Article Text

Neighbourhood inequities in the availability of retailers selling tobacco products: a systematic review
  1. Amanda Y Kong1,2,
  2. Joseph G L Lee3,4,
  3. Sarah M Halvorson-Fried4,5,
  4. Kerry B Sewell6,
  5. Shelley Diane Golden4,5,
  6. Lisa Henriksen7,
  7. Lily Herbert2,
  8. Kurt M Ribisl4,5
  1. 1Department of Family and Preventive Medicine, The University of Oklahoma Health Sciences Center, Oklahoma City, Oklahoma, USA
  2. 2Stephenson Cancer Center, The University of Oklahoma Health Sciences Center, Oklahoma City, Oklahoma, USA
  3. 3Department of Health Education and Promotion, College of Health and Human Performance, East Carolina University, Greenville, North Carolina, USA
  4. 4University of North Carolina Lineberger Comprehensive Cancer Center, Chapel Hill, North Carolina, USA
  5. 5Department of Health Behavior, University of North Carolina at Chapel Hill Gillings School of Global Public Health, Chapel Hill, North Carolina, USA
  6. 6Laupus Health Sciences Library, East Carolina University, Greenville, North Carolina, USA
  7. 7Stanford Prevention Research Center, Stanford University School of Medicine, Palo Alto, Stanford, California, USA
  1. Correspondence to Dr Amanda Y Kong, Department of Family and Preventive Medicine, The University of Oklahoma Health Sciences Center, Oklahoma City, Oklahoma, USA; Amanda-Kong{at}OUHSC.edu

Abstract

Objective To examine inequities in tobacco retailer availability by neighbourhood-level socioeconomic, racial/ethnic and same-sex couple composition.

Data sources We conducted a 10 November 2022 search of PubMed, PsycINFO, Global Health, LILACS, Embase, ABI/Inform, CINAHL, Business Source Complete, Web of Science and Scopus.

Study selection We included records from Organisation for Economic Co-operation and Development member countries that tested associations of area-level measures of tobacco retailer availability and neighbourhood-level sociodemographic characteristics. Two coders reviewed the full text of eligible records (n=58), including 41 records and 205 effect sizes for synthesis.

Data extraction We used dual independent screening of titles, abstracts and full texts. One author abstracted and a second author confirmed the study design, location, unit of analysis, sample size, retailer data source, availability measure, statistical approach, sociodemographic characteristic and unadjusted effect sizes.

Data synthesis Of the 124 effect sizes related to socioeconomic inequities (60.5% of all effect sizes), 101 (81.5%) indicated evidence of inequities. Of 205 effect sizes, 69 (33.7%) tested associations between retailer availability and neighbourhood composition of racially and ethnically minoritised people, and 57/69 (82.6%) documented inequities. Tobacco availability was greater in neighbourhoods with more Black, Hispanic/Latine and Asian residents (82.8%, 90.3% and 40.0% of effect sizes, respectively). Two effect sizes found greater availability with more same-sex households.

Conclusions There are stark inequities in tobacco retailer availability. Moving beyond documenting inequities to partnering with communities to design, implement, and evaluate interventions that reduce and eliminate inequities in retail availability is needed to promote an equitable retail environment.

PROSPERO registration number CRD42019124984.

  • Advertising and Promotion
  • Disparities
  • Priority/special populations
  • Socioeconomic status

Data availability statement

All data relevant to the study are included in the article or uploaded as supplementary information.

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Introduction

Globally, evidence on tobacco use prevention and cessation has shown reducing the retail availability of commercial tobacco products reduces supply and demand.1–3 Several studies report positive associations between the neighbourhood number or concentration of brick-and-mortar stores selling tobacco products (ie, tobacco retailer availability [TRA], supply or density) and tobacco use and tobacco-related health outcomes.4–14 Living in neighbourhoods with higher TRA reduces travel costs to obtain tobacco products15 16 and increases the likelihood of observing tobacco-related marketing, cueing tobacco use and relapse behaviours.17–19 Several reviews document associations between TRA and tobacco use among youth or young adults4–7 and adults.8 9 A meta-analysis of 11 studies from six countries documented a significant positive association between past-month adolescent smoking behaviours and greater TRA near homes, but not near schools.4 Furthermore, a meta-analysis of 27 studies from six countries found that lower availability was associated with a 2.5% reduction (95% CI 1.95 to 3.02) in the relative risk of tobacco use among adults.9

