Improving Physical Activity among Residents of Affordable Housing: Is Active Design Enough?

Physical inactivity increases risk of chronic disease. Few studies examine how built environment interventions increase physical activity (PA). Active design (AD) utilizes strategies in affordable housing to improve resident health. We assessed how AD housing affects PA among low-income families in Brooklyn, New York. Participants were recruited at lease signings in 2016 from a new AD apartment complex and two recently renovated comparison buildings without AD features. Eligibility included age ≥18 years with no contraindications to exercise. Anthropometric data were collected. PA was self-reported using the Recent and Global Physical Activity Questionnaires. Smartphone users shared their tracked step. Data collection was repeated one year after move-in. All data were analyzed using SPSS. Eighty-eight eligible participants completed the initial questionnaire (36 AD and 52 from 2 comparison buildings) at baseline (T0). There were no differences between AD and comparison cohorts in: stair use, PA, sitting time or, mean waist-to-hip ratio (WHR) at T0. However, the AD cohort had a lower baseline BMI (27.6 vs. 31.0, p = 0.019). At one-year follow-up (T1), 75 participants completed our survey including a 64% retention rate among those who previously completed the T0 questionnaire. Among T0 questionnaire respondents, mean daily steps increased at T1 among AD participants who moved from an elevator building (∆6782, p = 0.051) and in the comparison group (∆2960, p = 0.023). Aggregate moderate work-related activity was higher at T1 in the AD building (746 vs. 401, p = 0.031). AD building women reported more work-related PA overall but AD men engaged in more moderate recreational PA. Living in an AD building can enhance low-income residents’ PA. More research with objective measures is needed to identify strategies to sustain higher PA levels and overall health.


Background
Sedentary behavior and obesity have dramatically increased in the United States overtime [1]. In 2012, the prevalence of obesity among children and adults was 17% and 34.9% respectively in the United States [2]. Inadequate physical activity is a major contributor to the obesity epidemic which is associated with premature mortality and economic loss [3,4].
The U.S. Department of Health and Human Services issued the Physical Activity Guidelines for Americans in 2008 which suggest 60 min of physical activity a day for children and adolescents and at least 150 min of moderate or 75 min of vigorous-intensity exercise for adults that includes at least two or more days of muscle strengthening activities [5]. These guidelines have been shown to improve

Study Setting
A new affordable housing development comprised of five 6-story elevator buildings with 179 units across three adjacent blocks was identified for this study (AD). Occupancy began in April 2016. Active design features include accessible, prominent, and well-lit stairwells with locally-produced artwork; point-of-decision prompts at elevators demonstrating locations of stairs; proposed outdoor community garden areas; outdoor fitness areas at each building with playground facilities for children, and a community gym onsite. Eligibility for housing in the AD building was determined via income criteria based on apartment size and a housing lottery that was publicized to the community through multiple avenues including: newspapers, the internet, and telephone housing hotlines. Current neighborhood residents and civil servant applicants were given priority. For comparison, we identified two apartment complexes in low-income North and Central Brooklyn neighborhoods without the active design features (Non-AD).

Recruitment
Participants were recruited at lease signings in 2016 from AD and two recently renovated, non-AD buildings with a similar socioeconomic background through partnership with building management. Participants were eligible if they were at least 18 years old without contraindications to exercise, were non-pregnant by self-report (women), and were able to communicate in English. Participation in the study was not associated with the apartment rental process which was established prior to the start of the project. Anthropometric data including weight, body mass index (BMI) and waist to hip ratio (WHR) were collected from participants prior to move-in at baseline (T0) by trained research assistants. At T0, physical activity was measured using the Recent Physical Activity Questionnaire and the Global Physical Activity Questionnaire. Smartphone users shared data on the average number of daily steps and flights of stairs climbed over the previous one-month period. The questionnaires are available online as a supplement (Supplementary Tables S1 and S2). Repeat assessments were conducted after participants had resided in their apartments for 12 months (T1) from both the intervention (AD) and controls (Non-AD). T0 data were collected on the same day as study consent whenever possible. All participants signed informed consent forms. The research protocol was approved by the Institutional review board of Icahn School of Medicine at Mount Sinai (ethical code: HS#: 15-01159).

