Background In 2018, New York City (NYC) implemented a tobacco-free pharmacy law as part of a comprehensive policy approach to curb tobacco use. This study models the reduction in tobacco retailer density following the ban to examine differences in the policy’s impact across neighbourhoods.
Methods Tobacco retailer density per 1000 residents was calculated in July 2017 for each of NYC’s Neighborhood Tabulation Areas (NTAs, n=188) before and after removing pharmacies as licensed tobacco retailers. Pearson correlations and linear regression (with predictors scaled to 10 unit increments) measured associations between the projected change in retailer density after the ban and NTA demographic characteristics.
Results On average, retailer density decreased by 6.8% across neighbourhoods (SD: 6.3), with 17 NTAs experiencing reductions over 15%. Density reduction was greater in NTAs with higher median household income (r: 0.41, B: 1.00, p<0.0001) and a higher proportion of non-Hispanic white residents (r: 0.35, B: 0.79, p<0.0001). NTAs with a higher percentage of adults with less than a high school education (r: −0.44, B: −2.60, p<0.0001) and a higher proportion of Hispanic residents (r: −0.36, B: −1.07, p<0.0001) benefited less from the policy. These relationships held after assessing absolute changes in density (vs per cent change).
Conclusions NYC’s tobacco-free pharmacy law substantially reduces tobacco retailer density overall, but the impact is not equal across neighbourhoods. In order to minimise disparities in the tobacco retail environment, local governments considering a similar ban should supplement this strategy with other retailer restrictions to achieve equitable outcomes.
- public policy
- socioeconomic status
Statistics from Altmetric.com
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.
The sale of cigarettes and other tobacco products in pharmacies has been criticised by public health advocates for decades.1 2 Indeed, marketing an addictive and deadly substance contradicts the claim that pharmacies promote health and wellness in communities. Moreover, some consumers may misperceive the health risks of tobacco when the products are advertised and prominently displayed in retailers that also provide health services and sell medications.2 Recognising this paradox, many public health agencies have urged pharmacies to stop selling tobacco and have publicly supported legislative bans on the practice. Consumer support for this tobacco control strategy is also high. A population-based survey in California found that nearly three-quarters of respondents disagreed that tobacco should be sold in pharmacies; 97% would shop at their usual drug store just as often or more frequently if tobacco products were no longer sold.3 In 2012, nearly 60% of New York City (NYC) residents reported that they would support a policy prohibiting pharmacies from selling tobacco.4 Even the American Pharmacists Association, the largest association of pharmacists in the USA, calls for the discontinuation of tobacco sales in their policy manual.5 With the exception of CVS, which voluntarily stopped selling tobacco in 2014,6 most chain pharmacies and some independent drug stores have ignored these appeals and continue to sell tobacco products, which generate enormous sales revenue. CVS is estimated to have lost nearly US$2 billion in annual sales after discontinuing tobacco sales.7
Emerging evidence suggests that the removal of tobacco from pharmacies could have a positive impact on health. A state-level analysis using point-of-sale purchase data indicated that after CVS discontinued tobacco sales, households in states with CVS pharmacies were significantly more likely to stop purchasing tobacco compared with control households in states with no CVS locations.8 In Massachusetts, the first state with municipalities that banned tobacco sales in pharmacies, there was a statistically significant decrease in smoking prevalence among cities with tobacco-free pharmacy laws compared with those without such laws.9 One of the most beneficial effects of this policy is that it would reduce the density of tobacco retailers in communities, a strong correlate of youth and adult smoking behaviours.10–12 In California and Massachusetts, Jin et al documented a significantly larger reduction in tobacco retailer density over time in cities that implemented a tobacco-free pharmacy law versus cities without such a law.13 Computational modelling confirms that removing tobacco from pharmacies would reduce tobacco retailer density, thereby decreasing accessibility of tobacco products by driving up search and purchase costs.14 15
