UBC Faculty Research and Publications

Cross-validation of the factorial structure of the Neighborhood Environment Walkability Scale (NEWS)… Cerin, Ester; Conway, Terry L; Saelens, Brian E; Frank, Lawrence D; Sallis, James F Jun 9, 2009

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


52383-12966_2008_Article_245.pdf [ 206.61kB ]
JSON: 52383-1.0074657.json
JSON-LD: 52383-1.0074657-ld.json
RDF/XML (Pretty): 52383-1.0074657-rdf.xml
RDF/JSON: 52383-1.0074657-rdf.json
Turtle: 52383-1.0074657-turtle.txt
N-Triples: 52383-1.0074657-rdf-ntriples.txt
Original Record: 52383-1.0074657-source.json
Full Text

Full Text

ralInternational Journal of Behavioral ssBioMed CentNutrition and Physical ActivityOpen AcceResearchCross-validation of the factorial structure of the Neighborhood Environment Walkability Scale (NEWS) and its abbreviated form (NEWS-A)Ester Cerin*1, Terry L Conway2, Brian E Saelens3, Lawrence D Frank4 and James F Sallis5Address: 1USDA/ARS Children's Nutrition Research Center, Baylor College of Medicine, 1100 Bates Street, Houston, Texas 77030, USA, 2Graduate School of Public Health, San Diego State University, San Diego, California, USA, 3Seattle Children's Hospital Research Center, University of Seattle, UW-CHI NE 74th St., Box 354920, Seattle, Washington 98101, USA, 4School of Community and Regional Planning, University of British Columbia, 231-1933 West Mall, Vancouver, BC Canada V6T 1Z2 and 5Department of Psychology, San Diego State University, 3900 Fifth Avenue, San Diego, California 92103, USAEmail: Ester Cerin* - ecerin@hku.hk; Terry L Conway - tconway@mail.sdsu.edu; Brian E Saelens - brian.saelens@seattlechildrens.org; Lawrence D Frank - ldfrank@interchange.ubc.ca; James F Sallis - sallis@mail.sdsu.edu* Corresponding author    AbstractBackground: The Neighborhood Environment Walkability Scale (NEWS) and its abbreviated form (NEWS-A) assess perceivedenvironmental attributes believed to influence physical activity. A multilevel confirmatory factor analysis (MCFA) conducted ona sample from Seattle, WA showed that, at the respondent level, the factor-analyzable items of the NEWS and NEWS-Ameasured 11 and 10 constructs of perceived neighborhood environment, respectively. At the census blockgroup (used by theUS Census Bureau as a subunit of census tracts) level, the MCFA yielded five factors for both NEWS and NEWS-A. The aim ofthis study was to cross-validate the individual- and blockgroup-level measurement models of the NEWS and NEWS-A in ageographical location and population different from those used in the original validation study.Methods: A sample of 912 adults was recruited from 16 selected neighborhoods (116 census blockgroups) in the Baltimore,MD region. Neighborhoods were stratified according to their socio-economic status and transport-related walkability levelmeasured using Geographic Information Systems. Participants self-completed the NEWS. MCFA was used to cross-validate theindividual- and blockgroup-level measurement models of the NEWS and NEWS-A.Results: The data provided sufficient support for the factorial validity of the original individual-level measurement models, whichconsisted of 11 (NEWS) and 10 (NEWS-A) correlated factors. The original blockgroup-level measurement model of the NEWSand NEWS-A showed poor fit to the data and required substantial modifications. These included the combining of aspects ofbuilding aesthetics with safety from crime into one factor; the separation of natural aesthetics and building aesthetics into twofactors; and for the NEWS-A, the separation of presence of sidewalks/walking routes from other infrastructure for walking.Conclusion: This study provided support for the generalizability of the individual-level measurement models of the NEWS andNEWS-A to different urban geographical locations in the USA. It is recommended that the NEWS and NEWS-A be scoredaccording to their individual-level measurement models, which are relatively stable and correspond to constructs commonlyused in the urban planning and transportation fields. However, prior to using these instruments in international and multi-cultural studies, further validation work across diverse non-English speaking countries and populations is needed.Published: 9 June 2009International Journal of Behavioral Nutrition and Physical Activity 2009, 6:32 doi:10.1186/1479-5868-6-32Received: 31 December 2008Accepted: 9 June 2009This article is available from: http://www.ijbnpa.org/content/6/1/32© 2009 Cerin et al; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.Page 1 of 10(page number not for citation purposes)International Journal of Behavioral Nutrition and Physical Activity 2009, 6:32 http://www.ijbnpa.org/content/6/1/32BackgroundEcological models postulate that health behavior changesare a function of psychological, social, policy, and physi-cal environmental factors [1,2]. Numerous authors andagencies have identified environmental and policy inter-vention as promising strategies for creating population-wide changes in physical activity and obesity [3-6]. Cur-rent evidence of a relationship between the built environ-ment and physical activity is generally supportive butthere are limitations [7]. An important limitation is thatvirtually all studies have been conducted in settings withrestricted environmental variability. Restricted variabilityyields attenuated estimates of associations [8], so the mag-nitude of associations between environmental character-istics and physical activity is likely underestimated [9,10].An accurate assessment of such associations requiresgreater environmental variability than any one country orregion can offer. The International Physical Activity andthe Environment Network (IPEN; http://www.ipenproject.org) has set out to support coordinated data col-lection in countries with diverse environments and popu-lations. IPEN uses common study design andmeasurement to produce more reliable and valid effectsize estimates.The Neighborhood Environment Walkability Scale,(NEWS) [11] and its abbreviated form (NEWS-A) [12], arethe measures of perceived neighborhood environmentselected for the IPEN initiative. The NEWS and NEWS-Aassess perceived environmental characteristics, stemmingin part from the urban planning literature [13], believedto influence walking and other forms physical activity.Initial evidence for their criterion validity and reliabilityhas been documented in four countries across four conti-nents [11,14-17]. Except for the neighborhood character-istic of not many/any cul-de-sacs, test-retest reliability ofthe individual items of the NEWS and NEWS-A was mod-erate to high [14]. Significant associations were observedbetween the NEWS subscales and objective [18] as well asself-report measures of physical activity and walking[12,16,17]. Additionally, scores on the NEWS werestrongly associated with corresponding objectively-meas-ured constructs of neighborhood environments[11,16,19].Two recent studies examined the measurement models ofthe original and Australian versions of the NEWS [12,16].The measurement models described the relationships ofthe items to the theoretical constructs measured by thescales [20]. In other words, they identified groupings ofitems measuring distinct perceived neighborhood-envi-ronment constructs (e.g. environmental aesthetics, traffichazards, and access to services).To maximize the variability in environmental attributes,both studies adopted a stratified two-stage cluster sam-pling strategy whereby participants were recruited fromspecific areas (here, census blockgroups, the smallest geo-graphical units for which census bureaus publish demo-graphic data) selected according to their objectively-measured walkability and socio-economic status (SES; i.e.median income). Stratification by SES likely enhanced therepresentativeness of the sample because, otherwise, lowSES respondents might have been underrepresented [21].Two distinct measurement models of the NEWS andNEWS-A, one for each level of variation in the data, wereexamined to address violations of the statistical assump-tion of independence of observations resulting from theadopted sampling strategy [20]. Thus, measurement mod-els were defined at the individual (based on within-censusblockgroup variations in the responses to the items) andblockgroup levels (based on the between-census block-group variations) [12,16].The individual-level measurement models were based ondifferences in responses between study participants livingin the same blockgroups and described the way perceivedenvironmental attributes (represented by the NEWSitems) covaried within census blockgroups. The differ-ences in responses may have resulted from actual environ-mental differences within a blockgroup (e.g. differences intraffic load or aesthetics across locations within a block-group), response biases (e.g. tendency to provide extremeratings), and/or perceptual biases (e.g. anxious respond-ents' tendency to overestimate the risk of crime in theirneighborhood) [12]. In contrast, the blockgroup-levelmeasurement models were based on the blockgroup aver-age ratings of the items and indicated how perceived envi-ronmental attributes covaried between blockgroups.These models likely reflected the way environmentalattributes clustered objectively across blockgroups. In fact,the average rating of a blockgroup characteristic can beconsidered a relatively reliable and valid indicator of theobjective environment. This is because response and per-ceptual biases are likely to be random effects that, by def-inition, cancel out when summed across respondents and,hence, have no impact on the average rating for a block-group. Importantly, blockgroup-level factors were foundto be strongly correlated with corresponding objectivemeasures [16].The authors recommended scoring the NEWS accordingto the individual-level measurement model for three mainreasons: (1) the individual-level factors more accuratelyrepresented constructs commonly used in the urban plan-ning and transportation fields [13]; (2) they likely indi-cate how perceptions of environmental attributes groupPage 2 of 10(page number not for citation purposes)together into factors, while blockgroup-level factors likelyrepresent patterns of associations between objective envi-International Journal of Behavioral Nutrition and Physical Activity 2009, 6:32 http://www.ijbnpa.org/content/6/1/32ronmental attributes; (3) they are likely to be more gener-alizable across locations and populations than areblockgroup-level factors [12].The measurement model of the Australian NEWS mostlyresembled that of the original version, but differed in sig-nificant ways [16]. For instance, while traffic-related itemsformed a unique individual-level latent factor in the orig-inal NEWS tested in some USA cities, in the Australian ver-sion they split into two weakly correlated factors –namely, traffic safety and traffic hazards. Although dissim-ilarities in factorial structures were partly attributed tosubstantive item-content differences between the two ver-sions of the NEWS [16], they also raise concerns about thereliability and generalizability of the original measure-ment model to different geographical and cultural set-tings. Hence, it was necessary to cross-validate the originalNEWS and NEWS-A in a geographical location and popu-lation different from those used in the original validationstudy (i.e. Seattle, Washington region). Such informationis important for establishing common, valid scoring pro-tocols, which in turn can provide a more accurate estima-tion of a dose-response relationship of the perceived builtenvironment with physical activity and obesity. Thus, thecurrent paper reports the individual- and blockgroup-level factor structures of the NEWS and NEWS-A tested inthe Baltimore, Maryland – Washington, DC region, whichis a demographically- and environmentally-dissimilar cityto Seattle, WA.Specifically, according to the 2003 American CommunitySurvey, Seattle was the most educated larger city in theUSA, with 52% of residents aged 25 and over havingattained at least a bachelor's degree [22]. In contrast, thepercentage of highly educated residents in Baltimore was35.6. The 2000 median household income in Baltimorewas approximately $32,500, while in Seattle it was$49,500. Seattle had 70.5% of White and only 7.8% ofAfrican American residents, while the percentage of Afri-can Americans in Baltimore was 63.8, and that of Whites31.4. Baltimore is located on the East coast of the UnitedStates, while