Data Analysis
All data analysis was conducted using SPSS software version 23 (IBM Corp, Armonk, NY, USA). Categorical data were analyzed using Chi-square and McNemar's tests. Continuous data were analyzed using T-tests and Mann Whitney U test as appropriate. Normality of the data distribution of individual outcome variable was assessed using the Shapiro-Wilk test to identify the appropriate statistical analytic method for each one.
T0 vs. T1 Comparison: Descriptive statistics were obtained for both AD and non-AD participants. Univariate analysis was conducted to assess the differences between the two groups and identify potential confounders at T0.
Individual level changes in anthropometric measures, were assessed at time T0 and T1 using paired T-tests for both groups. Data were stratified by whether the participants previously lived in walk-up building at the time of study enrollment. McNemar's and chi-square testing was used to assess for change in BMI category, and frequency of stair usage during the week and on weekends.
Cross Sectional Analyses at T1: Cross sectional, descriptive analyses were performed at T1 to assess group level differences of anthropometric measures and PA between residents of the AD and the non-AD buildings. The questionnaire responses and anthropomorphic data measures of both groups of participants were compared using t-tests for normally distributed continuous variables, Mann Whitney U tests for non-normally distributed continuous variables, and Chi-square tests for binary data.

Baseline Demographics
A total of 118 participants were screened for study inclusion at T0, 41 from AD and 67 from the non-AD buildings. Of these, 36 AD building and 52 non-AD participants were eligible, consented and completed the survey at T0. The remaining 5 AD and 15 non-AD individuals did not meet criteria for study enrollment. Reasons for exclusion from the study were: being non-English speaking, having mobility issues as identified through our screening questionnaire, and being visibly pregnant. The majority of enrolled participants were women (76.9% in AD vs. 80.6% non-AD, p = 0.864). AD residents were younger and had lower BMI at T0 compared to the non-AD group (34.9 years vs. 41.8 years, p = 0.027; 27.6 BMI vs. 31.0 BMI, p = 0.019). However, mean WHR was comparable in both groups at baseline (0.831 vs. 0.868, p = 0.083). Table 1 summarizes demographic data and the responses to the physical activity questionnaires at T0. Only about 30% of individuals in either group reported any work-related vigorous activity. About two thirds of AD participants reported any work-related moderate activity compared to 45% of non-AD participants. Similarly, the AD group was more likely to report recreational activity compared to the non-AD group, but the difference was not statistically significant. Generally, both groups had similar mean minutes of work and recreational PA, and sitting time. However, the non-AD group climbed more flights of stairs per day compared to AD (9.4 vs. 5.0, p = 0.032); and the AD group reported more minutes of walking or cycling during the week (717 vs. 450 p = 0.019).

One Year Follow-Up
Of those who completed the initial enrollment survey, 21 AD and 35 non-AD participants completed the one-year (T1) assessments which corresponds to 58% and 67% completion rates, respectively. Completion rates were similar in men and women. Of the 15 AD participants who were lost to follow-up, 6 moved away, 7 participants were unreachable, 1 was unable to complete survey due to work schedule conflict and 1 participant died prior to completing the follow-up. In the non-AD group, 17 participants were lost to follow-up. Two participants moved away and the others were unreachable using a combination of telephone, mail, and in-person efforts which were identical in both groups. Those who completed the T1 assessments were an average of 2 years older and had more daily steps and flights climbed at baseline than those who were lost to follow-up. More specific data are available in the Supplementary Materials (Table S3). However, there was no difference in PA questionnaire responses, BMI, or WHR between the two groups.

Individual Participant Level Analyses
Both AD and non-AD building participants reported an increase in mean daily steps, however the increase was greatest among the AD group residents who previously resided in an elevator building. (∆6782, p = 0.051), compared to the non-AD group (∆2960, p = 0.023) as outlined in Tables 2 and 3. There was also a statistically significant, greater decline in mean daily flights that was present in the AD residents with prior residence in an elevator building only. (∆−1.5, p = 0.05 vs. ∆−1.0, p = 0.46). There were no significant changes in self-reported recreational activity, time spent walking/cycling for travel, or sitting time, in either participant group, or within the sub-group analysis of AD building residents who previously resided in an elevator building. There was no difference in BMI category shift, or change in weekday and weekend stair use frequency between the two groups.
Work related moderate activity minutes per week declined from T0 to T1 in both groups, with a larger decline in the AD group (∆−705, p = 0.027; ∆−436, p = 0.016). Mean BMI increased in the AD building but not in the non-AD building (∆1.46, p = 0.007; ∆−0.17, p = 0.633). WHR remained stable in both groups at follow-up.