It is well documented that tobacco retailer density and exposure to tobacco marketing are significantly greater in neighbourhoods with lower income levels and a higher percentage of African American and Hispanic/Latino residents.16 17 Policies that reduce tobacco retailer density, such as tobacco-free pharmacy laws, could conceivably minimise tobacco use disparities if they positively impact communities that have historically been targeted by the tobacco industry. To date, however, only one known study evaluated whether banning tobacco sales in pharmacies has an equitable impact on retailer density across diverse neighbourhoods. Tucker-Seeley et al investigated the association between neighbourhood sociodemographic characteristics and tobacco retailer density in Rhode Island before and after removing pharmacies from the sample.18 Consistent with other studies on the tobacco retail environment,16 17 census tracts with high poverty levels and a greater proportion of African American and Hispanic/Latino residents had higher tobacco retailer density. When pharmacies were excluded as tobacco retailers, however, there were no observed changes in these associations. The authors concluded that a tobacco ban in pharmacies may not address disparities in tobacco retail.18
In August 2017, NYC passed one of the most progressive tobacco control laws in the nation, supplementing its already strong policy measures. In addition to raising minimum prices on cigarettes and little cigars, setting a minimum price for all other tobacco products, increasing retailer licensing fees, creating a retail licence for electronic cigarettes and expanding a ban on smoking in common areas of multiunit buildings, the new legislation aims to drastically reduce the number of tobacco retailers in each of the city’s 59 community districts.19 To achieve this goal, the bill caps the tobacco retail licences in each district at 50% of the current number of licences. No new licences will be issued until the total falls below that cap through retailer attrition. The effects of this particular policy, however, may not materialise for several years since no current tobacco retail dealers will lose their licence as a result of the proposal. Existing retailers may continue to apply for licence renewals, as can new businesses opening in locations where the previous owner held a tobacco licence.19 Indeed, city officials suggest that a 40% reduction in tobacco retailers as a result of this policy is expected to occur approximately 10 years after implementation.19 In contrast, one of the bill’s most notable and celebrated components, a city-wide ban on tobacco sales in pharmacies, will have a more immediate effect. Pharmacies without a current tobacco licence are prohibited from applying for one and, by the end of 2018, all of the city’s pharmacies—both chain and independent—must cease selling tobacco products altogether.19 The goal of this study is to model the reduction in tobacco retailer density across NYC neighbourhoods following the ban on tobacco sales in pharmacies. Additionally, we will examine demographic and socioeconomic correlates of changes in retailer density to examine the equity of the tobacco-free pharmacy law and identify communities that may be more likely to benefit.
Retailer and pharmacy identification
We obtained a list of all licensed tobacco retailers in NYC (n=8291) and their latitude and longitude coordinates from the NYC Open Data Portal in July 2017.20 Pharmacies selling tobacco were identified by searching the retailer list for the names of the three major drug store chains operating in NYC (ie, Duane Reade, Rite Aid, Walgreens), in addition to independent drug stores using the following search terms: ‘health’, ‘pharm’, ‘drug’, ‘rx,’ ‘chemist’ and ‘prescription’. We verified that these stores were in fact pharmacies through online searches or phone calls, if necessary. A total of 510 pharmacies with tobacco licences, 6.2% of licensed tobacco retailers in NYC, were identified using these methods. All retailer locations were then geocoded using Google Earth Pro, which has been used to geocode retailers in previous studies on tobacco retailer density.21 22 Google Earth Pro’s ability to display current satellite imagery and ‘street view’ images of storefronts allowed the research team to verify the accuracy of the geocoded results.
Neighbourhood demographic and socioeconomic characteristics
NYC’s Neighborhood Tabulation Areas (NTAs), subdivisions of the 59 broad community districts, were used as the operational definition of ‘neighbourhood’ for this study. NTAs are groups of census tracts, have a minimum population of 15 000 people and often approximate boundaries of historical neighbourhoods (eg, Upper West Side, Chinatown, Williamsburg).23 Non-residential NTAs (eg, airports, parks/cemeteries) were excluded from analyses (n=7), resulting in a final sample size of 188 NTAs.
Demographic data at the census tract level were gathered using the US Census Bureau’s 2015 American Community Survey 5-year estimates.24 The following variables were used in the analysis: total population size, proportion of residents from major racial/ethnic subgroups (ie, non-Hispanic black, non-Hispanic white, Hispanic and non-Hispanic Asian), proportion of residents with no health insurance, median household income (calculated as the average median household income of all census tracts in the NTA), proportion of the population 25 years and older with less than a high school education and proportion of residents under 18 years old (ie, youth). The research team manually aggregated the raw tract-level data into their respective NTAs to compute these variables.
