Journal Logo

New Methods for Assessing and Modeling Walking Behavior

New Techniques and Issues in Assessing Walking Behavior and Its Contexts

FREEDSON, PATTY S.1; BRENDLEY, KEITH2; AINSWORTH, BARBARA E.3; KOHL, HAROLD W. III4; LESLIE, EVA5; OWEN, NEVILLE6

Author Information
Medicine & Science in Sports & Exercise: July 2008 - Volume 40 - Issue 7 - p S574-S583
doi: 10.1249/MSS.0b013e31817c71e7
  • Free

Abstract

The purpose of this article was to present provocative and innovative ideas for addressing assessment and research issues associated with walking. Our challenge was to examine metabolic features of walking and to explore factors characterizing and assessing walking behavior and relate these to the context and domain of walking. In addition to considering traditional physiologic domains of walking (energy expenditure, time, intensity, and duration), we have addressed less conventional issues such as assessment or measurement of walking behavior in the context of the built environment and the role of particular physical attributes of urban environments and how they may influence walking. Although the topics included in this article are broad, we believe that blurring of traditional boundaries may help to stimulate efforts for interdisciplinary research related to determinants of walking behavior and how walking relates to health.

METABOLIC ISSUES IN ASSESSING WALKING BEHAVIOR

The metabolic responses to walking have been studied extensively. One of the major conclusions about the metabolic energetics of walking is that, in a natural setting, individuals self-select a preferred walking speed where the energy cost per unit distance is minimized (21) and energy efficiency is maintained at approximately 25%. During walking at a preferred pace, arm, leg, and trunk motions are coordinated in such a manner that keeps vertical displacement relatively small, thereby maximizing metabolic economy (39). In contrast, when individuals are forced to walk at slower and faster speeds than preferred, energy efficiency is reduced to 15% (39).

Energy Cost of Walking

The predominant source for adenosine triphosphate during walking is derived from the aerobic energy system. Therefore, examination of the oxygen consumption responses to walking provides a relatively accurate picture of the energy cost that is involved. Studies have examined the energy cost of walking during controlled treadmill exercise and during over-ground walking protocols. In general, the energy cost of treadmill and over-ground walking on a firm surface is similar. Hall et al. (18) reported that the net caloric expenditures for walking 1600 m on a track and treadmill were similar (43 and 45 kcal, respectively). Figure 1 presents data from Bassett et al. (5) and Hendelman et al. (20) for over-ground walking over a wide range of walking speeds where oxygen consumption was measured. Additional data points from the compendium of physical activities (2) are included. The pattern for oxygen consumption (V˙O2) is linear between 80 and 100 m·min−1 with curvilinear components evident at slow and fast walking speeds. The MET level is considered moderate intensity (3-6 METs) at walking speeds between 80 and 115 m·min−1. Durnin and Passmore (12) reported that for a 73-kg person, energy expenditure ranges from 3.2 to 5.8 kcal·min−1 at speeds from 3.22 to 6.44 km·h−1. This translates to an approximate equivalent energy cost per unit distance walked, because the caloric expenditures to walk a mile at 3.22 and 6.44 km·h−1 are 96 and 87 kcal, respectively.

FIGURE 1
FIGURE 1:
METs at various walking speeds.

Effects of gender and weight status.

A study conducted at the University of Massachusetts examined how gender and weight status affected the energy cost of treadmill walking at 4.02, 4.83, and 5.63 km·h−1 in 22 men and 25 women (12 min of walking for each condition). We observed a significant gender effect and a significant interaction. At the slowest speed, there was no gender difference, but as the speed increased, the relative energy cost (kcal·kg−1) became 8% to 11% higher for women. In this group of subjects, 35 were classified as normal weight (BMI < 25) and 12 were classified as obese (BMI > 29.9). At all speeds, the obese subjects exhibited a 5% lower relative energy cost(kcal·kg−1) in comparison to the normal-weight subjects. The causes for these differences are not known, but these data indicate that men and obese subjects are more metabolically economical than women and normal-weight individuals. Some of the possible factors that might explain these results include biomechanical factors and differences in substrate use.

Energy cost of walking in children.

There is a lower metabolic economy during walking among children (40). Sallis et al. (43) compiled locomotion economy data from several studies for children and adults and reported economy differences between children and adults. For example, at the age of 5 yr, walking economy is 37% lower, and at the age of 7 yr, economy is 19% lower than adults. As children grow and develop, the difference continues to decline, so that by the age of 17 yr, walking economy is equivalent to that seen in adults. Several factors have been suggested to contribute to these differences in economy: a higher resting V˙O2, a higher oxygen cost of breathing (higher frequency of breaths in children), motor skill development, and biomechanical factors associated with their smaller body size. It is likely that body size is a major contributing factor, because it has been shown that when speed is set according to leg length, the child-adult difference in economy is eliminated (31).

