We conducted a cross-sectional study with two cut-off points. Our study universe consisted of non-institutionalized individuals aged more than 15 years residing in Barcelona (26, 27). The main data sources were the Barcelona Health Survey for the years 2011 and 2016/2017. This survey is conducted every 5 years by the Barcelona Public Health Agency in a representative sample of the population of the city to obtain data on their health, disabilities, health-related behaviors, and their physical and socioeconomic context. Participants are selected in advance and are contacted at home by professional interviewers. The survey is administered through personal, home-based interviews conducted by the professional interviewers.

Dependent variables consisted of self-rated health and the risk of poor mental health. To assess self-rated health, participants were asked “How is your health in general?” Response options were “excellent”, “very good”, “good”, “fair”, and “poor”. We created a dichotomous outcome variable, with the responses “fair” or “poor” coded as “poor self-rated health” (28). Mental health was evaluated using the 12-item version of the General Health Questionnaire (12-GHQ). Participants with a score of ≥3 were classified as having a high probability of poor mental health (29–31).

Explanatory variables consisted of individual socioeconomic position, the level of gentrification, and the year.

To assess socioeconomic position, we used the social class classification of the Spanish Society of Epidemiology, which comprises five categories, based on the individual's current or last occupation (32). These five categories were recoded as follows: (a) social class I (I + II), the highest social class; (b) social class II (III), and (c) social class III (IV + V), the lowest social class.

To evaluate the neighborhoods’ gentrification process, we used the Gentrification Index (GI) (33). The GI is a composite index comprising seven theoretical dimensions associated with various axes of neighborhood socioeconomic transformation: (a) population rejuvenation; (b) changes in the origin of the population, understood as the growth of the population born in countries with a high Human Development Index; (c) changes in family arrangements/values, associated with the growth of single-person adult households; (d) attraction to the population with a university degree; (e) population substitution, understood as a loss of the population without a university degree due to the effect of migration and changes of residence; (f) speed of change, which introduces the transformative capacity of migratory and residential flows to modify the composition of the population; (g) transformations in the housing market, measured through price increases.

The GI yields a final score ranging from 0 to 1, where 0 indicates the lowest degree of gentrification between 2011 and 2017 and 1 indicates the highest degree. We categorized the neighborhoods in three groups based on quartiles (Figure 1): C1 (GI => 0.51), neighborhoods that underwent intense gentrification processes between 2011 and 2017; C2-C3 (GI > 0.30 and GI < 0.51), neighborhoods with moderate or mild processes associated with gentrification; and C4 (GI <= 0.31), neighborhoods not undergoing any significant gentrification processes between 2011 and 2017 (33).

Level of gentrification and sex (i.e., women/men) were used as stratifying variables. We made this decision, based on two hypotheses. Firstly, we hypothesized that health inequalities would increase in neighborhoods with high or moderate-low gentrification processes, while these inequalities would remain stable in non-gentrified neighborhoods. Secondly, we chose to stratify by sex (as a proxy of gender), because we hypothesized that inequalities would be more pronounced among women, as they are more affected by changes related to gentrification, such as changes in businesses, housing prices, and the breakdown of social networks).

Age and the average household disposable income index were included as adjusting variables, to avoid potential confounding factors. The household disposable income index was obtained from data provided by the Barcelona City Council Department of Statistics (34). This index is a theoretical indicator that is constructed from gross household disposable income and gross household disposable income per capita, while also including variables related to the level of education of the resident population, employment status, the characteristics of the motor fleet, and housing prices. The indicator reveals the relative position of each territorial unit in relation to the average of the city, set at a value equal to one hundred.

Firstly, we performed a descriptive analysis of the dependent and sociodemographic variables, by the level of gentrification (i.e., intense, moderate/mild or none) of the neighborhood, sex, and year.

Then, we combined the databases of 2011 and 2017 to estimate if there were any differences in the magnitude of health-related inequalities between 2011 and 2017. We stratified the analysis by sex and level of gentrification.

We estimated the prevalence ratios (PR) and their 95% confidence intervals (95% CI) for poor self-rated health and poor mental health when comparing class III with class I. We fitted three Poisson regression models with robust variance, introducing the interaction between year and social class. The inclusion of this interaction allowed us to identify whether variations in health inequalities during the period were statistically significant. The result of the interaction can be interpreted as a relative difference between the PR of each year (i.e., as a ratio of prevalence ratios) or,PRinteraction=PRofclassIIIvsIin2017/PRofclassIIIvsIin2011The three models were: (1)

Model 1: crude model including only the interaction between year and social class.

In neighborhoods with moderate or mild gentrification processes, the absolute difference in the percentage of women with poor mental health between class III and class I was 1.3 percentage points (Figure 2A). In 2017 this difference was 14.2 percentage points. Thus, inequalities increased during the period by 19.9 points, which was statistically significant. The increase in inequalities was also reflected in the relative difference in PRs (PRi) between 2011 and 2017, which more than doubled in all models [e.g., PRi = 2.51 (95% CI = 1.53–4.17)] (for more details see Supplementary Material Appendix 2). The same pattern was observed in neighborhoods without gentrification processes, but the results were not statistically significant. In neighborhoods with intense gentrification processes, the absolute difference between 2011 and 2017 decreased by 0.8 percentage points, but this decrease was not statistically significant.