The burden of tobacco-related morbidity and mortality is not equally shared by socioeconomic status (SES),20–22 ethnicity or race,20 23 or sexual orientation24 25 which may reflect environmental injustices in TRA. In 2002, Laws et al assessed neighbourhood inequities in TRA in 10 predominately Latine business districts in Boston, Massachusetts.26 Hyland et al subsequently examined availability in Erie County, New York.27 Numerous studies have subsequently documented neighbourhood inequities in TRA in the USA,28–34 Australia,35 Canada,36 Germany,37 New Zealand38 and Scotland.39

While a systematic review of 43 studies across eight countries documented greater point-of-sale tobacco marketing in neighbourhoods with lower SES and a greater percentage of Black residents,40 to our knowledge, no systematic review has examined neighbourhood inequities in TRA. This gap hinders efforts to reduce health inequities related to tobacco use and the retail environment. With a focus on health equity, we conducted a systematic review examining place-based differences in TRA by SES, ethnicity and race of neighbourhood residents and same-sex household composition. With concern for methodological gaps, we also assessed how TRA was operationalised and the number/type of data sources used to identify tobacco retailers.

Methods

Search and eligibility

A professional health sciences librarian (KBS) iteratively developed the search strategy (online supplemental file A). Our search was limited to records published in 2000 and last updated on 10 November 2022. We aimed to include peer-reviewed or grey literature with the following characteristics: records that (a) were from an Organisation for Economic Co-operation and Development (OECD) member country; (b) included an area-level measure of TRA; (c) included an area-level sociodemographic characteristic of interest (ie, area unit characteristics or composition by SES, ethnicity, race or sexual orientation); and (d) used statistics to test associations (eg, correlations, regression). We limited our records to OECD countries to reduce heterogeneity in country-level socioeconomic resources.

Supplemental material

As our systematic review is focused on answering questions about area-level or neighbourhood-level inequities, we excluded records that only measured person-level or individual-level proximity measures (eg, the distance a person lived to a tobacco retailer) or individual-level sociodemographic characteristics. We further excluded school-based studies (eg, those that measured school-level inequities in TRA), as this study focused on neighbourhood inequities, and school and neighbourhood boundaries and sociodemographic characteristics do not always align. Our study is included as research question 2 in a registered protocol (PROSPERO CRD42019124984).41 We defined tobacco retailers as physical or brick-and-mortar locations that sell any tobacco products and excluded records that sell nicotine vape products exclusively from the synthesis (n=5),42–46 given evidence these retailers may be patterned differently (online supplemental file B).

Supplemental material

Inclusion coding

All screening was performed in Covidence. First, two independent coders reviewed each record’s title and abstract for inclusion. For each eligible record identified, two coders (AYK, JGLL) independently reviewed the full text. Disagreements were resolved by consensus. Figure 1 shows the Preferred Reporting Items for Systematic Reviews and Meta-Analyses 202047 flow diagram. In January 2024, one author (SMH-F) used Google Scholar to search for full manuscript publications by lead and senior authors of eligible dissertations and conference abstracts to help ensure full articles were examined for data extraction.

Figure 1

Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow diagram of study coding and inclusion. OECD, Organisation for Economic Co-operation and Development.

Data extraction and abstraction

One author (AYK) conducted data extraction in Microsoft Word, which was confirmed by a second author (SMH-F). Data extraction items included study design; study location; unit of analysis (ie, neighbourhood operationalisation) and sample size; statistical approach; data source and sociodemographic composition variables; data source, operationalisation and sample size of tobacco retailers; TRA operationalisation; and unadjusted effect sizes (online supplemental file C). We extracted and synthesised unadjusted effect sizes, or those associations that examined a relationship between a single sociodemographic characteristic and TRA (eg, the relationship between neighbourhood-level median household income and TRA). Adjusted results (eg, the relationship between neighbourhood percentage of Black residents and TRA controlling for neighbourhood median household income) were not synthesised because they represent what inequities would be observed in a counterfactual scenario and thus do not directly answer our research question about inequities. We included studies that report adjusted results in the evidence table, but for synthesis and analysis purposes, we focused on unadjusted results.