Tobacco retailer density
Tobacco retailer density was operationalised as the number of retailers per 1000 residents in each NTA, a common metric to calculate retailer density in the tobacco control literature.13 25 26 We first calculated a baseline measure of tobacco retailer density before the pharmacy ban (ie, as of July 2017) (Density1). After removing the retailers identified as pharmacies, we recalculated a second density variable that represented density per 1000 residents after the implementation of the pharmacy ban (Density2). In order to evaluate the impact of the pharmacy ban in each NTA, we computed the per cent reduction in tobacco retailer density . The per cent reduction in tobacco retailer density was the primary outcome of interest and is hereafter referred to as the ‘impact’ of the tobacco-free pharmacy law. Recognising that a relative measure of change may be inflated for NTAs with lower baseline density values, we also calculated the absolute change in density (Density1 – Density2) as a second and confirmatory measure of impact. Higher, positive values for both density change measures represent a greater policy impact on the reduction in tobacco retailer density.
Descriptive statistics were computed for each demographic and retailer density variable. Pearson correlation coefficients measured the presence and strength of linear associations between the modelled change in tobacco retailer density and each demographic variable across all NTAs. Since Pearson correlation tests can be sensitive to outliers and non-normally distributed data, we also ran sensitivity analyses using Spearman’s correlation coefficients, which make fewer assumptions about the data distribution structure. As a measure of effect size, we fit a linear regression model to calculate the mean change in retailer density postban per 10 unit change in an NTA’s demographic characteristics (ie, 10% increase for each demographic variable, US$10 000 increase in median household income). Because measures of relative change (ie, per cent change) can be impacted by baseline values, we also tested associations between NTA demographics and retailer density at baseline to examine whether confounding may contribute to observed outcomes.
We decided to use unadjusted regression models in our primary analyses, a choice dictated by both conceptual and methodological reasons. While adjusted models have the advantage of controlling for confounding, this may not be as important in this type of policy evaluation compared with other types of epidemiological research. In our study, the unadjusted models are more relevant because they indicate what actually happened using a complete census of retailers and neighbourhoods, not the predicted policy impact had all other factors been held constant. Further, we believe that the correlations between our covariates are conceptually meaningful and may not need to be controlled for in a fully adjusted regression. That is, neighbourhoods in NYC that are low income also typically have more racial/ethnic minority residents, lower rates of health insurance, lower levels of education and more youth residents. Attempting to parse out characteristics that singularly ‘explain’ the relationship ignores the important and frequent intersection of these neighbourhood characteristics in NYC. Indeed, there were strong, significant correlations between nearly every covariate in the model, which also introduces the statistical threat of multicollinearity. Model diagnostic testing revealed that SEs are severely inflated in an adjusted model, given the high correlation between virtually all predictor variables.
Finally, geographical analyses are sometimes prone to ‘spatial autocorrelation’ (ie, the values of the dependent variables are related to their spatial distance), which may violate the independence assumption of regression. Using Moran’s I test, we found no statistical evidence of spatial autocorrelation (per cent change in density: I=0.031, p=0.23; absolute change in density: I=−0.015, p=0.43) and proceeded with a standard regression approach. SAS V.9.4 (SAS Institute) was used for all analyses.
Table 1 displays descriptive statistics for neighbourhood demographic characteristics and tobacco retailer density. Although NYC is a racially, ethnically and socioeconomically diverse city, the wide range of values for the demographic variables in each NTA indicates that some neighbourhoods are considerably homogenous. For example, one NTA had 0.12% non-Hispanic black residents (Great Kills, Staten Island), compared with 90.2% in another NTA (Rugby-Remson Village, Brooklyn). Similarly, median annual household income ranged from approximately US$21 000 (East Tremont, Bronx) to over US$170 000 (Upper East Side, Manhattan). Before the tobacco-free pharmacy law, the average number of tobacco retailers in each NTA was 43.1 (SD: 31.0, range: 3–272). Adjusting for population size, the average retailer density per 1000 residents was 0.97 (SD: 0.76, range: 0.11–9.68). After removing all pharmacies from the sample (n=510), average retailer density reduced to 0.91 (SD: 0.68, range: 0.11–8.47). On average, the per cent reduction in tobacco retail density postban across all NTAs was 6.8% (SD: 6.3), but ranged from 0% to 50%. Seventeen NTAs experienced density reductions over 15%.