Seattle is located on the West coast. Balti-more and Seattle are similar in size, terrain, urban layout(grid pattern), and are both considered "cities of neigh-borhoods". However, with its climate and geographicallocation, Seattle provides more ample access to a varietyof outdoor activities. Also, Seattle has higher populationdensity, more traffic congestion problems, but lowercrime rates than Baltimore [22,23].We hypothesized that the individual-level factor struc-tures of the NEWS and NEWS-A, derived from the originalvalidation sample (Seattle, Washington, USA), woulddue to them being in part a function of psychologicalprinciples that apply across diverse subgroups. We alsohypothesized that the original blockgroup-level factorstructures of the NEWS and NEWS-A would show poorerfit to the data from the cross-validation sample than theirindividual-level counterparts due to them reflecting pat-terns of associations between objective environmentalfactors, which likely vary across geographical locations.MethodsParticipantsThis study used cross-sectional survey data from theNeighborhood Quality of Life Study (NQLS) conductedin the Baltimore, MD – Washington, DC region. Neigh-borhoods, defined as clusters of blockgroups, wereselected to vary in walkability characteristics and socio-economic status (SES). Median household income wasused to define blockgroup SES, while data within a Geo-graphic Information System (GIS) on residential density,street connectivity, land-use mix, and retail floor area ratiowere used to operationalize blockgroup walkability [24].Blockgroups were deciled based on their walkability levelsand the lowest (deciles 1–3) and highest (deciles 7–10)were selected for potential recruitment. Income was alsodeciled within these selected blockgroups and block-groups within deciles 2–4 and 7–9 were selected forpotential recruitment. The end result was the selection ofparticipants from 16 neighborhoods (116 census block-groups) whose blockgroups met the specified walkabilityand income criterion [24].Households within selected blockgroups were identifiedby a marketing firm, sent an invitation letter, and thencalled within 2 weeks of the expected receipt of this letter.An adult in the household was asked about interest andstudy eligibility. A sample of 912 (19%; 912 participants/4,816 eligible people contacted) English-speaking adults,aged 20–65, able to walk without assistance, and living inprivate dwellings, was recruited. Participants' socio-demo-graphic characteristics are shown in Table 1.MeasuresSocio-demographic characteristicsParticipants self-reported gender, age, educational attain-ment, annual household income, marital status, andnumber of children (≤ 18 years old) in the household.Neighborhood Environment Walkability Scale (NEWS and NEWS-A)The NEWS and NEWS-A (abbreviated version) consist of67 and 54 items, respectively http://www.drjamessallis.sdsu.edu/measures.html[12]. These are grouped intoeight multi-item subscales (representing distinct con-structs or latent factors) including perceived residentialPage 3 of 10(page number not for citation purposes)show a sufficient level of fit to the data from the cross-val-idation sample (Baltimore, MD – Washington, DC, USA)density; proximity to nonresidential land uses (land usemix – diversity); ease of access to nonresidential usesInternational Journal of Behavioral Nutrition and Physical Activity 2009, 6:32 http://www.ijbnpa.org/content/6/1/32(land use mix – access); street connectivity; infrastructurefor walking and cycling; aesthetics; traffic safety; andsafety from crime. The first two subscales are not factor-analyzable and, hence, represent constructs rather thanlatent factors. Five single-item subscales (four in theNEWS-A) assess perceived major physical barriers to walk-ing; hilly streets; difficult car parking in shopping areas;absence of cul-de-sacs; and presence of people beingactive in the neighborhood (not included in the NEWS-A). All subscales, with the exception of residential densityand land use mix – diversity, are rated on a 4-point Likertscale. Residential density items are rated on a 5-pointscale, and ratings are weighted relative to the average resi-dential density that a specific item represents [11]. Theweighted ratings are summed to create a perceived resi-dential density score. Land use mix – diversity is assessedby the perceived walking proximity from home to varioustypes of destinations, with responses ranging from 1- to 5-minute walking distance (coded as 5) to >30-min walkingdistance (coded as 1).ProcedureOne interested and eligible adult per household was sentthe consent form and, upon its return, was sent question-naires with instructions and postage paid return envelopeor was sent a link via e-mail to complete the survey online.The study was approved by the ethics committee of partic-ipating research institutions.Data AnalysesMultilevel confirmatory factor analysis (MCFA) was(all but residential density and land use mix – diversityitems; i.e. six subscales in total) of the NEWS and NEWS-A. The analyses were multilevel because the study adopteda two-stage cluster sampling design and substantial intra-class correlation coefficients (ICCs; denoting the propor-tion of total item variance due to differences betweenblockgroups) were observed at the blockgroup level. Theaverage ICC was .22 (range: 0.02 to 0.42).MCFA was conducted using Bentler and Liang's MaximumLikelihood Estimation (MLE) method, applicable to mul-tilevel samples with clusters (e.g. blockgroups) varying insize [25]. Empirically-derived a priori two-level measure-ment models of the NEWS and NEWS-A were tested [12].The a priori models consisted of six individual-level corre-lated factors (see Measures section) and five blockgroup-level correlated factors [12]. Additionally, the model ofthe NEWS had five, and that of the NEWS-A four, singleitems. A well-fitting individual-level model would suggestthat the pattern of correlations between individual differ-ences in perceived attributes of the neighborhood envi-ronment observed in a sample of Seattle residents isgeneralizable to residents of the Baltimore – Washingtonregion. A well-fitting blockgroup-level model would indi-cate that the pattern of