Supplemental material

To aid in synthesis and analysis, we recoded extracted neighbourhood sociodemographic characteristics. The combined categories were SES composition; racially and ethnically minoritised (REM) population composition; White population composition; and same-sex household composition. We additionally recoded SES composition into subcategories (eg, income, education, employment, health insurance status) and REM population (eg, Black population composition). We present results by these larger combined categories as well as the subcategories that make up each. As Hispanic/Latine ethnicity was not consistently described in studies, we report racialised associations including any ethnicity (eg, White composition inclusive of non-Hispanic and Hispanic White composition).

We additionally coded extracted record characteristics for study location type (ie, country, county/city or equivalent, state/province); neighbourhood unit of analysis (ie, census tracts, meshblocks, data zones, dissemination area); data source of tobacco retailers (ie, ground truthing [calling or visiting retailers to confirm location and tobacco product sales]; licensing list or government registry; secondary business establishment database or phone book; combination of strategies); and the operationalisation of TRA (ie, total count of tobacco retailers; tobacco retailers per land area; tobacco retailers per population; tobacco retailers per roadway; presence vs absence of tobacco retailer[s]; other measure). Some studies investigated inequities in multiple places, and we retained all effect sizes for synthesis and analysis. Additionally, some records investigated inequities by measuring TRA multiple ways (eg, tobacco retailers per 1000 population vs tobacco retailers per square mile). As there is no consensus on which measure of TRA may be most appropriate or valid and some measures may capture different geographic constructs,8 9 31 we include all effect sizes reported.

Study risk of bias assessment

One author (AYK) used a modified Downs and Black checklist to assess the risk of bias48 (see online repository protocol and online supplemental file C), and when uncertain, confirmed with a second author (JGLL). We created a risk of bias index (0–7, with higher numbers indicating a higher risk of bias) and a priori planned to exclude studies with a score of 4 or higher. All studies had a risk of bias assessment under 4 (no studies excluded).

Analysis of inequities

Due to social and systemic processes of discrimination and racism used to create and sustain group-based hierarchies to advantage and minoritise specific populations, we define an inequity as having greater TRA in neighbourhoods with lower (vs higher) SES, a greater (vs lesser) concentration of REM residents and more (vs fewer) same-sex households. We coded the indication of inequity for each eligible effect size (ie, presence of inequity; no inequity [effect size was zero or the same for all groups]; or opposite [counterhypothesised] direction, such as having greater TRA in neighbourhoods with a higher proportion of White residents).

Online supplemental file C shows the full evidence extracted, and we present and visualise results with modified harvest plots.49 50 Harvest plots, as we have operationalised them, show the weight of the evidence by the directionality of the result. We additionally coded the statistical significance of each eligible effect size (ie, statistically significant at the traditional p<0.05 threshold; not statistically significant; statistical significance not specified). Records sometimes used multiple statistics to test the same associations: in these instances, two authors (AYK, JGLL) unanimously decided which primary statistical test best answered the research question and included this test for analysis. For example, we prioritised regression coefficients over analysis of variance; spatial regression coefficients over ordinary least squares regressions; and continuous operationalisation of variables over categorical/dichotomous. Following a standard distribution of significance at the p<0.05 level when there is no association, we would expect just 2.5% of results to show a significant negative association, 95% of results to show no association and 2.5% of results to show a significant positive association. Unless otherwise stated, throughout the Results section, we discuss effect sizes for statistically significant results. However, consistent with recommendations not to rely solely on statistical significance when discussing meaningful differences and effect sizes,51 we also present effect sizes where statistical significance was not specified or effect sizes were not deemed to be statistically significant (using 95% CI and p<0.05 thresholds) and note when statistical uncertainty is present. We note that some percentages do not perfectly add to 100% due to rounding.

