An online tool for obesity intervention and public health

Fruit and vegetables have a low energy density because of their high content of water, low content of energy [19] and they have high content of dietary fiber, which is considered to increase satiety and reduce feelings of hunger [20]. Fruit and vegetables also contain flavonoids, a group of nonnutritive phytochemicals that may have antiobesity effects [21, 22]. High fruit and vegetable intakes were shown in some studies inversely associated with weight gain [22] and might lead to eating fewer high-fat foods [23]. To serve the purpose of possible weight gain intervention, we geocoded fruit & vegetable stores and farmers’ markets for the Los Angeles Metropolitan Area, including the Los Angeles County, western part of the San Bernardino County and the northern part of the Orange County. Data were acquired from the Environmental Systems Research Institute (ESRI, Redlands, CA) and the California Department of Public Health. Fruit basket markers are used to represent fruit & vegetable stores and each popup window shows the street address and contact information of a store (Fig. 2, left). For farmers’ markets (strawberry markers in Fig. 2), additional information includes time of opening and if available a link to the website of a farmers’ market. Figure 2 on the left shows the locations of fruit & vegetable stores and farmers’ markets within 5 miles of the Kaiser Permanente Bellflower Medical Center. In addition to fruit & vegetation stores and farmers’ markets, grocery stores including supermarkets are also considered as important contributors to nutrition among residents of neighborhoods [24]. We also included grocery stores in the region (Fig. 2 right) for the online obesity intervention program, including convenient stores, cooperative food stores, drug stores, grocery stores, supermarkets and warehouse club stores with annual sales ranging from less than $50,000 to over 1 million dollars. Similar to fruit & vegetable stores and farmers’ markets, a grocery popup window includes its street address and contact information. To show on a map the healthy food stores within a distance of location of interest, data are extracted from an Access database using Asynchronous JavaScript and eXtensible Markup Language (AJAX) techniques and web services in C#.

One way to help address the epidemic of obesity in the United States is an ambitious parks and recreation program [25]. Parks and green spaces provide many environmental, social and psychological services that are of significance for the livability of modern cities and the wellbeing of urban dwellers [26, 27]. In addition, parks and green space often serve as sites of physical activity and may be an effective health promotion target for increasing active living in youth [28]. Wolch et al. [29] found that young children with better access to public recreational programs are less likely to become obese. Though some studies showed that reducing dietary energy density or increasing physical activity alone did not result in a weight loss or obesity rates [30‐33], research found a significant greater survival rate for breast cancer participants who followed a healthy lifestyle that included both recommended intakes of vegetables-fruits and moderate levels of physical activity [34].

The strong protective effect of interaction led us to test if in addition to dietary interventions, integrating public parks and recreational services at the community neighborhoods would help prevent childhood obesity. The public parks used in this study were created by pooling together data from the following sources: ESRI’s Business Analyst, land use/land cover data from the Southern California Association of Governments (SCAG, Los Angeles, CA), coastal access information from the California Coastal Commission (Long Beach, CA), and Thomas Brothers Maps (Rand McNally, Irvine, CA), with the latter used mainly for cross-referencing and verification. Recreational programs included both general and detailed categories. The programs in the general category was derived from the ESRI’s Business Analyst and it includes the name of a recreational program, the types of programs available, and the city it belongs to (Fig. 3 left). Detailed recreational programming data (Fig. 3 right) were collected by means of an audit of public recreation courses performed during the summer 2006. Recreational offerings were gathered from municipal park and recreation brochures and materials. Most data were collected via internet search (though some direct telephone contact/mailings were necessary). Each listing for sports, fitness, or other recreational programs that required a metabolic equivalent (MET) value greater than 3.0 (physical exertion equivalent to moderate walking) was included in the database. All courses had to be sponsored by the city, but could occur at either a city park or community recreation center, or a non-city owned site. Addresses were geocoded in ESRI ArcView 9.2 and verified with Teleatlas.

A participant has options to select amenities of specified distance around his/her home or at a place of his/her interest as shown in Figs. 2 and 3. A participant can obtain detailed information for each amenity by left-clicking the amenity for a popup window.