correlations between average per-ceived attributes of the neighborhood environmentobserved in a sample of Seattle neighborhoods is general-izable to the selected neighborhoods from the Baltimore-Washington region.Re-specification of the a priori models was based onTable 1: Socio-demographic characteristics of the sample (N = 912)Characteristic Estimate Characteristic EstimateGender, % Age, mean (SD), y 46.6 (10.7)Female 52.3 Missing values, % 0.1Missing values 0.0 Marital status, %Ethnicity, % Married 54.4Caucasian 61.9 Widowed/divorced/separated 18.9African-American 27.2 Single/never married 20.4Asian-American 3.0 Living with partner 5.7Pacific Islander 0.1 Missing values 0.6Amer. Indian/Alaskan Native 0.4 Children in household, %Hispanic 3.2 Yes 39.6Other 3.1 Missing values 0.5Missing values 1.1 Annual household income, %Educational attainment, % < $19,500 5.2Some high school or less 2.1 $19,500 – $39,500 12.9Completed high school 7.3 $39,500 – $59,500 17.8Some college 22.9 $59,500 – $79,500 15.9Completed college 30.5 $79,500 – $99,500 13.8Completed graduate degree 36.7 > $99,500 27.3Missing values 0.5 Missing values 7.1Page 4 of 10(page number not for citation purposes)employed to estimate the individual- and blockgroup-level measurement models of the factor-analyzable itemsJöreskog and Sörbom's iterative model-generatingapproach [26], whereby inadequate fit of the data to theInternational Journal of Behavioral Nutrition and Physical Activity 2009, 6:32 http://www.ijbnpa.org/content/6/1/32model is followed by re-specification of the model so toachieve a statistically acceptable fit and a theoreticallymeaningful interpretation of the data. Model re-specifica-tion was guided by the analysis of standardized factorloadings, standardized residual covariances, univariateLangrage multiplier tests, and Wald tests, and theoreticalissues [26]. Factor loadings equal or greater |.30| wereconsidered to be significant [27].The measures of model fit included the Bentler-Liang like-lihood ratio (LR) statistic, the Goodness-of-Fit Index(GFI), the Root Mean Square Error of Approximation(RMSEA), the Non-Normed Fit Index (NNFI), the Com-parative Fit Index (CFI), and the Standardized Root MeanSquared Residual (SRMS) [20,28]. We also used theAikake Information Criterion (AIC), including a penaltyfor model complexity and allowing the comparison ofnon-nested models [20]. The following cut-off values ofacceptable model fit were adopted: >.95 for CFI, NFI, andGFI; <.08 for SRMR; and <.06 for RMSEA [29]. Analyseswere performed using EQS 6.1 (Multivariate SoftwareInc., Encino, CA, 2004).ResultsAll but two items had acceptable values of univariateskewness (< 2.0) and kurtosis (< 7.0) for the use of maxi-mum likelihood estimation [30]. The average skewnessand kurtosis of items were 0.32 and 0.57, respectively. Theitems 'There are canyons/hillsides in my neighborhood'(item 7) and 'The crime rate in my neighborhood makesit unsafe to go on walks during the day' (item 36) hadskewness 2.67 and 2.26, respectively. Although higherthan recommended, these two values are likely to yieldignorable estimation biases [30].Table 2 shows the model fit indices for the a priori and re-specified measurement models of the NEWS and NEWS-A. The application of the present data to the a priori meas-urement model of the NEWS, based on its empirically-derived model [12], met three (GFI≥0.95; RMSEA<0.06;SRMR < 0.08) out of five goodness-of-fit criteria (Table 2).The a priori measurement model of the abbreviated ver-sion of the NEWS (NEWS-A) showed an even better fit,with four (GFI≥0.95, RMSEA<0.06, NNFI≥0.95, andCFI≥0.95) out of five indices meeting the adopted cut-offvalues.Individual-level measurement modelsThe individual-level models of the NEWS and NEWS-Ayielded acceptable standardized factor loadings for all butone item ('Sidewalks are separated from the road/traffic inmy neighborhood by parked cars'; Table 3). As this itemshowed the highest standardized loadings (i.e. .25 for theNEWS and .26 for the NEWS-A) with the factor it was sup-posed to measure ('Infrastructure and safety for walking/cycling') followed by 'Land use mix – access' (.21 for theNEWS and .18 for the NEWS-A), no modifications weremade to the a priori individual-level model of the NEWSand NEWS-A. The correlations between individual-levelfactors of the NEWS and NEWS-A are reported in Tables 4and 5, respectively (above the diagonals). All items' stand-ardized loadings were significant at the 0.001 level.Blockgroup-level measurement modelsAlthough the values of the fit indices were similar to thoseobtained for the final model for the NEWS in the priorMCFA [12], an analysis of standardized item loadings ofthe blockgroup-level measurement model revealed lowerthan acceptable loadings (i.e. <|.30|) for eight items (# 6,7, 13, 16, 19, 26, 28, and 34) (data not shown).Based on analyses of indices of poor model fit and theo-retical considerations, the blockgroup-level model of theNEWS was modified. While the a priori model includedfive correlated factors (i.e. land use mix access and infra-structure for walking; physical obstacles for walking/cycling; aesthetics and friendliness; traffic hazards; andcrime), the data in the present study supported the exist-ence of six correlated factors (Table 3). Items describingnatural physical obstacles to walking (items 6 and 7)Table 2: Results of the multilevel CFAs of the NEWS and NEWS-A on the cross-validation sample: Model fit indicesModel χ2 df GFI RMSEA(90% CI)SRMR NNFI CFI AICa) NEWSModel 1:A priori2881 1135 1.00 .038(.035 – .041).066 .91 .92 609Model 2:Re-specified2801 1135 1.00 .036(.033 – .040).070 .92 .92 531b) NEWS-AModel 1a:A priori1099 445 1.00 .030(.024 – .035).113 .95 .96 211Model 2a: 1066 445 1.00 .026 .076 .96 .97 178Page 5 of 10(page number not for citation purposes)Re-specified (.020 – .032)International Journal of Behavioral Nutrition and Physical Activity 2009, 6:32 http://www.ijbnpa.org/content/6/1/32formed a factor ('Physical obstacles to walking') separatefrom items related to lack of facilities for cycling. Items 12and 13, describing sidewalks, combined into a unique fac-tor, while, in the prior MCFA [12], they loaded on thelatent factors 'Obstacles to walking/cycling' and 'Land usemix – access and infrastructure for walking'. Contrary tothe a priori model definition, items related to greennessand cleanliness). Item 17 ('It is safe to ride a bike in ornear my neighborhood') was related to environmentalaesthetics, as were items describing level of crime safety inthe neighborhood ('Aesthetics and crime safety'). Similarto the a priori model, all traffic-related items loaded ontoa single factor ('Traffic safety and presence of pedestrian').While the a priori model defined presence of pedestriansTable 3: Standardized factor loadings and uniquenesses for final re-specified individual-level and neighborhood-level measurement models of the NEWS and NEWS-A (in brackets).Individual-level Blockgroup-levelItem SL SU LF SL SU LF1 Shopping at local stores .66 (-) .56(-) IL1 (-) .75 (-) .44 (-) BL1 (-)2 Stores within easy walking distance .85 (.63) .28 (.61) IL1 (IL1A) .95 (.95) .09 (.09) BL1 (BL1A)3 Parking difficult Single item .46 (.40) .79 (.84) BL1 (BL1A)4 Many places within walking distance .68 (.91) .54 (.17) IL1 (IL1A) .90 (.90) .19 (.19) BL1 (BL1A)5 Easy to walk to a transit stop .31 (.30) .90 (.90) IL1 (IL1A) .83 (.83) .31 (.31) BL1 (BL1A)6 Hilly streets Single item .99 (.70) .02 (.51) BL2 (BL2A)7 Major barriers to walking Single item .34 (.49) .89 (.76) BL2 (BL2A)8 Few cul-de-sacs Single item .60 (.56) .64 (.69) BL1 (BL1A)9 Short distance between intersections .34 (.37) .88 (.86) IL2 (IL2A) .86 (.88) .26 (.22) BL1 (BL1A)10 Four-way intersections .57 (-) .67 (-) IL2 (-) .88 (-) .23 (-) BL1 (-)11 Many alternative routes .49 (.43) .76 (.82) IL2 (IL2A) .92 (.92) .15 (.15) BL1 (BL1A)12 Sidewalks .50 (.41) .75 (.83) IL3 (IL3A) .98 (.74) .04 (.45) BL3 (BL1A)13 Well-maintained sidewalks .58 (-) .66 (-) IL3 (-) .60 (-) .64 (-) BL3 (-)14 Bicycle or pedestrian trails .41 (-) .83 (-) IL3 (-) .38 (-) .86 (-) BL4 (-)15 Cars dividing sidewalk and traffic .25 (.26) .94 (.93) IL3 (IL3A) .81 (.84) .35 (.29) BL1 (BL1A)16 Grass/dirt dividing sidewalk and traffic .30 (.31) .90 (.90) IL3 (IL3A) .90 (.73) .19 (.47) BL4 (BL3A)17 Safe to ride .64 (-) .56 (-) IL3 (-) .84 (-) .29 (-) BL5 (-)18 Trees .43 (.42) .81 (.82) IL4 (IL4A) .85 (.99) .27 (.02) BL4 (BL3A)19 Trees give shade .41 (-) .83 (-) IL4 (-) .64 (-) .59 (-) BL4 (-)20 Many interesting things to look at .72 (.77) .48 (.41) IL4 (IL4A) .55 (.64) .69 (.60) BL5 (BL4A)21 No litter .51 (-) .74 (-) IL4 (-) .92 (-) .16 (-) BL5 (-)22 Many attractive natural sights .72 (.71) .49 (.50) IL4 (IL4A) .91 (.86) .18 (.26) BL5 (BL4A)23 Attractive buildings/homes .71 (.69) .50 (.52) IL4 (IL4A) .63 (.71) .60 (.50) BL5 (BL4A)24 Heavy traffic along the street .73 (-) .47 (-) IL5 (-) -.71 (-) .50 (-) BL6 (-)25 Heavy traffic along nearby streets .72 (.66) .48 (.57) IL5 (IL5A) -.64 (-.56) .59 (.69) BL6 (BL5A)26 Slow traffic speed on the street -.50 (-) .75 (-) IL5 (-) .87 (-) .25 (-) BL6 (-)27 Slow traffic speed on nearby streets -.48 (-.55) .77 (.69) IL5 (IL5A) .75 (.71) .44 (.50) BL6 (BL5A)28 Speeding drivers .42 (.42) .82 (.82) IL5 (IL5A) -.45 (-.60) .80 (.64) BL6 (BL5A)29 Street lights .49 (.60) .76 (.64) IL3 (IL3A) .68 (.71) .54 (.50) BL1 (BL1A)30 Walkers and bikers easily seen .49 (.61) .76 (.63) IL3 (IL3A) .81 (.87) .34 (.25) BL6 (BL5A)31 Crosswalks and pedestrian signals .34 (.30) .88 (.90) IL3 (IL3A) .90 (.82) .19 (.32) BL1 (BL1A)32 Crosswalks help walkers feel safe .49 (-) .76 (-) IL3 (-) .89 (-) .21 (-) BL1 (-)33 Exhaust fumes .50 (-) .75 (-) IL5 (-) -.84 (-) .30 (-) BL5 (-)34 Seeing/speaking to other people Single item (NEWS only) .61 (-) .63 (-) BL6 (-)35 High crime rate .76 (.75) .43 (.44) IL6 (IL6A) -.98 (-.99) .04 (.02) BL5 (BL4A)36 Unsafe to walk during the day .55 (.52) .70 (.73) IL6 (IL6A) -.92 (-.95) .17 (.10) BL5 (BL4A)37 Unsafe to walk at night .80 (.81) .35 (.34) IL6 (IL6A) -.97 (-.98) .06 (.03) BL5 (BL4A)38 Safe for children to walk alone -.39 (-) .85 (-) IL6 (-) .78 (-) .39 (-) BL5 (-)(-) = not applicable. SL = standardized loadings; SU = standardized uniqueness; LF= latent factor. Latent individual-level factors: IL1 and IL1A = Land use mix – access; IL2 and IL2A = Street connectivity; IL3 = Infrastructure and safety for walking/cycling; IL3A = Infrastructure and safety for walking; IL4 and IL4A = Aesthetics; IL5 and IL5A = Traffic hazards; IL6 and IL6A = Crime. Latent blockgroup-level factors: BL1 and BL1A = Land use mix access and infrastructure for walking; BL2 and BL2A = Physical obstacles to walking; BL3 = Sidewalks; BL4 and BL3A = Green areas; BL5 and BL4A = Aesthetics and safety from crime; BL6 and BL5A = Traffic safety and presence of pedestrians. Autocorrelated error terms were modeled for items 18 and 19 (r = 0.55; t = 14.4; P < 0.001), and 31 and 32 (r = 0.59; t = 15.6; P < 0.001).Page 6 of 10(page number not for citation purposes)('Green areas') formed a factor separate from other aes-thetic aspects of the environment (e.g. attractive buildingsas a characteristic associated with aesthetics and trafficsafety, the data supported associations with traffic safetyInternational Journal of Behavioral Nutrition and Physical Activity 2009, 6:32 http://www.ijbnpa.org/content/6/1/32only. Four goodness-of-fit indices (χ2, RMSEA, NNFI, andAIC) indicated a better fit of the re-specified than the a pri-ori model to the data (Table 2).Similar to the model of the NEWS, an analysis of stand-ardized factor loadings indicated a certain degree of misfitat the blockgroup-level of the NEWS-A model, wherebytwo items (16 and 25) had unacceptably low loadings (-.10 and -.21). An analysis of indices of poor model fit ledto several modifications of the blockgroup-level model.The re-specified model showed excellent fit, with all indi-ces meeting the goodness-of-fit criteria (Table 2). Bothblockgroup-level a priori and re-specified models con-sisted of five correlated factors including 'Land use mix –access and infrastructure for walking' and 'Physical obsta-cles to walking'. However, while the a priori blockgroup-level model included the factors 'Aesthetics', 'Traffic haz-ards', and 'Safety from crime', the re-specified modelincluded 'Green areas', 'Aesthetics and safety from crime',and 'Traffic safety and presence of pedestrians'. The corre-lations between blockgroup-level factors of the NEWS andNEWS-A are shown in Tables 