Summary

Understanding the metabolic characteristics of walking has substantial value both from a practical and theoretical perspective. Understanding how metabolic responses to walking are affected by speed, gender, weight status, and age is important information for clinical applications in exercise prescription and in developing a sound knowledge base on which to build physical activity recommendations for public health. The underlying mechanisms related to the minimization of energy cost of walking remains a subject of debate and investigation in an effort to improve our understanding of control and regulation systems.

MEASURING NONMOTORIZED TRAVEL WITH A FOCUS ON WALKING

Background.

The term "transportation engineer" is understood to mean motorized transportation and the road network that supports it. Other forms of travel are covered by fields such as railroad engineers, aeronautical engineers, naval engineers, and even aerospace engineers. This would seem to cover all the possibilities from space to deep underground, and it does-at least for mechanized forms of travel. Curiously, the oldest form of transportation, walking, is hardly noticed by any field of engineering. You will not find a single class on the topic in any engineering school, at least outside the fields of architecture and urban planning. The current state of affairs is perhaps best summarized by Frank and Engelke (14).

"Though motorized travel has been the subject of much research, nonmotorized travel has not. This disparity reflects a research and cultural bias that conceptualizes travel as an automobile-dependent phenomenon. Much of the work in transportation focuses on congestion and emissions reductions. The resulting data collection regime has therefore generated much information on automobile transportation and relatively little on nonmotorized modes."

Yet, walking is a fundamental mode of getting from one place to another, and for the ambulatory, every trip begins and ends with a walk. There is clearly a renewed focus on walking within the health policy and urban planning communities, with the former focused on its positive health benefits and the latter for its potential for reducing traffic congestion. Both of these communities are hampered by how little is really known about walking patterns and behaviors, as they are used for transportation. This section addresses technologies that could be applied to this problem.

Who cares about human motion?

Outside engineering, there are many fields interested in human motion to include walking, with a few of these listed in Table 1. Of these, the fields from which one could expect to find technologies of most interest would be defense, security, and health policy, the former two because of the size of investment and the latter because it is somewhat more applicable to the nonmotorized transportation problem.

TABLE 1
TABLE 1:
Motion measurement across different fields of study.

Whereas many technologies within the defense and security, fields detect and track people, few are affordable for health policy studies, are as accurate as is required, or are amenable for use in crowded urban environments. The basic incompatibility is that military systems are mostly concerned with classifying detections of walking into broad groups (individual, team, squad, platoon, etc.) and then tracking their direction of motion over the widest possible area. Transportation systems, on the other hand, typically count objects of interest moving through a specified area, such as an intersection, with high accuracy. There are some promising technologies being developed in areas such as urban surveillance and networked sensors that may become available in the future.

Emerging Technologies for Characterizing Nonmotorized Transportation

For most applications, one can assume that data must be collected for at least a month and perhaps for as long as a year if seasonal variations are of interest. Requirements defining the data collection systems include the following: 1) remote sensors must be easy to set up and require minimal maintenance; 2) personnel sensors must be inexpensive and easily worn; 3) devices, if battery-powered, must have sufficient duration to support the experiment; and 4) any individual apparatus must be low in cost. In addition, validation data must be available at selected sites to ensure data integrity.

The apparatus to be used in characterizing nonmotorized transportation would fall into three categories plus monitoring of the environment (primarily the weather but may also include sidewalk conditions, such as icing, etc.). Tracking of individual movements would include radiofrequency (RF) identification tags (RFID tags), RF tracking, and global positioning system (GPS) tracking. Monitoring of important control spaces would be accomplished with beam breakers and a newer technology called Smartmat that is essentially a carpet of sensors. A few of the more novel technologies envisioned for use are described as follows.

Smartmat.

The Smartmat Area Monitoring and Analysis System is a tool for measuring physical activity levels in a defined area. The system, which resembles an industrial carpet in appearance, counts pedestrians and other objects passing across it; notes their direction, time; and speed; computes their stride length; and estimates their weight. The system works with groups of people, and it differentiates between walkers, cyclists, animals, and other sources of error. Smartmat is portable, affordable, rugged, quick to set up, easy to use, works in- or outdoors, and operates for long periods. As with all data collection apparatus, the generated data must be stored, processed, and managed using tools such as geographic information systems (GIS) and commercial databases.

RF tracking.

GPS is already being used within the health policy community. It presents several problems. First, it consumes much power, thereby requiring either sparse sample rates or large batteries. Second, its contact with satellites is tenuous in urban settings and largely nonexistent within buildings. Not only does this make it difficult to track people in cities but also that part of the solution for increasing reliability is to put the GPS antenna as high as possible on the body, typically on the shoulder, thus making the device inconvenient. Finally, GPS devices are expensive. A lower cost and more reliable approach would be to use a miniature RF transmitter and dispersed RF direction finding receiver stations. This provides sufficient accuracy (approximately 30 m depending on receiver station placement) and requires only that subjects carry a pen-sized transmission device. The downside of this approach is that receiver stations must be established, but these should not be overly large and can even be moved from place to place to permit the sample area to be shifted over time.

