Social and environmental predictors of walking among older adults

The target population includes community-dwelling older adults (age ≥ 60) drawn from a large integrated healthcare system in two counties (Bell and Brazos) in Central Texas. This analysis is part of a larger study that examined environmental safety factors associated with various health behaviors and health outcomes among adults 50 years or older. Four cities (Temple and Killeen from Bell County, Bryan and College Station from Brazos County) were chosen to provide adequate subject samples and environmental variability for this study. The counties were peri-urban with no large metropolitan cities, and represent understudied environmental settings in walkability research. With our interest in understanding factors associated with walking behaviors in older adults, persons 60 years or older were selected as the population group of interest for this analysis. We chose this age because this is the age when older adults are eligible for Older American Act benefits for health promotion and disease prevention [19].

Electronic medical records of patients who lived in our study areas were utilized to conduct initial subject screening by age and geographic residence. To target active individuals, only those that had at least one clinical encounter in the Family Medicine Department during the past 3 years were eligible to be included in our sampling frame. From the selected list, individual names were randomly drawn, reviewed, and approved by their primary care physician. Approximately 1000 letters were mailed out every month except during the winter months when people are less likely to walk outside. We further restricted our study sample to those who: a) did not have any difficulty to read, write, and speak English; b) were able to walk at least three city blocks (or for about five minutes); c) were not terminally ill; and d) did not live in a nursing home or assisted living community. In order to ensure the individual’s capability of walking or engaging in PA, screening questions for the ability to walk were included at the beginning of the survey instrument.

Participants were given the option to choose whether they wanted online (69.5 %) or paper surveys (30.5 %). The survey took approximately 20 min to complete, and included questions on demographic, health, and environmental factors potentially related to walking. A $10 gift card was offered at the completion of the full survey. Further, power analyses were conducted prior to recruitment indicating a sample of at least 250 was needed to detect differences (80 % power) in major outcomes. Assuming 60 % as those meeting the PA recommendation and 40 % as those not meeting the recommendation [20], a total of 250 patients can achieve 80 % power to detect a difference of 0.36 standard deviation with a significance level (alpha) of 0.05 using a two-sided, two-sample t-test. This metric was exceeded in the current analysis.

Ethical approval was sought and obtained from the institutional review board or IRB (ethics committee) of Baylor Scott & White Health before the study began. Written informed consent was gained from participants via online surveys or paper surveys. Consent was recorded/documented via online through Qualtrics secured survey management and via in-person (paper) retrieved from participants by research staff and kept in a secure file to be destroyed at the end if the study. This consent procedure was approved by the Baylor Scott & White Health IRB.

Recent literature differentiates different types of walking (e.g., walking for transportation and walking for recreation). Walking for “any purpose” was viewed as the most appropriate construct for older adults whose days may not be as structured around work or recreation as younger persons. To ascertain frequency of walking behavior, we defined PA based on the number of days an individual walked during a ‘typical’ week. We used the question: ‘How many days in a typical week do you walk in your neighborhood for any purposes?’ for this measure. Based on prior literature [21‐23], we created a variable to assess whether an individual participated in walking for any purposes for less than 3 days during a typical week (defined as non-frequent walking) versus 3 or more days (defined as frequent walking).

To ascertain whether patients were meeting the Surgeon General’s recommendations through walking, we assessed the average minutes of PA during a typical week. To evaluate whether individuals were meeting the recommended level of PA set by the CDC (i.e., 150 min of moderate activity) [14], we used a combination of the self-reported average number of minutes of PA through walking per day and the average number of days an individual participated in PA through walking per week in one’s neighborhood. The result was the average weekly minutes of PA through walking. Previous evidence suggests that walking at a normal pace may be equivalent to moderate intensity PA [24]. Therefore, walking will likely capture at least moderate PA with regard to the national guidelines.

As noted below, the survey items employed came from previously validated survey instruments (i.e., the Behavioral Risk Factor Surveillance System or BRFSS [25, 26] and the Neighborhood Environment Walkability Scale or NEWS [27, 28]), or reflect an assessment battery used by our study team in its health and environmental studies. Although the study targeted those 60 years or older, we wanted to examine differences among this group. Age was coded as 60–69 years (younger) and 70 years or older (older). Education was coded as having a high school education or less versus higher. Gender was coded as male or female. Race/Ethnicity was coded as White-non Hispanic or minority individuals (e.g., encompassing African Americans, Hispanic, and other racial groupings).