Environmental changes and heat waves have significant effects on plants, communities and ecosystems including human being [35, 36]. At the same time, the presence of vegetation provides shading and converts incident solar radiation to latent heat [37, 38]. Normalized difference vegetation greenness (NDVI) is used to represent vegetation greenness. Live green plants absorb solar radiation in the photosynthetically active radiation (PAR) spectral region and scatter (i.e., reflect and transmit) solar radiation in the near-infrared spectral region. Live green plants appear relatively dark in the PAR and relatively bright in the near-infrared. The NDVI is calculated as follows:

where RED and NIR stand for the spectral reflectance measurements acquired in the red and near-infrared regions, respectively. These spectral reflectance take on values between 0.0 and 1.0 and NDVI itself varies between −1.0 and +1.0. A higher NDVI index represents a greater vegetation greenness. The NIR and RED data for the study region were acquired from the US Geological Survey Landsat Enhanced Thematic Mapper Plus (ETM+) for 2001. Derived NDVI surface (Fig. 4a) were re-coded from low to high on the online intervention legend (Fig. 5a).

Increasingly, air pollution is found associated with disease and mortality [39‐42]. Evidence is starting to show that air pollution exaggerates adipose inflammation, insulin resistance and other symptoms to people with diabetes, obesity and hypertension [43‐46]. To help participants aware of the levels of air pollution in areas of their interest and if possible to avoid high polluted areas, we modeled traffic related air pollution (the main source of air pollution in the region) for the study region (Fig. 4b) using a land use regression modeling strategy [47]. The land use regression method seeks to predict pollution concentrations at a given site based on surrounding land use, road network, traffic, physical environment and population characteristics using a series of buffers [48]. Similar to vegetation greenness, we recoded the pollution values from low to high on the online intervention legend (Fig. 5a).

To speed up the display of the vegetation greenness surface and the air pollution surface, each surface is dissected into 80 rectangular segments. Each time only the selected image that intersects the scope of display will be called from the Access database and be overlaid on top of the current map. For safety awareness, we also incorporated traffic accident data that involved at least a bicyclist or a pedestrian from year 2005 to 2007. The data were acquired from the Statewide Integrated Traffic Records System (SWITRS) by the California Highway Patrol. Data were gecoded and stored in an Access database. Similar to the healthy food amenities, AJAX and web services were used to display locations and counts of accidents on the Google map. Other amenities selected for the online intervention tool included locations of restaurants (external link), schools and educational institutions, and licensed healthcare facilities (Fig. 5a).

To enable a layer to be displayed on the Google map, a user needs to select types of amenity and then click the “Update Selection” button. Because of overlapping issues associated with public parks, schools and educational institutions, vegetation greenness and traffic-related air pollution, only one of the four layers was allowed to be shown at a time. A user can also check with amenities within a specified walking, bicycling or driving distance by specifying travel speed and time spent in travel or simply by specifying a distance and then click the “Search Amenities Within” button.

The findings of the Cycling in Cities survey [49] and the growing number of general online route planners (e.g., Google Maps, Google Transit, Microsoft Live Search, MapQuest and Yahoo Maps) suggest an increasing demand for spatial decision tools for everyday transport. Some research [49, 50] suggests that bicyclists would use bicycles for more trips if they had trip information ahead of time. The Google-based bicycling trip planner developed for Metro Vancouver [51] allows individuals to choose routes based on user-stated preferences regarding safety, distance, elevation gain, air quality, and areas featuring trees and other vegetation. The trip planning tool for the OPT Health II is a tailored and enhanced tool that includes features not only identifying locations of general public’s interests, but also seeking locations of the healthy food stores and recreational program amenities in the region. To assist a participant in getting to an amenity or go to place after visiting an amenity, the “To Here” and “From Here” options are provided on the amenities popup windows (Figs. 2 and 3). To help participants and their communities achieve a significant shift from environmentally harmful and sedentary travel to environmentally friendly and active travel, this intervention tool also advocates for transportation that encourages and promotes safety, physical activity, health, recreation, and resource conservation. The optimized trip planner includes options not only for bicycling, but also for walking and public transit for the purpose of increasing active transport and reducing dependence on automobiles. In case a participant can only drive by car from or to an amenity, car driving is also provided as an option. To promote travel safety, traffic accidents involving at least a bicyclist or a pedestrian are included in the trip planning options. To maximize a participant’s participation in active transport, the tool also provides flexible weighting options for walking and bicycling. A participant can weight a bicycling choice using a scale from 1 to 10 for roadways to ride on, including bicycling designated, bicycling alternative, connecting roadway, local road and highways. For walking, the weighing include options for connecting road, local road and highways, also at a scale 1–10.

