OUP user menu

The role of socio-economic factors on prevalence and health outcomes of persons with diabetes in Rome, Italy

Valeria Fano, Patrizio Pezzotti, Roberto Gnavi, Katia Bontempi, Maria Miceli, Eugenia Pagnozzi, Maria L. Giarrizzo, Antonio Fortino
DOI: http://dx.doi.org/10.1093/eurpub/cks168 991-997 First published online: 15 December 2012


Background: We investigated the role of socio-economic status on diabetes prevalence, on mortality and hospitalization in a large population-based cohort enrolled in Rome, Italy. Methods: Diabetic residents aged ≥35 years in 2007 were identified using multiple data sources. The effect of the deprivation of the area of residence on diabetes prevalence and on mortality and hospitalization (years 2008–10) was investigated by multilevel regression models, both among diabetic and non-diabetic populations. Results: Prevalence of diabetes (8.3%) was directly related to the deprivation of the area of residence, especially for women. Diabetes increased the risk of mortality and hospitalization, mainly for cardiovascular complications, compared with non-diabetic subjects, with increasing relative risks in more deprived areas. The social gradients observed among diabetic patients are modest compared with non-diabetic subjects, both for some acute complications (myocardial infarction, stroke) and chronic complications (ischaemic heart disease, nephropathy, retinopathy and amputation). Conclusions: Prevalence of diabetes is directly related to deprivation, especially for women. Diabetes increases the risk of mortality and hospitalization for cardiovascular complications. However, similar to another study conducted in Northern Italy, we found that social differences in health outcomes do not differ between people with and without diabetes, suggesting that the care for diabetic patients living in Rome is provided without social disparities, and in some cases, it protects against the adverse effects of social inequalities. The Italian care system for diabetes deserves to be further investigated, as it could represent a model for the care of other chronic conditions and for contrasting social inequities in health.


Diabetes is a serious illness with multiple complications, it can lead to premature mortality, and it has a major impact on national health care expenditure.1 Recent estimates by the International Diabetes Federation2 report a crude prevalence of 8.5% in Europe and of 8.8% in Italy.

An inverse relation between socio-economic status and diabetes prevalence has been reported for many European countries including Italy.35 Obesity, diet, physical activity, lack of diabetes control and access to health services have been reported among socio-economic-related factors that contribute to observed inequalities.6

It has been shown that diabetic population is at higher risk for mortality and incidence of cardiovascular diseases compared with the general population.7 However, although mortality risk increases with worsening of material deprivation,8 the role of socio-economic factors on the incidence of hospitalizations is still controversial.9,10

A recent study10 conducted in two Northern Italian cities showed that diabetes increases both mortality and the incidence of cardiovascular diseases, but it reduces the inverse association with education found among non-diabetics. Authors ascribe the reduced social differences in adverse health outcomes among persons with diabetes to the high accessibility and good quality of health care models implemented in the two areas under study.

A survey conducted in 2004 to assess the quality of diabetes care in Italy11 showed that diabetic patients receive less than optimal care, they are engaged in unhealthy behaviours and receive inadequate treatment for comorbidities.

Quality of diabetes care in central Italy seems to be lower compared with Northern regions and heterogeneous within the region.12,13 Many community services are performed by hospital outpatient departments with low integration between hospital and community and between primary and secondary care. Even though hospital-affiliated centres see their diabetic patients often and provide them with a large number of services, most of these services are not guideline recommended and many patients do not receive high-quality care.

The aim of this study was to investigate the role of socio-economic status on diabetes prevalence, on mortality and hospitalization in a population-based cohort of adult individuals with diabetes in Rome, Italy.


Study population

The study base was of all residents in the Local Health Authority ‘Roma D’ (hereafter defined as RMD), which includes the southern part of the municipality of Rome (2 844 821 inhabitants) and the municipality of Fiumicino (66 510 inhabitants), for 575 912 inhabitants. Residents aged ≥35 years on 1 January 2008 with a diagnosis of diabetes were included in the study. Four data sources were used to identify cases of diabetes:

  1. The Hospital Discharge Registry, which collects information about residents discharged from all national hospitals. We selected all individuals with at least one discharge with a primary or secondary diagnosis of diabetes (ICD9-CM code = 250; period 2003–07).

  2. The Drug Prescription Registry; we selected all individuals filling at least two prescriptions of anti-diabetic drugs (codes of the Anatomical Therapeutic Chemical Classification System, ATC = A10A, A10B) on two different days during 2007.

  3. The Exemption Registry, which includes all subjects who obtained exemption from payment of drugs and laboratory testing because of a diagnosis of a chronic disease. We selected all individuals with the exemption code for diabetes in 2007.