RFID tracking.

RFID tags are a relatively new technology that have gained rapid acceptance in the retail community for tracking packages. It is envisioned that RFID tags will some day largely replace the ubiquitous bar code for identifying individual products on shelves. It seems plausible that this technology could be applied to tracking people. The advantage of this approach is that the tag itself is very low in cost, very small, and has a duration measured in months if not years, thus enabling very large samples of the population to be tracked in space for long periods and with no inconvenience to the user. In addition, tag reader density may be tailored to monitor specific patterns of movement, and the readers can be shifted to sample different areas at different times, all having no effect to the readers themselves.

Data analysis.

Analysis of geospatial data has long been the domain of GIS. Although these systems excel in associating data with locations and permitting researchers to process data across areas in different ways, they are poor tools for estimating how objects move through space and interact with their environment. Because the scope of data collected to assess nonmotorized transportation data is broad and may include pedometer logs, survey results, RFID samples, and data from a GPS or RF tracker, a modeling analytic approach is recommended. Because one can envision that many different models will fit the data, the research will entail finding the spread of models and thus the range of plausible patterns described by the data. This sort of analysis could be accomplished using a specialized type of simulation called an agent-based model (ABM). The advantage of an ABM over other approaches is that many different models corresponding to the data may be generated automatically and varied parametrically to test model resilience (see the paper by King et al. in this supplement for more information on ABM).

Summary.

Several approaches for measuring and analyzing nonmotorized travel were presented. The more novel technologies discussed included RFID tags, RF direction finding, and an electronic carpet called Smartmat. A means for analyzing very large and disparate data sets using an ABM simulation was introduced. In conclusion, although engineers have largely ignored the problem of measuring walking and other forms of nonmotorized travel, there are ample technologies that have yet to be exploited, which could provide researchers in the field with much better characterizations of these behaviors.

SELF-REPORT MEASURES OF WALKING BEHAVIORS

Self-report measures used to assess walking behaviors include questionnaires, logs, and diaries. Information obtained from these instruments includes the frequency, duration, intensity, and context for walking. The context for walking includes walking for leisure and exercise, transportation, and walking in occupation and in household settings. Walking may be deemed as social, as in walking with friends or with family members, or for specific purposes, such as walking the dog, walking while shopping or to travel between two places.

Because of their simplicity, questionnaires are used most often in epidemiologic and clinical studies to determine the associations between physical activity and health outcomes and in public health surveillance settings to assess the prevalence of walking. Questionnaire items may range from a single context (e.g., walking for exercise) to combined contexts within a single question (e.g., walking for exercise, or walking to and from work, or walking at work). Questionnaire items may also have nonspecific contexts, such as "how many city blocks do you walk each day?" Most questionnaire items assess the frequency and duration of walking. Some also identify the pace or intensity of walking. Logs and diaries provide detailed information about the frequency, duration, intensity, and context of walking behaviors and are often used to establish content validity of physical activity questionnaires. Because logs and diaries are more detailed and time-consuming to complete and score, they are seldom used to assess walking behaviors in large research studies.

Physical activity questionnaires.

Physical activity questionnaires are categorized as global, short-term recall, and long-term quantitative history questionnaires depending on the amount of detail obtained and the timeframe queried. Global questionnaires provide a general impression of a physical activity behavior, whereas short-term recall and long-term quantitative history questionnaires often reflect the volume of activity obtained in general or specific contexts. Table 2 provides a summary of questionnaires that assess walking behaviors in a variety of contexts.

TABLE 2
TABLE 2:
Selected physical activity questionnaires to assess walking behaviors by questionnaire type and context of walking activity measured.

Examples of walking questions.

Global questionnaires have been used to provide a general impression of walking behaviors. For example, the 2001 Behavioral Risk Factor Surveillance System (BRFSS) physical activity module contains one question that asks respondents to select a response that best describes their physical activity at work: (a) mostly sitting or standing, (b) mostly walking, or (c) mostly heavy labor or physically demanding work. In a nationally representative study of 5847 US men and women, approximately 25% reported "mostly walking" during their working time (51). The Stanford Usual Physical Activity Questionnaire focuses on lifestyle walking. Respondents were asked if they "walk instead of drive short distances," "park away from their destination so they have to walk more," "walk on their lunch our or after dinner," and "get off at a bus stop which is not the one nearest their destination and walk." In a validation study, these questions were low to moderately correlated (r ≥ 0.28) with aerobic fitness, lean body mass, and light- and vigorous-intensity activity recorded on a physical activity record (25).

Short-term recall questionnaires have been used to determine the frequency and duration of walking activities occurring in the past week or month. The questionnaires vary in context from asking about multiple contexts for walking within one question to no context at all. The short version of the International Physical Activity Questionnaire (IPAQ) combines several walking contexts in a single question, "Think about the time you spent walking in the past 7 days at work and at home, walking to travel from place to place, and any other walking solely for recreation, sport, exercise or leisure." Respondents are instructed to estimate the days per week and time per day they perform the combined walking activities (www.ipaq.ki.se).