The impact of the tobacco ban in pharmacies on retailer density was not equal across neighbourhoods. Table 2 displays unadjusted regression coefficients (B) between neighbourhood demographic characteristics (in 10 unit increments) and the per cent and absolute reduction in tobacco retailer density following the ban. Figure 1 presents scatter plots of neighbourhood demographic variables and the per cent reduction in tobacco retailer density with their Pearson correlation coefficients (r) to graphically illustrate the direction and strength of the linear relationships. Racial and ethnic neighbourhood composition was significantly associated with policy impact. Impact was greater in neighbourhoods with a higher percentage of non-Hispanic white residents (r: 0.35, B: 0.79, p<0.0001) and non-Hispanic Asian residents (r: 0.18, B: 0.78, p=0.015). An increasing proportion of non-Hispanic black and Hispanic residents was correlated with lower policy impact (r: −0.18, B: −0.46, P=0.012 and r: −0.36, B: −1.07, p<0.0001, respectively). Socioeconomic indicators produced even larger effect sizes for reductions in retailer density. For example, increases in median household income corresponded with a significantly greater reduction in retailer density postban (r: 0.41, B: 1.00, p<0.0001). On average, for every US$10 000 increase in median household income, tobacco retailer density decreased by 1%. Neighbourhoods with a higher percentage of uninsured residents (r: −0.33, B: −3.75, p<0.0001) and a higher percentage of adults with less than a high school education (r: −0.44, B: −2.60, p<0.0001) benefited less from the policy, as did neighbourhoods with a higher percentage of youth residents (r: −0.42, B: −4.48, p<0.0001). For every 10% decrease in the proportion of youth residents, retailer density decreased by approximately 4.5%. To account for potential biases from the non-normal data distribution and any outliers in the dataset, we replicated the correlation analyses using Spearman’s correlation coefficients; though there were slight differences in the coefficient values, neither the statistical significance nor the magnitude of our findings changed.
Higher median household income was moderately associated with greater retailer density at baseline, perhaps influenced by wealthier residents living in NYC’s dense business districts (eg, Midtown, Soho). Conversely, a greater proportion of youth residents was associated with lower retailer density at baseline. No other demographic or socioeconomic characteristics were associated with baseline retailer density, suggesting that a confounding relationship does not explain the major findings.
When the per cent and absolute changes in retailer density are mapped using geographic information system (GIS) software, geographical patterns emerge (figure 2). High-income neighbourhoods in Manhattan and the more suburban outskirts of other boroughs are more likely to benefit from the new policy. For example, among the seven NTAs experiencing a density reduction greater than 20% include: Brooklyn Heights, Brooklyn (73% non-Hispanic white, median household income: US$104 500, Density1: 0.89); West Brighton, Brooklyn (96% non-Hispanic white, median household income: US$38 100, Density1: 0.20); Yorkville, Manhattan (77% non-Hispanic white, median household income: US$101 700, Density1: 0.67); and Rossville-Woodrow, Staten Island (85% non-Hispanic white, median household income: US$85 000, Density1: 0.33). More disadvantaged neighbourhoods in the city frequently experienced no change in retailer density. Among the NTAs with a 0% change in retailer density following the pharmacy ban include: Crotona Park East, Bronx (30% non-Hispanic black, 67% Hispanic, median household income: US$24 900, Density1: 1.21); Morrisania-Melrose, Bronx (34% non-Hispanic black, 63% Hispanic, median household income: US$27 400, Density1: 1.18); East New York, Brooklyn (71% non-Hispanic black, 25% Hispanic, median household income: US$32 100, Density1: 1.01) and South Jamaica, Queens (66% non-Hispanic Black, 18% Hispanic, median household income: US$53 500, Density1: 0.84). Absolute changes in retailer density followed similar patterns, with the greatest reductions occurring in Midtown Manhattan and the city’s more suburban outskirts (figure 2).
NYC’s newest legislative efforts to reduce tobacco retailer density, including a ban on tobacco sales in pharmacies, have received deserved praise. It is important, however, to evaluate whether policies applied equally across the city will have an equitable impact on neighbourhoods and populations. Strong evidence exists that tobacco tax increases have a proequity effect on the socioeconomic status gradient in smoking,27 but research evaluating the equity impact of other tobacco control measures, including reductions in retailer density, is lacking.28 This study demonstrated that the overall effect of the pharmacy ban on tobacco retailer density is substantial. On average, retailer density decreased by nearly 7% across neighbourhoods, with some NTAs experiencing reductions as high as 50% after policy implementation.