The details of the trip planning component are shown in Fig. 5b. In summary, there are four modes of transport: travel by car driving, bicycling, walking and public transit. For bicycling, a participant can further select using either the most bicycle friendly lanes or balanced routes. The most bicycle friendly option put those designated and high quality bicycle roadways at a higher priority compared to the balanced option. A participant can also specify travel speed and limit maximum slope gradient for both bicycling and walking options. The slope gradient is defined as the ratio of the altitude change to the horizontal distance between any two intersections of a road segment. When the ratio equals to 1 (i.e. 100 %), the slope in degree is 45°. Under each mode of transport, a participant can also select environmental preference for a route, including no restriction, avoid elevation gain, prefer low traffic pollution, prefer vegetated route and avoid sites of previous traffic accidents. Elevation gain is the total elevation increased during the whole trip. Traffic accidents are the total number of accidents with at least a pedestrian or a bicyclist during the last 3 years (2005–2007) on a chosen trip. No route length is taken into consideration for elevation gain and traffic accident preferences. For traffic pollution and vegetation greenness preferences; however, road length is considered. For example, total traffic-related air pollution is calculated based on the possible total inhalation on a chosen route, i.e. pollution = sum(road length_i * mean pollution on road segment_i). Accidents involving a pedestrian or a bicyclist are provided only as a guidance for bicycling and walking safety.

For the trip planner component, asynchronous web services were developed using C#. One of the oldest and most widely used approaches in network optimization is the shortest path analysis (SPA) [52‐54]. In our application, we are interested in searching for the shortest path with cost impedances for attributes based on the user-defined preferences. Many optimization problems are solved using a matrix of dimension n by n (e.g., [52] by various constraints [55‐57]). These algorithms usually search more neighboring nodes or roads than those physically adjacent to the node of interest. An alternative to SPA that saves processing time is the use of topology. Since networks usually have a topological structure that identifies how the various features relate to one another (e.g., position, orientation, adjacency and connectivity), Su et al. [51] used the topological relationship of nodes and roadways and efficiently stored them in an index table and loaded them into memory when a client launches the bicycling trip planner. This works effectively (3–4 s to calculate a route) for the Greater Vancouver area in which 66,000 vertices (used for color-coded route output) and 22,500 roadway segments were used for the analysis; however, when applied to the Los Angeles Metropolitan Area, the index table was seen very inefficient. It took about 3 min to get a route for the 648,510 roadways and 2,025,053 vertices. To speed up time used for trip planning, a rectangular box with a half mile buffer distance around the start and destination addresses a participant specified is used for initiating the routing process. This algorithm significantly reduces the number of roadways needed in the computing process and thus significantly reduced the time required for calculating a route. Up to date, almost all the routing algorithms including those topology-based ones use nodes (i.e., road intersections) to guide a route selection process. The disadvantage of this method is its reliance on two separate tables: nodes and links (i.e., roadways) [51]. We further simplified the algorithm by using links as the basic selection process. In simple terms, the topology-based optimal route algorithm first identifies the starting roadway of a trip and marks it. All the physically adjacent roadways of unmarked within the buffered rectangular box a participant can travel to are then marked, and their impedances calculated and added to the corresponding total impedance. The marked roadway with the least total impedance to the original starting roadway is identified and used as a new starting roadway to find its physically adjacent and unmarked travelable roadways within the rectangular box. This process continues until the newly specified start roadway meets the destination roadway (i.e., a route is found) or until the two never merge after assessing all the travelable roadways within the rectangular box (no direct route). The link based algorithm does not require the usage of nodes and thus significantly improves the performance of the trip planner.

In the trip planner output window, carbon dioxide (CO₂) emissions from driving by car and “savings” from cycling and walking are calculated and displayed. We assume that a typical passenger car emits 5 tons of CO₂ from an annual travel distance of 20,000 km, which is equivalent to 0.25 kg km⁻¹ travelled [51]. The estimated direct CO₂ emissions or savings (in kg) during a bicycling/walking trip are therefore equal to the distance traveled in kilometer multiplied by 0.25 kg km⁻¹. In energy expenditure, on average a person uses 35 Cal when traveling on a bicycle for a mile [51]. The calories burned during a bicycle trip are estimated as the distance traveled in km multiplied by 21.75 Cal km⁻¹ (the metric equivalent). Similarly, energy expenditure on walking is based on bodyweight in pounds × 0.853 × distance in kilometers, assuming an average body weight of 150 lbs. The outputs also include estimated trip time, route length, total elevation gain, mean traffic pollution concentrations and degree of vegetation greenness of the chosen trip. Besides, an optimized route is linked to the slope gradient data and each roadway is color coded accordingly. Roadway by roadway directions and road length are displayed on the right side of the interface below the “Suggested Route”.

The associated software can be downloaded from https://​drive.​google.​com/​uc?​export=​download&​id=​0B9v9GRtvwvfoUDF​PeGZ2SVNfUnc.