Results

There were 58 records (figure 1) from six countries (Australia,35 52–56 Canada,36 57 Germany,37 New Zealand,38 58 UK,39 59 USA14 26–34 60–94) that met the inclusion criteria. Two studies84 94 used similar methods to examine identical relationships with the same data, so only the first published study84 was included for analysis (n=57 records). Records were excluded from synthesis and analysis if they only included model-adjusted effect sizes (n=16),29 30 33 34 36 55 56 60 86–93 resulting in 41 records and 205 effect sizes for synthesis and analysis.

Country

Most effect sizes were from studies conducted in the USA (n=172),14 26–28 31 32 61–85 followed by Australia (n=21),35 52–54 Germany (n=5),37 New Zealand (n=4),38 58 UK (n=2, both from Scotland)39 59 and Canada (n=1).57 All or most effect sizes documented inequities in each country (Australia [n=12, 57.1%35 52 53]; Canada [n=1, 100%57]; Germany [n=3, 60.0%37]; New Zealand [n=4, 100%38 58]; UK [n=2, 100%39 59]; USA [n=107, 62.2%14 26–28 31 32 62 64 66 68–71 73 75–85]). Some effect sizes (n=25, 14.5%) were in the opposite direction in the USA31 32 62 69 71 72 75 78 and Australia (n=1, 4.8%)54 (online supplemental file D). Across all records, a total of 72 (35.1%) effect sizes focused on investigating inequities in neighbourhoods across the entirety of the country,28 31 32 35 38 39 52–54 58 59 61 73 74 108 (52.7%) in a county or city equivalent14 26 27 37 57 62 64 66–72 75 76 80–82 and 25 (12.2%, from the USA) within specific states/provinces.63 65 77–79 83–85

Supplemental material

Neighbourhood SES composition

Across studies, 124 effect sizes (60.5% of the total) included a measure of SES.14 26 27 31 32 35 37–39 52–54 57–59 61–64 66–78 80–85 Regardless of statistical significance, 101 (81.5%) effect sizes documented greater TRA in areas with lower SES14 26 27 31 32 35 37–39 52–54 57–59 61–64 66–73 75–78 80–85 while just four effect sizes (3.2%) documented no inequity (figure 2).62 71 However, 81 (65.3%) of documented inequities were statistically significant14 26 27 31 32 35 37–39 52 53 57–59 62 64 66 68–71 73 75–78 80–85 (statistical significance was not specified for 3 [2.4%] of effect sizes).61 63 Specifically, 90 effect sizes examined inequities by poverty status or some index measure of SES, socioeconomic advantage/disadvantage or socioeconomic deprivation (eg, Australian Bureau of Statistics Index of Relative Socioeconomic Advantage and Disadvantage; Scottish Index of Multiple Deprivation). Of these 90 associations, 67 (74.4%) documented greater TRA in neighbourhoods with greater socioeconomic disadvantage (figure 3).14 26 27 31 32 35 37–39 52–54 57–59 61–64 66–78 80–85 Some studies examined inequities by specific measures of SES, such as education42 66 69–72 75 83 (6/13 [46.2%] documented inequities66 71 75 83) and employment37 62 69 71 75 (3/10 [30.0%] documented inequities69 71). One study documented greater TRA in neighbourhoods with a greater proportion of residents without health insurance (although not significant).72 Finally, 10 effect sizes examined inequities by some indicator of housing or neighbourhood structure (eg, vacant housing units, owner-occupied housing),31 32 69 75 with half of these suggesting inequities in the hypothesised direction31 32 75 and half in the opposite direction.31 69

Figure 2

Count and percentage of effect sizes by direction of association and neighbourhood sociodemographic composition (n=205). Opposite Direction indicates that the effect size was in the opposite direction expected for an inequity (eg, greater tobacco retailer availbility in higher income neighbourhoods). No Inequity indicates that the effect size was zero. Presence of Inequity indicates greater tobacco retailer availability in neighbourhoods with a greater concentration of marginalised residents.