4 and 5, respectively (belowthe diagonals). All items' standardized loadings were sig-nificant at the 0.001 level.DiscussionThe main aim of this study was to cross-validate the facto-rial structure of the NEWS [11,12] and NEWS-A [12] bycomparing prior analyses in one USA metropolitan area toanother USA metropolitan area. The overall goodness offit of the a priori model of the NEWS was similar to thatreported in the original validation study, while that for theNEWS-A was slightly lower [12]. Support was found forthe validity of the current individual-level measurementmodels consisting of six correlated factors (land use mix –access; street connectivity; infrastructure and safety forwalking; aesthetics; traffic hazards; and crime) and five(NEWS) or four (NEWS-A) single items. In contrast, theblockgroup-level models of the NEWS and NEWS-Ashowed a poor fit to the data. Re-specification of theblockgroup-level models resulted in goodness of fit levelcomparable to those from the original validation study forboth NEWS and NEWS-A [12]. The individual- and block-group-level measurement models are discussed in detailbelow.Individual-level measurement modelsAs noted earlier, the a priori individual-level measure-ment model fitted the data well. Only the item 'Sidewalksare separated from road/traffic in my neighborhood byparked cars' insufficiently, yet maximally, loaded on thefactor it was supposed to represent ('Infrastructure andsafety for walking/cycling'). It is noteworthy that the sameitem had a relatively low loading in the original validationstudy of the NEWS (0.38) [12] and did not sufficientlyload on any factors in a validation study of the Australianversion of the NEWS [16]. In the present study, this itemalso tended to correlate with the factor representing accessTable 4: Correlations between individual-level latent factors of the NEWS (above the diagonal) and NEWS-A (below the diagonal).NEWS-A factors IL2 IL3 IL4 IL5 IL6 NEWS factorsLand use mix – access (IL1A) .29 .21 .19 < .10a < .10a Land use mix – access (IL1)Street connectivity (IL2A) .45 .32 .23 -.10 < .10a Street connectivity (IL2)Infrastructure & safety for walking (IL3A) .17 .41 .43 -.59 -.40 Infrastructure & safety for walking/cycling (IL3)Aesthetics (IL4A) .29 .31 .30 -.42 -.31 Aesthetics (IL4)Traffic hazards (IL5A) < .10a -.34 -.46 -.31 .51 Traffic hazards (IL5)Crime (IL6A) < .10a < .10a -.37 -.21 .46 Crime (IL6)IL1A IL2A IL3A IL4A IL5Aa constrained to zero in the final model as correlation coefficients smaller than |.10|. The subscript A stands for NEWS-A.Table 5: Correlations between blockgroup-level latent factors of the NEWS (above the diagonal) and NEWS-A (below the diagonal).NEWS-A factors BL2 BL3 BL4 BL5 BL6 NEWS factorsLand use mix – access and infrastructure for walking (BL1A)-.29 .66 < .10a -.51 .19 Land use mix – access and infrastructure for walking (BL1)Physical obstacles to walking (BL2A) -.54 < .10a < .10a < .10a - .14 Physical obstacles to walking (BL2)Green areas (BL3A) < .10a < .10a .41 -.46 .26 Sidewalks (BL3)Aesthetics and safety from crime (BL4A) -.46 < .10a .47 .42 .36 Green areas (BL4)Traffic safety and presence of people (BL5A) .52 -.45 .57 .18 .39 Aesthetics and safety from crime (BL5)BL1A BL2A BL3A BL4A Traffic safety and presence of people (BL6)Page 7 of 10(page number not for citation purposes)a constrained to zero in the final model as correlation coefficients smaller than |.10|. The subscript A stands for NEWS-A.International Journal of Behavioral Nutrition and Physical Activity 2009, 6:32 http://www.ijbnpa.org/content/6/1/32to facilities, while in the Australian sample it showed astrong positive association with walking for transporta-tion at the blockgroup level.It appears that the separation of sidewalks from traffic byparked cars may be indicative of access to destinations(i.e. parked cars at nearby services) as well as infrastruc-ture for walking (i.e. sidewalks). However, this environ-mental characteristic is less likely to be associated withpedestrian safety. Cars parked along sidewalks are a signof local motorized traffic that may pose problems topedestrians wishing to cross a road, especially in theabsence of crosswalks. This may explain why this particu-lar environmental characteristic weakly or inconsistentlyloaded on the infrastructure and safety dimension. Atpresent, before more information on the factorial validityof the NEWS across various geographical locations is gath-ered, it is suggested the item 'Sidewalks are separated fromroad/traffic in my neighborhood by parked cars' be con-sidered part of the infrastructure and safety factor, partic-ularly given that both item and factor (excluding andincluding the item) were found to be significantly posi-tively related to weekly minutes of walking for transport[12,16] and, marginally, positively related to walking forrecreation [16].Blockgroup-level measurement modelsIn this study, the a priori blockgroup-level models of theNEWS and NEWS-A did not show a sufficient level of fitto the data. The main differences between the poorly-fit-ting a priori and the well-fitting re-respecified blockgrouplevel models were (1) the separation of natural (i.e. treesand natural sights) from building aesthetics; (2) a strongerassociation between building aesthetics and crime; (3)and the separation of sidewalks from other infrastructurefor walking (i.e. crosswalks, grass strips, and street connec-tivity).As noted earlier, blockgroup-level factors were expected tobe less stable and generalizable across locations than indi-vidual-level factors for two main reasons. First, block-group-level associations depend on the criteria for theselection of study areas. Second, blockgroup-level factorsare more likely to represent patterns of associationsbetween objective environmental factors, which can sub-stantially vary across geographical locations. For example,high levels of household density and access to servicesmay, in certain urban environments, be associated withlower socio-economic status and higher crime [31], whilein others this pattern may be related to higher socio-eco-nomic status and higher levels of aesthetics [32,33]. Also,a comparison of the results from the present study withthose from the original validation study