Pharmacy Deserts Mapped: New Study Reveals Stark Inequalities in Milan and Taranto

Italy is famous for the sheer density of its pharmacies. With roughly one pharmacy for every 3,000 inhabitants, the country sits comfortably above the European average, far ahead of Denmark, where a single pharmacy may serve more than 15,000 people, and behind only Greece, where the ratio approaches one per 1,200. But a new study published in Social Indicators Research argues that this national average conceals a far more troubling picture at street level. By combining census data, municipal open data and geographic information systems, researchers have shown that the distribution of pharmacies within two very different Italian cities, Milan and Taranto, produces sharp local inequalities in access to one of the most frequently used points of contact with the health system.

The research team, led by Stefano Cervellera and Carlo Cusatelli of the University of Bari Aldo Moro, together with colleagues from the University of Naples Federico II and the University of Campania Luigi Vanvitelli, built a spatial-statistical framework that works at an unusually fine scale. Instead of treating a city as a single unit, they analysed ISTAT census sections, the smallest geographical units for which Italian official statistics are released, each typically home to only 200 to 400 residents. Milan alone contains more than 6,000 such sections, covering a population of about 1.24 million. This granularity matters because averages can hide deserts: a city can appear well served overall while entire neighborhoods lack any pharmacy within walking distance.

The choice of study sites was deliberate. Milan is Italy’s economic powerhouse, a compact, densely built metropolis with a population that has grown 7.5 percent since 2011, driven largely by migration. Nearly one resident in five is a foreign citizen, and the city’s ageing index of 197.5 means there are almost two people over 65 for every child under 15. Taranto, by contrast, is a Southern Italian port city shaped by heavy industry and environmental controversy, with a population that has shrunk 7.2 percent over the same period, a foreign population of just 3.8 percent, and an urban fabric fragmented into physically separate islands such as the historic Borgo, the working-class Paolo VI district, and Tamburi, which sits near the steelworks. Comparing the two cities allowed the researchers to test whether the relationship between deprivation and pharmacy access is universal or context-dependent.

At the heart of the analysis lies a modified version of the Caranci deprivation index, a composite measure well established in Italian public health research and used in the national SENTIERI project on polluted sites. The researchers computed a standardized score for each census section by summing four components: male unemployment, lack of car ownership, household overcrowding, and the share of residents in low-skill occupational classes. Because each component was converted to a Z-score before summing, the index is comparable across neighborhoods and cities. The team also constructed a set of demographic predictors from the 2011 census, including the percentages of single-member households, three-member households, households with four or more members, illiterate residents, and foreign residents.

The statistical machinery consisted of two distinct models. Pharmacy presence in each census section was modeled with a binomial logistic regression, estimated separately for Milan and Taranto using the same six predictors. High socio-economic deprivation, defined as falling in the upper quartile of the modified Caranci index, was modeled with a spatial autologistic regression that included a spatial-lag term built from a row-standardized Queen-contiguity matrix. In plain terms, this term asks whether a section’s neighbors are also highly deprived, capturing the tendency of disadvantage to cluster geographically. The authors deliberately kept the two models separate so that descriptive mapping, pharmacy-presence inference and spatial dependence in deprivation would not be conflated.

The results are striking for their asymmetry. In Milan, the deprivation index was positively associated with pharmacy presence, with a coefficient of 0.1449 that was highly significant. In Taranto, the association flipped sign, with a coefficient of minus 0.298 that was also statistically significant. In other words, more deprived sections in Milan are more likely to host a pharmacy, while in Taranto deprivation predicts absence. The authors stress that no single cross-city interpretation of the deprivation effect is supported by the data. Meanwhile, two predictors behaved consistently in both cities: the share of large households and the share of illiterate residents were both negatively and significantly associated with pharmacy presence, suggesting that family structure and educational attainment mark neighborhoods at a disadvantage in access regardless of context.

The spatial model of high deprivation delivered perhaps the study’s most dramatic finding. The spatial-lag term, capturing whether neighboring sections are also highly deprived, was by far the strongest predictor in both cities, with odds ratios of 3.95 in Milan and 4.21 in Taranto. Deprivation, in other words, is intensely clustered: living next to a deprived area nearly quadruples the odds that a section is itself deprived. Other patterns diverged between the cities. Single-member households raised the odds of high deprivation in Milan, with an odds ratio of 1.45, but lowered them in Taranto, at 0.78. The authors attribute this to Milan’s demography: its single-person households include many elderly residents living alone and young migrants in precarious housing, whereas in Taranto they are more likely to be independent adults without dependents. Older housing stock was positively associated with deprivation in Milan but negatively in Taranto, and foreign-resident concentration predicted deprivation only in Milan.

The descriptive maps reinforce the statistical story. In Milan, pharmacy density follows a monocentric pattern, concentrated heavily in the historic and commercial core and thinning toward the fringes, while several densely populated peripheral districts show inhabitants-to-pharmacy ratios exceeding 4,000. In Taranto, the landscape is fragmented, with populated clusters such as Paolo VI and Tamburi showing significant spatial gaps interspersed with isolated, high-burden sections. The authors conclude that market-driven distribution does not align perfectly with demographic density in either city, creating localized pockets of constrained accessibility in both the continuous fabric of the metropolis and the discontinuous layout of the post-industrial port.

Beyond the statistics, the framework had a real-world payoff in Taranto. The municipality’s effort to decentralize pharmaceutical services had been hampered by an older zoning approach based on street names and civic numbers, which produced ambiguities such as duplicate toponyms and contested population assignments in areas labeled simply as buildings without numbers. By redefining pharmacy service areas using official census-section boundaries and assigning population by aggregating those sections, the researchers provided a stable, reproducible and legally defensible demographic base. The census-section methodology supported the decentralization plan through the Conferenza dei Servizi and subsequent administrative proceedings, covering the three decentralization areas of Talsano Sud, Taranto 2 and Paolo VI.

The policy implications reach well beyond Italy. The authors argue that national and regional pharmacy regulations, which rely on rigid demographic quotas such as the one-pharmacy-per-3,000-inhabitants rule, treat a necessary but insufficient condition as if it guaranteed equitable access. Spatial equity, they contend, should sit alongside demographic criteria in planning, and census sections offer a stable, fine-grained unit for linking population data with service locations. The study also acknowledges its limits: dichotomizing the deprivation index at the 75th percentile discards information, and a binary pharmacy-presence measure cannot distinguish a pharmacy serving 10,000 residents from one serving 100. Future work, the authors suggest, could model pharmacy counts directly with Poisson auto-models, retain the deprivation index as a continuous variable, and apply longitudinal data and formal out-of-sample validation. But the core message stands: the relationship between deprivation and pharmacy access is city-specific, and equitable planning of essential services demands fine-grained territorial evidence rather than national averages.

Subject of Research: Spatial inequalities in community pharmacy access and their association with socio-economic deprivation in Italian cities

Article Title: A Spatial-Statistical Framework for Analyzing Territorial Inequalities in Pharmacy Access: Integrated Evidence from Milan and Taranto

Article References: A Spatial-Statistical Framework for Analyzing Territorial Inequalities in Pharmacy Access: Integrated Evidence from Milan and Taranto. (n.d.). https://doi.org/10.1007/s11205-026-03941-6

Image Credits: AI Generated

DOI: 10.1007/s11205-026-03941-6

Keywords: pharmacy access, spatial analysis, territorial inequalities, deprivation index, GIS, logistic regression, Milan, Taranto, health services planning, census sections, spatial autologistic model, urban inequality