Figure 3

Count and percentage of effect sizes by direction of association and neighbourhood socioeconomic composition (n=124). Opposite Direction indicates that the effect size was in the opposite direction expected for an inequity (eg, greater tobacco retailer availability in higher income neighbourhoods). No Inequity indicates that the effect size was zero. Presence of Inequity indicates greater tobacco retailer availability in neighbourhoods with a greater concentration of marginalised residents.

Neighbourhood racial and ethnic composition

Of the 205 effect sizes, 69 (33.7%) tested associations between TRA and neighbourhood composition of residents from REM categories (figure 2), and all records were from the USA.27 31 32 61–63 65–68 70 72–83 Of these, 41 (59.4%)27 31 32 62 66 68 70 73 75–83 documented an inequity of greater TRA in neighbourhoods with a higher proportion of REM while 8 (11.6%)31 32 62 75 78 documented an association in the opposite (counterhypothesised) direction. Specifically, neighbourhoods with a greater composition of Black and Hispanic/Latine residents had greater TRA for 19/29 (65.5%)27 31 32 37 66 68 70 73 77 78 80 81 83 and 19/31 (61.3%)31 62 66 68 73 75 77–81 83 effect sizes, respectively (figure 4). Additionally, inequities were documented for Asian population composition (n=2/5, 40.0%), but statistical significance was not specified.63 Inequities were also observed for effect sizes in studies assessing immigrant status composition (n=1/1, 100%)75 and that combined REM populations (n=2/3, 66.7%).76 82 As seen in figure 4, there were significant effect sizes in the counterhypothesised direction for Asian (n=2/5, 40.0%),32 78 Black (n=4/29, 13.8%)31 62 75 and Hispanic/Latine (n=2/31, 6.5%)31 composition (ie, lower TRA in neighbourhoods with a greater composition of these population groups). A total of 10 effect sizes examined the association between White population composition and TRA,32 63 66 69 71 72 75 and 5 (50.0%) found that TRA decreased as the neighbourhood composition of White people increased32 66 69 71 75 while 3 (30.0%) effect sizes documented the opposite71 (figure 2).

Figure 4

Count and percentage of effect sizes by direction of association and specific neighbourhood racial, ethnic and immigrant status sociodemographic composition (n=69). Opposite Direction indicates that the effect size was in the opposite direction expected for an inequity (eg, greater tobacco retailer availability in higher income neighbourhoods). No Inequity indicates that the effect size was zero. Presence of Inequity indicates greater tobacco retailer availability in neighbourhoods with a greater concentration of marginalised residents.

Neighbourhood same-sex household composition

Two effect sizes tested associations with same-sex couple household composition, and both found greater TRA with greater rates of male and female same-sex households (figure 2).28

Measures of TRA and tobacco retailer data sources

We also examined whether there were differences in the documentation of neighbourhood inequities in tobacco availability by different measures of the construct and data sources used to locate tobacco retailers (figure 5). Per population measures of TRA (eg, number of tobacco retailers per 1000 people) were most common (n=102),14 28 31 32 35 39 53 61 62 65 66 70–73 76 82 followed by per roadway (n=40),27 31 63 68 69 74 77 79 80 83 count (n=18),31 37 38 52 54 58 84 85 land area (n=16),31 57 59 64 67 73 81 presence versus absence of retailers(s) (n=13),31 78 and other measures (eg, percentage of stores selling tobacco products;26 count of block faces with at least one retail outlet divided by the total number of observed block faces per census tract)75 (n=16).26 75 81 For per population measures, regardless of statistical significance, 79 (77.5%) effect sizes documented an inequity in TRA,14 28 31 32 35 39 53 61 62 65 66 70–73 76 82 19 (18.6%) documented associations in the opposite direction31 32 53 62 71 72 and 4 (3.9%) found no inequity.62 71 Most effect sizes for all other measures of TRA (ie, count,31 37 38 52 54 58 84 85 per land area,31 57 59 64 67 73 81 per roadway,27 31 63 68 69 74 77 79 80 83 other26 75 81) documented inequities except for measures indicating the presence versus absence of a tobacco retailer, where 8 (61.5%)31 78 of 13 total effect sizes documented an association in the opposite hypothesised direction, regardless of statistical significance (eg, greater likelihood of a tobacco retailer [vs absence] in neighbourhoods with higher SES).