of the NEWS [12]levels of building aesthetics), but a weaker associationbetween aesthetics and crime, in selected neighborhoodsof Seattle than Baltimore regions.In contrast, individual perceptions of environmental char-acteristics are likely to be in part a function of psycholog-ical principles that apply across diverse subgroups.Experimental studies indicate that the evaluation of con-cepts believed to be related substantially influence percep-tions of these concepts [34]. People tend to exaggeratedifferences among items/attributes that fall into differentconceptual categories, and minimize differences amongitems/attributes that fall into the same category [35]. Sucha mechanism would explain the respondents' tendency togroup environmental characteristics into meaningful, dis-tinct concepts (e.g. access to services, street connectivity,and crime) common to their culture and language irre-spective of their place of residence. Hence, we recommendusing the individual-level measurement models of theNEWS and NEWS-A in both single- and multi-site studies.Limitations and future researchThe relatively low response rate is one of the main limita-tions of the study. This is likely due in part to the extensivemeasurement protocol, including surveys and accelerom-eter monitoring on two occasions. Because recruitmentrates did not differ by walkability/income quadrants, dif-ferential selection bias seems unlikely. However, the factthat similar individual-level measurement models wereobserved across three geographical locations (Baltimore,Seattle, and Adelaide) assuages concerns about samplebias effects. Another limitation pertains to the adoptedsampling design that, while facilitating the recruitment ofa socio-economically balanced sample, precluded the der-ivation of blockgroup-level measurement models of theNEWS and NEWS-A representative of the geographicallocations. It is possible that a random sample of block-groups might have resulted in greater similarities betweenthe individual- and blockgroup-level measurement mod-els and higher levels of generalizability of the blockgroup-level measurement model across geographical locations.This is because the procedure for the selection of block-groups adopted in the three validation studies of theNEWS might have artificially inflated the blockgroup-level correlation between certain environmental features(street connectivity and land use mix – access).It is important to note that all three validation studies ofthe NEWS were conducted in the USA and Australia, twocountries with similar cultures and language, as well as apreponderance of low-density land uses. This may have inpart contributed to the observed similarities among theindividual-level measurement models. It is yet to be seenPage 8 of 10(page number not for citation purposes)suggests a stronger association between greenery andblockgroup socio-economic status (represented by higherwhether the current measurement model of the NEWS canbe replicated in populations outside Australia and theInternational Journal of Behavioral Nutrition and Physical Activity 2009, 6:32 http://www.ijbnpa.org/content/6/1/32USA, although we hypothesize that it is likely to be suffi-ciently generalizable to other countries and cultures. Thereason for this is that most the items of the NEWS depicttangible, physical neighborhood attributes, with theexception of safety-related items. We believe that theinterpretation of items dealing with physical attributes isless likely to differ across cultures than is that of itemsgauging socio-cultural and psychological concepts. Yet,empirical confirmation for our hypothesis is needed.A limitation on generalizability is likely to be incompleteassessment by the NEWS and NEWS-A of environmentalattributes found in other geographic locations. Differentforms of mixed use, different pedestrian and bicyclinginfrastructure, and different public transit access and facil-ities in other countries should be reflected in modifiedversions of the NEWS and NEWS-A. Efforts to make suchmodifications are currently ongoing.ConclusionThis study provided further support for the factorial valid-ity of the NEWS and its abbreviated version. At present, itis recommended that the NEWS and NEWS-A be scoredaccording to the individual-level model comprising eightmulti-item subscales and five (for the NEWS) or four (forthe NEWS-A) single-item subscales. As these subscales areclearly related to constructs used in urban planning andtransportation, findings based on these subscales caninform policies and interventions that may improve theactivity-friendliness of a neighborhood. Given that all fac-torial-validation studies of the NEWS and NEWS-A wereconducted on English-speaking populations, before theycan be 'comfortably' used in global or multi-cultural stud-ies further validation work across diverse populations isneeded.Competing interestsThe authors declare that they have no competing interests.Authors' contributionsEC analyzed the data, conceptualized and wrote all draftsof the manuscript. BES, LDF and JFS designed, organizedand conducted the Neighborhood Quality of Life Study(NQLS). TLC contributed to the coordination of NQLS,data management, and preliminary data analyses for thispaper. TLC, BES, LDF, and JFS critically reviewed andedited all drafts of the manuscript. All authors approvedthe final version of the manuscript.AcknowledgementsThis work was supported by grant HL67350 from the National Heart, Lung, and Blood Institute. Kelli Cain and James Chapman made important contri-butions to the study. This work is also a publication of the United States Department of Agriculture (USDA/ARS) Children's Nutrition Research under cooperative agreement No. 58-6250-6001 (Dr. Cerin). The contents of this publication do not necessarily reflect the views and policies of the USDA.References1. Sallis JF, Owen N, Fisher EB: Ecological models of health behav-iour.  In Health Behavior and Health Education: Theory, Research, andPractice Edited by: Glanz K, Rimer BK, Viswanath K. San Francisco CA:Jossey-Bass; 2008:465-482. 2. Stokols D: Establishing and maintaining healthy environ-ments. Toward a social ecology of health promotion.  