The long form of the IPAQ includes several questions that identify the frequency and duration walking within different contexts, occupation, transportation, and leisure and recreation (www.ipaq.ki.se). Alternatively, the College Alumnus Questionnaire has no context for the walking question, "How many city blocks or their equivalent do you normally walk each day?" However, respondents are asked to estimate their usual pace of walking, ranging from casual or strolling (<2 mph) to brisk or striding (≥4 mph) walking (28).

The Occupational Physical Activity Questionnaire (OPAQ) was developed to supplement the global occupational question used in the BRFSS Physical Activity surveillance system activities. The OPAQ identifies the frequency and duration of sitting and standing, walking, and heavy labor activities performed at work in a usual week. The walking items have a 1-wk test-retest reliability of r = 0.55 and a criterion-related validity of r = 0.62 when compared with similar items from a physical activity record (38).

Long-term quantitative history questionnaires assess physical activity behaviors from one or more years in the past and are similar in content and structure to short-term recall questionnaires. Table 2 also provides examples of quantitative history questionnaires used in epidemiological studies.

Factors to Consider When Using Questionnaires to Assess Walking Behaviors.

For a questionnaire to be an accurate estimate of walking behaviors, it must have acceptable test-retest reliability (r ≥ 0.70) and validity (r ≥ 0.80). The content should reflect the context of the walking behaviors assessed (e.g., occupation, transportation, leisure, and exercise). Questionnaires should be culturally, geographically, and seasonally relevant for the population measured. For example, it is of little value to ask about occupational walking in retired populations. Questionnaires used in diverse cultures should be evaluated for validity, reliability, and acceptability in a specific cultural population before it is used to assess walking behaviors in that population. Questionnaires should also be objective, yielding similar results when self- and interviewer-administered.

Last, knowing the level of measurement needed for data analysis can help to select an appropriate walking questionnaire. Global questionnaires yield nominal and ordinal scores that can be expressed as frequency distributions or can be used in nonparametric analyses. Short- and long-term recall questionnaires generally provide sufficient information to compute minutes per week, days per week, or a measure of intensity and duration, MET-minutes per week.

Surveillance.

Public health surveillance has been defined as the ongoing, systematic collection, analysis, and interpretation of health-related data essential to the planning, implementation, and evaluation of public health practice (48). An additional necessary component of surveillance is the timely dissemination of relevant data to those responsible for health promotion and disease prevention and control policies. Surveillance data are used for a wide variety of health applications (Table 3).

TABLE 3
TABLE 3:
Uses of public health chronic disease and behavioral surveillance systems.

Physical activity surveillance in the United States is accomplished by using several data tools. One of the most prominent is the state-based BRFSS. This behavioral health telephone survey of adults is conducted monthly in all 50 states, Washington, DC, and US territories. It is the largest telephone health survey in the world, with more than 264,000 respondents in 2004 (46). The design of the BRFSS (in which each state draws a representative sample) allows state-based estimates of behavioral health issues that can then be translated into public health planning efforts at the state level.

Physical activity assessments in the BRFSS allow for respondent reporting moderate-intensity and vigorous-intensity physical activity in leisure time, transportation, household, and occupational domains. These results are then translated into prevalence data for population estimates of the proportions of adults who meet or exceed the recommended minimum physical activity guidelines of 30 min·d−1 of moderate-intensity physical activity on most, preferable all days each week (34). Data available from the 2004 BRFSS indicate that 45.9 of US adults were sufficiently physically active in the United States in 2003 (46).

Trend data in BRFSS surveillance measures are more difficult to obtain because of a change in the BRFSS physical activity questions, which was enacted in 2001 (30). One question (assessing leisure time physical inactivity) has remained relatively stable for 15 yr in the BRFSS. Recent analyses of these data suggest a positive downward prevalence in the inactivity among US adults (27).

Because of the focus on total physical activity participation in the BRFSS, data on participation in specific physical activities are not available in the current BRFSS, although an optional (voluntary) set of walking questions does exist for use by states (Table 4). Before 2001, walking-specific surveillance data were available from the BRFSS. In 2003, Simpson et al. (47) published an evaluation of trends in US adult walking behavior from 1987 to 2000 from 31 states. Among all age groups, the most frequently reported leisure time physical activity was walking. The analyses suggested that 26.2% of men and 40.4% of women participated in walking as one of two frequent leisure time physical activities in 1987. Trend data derived from the three iterations of the survey (1987, 1994, and 2000) were strongly indicative of an increasing prevalence of walking (women improved 6.6% and men improved 3.8%) among adults in these 31 states. Increases were also seen in all population subgroups (Fig. 2).

TABLE 4
TABLE 4:
Optional BRFSS walking module.
FIGURE 2
FIGURE 2:
Prevalence of leisure time walking behavior by sex-31 states1 BRFSS 1987-2000.