  4. The Outpatient Specialty Care Registry, which includes all diagnostic and therapeutic procedures performed in the Lazio region. We selected all individuals with at least one referral for specialized care during 2007 with the exemption code for diabetes.

All data sources were matched by a deterministic linkage procedure using the tax code as unique identifier. We selected individuals recorded in at least one database. All other residents were considered as non-diabetic.


Subjects were followed up for mortality and hospital admissions occurred in years 2008–10, by record linkage with the Hospital Discharge Registry and the Mortality Registry. The following ICD9-CM codes of primary causes of hospitalization were considered: all cancers (140–239), retinopathy (362), cardiovascular diseases (390–459), myocardial infarction (410 in any position), stroke (398.91;402.01; 402.11;402.91;404.01; 404.03;404.11;404.13; 404.91;404.93;428), renal diseases (250.4;403–404;581;583–586;588; intervention codes in any position 38.95; 38.97; 39.42; 39.95; 54.93), lower limb amputations (895–897 and 250 in any position or DRG codes 113; 114; 285).

Deprivation index

A composite deprivation index developed at census tract level was used as a proxy of the socio-economic status for each resident. The study area comprises 3336 census tracts (inhabitants: mean = 195, SD = 254). The index includes five dimensions of deprivation (education, occupation, housing tenure, family composition and immigration) combined to create a composite indicator. Five categories of deprivation corresponding to the quintiles of the population have been obtained by summing up the standardized indicators: very low, low, medium, high and very high. When patient’s municipality of residence was the city of Rome (89%), a deprivation index standardized on the municipal population was used.14 For patients resident in Fiumicino (11%), a municipal-based index was not available, therefore the analogous deprivation index standardized on the regional population was used.15 Three intermediate categories were grouped into one category named ‘medium’, and all analyses were conducted referring to the three categories ‘low’, ‘medium’ and ‘high’ deprivation.

Individual information on 2001 census tract was retrieved through record linkage with the municipal databases to allow the attribution of the corresponding deprivation index. As for 222 census tracts (∼9000 inhabitants), the deprivation index was not available, an iterative imputation procedure was, therefore, applied to estimate the missing information on deprivation index, based on mean value of the index measured in contiguous census tracts weighed by resident population at 2010.

Statistical analyses

The number of missing cases and the completeness of ascertainment was estimated by applying capture–recapture methods.16

The start of follow-up was defined as 1 January 2008; cause-specific person-years at risk were cumulated until the day of hospitalization, date of death, moving date out of the area of residence or 31 December 2010, when appropriate. We considered as lost to follow-up people who moved out of the area of residence during the study period (3% diabetics, 4% non-diabetics).

The effect of the mean deprivation index on diabetes prevalence was investigated through multilevel logistic regression (Table 1), by grouping patients according to their census tract and by adjusting for gender, age (5 years age group) and nationality at birth (Italian vs. non-Italian). Contextual effects related to the area of residence were taken into account by adding to the regression model the corresponding covariates measured at census tract level (mean age, percentage of male subjects, percentage of non-Italian residents). An interaction term between gender and deprivation index was included in the model to estimate separate odds ratios (OR) for male and female subjects.

View this table:
Table 1

Demographic characteristics of diabetic and non-diabetic population of the Local Health Authority RMD, Italy

Non-diabetic population (n = 304 281)Diabetic population (n = 27 642)OR (CI 95%)
n (%)n (%)
    Women167 067 (54.9)13 694 (49.5)1.00
    Men137 214 (45.1)13 948 (50.5)1.57 (1.48–1.66)
Age, years
    35–64218 060 (71.7)10 086 (36.5)1.00
    65–7450 029 (16.4)9532 (34.5)4.14 (4.02–4.27)
    75+36 192 (11.9)8024 (29.0)4.98 (4.82–5.14)
    Italian279 823 (92.5)26 300 (95.5)1.00
    Non-Italian22 835 (7.5)1255 (4.5)0.87 (0.82–0.93)
        Low39 506 (23.7)2353 (17.2)1.00
        Medium96 003 (57.5)7835 (57.3)1.30 (1.23–1.37)
        High31 328 (18.8)3491 (25.5)1.78 (1.66–1.90)
        Low32 051 (23.4)2721 (19.5)1.00
        Medium78 528 (57.3)8049 (57.8)1.21 (1.15–1.27)
        High26 400 (19.3)3160 (22.7)1.45 (1.36–1.54)
  • OR of diabetes prevalence adjusted for gender, age, deprivation index, nationality, interaction between gender and deprivation; contextual variables are also included in models: mean age, percentage of male subjects, percentage of non-Italian residents.

