturing perceptions of health using criteria that are as broad and inclusive as the responding individual chooses to make them” (Idler & Benyamini, 1997, p. 22). That is, older adults’ perceptions of their health status are generally self-determining (DeSalvo, Bloser, Reynolds, He, & Muntner, 2006). Studies have shown that older adults with poor perceptions of health are less likely to engage in exercise routines (Rakowski & Mor, 1992) and self-care (Haug, Wykle, & Namazi, 1989) and are more likely to have health risk behaviors, such as smoking and heavy alcohol consumption (Hirdes & Forbes, 1993). Self-rating of health may hold the key to understanding the influences of various psychocognitive factors (e.g., self-efficacy, motivation, personal values, preferences) on health-promoting behaviors of older adults (DeSalvo et al., 2006).

Variations were found in the predictive strength and pattern of demographic characteristics toward different aspects of health-related lifestyle. In general, after the stepwise selection of the previous two predominant personal health factors, the demographic variables appeared to be less accountable for the variance of lifestyle behaviors (see β coefficients in Table 3), yet the levels of their predictive strength were still statistically significant. Although findings of population-based studies on various health outcomes (e.g., functional independence, life satisfaction, quality of life) have generally favored married older men and women (de Belvis et al., 2008; McCamish-Svensson, Samuelsson, Hagberg, Svensson, & Dehlin, 1999), our results did not suggest the effect of marital status on the outcomes of the lifestyle measure. Interestingly, as distinguished by Pender (1996) in her Health Promotion Model, personal psychocognitive factors, such as self-perception of health and self-efficacy, may have more direct influence on the pattern and frequency of health-promoting behaviors, whereas personal demographic characteristics, such as age, gender, race, and marital status, are considered modifying factors that have indirect effects on health behaviors.

The possible relationships between the demographic characteristics and different aspects of heath-related lifestyle after taking into account the contribution of personal health factors (i.e., the disease/impairment count, self-rated health) can be summarized as follows:

As discussed previously, given the multifaceted and dynamic features of the lifestyle context, caution in applying these findings is warranted.