More surveillance data on walking prevalence and trends are needed to better monitor this activity among US adults and children. This is particularly true for state and local analyses and planning efforts. Walking surveillance instruments should be tied to existing recommendations for physical activity and or to health planning documents (e.g., Healthy People 2010 (50)). Finally, the current growing interest in environmental influences on physical activity offers a unique opportunity to develop walking assessment instruments that incorporate ways to better understand how individuals interact with their environment in terms of walking behavior.

ASSESSING NEIGHBORHOOD AND COMMUNITY WALKABILITY

It is important to capture key aspects of the environmental contexts in which walking takes place to better understand one of its key modifiable determinants. This is particularly important at a time when the environmental and policy interventions are being strongly advocated as ways of enhancing physical activity in populations. Having a strong evidence basis for such public health strategies is crucial. Behavior-specific ecological models in population health (8,45) identify a key role for particular environmental attributes as determinants of specific health-related behavioral choices. In the case of adults' walking in local communities, the choice to walk may be shaped by the cueing properties of environmental attributes (e.g., sidewalks, shade, accessible destinations) and by the reinforcing consequences of walking (e.g., completion of errands, use of recreational facilities) that are likely to increase the future probability of choices to walk.

A specificity of focus (both on behaviors and on associated environmental attributes) should increase the predictive power in physical activity "determinants" studies. Understanding the environmental influences on walking for transport and on walking for recreation or exercise is a case in point. A recent review (49), and the comprehensive updates by Saelens and Handy in this supplement, found an increasing body of evidence on the correlates of specific behavioral physical activities and on the role of particular environmental attributes.

Studies of the relationships between the physical environment and physical activity behaviors have used three main types of measurement methods to identify environmental attributes as independent variables: microlevel ratings of relevant environmental attributes in specific areas by trained observers (10,35); self-report measures of attributes such as facilities, activity opportunities and esthetics (23); and the use of GIS data to derive spatial measures of particular environmental attributes in local areas (17). A recent review of public health research on the environmental determinants of physical activity in adults found the most consistent evidence for accessibility of facilities, opportunities for activity, and esthetics (23). However, the majority of studies (15 of 19 studies) were focused on perceived rather than objective measures, and only two of the studies reviewed used GIS approaches to construct objective measures of the relevant environmental attributes (23).

Combinations of relevant measurement methods can be highly informative in capturing the potential roles of environmental attributes. For example, Giles-Corti and Donovan (17) used a composite measure of spatial access to built and natural recreational facilities (using GIS methods), and observer ratings of the functional environment (existence of footpaths and shops) and of the appeal of the environment (street type, tree coverage). They found that spatial access to recreational facilities was positively associated with the likelihood of being more physically active. In other Australian studies, geographical location based on objectively determined place of residence (coastal as opposed to inland) was found to be associated with increased likelihood of being physically active (9,24).

There are two major strands of research focused on how the particular physical attributes of urban environments may influence walking: 1) health research (primarily epidemiological and behavioral studies of walking); and 2) transportation and urban planning studies (focusing on vehicular travel and its alternatives and examining environmental variables that influence the behavior of entire communities). Building on research findings from the transportation, urban design, and planning fields, Saelens et al. (42) argue that there are physical elements of local environments that influence residents' choices to walk. Two main dimensions of the way land is used seem to be important: proximity (distance) and connectivity (directions of travel). Thus, the "walkability" of a community or neighborhood may be conceptualized as the extent to which these characteristics of the built environment and land-use may, or may not, be conducive to walking-either for leisure, exercise, recreation, to access services, or to travel to work.

Proximity reflects two key land-use variables: density or compactness of land-use and land-use mix (the degree of heterogeneity with which functionally different uses are colocated in space). The more compact and intermixed an urban environment is, the shorter the distances between destinations. Walking, relative to other modes of travel, becomes less probable, as distances between origins and destinations increase (13).

Connectivity is the directness of routes between households, stores, and workplaces. Walking is facilitated where there is a lack of barriers (freeways and other physical obstacles) and where there are several options for travel routes. Interconnecting streets laid out in a regular grid pattern will act to facilitate walking for transport (15,42).

Using GIS to Measure Neighborhood and Community Walkability.

The walkability dimensions of proximity and connectivity described above may be operationalized using GIS methods (4,19,37). A spatial index of walkability was originally developed by Frank et al. (16) in the US, and it uses the most common measures of urban form identified in the transportation, urban design, and planning literature (42). The walkability index has been adopted for use with Australia GIS databases (29).

For GIS studies of the walkability attributes of local communities, the smallest available spatial units should be selected to minimize within-unit variability and to maximize the variation between units. It is common in such analyses to use the smallest unit for which population census data are made available, because such data are not provided at the unit record level and there is a degree of spatial aggregation to preserve individual privacy. In Australia, this unit is the Australian Bureau of Statistics Census Collection District (CCD). In the US, this would be a "block group." At the time of the 2001 Australian Census, there was an average of 225 dwellings in each CCD, but in rural areas, the numbers of dwellings per CCD decline as population densities decrease.