  • Adults aged ≥35 years, resident and alive as of 1 January 2008. Adjusted OR and 95% confidence intervals (CI) of diabetes prevalence.

  • a: Totals may vary because of missing information.

As for mortality and hospitalization, gender-specific relative risks (RR) by deprivation index (reference: high deprivation) were computed by means of multilevel Poisson regression; models were adjusted by age group and nationality. Analyses were conducted on diabetic vs. non-diabetic patients (Table 2), as well as separately for diabetic and non-diabetic patients (Table 3). When computing RRs for diabetic patients, models were also adjusted for type of therapy prescribed (none, oral drugs, insulin). Covariates obtained by grouping patients by census tract were added to models to assess the effect of a second-level variable. Finally, to test the hypothesis that the severity of the disease could have affected early mortality, further models for mortality were performed, by restricting the data set to patients who had never been hospitalized during the study period. All statistical analyses were performed with the statistical package STATA.17

View this table:
Table 2

Adjusted RR of death and of hospital admissions of diabetic vs. non-diabetic patients by deprivation index of the area of residence

DeprivationDiabetic vs. non-diabetic patientsP-value
nRR (95%CI)nRR (95%CI)
Death (all causes)Low1681.58 (1.34–1.86)2281.58 (1.361.83)1.00
Medium8642.65 (2.38–2.94)10412.99 (2.693.32)0.07
High3452.25 (1.97–2.56)3162.30 (2.002.64)0.81
All13811.59 (1.50–1.69)15951.50 (1.41–1.59)0.16
Hospitalization (all causes)Low7921.35 (1.25–1.45)10171.64 (1.531.75)<0.01
Medium25781.31 (1.26–1.37)30421.71 (1.641.79)<0.01
High11931.37 (1.29–1.46)11881.70 (1.591.81)<0.01
All45701.31 (1.27–1.35)52561.61 (1.56–1.66)<0.01
 All cancersLow1221.06 (0.88–1.28)1851.43 (1.221.67)0.02
Medium3650.96 (0.85–1.08)5851.60 (1.441.77)<0.01
High1440.85 (0.71–1.01)1951.36 (1.171.59)<0.01
All6340.96 (0.88–1.04)9701.43 (1.34–1.54)<0.01
 RetinopathyLow243.69 (2.31–5.90)293.08 (1.974.83)0.59
Medium632.84 (2.00–4.02)923.62 (2.615.01)0.32
High353.70 (2.45–5.58)384.05 (2.706.08)0.76
All1232.82 (2.28–3.48)1593.59 (2.934.39)0.17
 Cardiovascular diseasesLow1861.73 (1.48–2.02)3491.80 (1.602.02)0.69
Medium7121.94 (1.76–2.13)10691.95 (1.792.11)0.95
High3792.39 (2.12–2.69)4252.06 (1.852.31)0.07
All12771.84 (1.73–1.95)18461.81 (1.71–1.91)0.70
 Ischaemic diseasesLow453.24 (2.32–4.53)1352.51 (2.063.06)0.20
Medium1623.40 (2.71–4.26)3762.45 (2.112.83)0.02
High994.85 (3.74–6.28)1552.70 (2.233.26)<0.01
All3062.96 (2.59–3.38)6672.23 (2.04–2.44)<0.01
 Myocardial infarctionLow212.57 (1.59–4.14)512.12 (1.552.92)0.52
Medium893.21 (2.37–4.34)1742.57 (2.063.19)0.24
High584.89 (3.48–6.86)532.13 (1.562.92)<0.01
All1682.62 (2.20–3.13)2791.90 (1.65–2.18)<0.01
 StrokeLow382.08 (1.46–2.97)642.59 (1.923.48)0.36
Medium1722.63 (2.11–3.27)2303.30 (2.674.09)0.14
High1063.98 (3.10–5.11)964.04 (3.105.26)0.94
All3162.40 (2.10–2.74)3922.63 (2.32–2.98)0.32
 Renal diseasesLow213.38 (2.06–5.56)523.21 (2.304.49)0.87
Medium1125.22 (3.81–7.15)1813.97 (3.115.08)0.18
High626.93 (4.85–9.90)694.19 (3.085.70)0.04
All1964.04 (3.38–4.82)3023.11 (2.69–3.59)0.02
 Lower limb amputationsLow243.29 (2.08–5.19)555.11 (3.627.22)0.13
Medium963.84 (2.85–5.18)1444.65 (3.526.14)0.35
High423.70 (2.53–5.40)736.17 (4.478.52)0.04
All1623.85 (3.18–4.66)2724.11 (3.51–4.81)0.60
  • RR adjusted for gender, age, deprivation index and nationality; contextual variables measured at census tract level are also included in models: mean age, percentage of male subjects, percentage of non-Italian residents; P-value: P-value from χ2 test comparing RR increase of male vs. female subjects; total for All may vary due to missing values for deprivation index.