Health status was measured using two standard survey items assessing physical and mental well-being from the CDC health-related quality of life assessment battery used in the Behavioral Risk Factor Surveillance System or BRFSS [25, 29]. The following was used to assess physical well-being: ‘Now thinking about your physical health, which includes physical illness and injury, for how many days during the past 30 days was your physical health NOT good?’ The following was used to assess mental well-being: ‘Now thinking about your mental health, which includes stress, depression, and problems with emotions, for how many days during the past 30 days was your mental health NOT good?’ We wanted to determine differences in those with high versus low number of days in either physical or mental well-being (separately). Therefore, we dichotomized these variables into high (at or above the upper quartile) versus low (less than the upper quartile), which resulted in a split of less than 2 days versus greater than or equal to 2 days in the past 30 days for both physical and mental well-being in two separate variables.

Marital status (never married or divorced, widowed or separated versus married or living with a partner) was included to assess differences in social support influenced by having a spouse/partner. We also included items to tap social engagement for walking [30]. We examined whether individuals reported having someone to walk with to assess human social support for walking. We also included an item to reflect human-animal companionship as a potential motivator for walking. A new combined measure of social support was constructed using the questions: ‘Is there someone in your household you go walking with?’ and ‘Is there a dog in your household that you usually walk?’ Social support for walking was coded on a scale of 0 to 2, with 0 representing a negative (no) response to both items, 1 representing a single affirmative (yes) response, and 2 representing affirmative responses to both items.

In addition to individual respondent’s age, we wanted to reflect median neighborhood age. Median age at the participant’s Census Tract was included to identify differences across younger versus older neighborhood populations. This was coded as at or below the median versus higher than the median among those Census Tracts that were included in the study population.

We also included respondents’ perception of their neighborhood, measured with the Neighborhood Environment Walkability Survey (NEWS) instrument [27] which included multiple domains of variables capturing neighborhood features influencing walking. Neighborhood traffic safety was included to account for variation in perceived safety. This was coded as a scaled variable with the following 4 items: 1) There is so much traffic along the street I live on that it makes it difficult or unpleasant to walk in my neighborhood; 2) There is so much traffic along nearby streets that it makes it difficult or unpleasant to walk in my neighborhood; 3) There are sidewalks or protected walkways (e.g., trails) in my neighborhood; and 4) There are crosswalks and pedestrian signals to help walkers cross busy streets in my neighborhood. All questions were on a scale from 1 to 4, with 4 representing all responses indicative of the poorest safety and 16 indicative of the best safety. To preserve cell size, we combined this scale into at or below the lower quartile or low (4–10), moderate (11, 12), and at or above the upper quartile or high (13–16) according to quartile splits. Perceived neighborhood fall injury risk was also included and assessed on a single item: “I am worried about falling when I walk in my neighborhood.” The scores on a scale ranging from 1 (strongly disagree) to 4 (strongly agree) and were dichotomized into agree or disagree.

Finally, neighborhood cohesion was captured by combining responses from the following 5 questions: 1) I see and speak to other people when I am walking in my neighborhood; 2) Many people walk or bike in my neighborhood; 3) The streets and walkways in my neighborhood are clean and well maintained; 4) Walkers and bikers on the streets in my neighborhood can be easily seen by people from their homes; and 5) My neighbors could be counted on to help in case of need. All questions were on a scale of 1–4, with 5 representing all responses ‘strongly disagree,’ with a maximum of 20 representing all responses ‘strongly agree.’ To preserve cell size we combined this into at or below the lower quartile or low (5–14), moderate (15–17), and at or above the upper quartile or high (18–20) according to quartile splits. Similar variables have been used to describe neighborhood cohesion elsewhere [3, 16].

SAS 9.4 was used to run all analyses. Descriptive analyses were utilized to characterize the population and walking patterns. Chi square tests were used to test whether there were significant differences between groups across our outcomes of interest. Multivariate logistic regression was used to assess the likelihood of being frequent walkers vs. non-frequent walkers and meeting or failing to meet the Surgeon General’s PA recommendation. Odds ratios and 95 % upper and lower confidence intervals are reported.