Street centerline data, land-use, zoning data, shopping center location data, and census data can be spatially integrated within a GIS to create an environmental attribute (walkability) index for each CCD. For example, intersections may be identified from street centerline data and an index of street connectivity can be calculated based on the number of unique street connections at each intersection (or the potential for different route choices available). Density is measured from the number of intersections per square kilometer within each CCD. Intersection density is then calculated for each CCD (intersection count divided by CCD area), and the resultant densities are classified into deciles. This provides a standard score from 1 to 10, with 1 the least dense CCD and 10 the most dense CCD. A walkability index may then be derived, using such calculations with GIS data sets: 1-10 scores for the four relevant walkability attribute measures (dwelling density, street connectivity, land-use, and net retail area) may be summed for each CCD, with a possible score of 4 to 40. The resulting walkability index may be further classified, for example, into quartiles with the first quartile used to identify low-walkability CCD and the fourth quartile identifying high-walkability CCD (29).

Measuring Neighborhood and Community Walkability: Research Directions.

There is now a modest, but consistent and growing, body of evidence showing positive relationships of directly observed, perceived, and GIS-derived environmental attributes with particular types of walking. Although the measures of environmental attributes used account for modest proportions of the variance in activity, on a population-wide basis, these proportions can be substantial (33,42,44). Generally, stronger associations with environmental attributes are found for the more particular walking behavior indices. For example, Humpel et al. (24) found a greater likelihood of neighborhood walking (a quite specific walking behavior measure) for those who lived in a coastal location and had positive perceptions of neighborhood esthetics and convenience of, and access to places to walk. No significant associations were found for total walking and for total physical activity (24).

Reliability of environmental measures is essential. Saelens et al. (41) found that the majority of 1-wk test-retest values for items used in their Neighborhood Environment Walkability Scale to be 0.75 or above, a high level of consistency. Individual test-retest intraclass correlations were generally in the 0.60 to 0.80 range for residential density, land-use mix diversity, land-use mix access, street connectivity, walking/cycling facilities, esthetics, pedestrian/traffic safety, and safety from crime. Measures of perceived environmental attributes may, in some circumstances, be able to be gathered more economically than can objective measures. The inclusion of standardized, reliable self-report measures in multiple studies would help this research field to advance more rapidly (11).

Validity of these measures must also be addressed. Perceived environmental attributes should be objectively verifiable, either by independent observation or by objective indices derived from GIS databases (10). For example, Pikora et al. (36) developed a framework of potential environmental influences on the specific behaviors of walking and cycling for recreation and for transport. Items based on findings from the health, transport, and urban planning literature were used to develop an environmental audit instrument, the Systematic Pedestrian and Cycling Environmental Scan (SPACES (35)), which collected data through observational checklists used by trained observers. It is to be hoped that strong patterns of concordance will emerge among objectively observed, perceived, and GIS-derived indices of the same environmental attributes.

Causality is a central scientific concern. To conclude that walking is actually influenced by the environmental attributes that are now being measured by direct observation, self-report, and GIS systems, there is the need to move beyond the examination of cross-sectional associations, making use of prospective study designs with multiple observation points and of intervention study designs (7,26). It is possible, for example, that increased levels of walking might influence the participants' perceptions of the environment, rather than more positive environmental features prompting choices to walk (22).

The community and neighborhood walkability research agenda requires a behavior-specific approach, paying particular attention to objectively defined environmental attributes and to the reliability and validity of measures; prospective study designs are required (6) to allow the possibly causal nature of environment-behavior relationships to be considered seriously. It also requires multilevel modeling approaches to identify how potentially relevant environmental factors reflecting the walkability construct might be acting (32).

CONCLUSIONS

The extraordinarily broad scope of topics presented in this article includes information describing metabolic responses to walking and to how to quantify walking behavior and patterns using self-report and technology. We have considered walking behavior in the context of the built environment and how walking is influenced and measured in the built environment. This interdisciplinary approach illustrates several aspects of contemporary research innovation, directed at carefully examining the behavior that we all use every day for activity and self-transport. Our aim has been to provide insights into the tools and the conceptual frameworks that we need, to build the evidence base from which it will be possible to go about on more effectively promoting walking for health and well being of the population.

Information about specific products does not constitute endorsement of the products by the authors or ACSM. The findings and conclusions in this article are those of the authors and do not necessarily represent the views of the Centers for Disease Control and Prevention. Use of trade names and commercial sources is for identification only and does not imply endorsement by the US Department of Health and Human Services.