Figure 5

Count and percentage of effect sizes by direction of association and tobacco retailer data source and availability measure (n=205). Opposite Direction indicates that the effect size was in the opposite direction expected for an inequity (eg, greater tobacco retailer availability in higher income neighbourhoods). No Inequity indicates that the effect size was zero. Presence of Inequity indicates greater tobacco retailer availability in neighbourhoods with a greater concentration of marginalised residents.

Effect sizes from studies that used a tobacco retailer licensing list (n=120)27 35 39 53 54 59 65–72 74 76–83 to locate tobacco retailers were most common, and 76 (63.3%)27 35 39 53 59 66 68–71 76–83 of these documented the presence of an inequity while 13 (10.8%) documented an association in the opposite direction hypothesised.54 69 71 72 78 Most effect sizes for other tobacco retailer data sources (ie, secondary business establishment database28 31 32 61 62 73 84/phone book64 [n=45], ground truthing [n=1926 37 52 75 85]; combination of data sources [n=614 38 57 58]) documented inequities, regardless of statistical significance.

Discussion

Principal findings

This review provides strong evidence of inequities in TRA. The evidence is consistent across measures, methods and six countries. In this review, analyses focused on socioeconomic inequities were most common (n=124 total effect sizes),14 26 27 31 32 35 37–39 52–54 57–59 61–64 66–78 80–85 and greater TRA was documented in over 80% of effect sizes, regardless of statistical significance.14 26 27 31 32 35 37–39 52–54 57–59 61–64 66–73 75–78 80–85 Contrary to expectation, 12.1% of effect sizes documented significantly greater TRA in neighbourhoods with higher SES.31 54 62 71 72 One study in New York City posits that this observation may be due to wealthier residents living in business districts that are more likely to have retailers in general,72 while another in Australia visually explored this pattern and found more retail and entertainment businesses in areas with greater SES.54 It is possible that some neighbourhoods may be so disadvantaged that they have very few or no retailers, thus creating a counterhypothesised effect.

In this systematic review, all records were from cross-sectional studies, and two records were repeated or pooled cross-sectional.26 64 Future work that examines change over time in the associations of TRA and neighbourhood sociodemographic composition may provide new insights.30 For example, there may be shifts in inequities over time due to gentrification that may partially explain why some effect sizes were in the counterhypothesised direction (eg, the few effect sizes that indicated neighbourhoods with higher SES and White population composition had greater TRA). Specific measures of SES (ie, employment,37 62 71 75 educational attainment63 66 69–72 75 83 and health insurance72 rather than deprivation/disadvantage or poverty measures/indices) were uncommon and may warrant future attention.

An overwhelming majority (82.6%)27 31 32 61–63 65–68 70 72–83 of effect sizes pointed to greater TRA in neighbourhoods with a greater composition of REM residents. Lower TRA with an increasing composition of White residents was present in almost two-thirds of effect sizes.32 63 66 69 71 75 These findings also held when looking only at results that are statistically significant under historical thresholds of significance. Thus, this review and synthesis provide compelling evidence that TRA remains an important issue for racial/ethnic health equity and complements prior syntheses showing similarly pervasive inequities in tobacco product marketing.40 95

Of all racial and ethnic groups, associations of TRA with neighbourhood composition of Black (n=29 effect sizes)27 31 32 61 62 65–68 70 72–75 77 78 80 81 83 and Hispanic/Latine residents (n=31 effect sizes)31 32 61–63 65–68 70 72–75 77–81 83 were examined most frequently. Results from this review indicate that neighbourhoods with a greater composition of these population groups overwhelmingly face greater TRA, with very few effect sizes indicating null or counterhypothetical results.