Given the approximately 52,000 individuals (age ≥ 50) in the sample frame, we randomly sampled 10,000 subjects, who lived in our study areas and were seen in primary care (Family Medicine) clinics during the past 3 years. After primary care physicians identified potentially eligible subjects on the sampling lists, a total of 7336 recruitment letters were sent out from October 2013 to June 2014. A total of 496 individuals participated in the surveys. Of the 496, 80 participants who did not meet the study screening criteria and did not complete the survey were excluded. Based on our expanded eligibility criteria (e.g., living in our study areas, not living in assisted facilities), 22 of the remaining 416 were excluded. Of these 394, 122 younger than 60 years were excluded to focus on our target population group (60 or older). Analyses for this study are based on the remaining 272 subjects. Figure 1 illustrates the flow of survey response and data collection.

As indicated in Table 1, the median age of the study participants was 69 years (range: 60 to 92). They were of both genders (50.37 % female), primarily white-non Hispanic (84.87 %), educated (76.01 %) having more than a high school degree), and married (77.04 %). They generally reported good health, with the majority reporting relatively few days (≤1 day per month) with poor physical (73.80 %) or mental health (73.53 %). While only 19.85 % had both human and animal companions to walk with, almost half (49.26) had a least one source of social support for walking. Very few participants expressed worry about falling when walking in their neighborhood (16.54 %). The majority reported positive environmental perceptions of safety (73.16 %) and moderate or high neighborhood cohesion (64.71 %). Overall, the neighborhood median age measured at the Census Tract ranged from 22.8 to 43.3 with a median of 27.6. Table 1 also provides a descriptive view of how study characteristics differ by the two main walking outcomes. Table 1 further provides evidence of significant differences using Chi Square tests in the respondents characteristics between frequent walkers and non-frequent walkers, and between those who meet vs. fail to meet the Surgeon General’s recommendations for PA through walking.

As further indicated in Figs. 2 and 3, there was variability in walking behaviors. Overall, the median number of days individuals walked was 3, with the median number of minutes per week at 80 min. While the majority (59.55 %) walked at least three times a week, 14.61 % did not walk in their neighborhoods at least weekly (see Fig. 2). Further, only 27.86 % walked ≥150 min a week and 9.4 % walked 250 or more minutes per week (see Fig. 3).

Participants reported a variety of places where they walked in their neighborhood at least once a week. Overall, 75.38 % (196 out of 260 who answered this question) reported neighborhood streets, followed by green areas (n = 89, 34.23 %), malls (n = 67, 25.77 %), gyms (n = 19, 7.31 %), and schools (n = 15, 5.77 %). Other less frequently used places mentioned included treadmills at home, around large yards or properties, along the highway, or at golf courses (n = 31, 11.92 %). Since participants could indicate more than one place, percentages do not add up to 100 %.

Table 2 presents results of multivariate analyses. An examination of factors associated with frequent walking versus non-frequent walking behaviors reveals no significant demographic predictors. With regard to health-related variables, those reporting two or more poor mental health days (per month) are 65.5 % less likely to report frequent walking behaviors than those reporting fewer poor mental health days (OR = 0.345, 95 % CI = 0.185–0.645). Regarding environmental factors, those living in areas with younger population (census tract) median age are almost 80 % more likely to report frequent walking behaviors (OR = 1.799, 95 % CI = 1.034–3.131) than those in areas where the population is older. Compared to those reporting a high level of social cohesion, those reporting a low level are 66.2 % (OR = 0.338, 95 % CI = 0.159–0.715) less likely and those reporting a moderate level are 52.9 % (OR = 0.471, 95 % CI = 0.228–0.974) less likely to report frequent walking behaviors.

An examination of factors associated with meeting versus not meeting the Surgeon General’s recommendations for PA through walking reveals that age is a significant predictor. Those who are younger (60–69 years) versus older (70 years or older) are 46.2 % less likely to meet these recommendations (OR = 0.538, 95 % CI = 0.290–0.997). Those reporting poorer mental health are 56.8 % less likely to meet these recommendations than those reporting fewer poor mental health days (OR 0.432, 95 % CI = 0.198-.944). Those reporting no human or animal companions to walk with are 61.7 % less likely to meet these recommendations than those who have both sources of social support for walking (OR = 0.383, 95 % CI = 0.154–0.957).