Secondary Cross-Sectional Responses
Seventy-five participants completed the survey at T1, 27 from AD and 48 from our non-AD buildings, including the 21 AD and 35 non-AD building residents who also completed the T0 questionnaire. Six AD and 13 non-AD building participants who were not enrolled at T0 also completed the T1 questionnaire. The majority of T1 survey respondents were women (81.5% in AD and 68.7% non-AD, p = 0.707). Among all participants who completed the T1 survey, BMI and WHR were comparable in the two groups (30.9 vs. 31.6, p = 0.671, 0.881 vs. 0.847, p = 0.244). The majority of participants in both groups lived above the ground floors of their building at the time of survey (74.1% and 81.3%). However, non-AD building residents were less likely to live on the ground floor (4.2% vs. 22.2%, p = 0.05). Data on floor of residence were unavailable for 3.7% (n = 1) of AD and 14.5% (n = 7) of non-AD participants. Table 4 summarizes the T1 demographic information, as well as BMI and WHR stratified by gender, age, and whether participants completed the survey at T0. Data on daily steps and flights of stairs climbed were also stratified by whether or not the residents lived on a ground floor apartment. Our data showed no differences in questionnaire responses between AD and the non-AD buildings. However, among newly enrolled T1 participants, WHR was higher in the AD group (0.93 vs. 0.81, p = 0.03).  Participants who lived in the non-AD buildings reported more daily steps taken and flights stairs climbed compared to AD residents (5483 vs. 9151, p = 0.001; 3.8 vs. 7.2, p = 0.022) as outlined in Table 5.
These trends persisted in our subgroup analyses of participants who were enrolled at T0 and new participants, as well as after stratification by apartment type. Interestingly, new non-AD building participants spent more than twice as much time sitting compared to their AD counterparts (275 min vs. 112 min, p = 0.022).
AD men reported more moderate recreational activity minutes (1115 vs. 131, p = 0.044), but there were no significant differences among women. On the contrary, women who resided in the AD building reported more work-related vigorous and moderate activity minutes (404 vs. 32, p = 0.01; 865 vs. 371, p = 0.015), but no differences were found among men. Aggregate moderate work-related activity minutes were also higher in the AD group than in the non-AD group among previously enrolled participants despite the individual level declines that were discussed earlier (746 vs. 401, p = 0.031).

Variable AD Building (n = 27) Non-AD Building (n = 48) p-Value
Daily Differences in questionnaire responses stratified by age category were analyzed. Among older participants (age ≥ 50 years), those in AD were generally more active than those in the non-AD building including a statistically significant difference in weekly recreational moderate activity (1034 vs. 72 min, p = 0.029). Younger non-AD participants (age < 50 years) reported more than twice as many daily flights climbed as those residing in AD (7.8 vs. 3.3, p = 0.017). There were no other between-group differences in this subpopulation.