REFERENCES

1. Addy CL, Ainsworth BE, Kirtland KA, et al. Association of social and physical environmental supports for physical activity and walking behaviors. Am J Public Health. 2004;94:440-2.
2. Ainsworth BE, Haskell WL, Whitt MC, et al. Compendium of physical activities: an update of activity codes and MET intensities. Med Sci Sports Exerc. 2000;32(9 Suppl.):S498-516.
3. Ainsworth BE, Richardson MT, Jacobs DR Jr, Leon AS, Sternfeld B. Evaluation of occupational activity surveys. J Clin Epidemiol. 1999;52:219-27.
4. Aultman-Hall L, Roorda M, Baetz BW. Using GIS for evaluation of neighborhood pedestrian accessibility. J Urban Plann Dev. 1997;123:10-7.
5. Bassett DR Jr, Ainsworth BE, Swartz AM, Strath SJ, O'Brien WL, King GA. Validity of four motion sensors in measuring moderate intensity physical activity. Med Sci Sport Exerc. 2000;32(9 Suppl.):S471-80.
6. Bauman A. The physical environment and physical activity: moving form ecological associations to intervention evidence. J Epidemiol Community Health. 2005;59:535-6.
7. Bauman AE, Sallis JF, Dzewaltowski DA, Owen N. Towards a better understanding of the influences on physical activity: the role of determinants, correlates, causal variables, mediators, moderators, confounders. Am J Prev Med. 2002;23(Suppl.):5-14.
8. Bauman A, Sallis JF, Owen N. Environmental and policy measurement in physical activity research. In: Welk G and Dale D, editors. Physical Activity Assessments for Health-Related Research. Champaign (IL): Human Kinetics; 2002. p. 241-51.
9. Bauman A, Smith B, Stoker L, Bellew B, Booth M. Geographical influences upon physical activity participation: evidence of a 'coastal effect.' Aust N Z J Public Health. 1999;23:322-4.
10. Craig C. Exploring the effect of the environment on physical activity: a case study examining walking to work. Am J Prev Med. 2002;23(Suppl.):36-43.
11. De Bourheaudhuij I, Sallis JF, Saelens BE. Environmental correlates of physical activitymple of Belgian adults. Am J Health Prom. 2003;18:83-92.
12. Durnin JVGA, Passmore R. Energy, Work and Leisure. London (England): Heinemain Educational Books; 1967. p. 116-9.
13. Frank LD. Economic determinants of urban form. Resulting trade-offs between active and sedentary forms of travel. Am J Prev Med. 2004;27(Suppl.):146-53.
14. Frank LD, Engelke P. How land use and transportation systems impact health. Atlanta (GA): Georgia Institute of Technology and Center for Disease Control. Available from: http://www.cdc.gov/nccdphp/dnpa/pdf/aces-workingpaper1.pdf.
15. Frank LD, Engelke PO, Schmid TL. Community Design. The Impact of the Built Environment on Physical Activity. Washington (DC): Island Press; 2003.
16. Frank LD, Sallis JF, Conway TL, Chapman JE, Saelens BE, Bachman W. Many pathways from land use to health: Associations between neighborhood walkability and active transportation, body mass index, and air quality. J Amer Plan Assoc. 2006;72:75-87.
17. Giles-Corti B, Donovan RJ. The relative influence of individual, social and physical determinants of physical activity. Soc Sci Med. 2002;54:1793-812.
18. Hall C, Figueroa A, Fernhall B, Kanaley JA. Energy expenditure of walking and running: comparison with prediction equations. Med Sci Sports Exerc. 2004;36(12):2128-34.
19. Handy SL. Urban form and pedestrian choices: study of Austin neighborhoods. Transp Research Record. 1996;1552:135-44.
20. Hendelman D, Miller K, Baggett C, DeBold E, Freedson P. Validity of accelerometry for the assessment of moderate intensity physical activity in the field. Med Sci Sports Exerc. 2000;32(7 Suppl.):S422-59.
21. Holt KG, Hamill J, Andres RO. Predicting the minimal energy costs of human walking. Med Sci Sports Exerc. 1991;23(4):491-8.
22. Humpel N, Marshall AL, Leslie E, Bauman A, Owen N. Changes in perceptions of environmental attributes are associated with changes in neighborhood walking. Ann Behav Med. 2004;27:60-7.
23. Humpel N, Owen N, Leslie E. Environmental factors associated with adults' participation in physical activity: a review. Am J Prev Med. 2002;22:188-99.
24. Humpel N, Owen N, Leslie E, Marshall AL, Bauman A, Sallis JF. Associations of location and perceived environmental attributes with walking in neighborhoods. Am J Health Promot. 2004;18:239-42.
25. Jacobs DR Jr, Ainsworth BE, Hartman TJ, Leon AS. A simultaneous evaluation often commonly used physical activity questionnaires. Med Sci Sports Exerc. 1993;25(1):81-91.
26. King AC, Stokols D, Talen E, Brassington GS. Theoretical approaches to the promotion of physical activity. Am J Prev Med. 2002;23(Suppl.):15-25.