While we see significant contributions of this study in terms to adding to the knowledge base on the linkage between environmental factors and walking behaviors among older adults, there are several limitations which must be acknowledged.

While assessing electronic medical records provided an easy way to identify potential subjects, over 7000 patient records were needed to identify about 500 subjects. This low recruitment yield is unfortunately typical in survey-based studies in primary care settings with older adults [49, 50]. We are aware that we do not have a clinical population representative of our integrated health care system. Rather, given our screening criteria and limitation to individuals 60 years or older, our study population represents a subset of the larger population. The demographics of the study participants are described to frame the type of populations to which our findings may be most appropriately generalized. For example, participants were limited to four communities in central Texas and, as such, findings may not be generalized to other states or national populations with different socioeconomic characteristics.

Another limitation is the self-reported nature of the data, which may lead to misreporting of some of the variables. This is especially so for reporting walking behaviors, although we are more confident in the accuracy of reported days versus actual cumulative minutes. We acknowledge that our PA measures may be somewhat crude, and do not take into account that patients may be getting other types of exercise. Also, we did not specify walking at a moderate pace, so our reference to the Surgeon General’s recommendation must be considered approximate versus definitive. While our cut-point for non-frequent physically active behavior has been justified in previous literature [21‐23], due to our sample distribution, we were not able to examine those reporting zero days of PA. Examining other PA measures in a larger and more diverse population may result in different associations.

Further, the data set was limited to a fixed set of questions limiting the ability to examine relationships among a full set of psychosocial predictors or outcomes of interest (e.g., we did not have questions assessing patient motivations or other sedentary behaviors such as sitting or screen time). Further, we did not account for variation in employment status, which may influence one’s PA. In addition, we were not able to report on the relationship between walking and all potential neighborhood factors that may influence walking. For example, danger from others such as gun violence or mugging may have influenced our outcomes but was not measured in the current study. Finally, our data are cross-sectional and do not permit an understanding of the causal chains linking the different variables of interest.

Nevertheless, the current study adds to the literature on how place matters with regard to different aspects of PA among older adults and suggests further research, practice and policy action. Several factors deserve more attention. We note that the aggregate population age of one’s surrounding can act as a facilitator of PA, with older adults living in areas more populated by younger adults being more likely to be physically active. It is not clear whether older adults are influenced by seeing more active neighbors, or if neighborhoods catering to younger populations have more amenities promoting PA. This finding though supports the 8–80 philosophy of having activity friendly environments where persons of all ages can safely be active, and calls for more research into intergenerational influences on PA [32].

Further, identifying facilitators of PA in one’s environment provides guidance for interventions at the environmental level. Future analyses including objective measures of both the physical place and walking behaviors among older adults can help elucidate relationships between health and place. In future studies, we recommend longitudinal analyses to enable assessments of dynamic changes over time. In particular, we would want to test the interplay among the social and built environments, lifestyle factors, and health outcomes [40] to better understand the complex interactions among sociodemographic, health, environmental, and walking behaviors.

Additional longitudinal research with larger and more diverse populations should be conducted to identify how different aspects of the environment (e.g., natural, built or social) interact with personal factors to encourage PA and maximize population health for older adults [3]. This includes greater attention to the need for targeted research funding and an appreciation of the complexities of this type of environmental research.

In recognition of the health benefits of PA, we endorse clinical and sports medicine professional organizations dissemination of “exercise as medicine” messages to help Americans live healthier lives [51]. In line with the “exercise as medicine” philosophy, we encourage physicians to assess and advise their older patients on how they can be more physically active for health promotion and disease prevention. Additionally, we endorse efforts underway to inform physicians of the impact of environmental factors on their patient’s health [52] and to create and disseminate inventories of places where older adults can safely walk [53, 54]. Recognizing the important influence that mental health concerns can have on initiating and sustaining healthy lifestyles [37], consistent with clinical guidelines for older patients [55], we also recommend that physicians assess whether late-life depression or other mental health symptomatology are present and how these can be addressed to enhance activity levels and overall quality of life.