Discussion
We found that residing in an active design (AD) building can positively impact residents' stair usage. Our findings are consistent with previous literature that AD design strategies that target stair use can positively impact behavior change [25]. One study found that residents of buildings with active design components were significantly more likely to have reported increased stair use and less likely to have reported no change or decreased stair use than residents who did not [21]. Overall, we found that residents in our non-AD buildings had higher levels of objectively measured PA which highlights the multifactorial nature of PA-related health behaviors. A study by Zimmerman et al. found that poorer self-rated health and factors like smoking were predictors of decreased PA [26]. It is possible that our non-AD building residents had higher self-rating of their health which affected their self-efficacy for using the stairs and being more active during the day. In addition, the non-AD building residents had a stronger sense of community and better relationship with building staff which may translate to increase likelihood of pursuing more PA in the neighborhood. This is consistent with previous literature showing that residents with smaller social networks were less physically active than those with larger social networks after adjusting for employment status, gender, poverty level, current health status, age, and perceived neighborhood safety [27]. Although the overall work-associated and recreational PA were higher in the AD group at baseline and at T1, the differences were not statistically significant. Also, these data were self-reported which is an important limitation. Research shows that individuals tend to overestimate their PA which can contribute to the energy imbalance that leads to weight gain [28].
Women who lived in the AD building at T1 reported more moderate and vigorous work-related physical activity but the same trend was not noted for recreational physical activity. One possible explanation is that people who are more active at work may be less willing to engage in recreational physical activity. A 2004 study found that engaging in high levels of occupational PA was associated with more transport activity but less leisure time PA [29]. Furthermore, we did not assess participants' occupation or characteristics of the work environment. It is possible that the participants' work environments or neighborhoods of employment were more supportive of physical activity behaviors. This is consistent with literature suggesting that workplace built environment features such as presence of shops, bike and recreation facilities, and reduced crime rate can increase workplace PA [30].
Based on our data on gender differences in physical activity, AD women were more likely to be active compared to women in the non-AD building. However, those differences were not observed among men. Previous research has shown women report lower levels of physical activity than men, but the differences may be due to the types of activities they participate in rather than the true differences in activity levels [31]. Household work, child care, and other caregiving physical activity are reported more frequently by women compared to men, who report more sports and leisure-time exercise [31]. Predictors of physical activity also vary among women and include age, social roles, social support, and structural/environmental variables. Women in the non-AD building may have more social support and a closer community, which may serve as a predictor of exercise levels. Lack of information regarding social support and other socio-economic factors is a limitation in this study.
This study showed younger non-AD participants were more likely to use the stairs than younger AD participants, but older AD participants reported more recreational moderate physical activity than older non-AD participants. This is similar to what has been seen in previous research, where activities that require a higher exertion level decline consistently as individuals age; and relatively stable physical activity patterns appear in middle to late adulthood [32]. Older adults may take part in physical activity because of more concern for physical and psychological health issues, compared to younger participants who participate in sports for play [33]. Results from this study regarding age differences are not surprising, but lack of qualitative data is a limitation. Future studies should consider collecting qualitative data from younger and older participants regarding reasons for physical activity or lack thereof.
We found an increased BMI among AD residents at T1 and stable BMI among the non-AD building residents, as well as stable WHR in both groups during the same time period. These results are consistent with previous literature suggesting that BMI is impacted by factors other than physical activity. A 2006 Cochrane review showed that physical activity combined with diet resulted in a greater weight reduction than diet alone, with higher intensity resulting in a larger magnitude of weight loss [34]. However, we did not assess diet quality or changes in dietary patterns during our study which may have impacted our BMI and WHR findings.
Our study does have some limitations, such as the use of a non-random sample of participants which may have contributed to our study population being predominantly women with a noticeable age difference between the two groups at baseline. The lottery process used to identify potential residents for AD could have selected for individuals seeking an AD apartment resulting in a younger pool of applicants who were interested in the types of amenities that were offered. Stable employment was another criterion used for AD applicant screening that may have contributed to study bias. Though we were able to control for seasonal variation in PA by ensuring that our T1 follow-up data was collected during the same time of year as baseline, we were unable to account for differences in weather patterns that may have impacted participants' self-reported PA within the past week which was the time frame of the PA questions. Our follow-up surveys were scheduled at the convenience of building staff based on their availability to grant us access. As a result, most of our follow-ups were conducted in the late afternoons and early evening on weekdays. This may have prevented some working individuals from being able to complete our one-year assessments and contribute to our loss to follow-up, since the assessments needed to be conducted in person. Individuals who are employed regularly tend to be healthier than those who are not [35]. We hypothesize that since the AD building participants were younger, differential loss to follow-up due to lack of availability and competing priorities may have resulted in an underestimate of AD benefits. Our study sample was small due to recruitment challenges associated with varying level of participation interest based on competing priorities of both the real estate management and potential tenants. We acknowledge the skewed distribution of our PA outcome variables which could be addressed in follow-up studies by increasing the sample size and using additional objective measures to PA to reduce reporting bias. We also did not assess other socioeconomic variables that may also contribute to physical inactivity. Nevertheless, our study adds to the growing body of literature using community-based data collection methods which is important for assessing the impact of health interventions such as AD housing in a real-world setting.

Conclusions
Many low-income and minority families live in neighborhoods that are not designed to support physical activity and healthy eating, which contributes to these populations being disproportionately affected by chronic disease. Our finding that moving from a non-AD elevator building to an AD elevator building can positively impact daily steps demonstrates the potential of affordable housing environments to improve health among those who are most affected by health disparities related to exposure to an unhealthy environment. Broad adoption of affordable housing AD strategies is warranted to encourage behavior change that improves physical activity measures and is consistent with the Community Preventive Services Task Force recommendations for community engagement around this important lifestyle modification [36]. Additional research to identify health behavior and community-based interventions that can be implemented among affordable housing residents to supplement active design housing features is also important for informing stakeholders and public health advocates looking to promote health outcomes for neighborhoods in need.
Author Contributions: C.T. prepared the original manuscript draft, developed the analysis plan methodology, conducted the formal analysis, and was involved in project administration and development as well as data investigation. A.S. served as part of the data investigation team, contributed sections to the original draft, as well as in the review and editing process. M.G. contributed to data visualization, the analysis plan methodology and the manuscript editing process. D.P. was involved in the research project administration, data investigation, data curation, and the manuscript editing. S.R. was involved in the research project administration and provided critical feedback during the manuscript editing process. E.G. served as the senior principal investigator and was responsible for the project's conceptualization, overall supervision, funding acquisition, and manuscript revision.
Funding: This project was funded by the New York State Health Foundation, grant number 16-04051; and the National Heart, Lung, and Blood Institute Short-term Training Program, grant number 5R25HL108857.