Neighbourhoods that benefit most from this new policy, however, generally have higher income levels, greater educational attainment and a higher proportion of non-Hispanic white residents—demographic groups with a relatively low burden of tobacco use in NYC. In contrast, neighbourhoods in the South Bronx, East and Central Harlem, and North and Central Brooklyn—areas identified as ‘high risk’ for tobacco retail by the American Cancer Society29—experience little to no change in density after implementation of this policy. An underlying reason for this effect is that pharmacies that sell tobacco make up a greater proportion of tobacco retailers in higher income neighbourhoods in NYC, consistent with studies in other cities.30 In lower income areas, where non-chain convenience stores (eg, ‘bodegas’, corner stores) are historically more prevalent and popular for ‘convenience’ needs,31 including tobacco purchases, pharmacy bans are less likely to have a strong impact. Although NYC’s new tobacco control laws include another policy that aims to reduce tobacco retailer density by half in each community district, this is expected to occur over time through retailer attrition.19 It is critical to assess which neighbourhoods might be the first to benefit from comprehensive policies with both immediate and gradual impacts.
Several study limitations are noted. First, NTAs were used to designate city neighbourhoods, but these administrative boundaries may not accurately represent community members’ own conceptualisations of their neighbourhoods. Indeed, different definitions of ‘neighbourhood’ can strongly impact findings in studies examining tobacco retail and access.32 Similarly, the activity spaces of city residents are not confined to the neighbourhoods in which they live; districts where people work and spend free time are also important when examining the intersection between ‘place’ and health. Second, it is possible that not every pharmacy selling tobacco was identified on the licensing list. For example, some of the store names may not have contained our search terms or may have instead listed the name of the owner or parent company. We believe, however, that the sample had high validity due to statements released by the City of New York in which they estimated that approximately 550 pharmacies would be affected by the new policy.19 This number is close to the number of pharmacies we identified in the licensing dataset (n=510). Further, we do not believe that the omission of any unidentified pharmacies would change the consistent linear trends observed. Third, some neighbourhoods that did not benefit directly from the pharmacy tobacco ban may still experience decreases in retailer density through the aforementioned second retailer density policy or other circumstances (eg, store closings). Finally, this study identifies the differential effects of a policy on the tobacco retail environment in communities, but the impact of the policy on residents’ tobacco use behaviours is not yet known. Future studies should monitor the prevalence of tobacco use across neighbourhoods in the years following the city’s new policies to assess their effectiveness and equitable distribution.
A ban on tobacco sales in pharmacies, retailers that often claim to be centres of health and wellness, is a sensible public health strategy to curb tobacco use. Strong evidence exists that tobacco-free pharmacy laws have positive public health effects, including reductions in both smoking prevalence8 9 and tobacco retailer density.13–15 In NYC, the recent ban on tobacco sales in pharmacies decreases tobacco retailer density overall, but the impact of this policy is not evenly distributed across neighbourhoods. Neighbourhoods that historically have had higher exposure to tobacco promotions and the greatest burden of tobacco use are significantly less likely to benefit. NYC has supplemented this policy with a ‘retail agnostic’ plan to reduce all tobacco retail licences by half in each community district, but this is likely to occur gradually over time, and disparities may still persist if some neighbourhoods benefit earlier from the immediate ban on tobacco sales in pharmacies. Although most municipal policies are applied equally across neighbourhoods, their effects may not achieve health equity. In order to minimise health disparities related to the tobacco retail environment, local governments should consider targeted policy approaches that attempt to balance retailer density evenly across neighbourhoods.
What this paper adds
Tobacco-free pharmacy laws are a promising strategy to decrease tobacco retailer density.
Modelling projected changes in density after policy implementation can assess the ban’s impact and equity.
New York City’s tobacco-free pharmacy law is effective, but more likely to benefit neighbourhoods that are largely white and high income.
Targeted policy approaches that balance retailer density across neighbourhoods are needed to minimise disparities in tobacco retail.
Contributors DPG conceived of the study and led writing of the manuscript. TES completed analyses and contributed to the manuscript. CMM conceptualised the analytical design and made substantial revisions to the manuscript. DH contributed to data interpretation and made substantial revisions to the manuscript.
Funding This work was supported by the Office of The Director, National Institutes of Health of the National Institutes of Health (award number DP5OD023064).
Disclaimer The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Competing interests None declared.
Patient consent Not required.
Provenance and peer review Not commissioned; externally peer reviewed.