27. Kruger J, Ham SA, Kohl HW III. Trends in leisure time physical inactivity by age, sex and race/ethnicity, United States-1994-2004. MMWR Morb Mortal Wkly Rep. 2005;54:991-4.
28. Lee I-M, Paffenbarger RS Jr. Associations of light, moderate, and vigorous intensity physical activity with longevity. The Harvard Alumni Health Study. Am J Epidemiol. 2000;151:293-9.
29. Leslie E, Coffee N, Frank L, Owen N, Bauman A, Hugo G. Walkability of local communities: using geographic information systems to objectively assess relevant environmental attributes. Health Place. 2007;13:111-22.
30. Macera CA, Ham SA, Yore MM, et al. Physical activity prevalence in the United States: Behavioral Risk Factor Surveillance System, 2001. Prev Chronic Dis [Internet]. 2005;2:1-10. Available from: http://www.cdc.gov/pcd/issues/2005/apr/04_0114.htm.
31. Maliszewski AF, Freedson PS. Is running economy different between children and adults. Pediatr Exerc Sci. 1966;8:51-66.
32. Masse LC, Dassa C, Gauvin L, Giles-Corti B, Motl R. Emerging measurement and statistical methods in physical activity research. Am J Prev Med. 2002;23:44-55.
33. Owen N, Humpel N, Leslie E, Bauman A, Sallis JF. Understanding environmental influences on walking: review and research agenda. Am J Prev Med. 2004;27:67-76.
34. Pate RR, Pratt M, Blair SN, et al. Physical activity and public health: a recommendation from the Centers for Disease Control and Prevention and the American College of Sports Medicine. JAMA. 1995;273:402-7.
35. Pikora TJ, Bull FCL, Jamrozik K, Knuiman M, Giles-Corti B, Donovan RJ. Developing a reliable audit instrument to measure the physical environment for physical activity. Am J Prev Med. 2002;23(Suppl.):187-94.
36. Pikora T, Giles-Corti B, Bull F, Jamrozik K, Donovan R. Developing a framework for assessment of the environmental determinants of walking and cycling. Soc Sci Med. 2003;1693-1703.
37. Randall TA, Baetz BW. Evaluating pedestrian connectivity for suburban sustainability. J Urban Plann Dev. 2001;127:1-15.
38. Reis JP, DuBose KD, Ainsworth BE, Macera CA, Yore MM. Reliability and validity of the Occupational Physical Activity Questionnaire. Med Sci Sports Exerc. 2005;37(12):2075-83.
39. Rose J, Ralston HJ, Gamble JG. Energetics of walking. In: Rose J, Gamble JG, editors. Human Walking. 2nd ed. Philadelphia (PA): Williams and Wilkins; 1981. p. 42-72.
40. Rowland TW, Green GM. Physiological responses to treadmill exercise in females: adult-child differences. Med Sci Sports Exerc. 1988;20(5):474-8.
41. Saelens BE, Sallis JF, Black JB, Chen D. Neighbourhood-based differences in physical activity: an environment scale evaluation. Am J Public Health. 93:1552-8.
42. Saelens BE, Sallis JF, Frank LD. Environmental correlates of walking and cycling: findings from the transportation, urban design, and planning literatures. Ann Behav Med. 2003;25:80-91.
43. Sallis JF, Buono MJ, Freedson PS. Bias in estimating caloric expenditure from physical activity in children: implications for epidemiologic studies. Sports Med. 1991;11:203-9.
44. Sallis JF, Frank LD, Saelens BE, Kraft MK. Active transportation and physical activity: opportunities for collaboration on transportation and public health. Transp Res A. 2004;38:249-68.
45. Sallis JF, Owen N. Ecological models of health behaviour. In: Glanz K, Lewis FM, Rimer BK, editors. Health Behaviour and Health Education: Theory, Research, and Practice. 3rd ed. San Francisco (CA): Jossey-Bass; 2002. p. 462-84.
46. Sapkota S, Bowles H, Ham SA, Kohl HW. Adult participation in recommended levels of physical activity-United States 2001 and 2003. MMWR Morb Mortal Wkly Rep. 2005;54:1208-12.
47. Simpson ME, Serdula M, Galuska DA, et al. Walking trends among US adults. The Behavioral Risk Factor Surveillance System, 1987-2000. Am J Prev Med. 2003;25:95-100.
48. Thacker SB. Historical development. In: Teutsch SM, Churchill RE, editors. Principles and Practice of Public Health Surveillance. New York (NY): Oxford University Press; 1994.
49. Trost SG, Owen N, Bauman AE, Sallis JF, Brown W. Correlates of adults participation in physical activity: review and update. Med Sci Sports Exerc. 2002;34(12):1996-2001.
50. US Department of Health and Human Services. Healthy People 2010, Vols. 1 and 2: Understanding and Improving Health and Objectives for Improving Health. 2nd ed. Washington (DC): US Government Printing Office; 2000.
51. Yore MM, Ham SA, Jones DA, Macera CA, Ainsworth BE. Occupational physical activity: reliability and comparison of activity levels. J Phys Act Health. 2005;2:358-65.
Keywords:

METABOLIC RESPONSES; SELF-REPORT; BUILT ENVIRONMENT; NEW TECHNOLOGIES

©2008The American College of Sports Medicine