Am Psy-chol 1992, 47:6-22.3. Heath G, Brownson R, Kruger J, Miles R, Powell KE, Ramsey LT, theTask Force on Community Preventive Services: The effectivenessof urban design and land use and transport policies and prac-tices to increase physical activity: A systematic review.  J Phys-ical Act Health 2006, 3(Suppl 1):S55-S76.4. Koplan JP, Dietz WH: Caloric imbalance and public health pol-icy.  JAMA.  2000, 282(16):1579-1581.5. Kumanyika SK, Van Horn L, Bowen D, Perri MG, Rolls BJ, CzajkowskiSM, Schron E: Maintenance of dietary behaviour change.  HealthPsychol 2000, 19(Suppl 1):42-56.6. World Health Organization: Global Strategy on Diet, Physical Activity,and Health 2004 [http://www.who.int/dietphysicalactivity/]. GenevaSwitzerland, WHO7. Transportation Research Board and Institute of Medicine: Does theBuilt Environment Influence Physical Activity? Examining the Evidence. Spe-cial Report 282 Washington DC: National Academies Press; 2005. 8. Hunter JE, Schmidt FL: Methods of Meta-Analysis: Correcting Errors andBias in research Findings Newbury Park: Sage Publications; 1990. 9. Giles-Corti B, Donovan RJ: The relative influence of individual,social and physical environment determinants of physicalactivity.  Soc Sci Med 2002, 54:1793-1813.10. Hoehner CM, Ramirez LKB, Elliott MB, Handy SL, Brownson RC:Perceived and objective environmental measures and physi-cal activity among urban adults.  Am J Prev Med 2005,28(2S2):105-116.11. Saelens BE, Sallis JF, Black JB, Chen D: Neighborhood-based differ-ences in physical activity: An environmental scale evaluation.Am J Public Health 2003, 93:1552-1558.12. Cerin E, Saelens BE, Sallis JF, Frank LD: Neighborhood Environ-ment Walkability Scale: Validity and development of a shortform.  Med Sci Sports Exerc.  2008, 38(9):1682-1691.13. Frank LD, Engelke PO, Schmidt TL: Health and Community Design: TheImpact of the Built Environment on Physical Activity Washington DC:Island; 2003. 14. Brownson RC, Chang JJ, Eyler AA, Ainsworth BE, Kirkland KA, Sael-ens BE, Sallis JF: Measuring the environment for friendlinesstoward physical activity: a comparison of the reliability of 3questionnaires.  Am J Public Health 2004, 94:473-483.15. Cerin E, Macfarlane DJ, Ko HH, Chan KC: Measuring perceivedneighbourhood walkability in Hong Kong.  Cities 2007,24(3):209-217.16. Cerin E, Leslie E, Owen N, Bauman A: An Australian version ofthe Neighborhood Environment Walkability Scale: Validityevidence.  Meas Phys Educ Exerc Sci 2008, 12:31-51.17. De Bourdeaudhuij I, Sallis JF, Saelens BE: Environmental corre-lates of physical activity in a sample of Belgian adults.  Am JHealth Promot.  2003, 18(1):83-92.18. Atkinson JL, Sallis JF, Saelens BE, Cain KL, Black JB: The associationof neighbourhood design and recreational environmentswith physical activity.  Am J Health Promotion 2005, 19:304-309.19. Leslie E, Saelens B, Frank L, Owen N, Bauman A, Coffee N, Hugo G:Residents' perceptions of walkability attributes in objectivelydifferent neighbourhoods: A pilot study.  Health Place 2005,11:227-236.20. Kline RB: Principles and Practice of Structural Equation Modeling NewYork: Guilford; 1998. 21. Turrel G: Income non-reporting: implications for health ine-quality research.  J Epidemiol Community Health.  2000,54(3):207-214.22. U.S. Census Bureau: 2003 American Community Survey.Page 9 of 10(page number not for citation purposes)Center, Department of Pediatrics, Baylor College of Medicine, Houston, Texas, and had been funded in part with federal funds from USDA/ARS [http://factfinder.census.gov/].23. Federal Bureau of Investigation: Uniform Crime Report, 2006.[http://www.fbi.gov/ucr/cius2006].Publish with BioMed Central   and  every scientist can read your work free of charge"BioMed Central will be the most significant development for disseminating the results of biomedical research in our lifetime."Sir Paul Nurse, Cancer Research UKYour research papers will be:available free of charge to the entire biomedical communitypeer reviewed and published immediately upon acceptancecited in PubMed and archived on PubMed Central International Journal of Behavioral Nutrition and Physical Activity 2009, 6:32 http://www.ijbnpa.org/content/6/1/3224. Frank LD, Sallis JF, Saelens BE, Leary L, Cain K, Conway TL, Hess PM:The development of a walkability index: Application to theNeighborhood Quality of Life Study.  Br J Sports Med 2009 inpress.25. Bentler PM, Liang J: Two-level mean and covariance structures:Maximum likelihood via an EM algorithm.  In Multilevel Mode-ling: Methodological Advances, Issues, and Applications Edited by: ReiseSP, Dian N. Mahwah NJ: Lawrence Erlbaum; 2003:53-70. 26. Jöreskog KG, Sörbom D: LISREL 8: Structural Equation Modelling withthe SIMPLIS Command Language Chicago IL: SPSS; 1993. 27. Bryant FB, Yarnold PR: Principal components analysis andexploratory and confirmatory factor analysis.  In Reading andUnderstanding Multivariate Statistics Edited by: Grimm LG, Yarnold PR.Washington DC: APA; 1994:99-136. 28. Browne MW, Cudek R: Alternative ways of assessing model fit.In Testing Structural Equation Models Edited by: Bollen KA, Long JS.Newbury Park CA: Sage Publications; 1993:136-162. 29. Hu L, Bentler PM: Cutoff criteria for fit indices in covariancestructure analysis: Conventional criteria versus new alterna-tives.  Struct Equat Model 1999, 6:1-55.30. Curran PJ, West SG, Finch JF: The robustness of test statistics tononnormality and specification error in confirmatory factoranalysis.  Psychol Methods 1996, 1:16-29.31. Zhu X, Lee C: Walkability and safety around elementaryschools: Economic and ethnic disparities.  Am J Prev Med 2008,34:282-290.32. Burnley I, Murphy P: Sea Change: Movement from Metropolitan to Arca-dian Australia Sydney, Australia: UNSW Press; 2004. 33. Hui ECM, Chau CK, Pun L, Law MY: Measuring the neigboringand environmental effects on residential property value:Using spatial weighting matrix.  Building Environ 2006,42:2333-2343.34. Goldstone RL: Believing is seeing.  Am Psychol Soc Observer 2004,17:23-26.35. Goldstone RL: Perceptual learning.  Ann Rev Psychol 1998,49:585-612.yours — you keep the copyrightSubmit your manuscript here:http://www.biomedcentral.com/info/publishing_adv.aspBioMedcentralPage 10 of 10(page number not for citation purposes)


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items