Only four studies examined relationships between TRA and Asian population composition (n=5 total effect sizes),32 63 72 78 and while 2 (40.0%) effect sizes indicated greater TRA in neighbourhoods with a greater Asian population composition (though statistical significance was not specified),63 2 (40.0%) indicated significantly lower TRA in these neighbourhoods.32 78 One study also documented greater TRA in neighbourhoods with a higher percentage of immigrants.75 No studies examined associations by neighbourhood composition of Indigenous people. Future work examining racialised inequities with TRA may be warranted, especially within ethnic enclaves and by disaggregated ethnic categories of Asian and Hispanic/Latine populations, and by Indigenous people composition, though this work may be difficult due to small population sizes and data suppression.

On review of records, we recommend specificity in how racial and ethnic categories are defined, given evidence of substantial variation (online supplemental file C); for example, often ethnicity (eg, Hispanic/Latine) was not specified with race (eg, per cent Black vs per cent non-Hispanic/Latine Black). Additionally, language sometimes did not parallel the data source measures and further did not state how race/ethnicity was measured. Several studies report ‘Caucasian’ race as a synonym for White; authors should avoid the term as a synonym given its origins as part of a system of racial classification and hierarchy used to perpetuate white supremacy.96 97 All studies focused on REM were from the USA, and investigation of racialised inequities in other countries may be warranted.

Although not included in this synthesis, few studies examined other neighbourhood measures of sociodemographic composition, such as racialised (and economic) segregation98 and historical redlining,92 a process that delineated geographical areas as ‘hazardous’ for investment based on sociodemographic population composition. These measures may provide sociopolitical and historical context for how population groups have been segregated into neighbourhoods relative to one another, thereby creating tobacco-related environmental inequities.99

Only one study assessed inequities by same-sex households,28 documenting greater TRA in neighbourhoods with a higher proportion of male and female same-sex households. This evidence does not include single lesbian, gay and bisexual individuals, who may live in more urban areas; however, other research suggests same-sex couples may be a reasonable proxy for assessing inequities for the broader lesbian, gay and bisexual population.100

Secondary findings

While the harvest plots excluded effect sizes that focused exclusively on the retail availability of nicotine vape products (n=6),42–46 60 we briefly examined the findings from these records (online supplemental file B). Overall, there are inconsistent findings concerning differences in vape shop availability by neighbourhood racial, ethnic and socioeconomic composition, and these studies have been limited to the USA. Continued surveillance and monitoring of vape shop availability are needed to determine whether their locations will parallel inequities observed for more retailers selling more traditional tobacco products.

To complement the primary research question, this review abstracted data sources used to measure TRA, as well as how TRA was operationalised. Effect sizes from studies that used a ‘gold standard’ of licensing or government registry lists or ground truthing were more common than studies of secondary business establishment databases/phone books. Regardless of the data source used, most effect sizes indicated inequities. Licensing is needed to track enforcement and compliance of tobacco retailers, and jurisdictions with tobacco retailer licensing should leverage such data to prioritise equitable tobacco retail reduction to help eliminate place-based inequities in tobacco availability. Additionally, separate licences to sell e-cigarettes and vapour products may better allow the surveillance of the vape shop industry.101 102

Regardless of how or for whom TRA was measured, most effect sizes indicated the presence of inequities. However, there was variation in the count and percentage of statistically significant inequities (eg, 58.8% of effect sizes for retailers per population14 28 31 32 35 39 53 62 66 70 71 73 76 82 vs 75.0% of effect sizes for retailers per land area31 57 59 64 73 81 vs 38.5% for dichotomous measures [eg, any vs no tobacco retailers31 78]). Other reviews have discussed the use of different measures of TRA that may capture different aspects of the retail environment,8 9 which was evident in effect sizes that compared and noted differences in both the presence and statistical significance of inequities when using varying measures of TRA.31 Though we cannot conclude which measure of TRA may be most valid, considerations for measure selection should be taken when evaluating inequities in TRA. For example, land area or roadway measures may better capture the space where tobacco retailers are located or clustered; presence versus absence of retailers may not fully capture the concentration of tobacco retailers in a geographical area.31 74 Additionally, some measures of TRA tried to account for the weighted distance of tobacco retailers to some point39 59 61 62 (eg, kernel density estimation),103 as indicated in online supplemental file C. Finally, while this review was focused on TRA, we note that some literature has examined inequities in the proximity (or distance) of population groups to tobacco retailers.37 104