Among men (Figure 2B), in neighborhoods with moderate or mild gentrification processes, the absolute difference between class III and class I was −2 in 2011, increasing to 9.5 points in 2017 (11.5 percentage points), which was statistically significant. As among women, this was also reflected in the relative difference of PRs (PRi) between 2011 and 2017, which nearly doubled in all models [e.g., PRi = 1.99 (95% CI = 1.09–3.61)] (Supplementary Material Appendix 2). In neighborhoods with and without intense gentrification processes, we observed no significant differences between 2011 and 2017.

Inequalities in self-rated health among women (Figure 2C) showed the same pattern as in mental health for neighborhoods without gentrification and moderate or mild gentrification processes, although the results were not statistically significant. Among neighborhoods with intense gentrification processes, poor self-rated health increased between 2011 and 2017, but this increase was not statistically significant. In men (Figure 2D), the increase in inequalities between 2011 and 2017 was smaller than that in women, and was not statistically significant.

Gentrification processes lead to several changes in neighborhoods that may affect residents’ health, but their impact differs depending on the social group to which residents belong (35). For example, one of the processes occurring in gentrifying neighborhoods is the influx of private capital for reinvestment in the area, which can modify the neighborhoods’ housing market (36). Thus, the owners of both land and properties may view the urban transformation of neighborhoods as a business opportunity to attract individuals with greater purchasing power or tourists, driving up the prices of housing and commercial premises. Consequently, there may be instances of real estate mobbing, with forced displacement of the most disadvantaged residents to make room for more lucrative real estate opportunities (7).

This type of residential insecurity and fear of being driven out of the neighborhood can cause psychosocial stress (12), negatively affecting mental health (13, 14). In addition, the replacement of traditional stores by tourist-oriented or trendy shops that do not cater to the daily needs of long-term residents from lower social classes directly affects their quality of life and health (6).

These mechanisms, among others, contribute to increasing health inequalities among disadvantaged groups. However, the extent of these phenomena varies, depending on the stage of gentrification. They may be more pronounced in neighborhoods that are recently undergoing changes (mild or moderate gentrification), where most disadvantaged people from lower social classes have not yet been displaced. In contrast, in neighborhoods with intense gentrification (advanced stages of sociodemographic replacement), most residents would tend to have higher purchasing power and, in general, better health status. Even individuals from lower social classes (as measured by occupational classification) who manage to remain in intensely gentrified areas are likely to possess characteristics related to other axes of inequality, such as gender, immigration status, or income, which enable them to adapt to the changes in their neighborhood or even benefit from them. Consequently, health inequalities in intensely gentrified areas may persist or even decrease. Indeed, our findings show that health inequalities were maintained in neighborhoods with intense gentrification between 2011 and 2017, although there was a slight tendency for them to decrease in women.

Finally, in non-gentrified neighborhoods, we observed a non-significant increase in health inequalities between 2011 and 2017. Most of these neighborhoods are inhabited by people from disadvantaged social classes, who may have been particularly affected by the consequences of the economic crisis. Since 2011, Spain implemented austerity policies that affected the viability of the welfare state. These policies included labor market reforms, such as making dismissal cheaper and reducing unemployment benefits, as well as increasing value added tax from 18% to 21% and from 8% to 10%, and implementing health co-payment (37), among other measures. These changes in the economic landscape increased job instability and the cost of living, which had direct repercussions on the finances of the most disadvantaged people, affecting their health.

Although inequalities in mental health were similar in women and men, gentrification could have a more negative effect on women due to the different ways in which they use city spaces, which are often linked to their assigned social roles (19). Even in contemporary times, women still bear a disproportionate burden as caregivers, particularly those in the most disadvantaged social classes, as women in more privileged backgrounds are more likely to hire individuals to perform caregiving tasks. To fulfill these caregiving responsibilities, basic services and facilities must be available in the neighborhood. However, gentrification leads to the loss of local businesses, nearby leisure and care spaces, and neighborhood support networks, making it challenging to carry out daily activities and caregiving and resulting in longer and more time-consuming journeys (19). Gentrification tends to reinforce a city model that prioritizes economic performance and caters to the productive sphere, often neglecting the public reproductive sphere where women, especially those from the most disadvantaged social classes, engage in their daily activities (19).

This study has several limitations. The main limitation is that the sample is not representative at the neighborhood level because the information source (the Barcelona Health Survey) is only representative at the district level. However, this issue may be minimized to some extent by the grouping of neighborhoods. Another limitation is the small sample size of some groups of neighborhoods, which became more pronounced on stratification by different variables. Moreover, the generalisability of the results is limited because the processes related to gentrification are highly influenced by the unique social and political context of each city or territory. Therefore, our results can only be applied to similar contexts. Finally, the GI may not fully capture the gentrification process of the city's neighborhoods due to the complexity of the phenomenon.

Despite these limitations, to our knowledge, this is the first study on health inequalities due to gentrification in Spain and it was carried out by a multidisciplinary public health team and experts in neighborhood change and gentrification. In addition, this study may be useful for local policy makers- or those in similar contexts -to design and implement public policies aiming to reduce social and health inequalities. For future research, it would be interesting to study gentrification with mixed methodologies to further deepen understanding of the relationship between gentrification and health. Similarly, the gender and intercultural perspective should be incorporated, in addition to other axes of inequality such as age.