Health Impacts of Increased Physical Activity from Changes in Transportation Infrastructure: Quantitative Estimates for Three Communities

Recently, two quantitative tools have emerged for predicting the health impacts of projects that change population physical activity: the Health Economic Assessment Tool (HEAT) and Dynamic Modeling for Health Impact Assessment (DYNAMO-HIA). HEAT has been used to support health impact assessments of transportation infrastructure projects, but DYNAMO-HIA has not been previously employed for this purpose nor have the two tools been compared. To demonstrate the use of DYNAMO-HIA for supporting health impact assessments of transportation infrastructure projects, we employed the model in three communities (urban, suburban, and rural) in North Carolina. We also compared DYNAMO-HIA and HEAT predictions in the urban community. Using DYNAMO-HIA, we estimated benefit-cost ratios of 20.2 (95% C.I.: 8.7–30.6), 0.6 (0.3–0.9), and 4.7 (2.1–7.1) for the urban, suburban, and rural projects, respectively. For a 40-year time period, the HEAT predictions of deaths avoided by the urban infrastructure project were three times as high as DYNAMO-HIA's predictions due to HEAT's inability to account for changing population health characteristics over time. Quantitative health impact assessment coupled with economic valuation is a powerful tool for integrating health considerations into transportation decision-making. However, to avoid overestimating benefits, such quantitative HIAs should use dynamic, rather than static, approaches.


Greenville MPO Bicycle and Pedestrian Master Plan, Winterville, NC
In 2011, the Greenville MPO completed a Bicycle and Pedestrian Master Plan for the Greenville Metropolitan Area, which includes Winterville. We consider the impact of building out the pedestrian network as specified in the plan compared to a no-build scenario ( Figure S1).

Downtown Streetscape Master Plan, Sparta, NC
In 2012, the town of Sparta, NC completed a Downtown Streetscape Strategy, which proposes a number of improvements to the pedestrian environment in downtown. We conducted an HIA on the implementation of the plan and compared the results to the status quo scenario. The project contains streetscape and street crossing improvements along Main Street, which runs through downtown Sparta, as well as complementary improvements to several side streets ( Figure 3). ! Figure S3. Sparta proposed downtown streetscape improvements

Community Context
Descriptive statistics for each case study location is summarized in Table S1. Summary information for meetings held in each community are presented in Table S2 (scoping meetings) and Table S3 (post-analysis meetings). Age-and sex-specific population distributions for each community are provided in Figure S4.

S.2 Baseline Health Information
Additional details are presented below regarding our procedure to estimate continuous disease prevalence and incidence functions for CHD, diabetes, hypertension, and stroke as a function of age in each case study location (Table S4). Detailed vital statistics (baseline death rate, birthrate, and gender ratio) are presented in Table S5.

S.2.1 Disease Prevalence and Incidence Functions
To develop continuous age-and sex-specific prevalence functions for CHD, diabetes, hypertension, and stroke, we use data from the 2009 North Carolina BRFSS survey. The survey asks whether or not a respondent has been diagnosed with these conditions and reports prevalence by age group. In each community, we fit a second-order function to these data assuming that the prevalence reported for each age group represented the actual prevalence of that disease at the population-weighted midpoint of the age group. Using these prevalence

Age, Years
Males Females estimates, we then derive the age-specific rate at which individuals would have had to develop a disease in order for the observed prevalence to occur. To do so, we define second-order agespecific prevalence functions, p(x), and take the derivative: x = age (years) α = derived parameter for second-order term β = derived parameter for first-order term γ = derived constant And define c(x): c(x) = number of cases at age x ! And define the incidence function, i(x): i(x) = Incidence rate at age x m(x) = All-cause mortality at age x R(x) = Relative risk of all-cause mortality associated with the disease for which incidence is being derived at age x Estimated disease prevalence and incident functions are presented in Table S4.

S.3 Baseline Transportation Behavior
In Winterville and Sparta, we use data from the 2009 BRFSS survey. In 2009, North Carolina included an additional question regarding walking for transportation. Specifically, the survey asked "In the past week, how much time did you walk or bicycle for transportation, such as to and from work or shopping, or walk to the bus stop?" Respondents replied in one of five categories: No time, Less than 30 minutes, 30 minutes to 1 hour, 1 to 2 hours, or 2 hours or more. 34 In Winterville, we use county-level data (Pitt County) whereas in Sparta we use data aggregated across the Northwest Area Health Education Center (HEC), a ten-country area (Alleghany, Ashe, Davie, Davidson, Forsyth, Stokes, Surry, Watauga, Wilkes, and Yadkin counties). In BRRC, we use data from a survey conducted in 2012 by MacDonald Gibson et al.
The survey used the International Physical Activity questionnaire, a previously validated survey instrument. 37 The survey asked two questions from which estimates of weekly walking for transportation were derived: "During the last 7 days, on how many days did you walk for at least 10 minutes at a time to go from place to place?" immediately followed by "How much time did you usually spend on one of those days walking from place to place?" These estimates were then used to develop a distribution of walking for transportation time by placing each in one of 20 transportation physical activity time bins to: one for no walking, a series of twenty-minute bins up to 360 minutes per week (i.e., 0-20 minutes, 20-40 minutes, etc.), and a top bin for greater than 360 minutes per week. 36 Survey characteristics are summarized in Table S6.

S.4 Economic Valuations
To account for uncertainty inherent in selecting an appropriate discount rate, we consider three discount rates: 7%, 5%, and 3.5%. Benefit-cost ratios for the central estimate of health outcomes for each case study location at each of these three discount rates are plotted in Figure S2.  Value of avoided mortality Value of avoided disease