Strengths and limitations

There are several limitations of this systematic review. First, our review is limited to OECD countries, and the literature search ended in 2022. Notably, there are non-OECD country analyses of inequities in TRA,105–107 and continued evaluation of inequities in these countries is needed. We also focused our review on place-based inequities defined by neighbourhood SES, ethnicity and race of neighbourhood residents and same-sex household composition; however, there may be other neighbourhood factors to consider, such as rurality.30 31 33 Finally, there was great heterogeneity across records, including time, policy environment, history, land use planning, country and differences in measurement for both the predictor and outcome variables. Thus, we did not statistically combine records in a meta-analysis. This analysis compellingly documents inequities: however, it does not identify their origins nor are we able to provide an intersectional approach to look at combinations of neighbourhood characteristics manifesting from overlapping systems of oppression that might exacerbate inequities in TRA. Some studies did document inequities by stratification or interactions of neighbourhood racial and SES composition79 while others compared differences in TRA by counties that were matched on racial or socioeconomic composition.69–71

Conclusion

This review of 58 publications and synthesis of 41 studies (2002–2022) from six high-income countries found consistent evidence of widespread neighbourhood inequities in TRA by SES, ethnicity and race. These place-based inequities may contribute to persistent inequities in exposure to tobacco marketing, tobacco use, as well as tobacco-related morbidity and mortality. Interventions to reduce TRA108–110 include restricting retailer location (eg, distance from schools as has been done in Philadelphia111), limiting types of retailers that can sell tobacco (eg, banning pharmacy sales as has been done in Massachusetts112) and limiting the number of licensed retailers in a given area as has been done in San Francisco.113 However, analyses indicate that some of these interventions may widen inequities in TRA (eg, prohibiting the sales of tobacco products in retailer types that are more common in certain neighbourhoods, such as pharmacies).31 72 90 110 Local health equity assessments to determine the impact of different interventions are needed to promote an equitable reduction in TRA and environmental justice.

Overall, we document widespread and persistent racialised and socioeconomic inequities in TRA. We challenge public health and tobacco control researchers, practitioners and policymakers to move beyond merely documenting inequities to partnering with communities to design, implement and evaluate policies and interventions to reduce and eliminate inequities in TRA.

Data availability statement

All data relevant to the study are included in the article or uploaded as supplementary information.

Ethics approval

Not applicable.

Acknowledgments

The authors thank the students who helped conduct the initial screening of record abstracts. We also thank Dr Lance Ford for his assistance with data management.

References

Supplementary materials

Footnotes

  • X @AmandaYKong, @KurtRibisl

  • Contributors AYK conceptualised the study, developed the protocol, coded the data, abstracted the data, conducted the analyses, created the figures and led the drafting of the manuscript. JGLL conceptualised the study, developed the protocol, coded the data, assisted with data abstraction and assisted with drafting of the manuscript. SMH-F confirmed the data extraction. KBS developed and implemented the literature search. SDG, L Henriksen and KMR conceptualised the study and developed the protocol. L Herbert created the figures. All authors provided critical feedback, edited the manuscript and approved its final submission. AYK accepts full responsibility for the work and/or the conduct of the study, had access to the data, and controlled the decision to publish.

  • Funding Research reported in this publication was supported by the National Cancer Institute (NCI) of the National Institutes of Health for Advancing Science & Practice in the Retail Environment (ASPiRE) (P01CA225597) and P30CA225520. Additional support was provided by the Oklahoma Tobacco Settlement Endowment Trust (STCST00400_FY24). SMH-F was further supported by the Cancer Control Education Program, a grant from the UNC Lineberger Comprehensive Cancer Center (NCI: T32CA057726). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health and funders.

  • Competing interests AYK and KMR serve as paid expert consultants in litigation against the tobacco industry. JGLL and KMR hold a royalty interest in tobacco retailer mapping system owned and licensed by the University of North Carolina at Chapel Hill. The software was not used in this research.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.