View this table:
Table 3

Adjusted RR of death and of hospital admissions of diabetic and non-diabetic patients by deprivation index of the area of residence

DeprivationNon-diabetic patientsP-valueDiabetic patientsP-value
nRR (95%CI)nRR (95%CI)nRR (95%CI)nRR (95%CI)
Death (all)19091.007851.001681.002281.00
234261.72 (1.58–1.88)31532.04 (1.86–2.24)<0.018641.66 (1.40–1.98)10411.93 (1.66–2.25)0.28
39221.34 (1.21–1.48)8751.57 (1.41–1.75)0.023451.42 (1.17–1.71)3161.46 (1.22–1.74)0.86
Hospitalization (all causes)184651.0063091.007921.0010171.00
220 7911.02 (0.99–1.05)15 9341.07 (1.04–1.10)0.0225780.97 (0.89–1.05)30421.05 (0.97–1.12)0.19
369741.04 (1.01–1.08)54231.08 (1.04–1.12)0.1811931.01 (0.92–1.10)11881.03 (0.94–1.12)0.74
 All cancers114361.0011141.001221.001851.00
234961.00 (0.94–1.07)28421.09 (1.01–1.17)0.093650.90 (0.74–1.11)5851.12 (0.95–1.32)0.12
310820.95 (0.87–1.02)9171.04 (0.95–1.14)0.101440.80 (0.63–1.02)1950.95 (0.78–1.17)0.28
21891.16 (0.87–1.53)1541.07 (0.80–1.44)0.72630.74 (0.46–1.19)921.18 (0.77–1.81)0.19
3661.28 (0.91–1.80)471.02 (0.69–1.49)0.38350.96 (0.57–1.62)381.28 (0.79–2.09)0.46
 Cardiovascular diseases111241.0013711.001861.003491.00
230521.10 (1.03–1.18)34741.09(1.02–1.16)0.787121.11 (0.95–1.31)10691.09 (0.96–1.23)0.73
310401.17 (1.08–1.28)11811.12 (1.03–1.22)0.443791.37 (1.15–1.64)4251.14 (0.99–1.32)0.11
 Ischaemic diseases11461.003791.00451.001351.00
24571.27 (1.05–1.54)9991.13 (1.00–1.27)0.281621.04 (0.74–1.44)3760.98 (0.80–1.19)0.71
31761.52 (1.21–1.90)3811.30 (1.12–1.50)0.24991.48 (1.04–2.10)1551.06 (0.84–1.34)0.12
 Myocardial infarction1831.001621.00211.00511.00
22861.41 (1.10–1.81)4701.27 (1.06–1.52)0.48891.23 (0.76–1.98)1741.21 (0.89–1.66)0.91
3981.50 (1.11–2.02)1921.56 (1.26–1.94)0.82581.87 (1.14–3.09)530.99 (0.67–1.46)0.05
24661.15 (0.96–1.38)4061.29 (1.06–1.56)0.431721.25 (0.88–1.78)2301.30 (0.98–1.71)0.96
31851.53 (1.24–1.90)1521.53 (1.21–1.94)1.001061.90 (1.31–2.76)961.56 (1.13–2.15)0.42
 Renal diseases1611.001061.00211.00521.00
21991.31 (0.98–1.75)3281.34 (1.07–1.67)0.921121.50 (0.94–2.40)1811.26 (0.92–1.71)0.44
3831.76 (1.26–2.45)1001.26 (0.95–1.66)0.13621.99 (1.21–3.27)691.28 (0.89–1.84)0.15
 Lower limb amputations1811.00801.00241.00551.00
21710.84 (0.64–1.10)2581.34 (1.04–1.73)0.01961.14 (0.73–1.78)1440.92 (0.67–1.25)0.37
3851.30 (0.95–1.76)801.21 (0.88–1.66)0.76421.09 (0.66–1.81)731.18 (0.83–1.69)0.82
  • RR adjusted for gender, age, deprivation index, nationality and therapy (only for diabetic patients); contextual variables measured at census tract level are also included in models: mean age, percentage of males, percentage of non-Italian residents; deprivation index: 1 = low, 2 = medium, 3 = high; P-value: P-value from χ2 test comparing RR increase of male vs. female subjects.

The model building procedure included the testing of significance of the random effect for deprivation, as well as the comparison between models that included parameters measured at contextual level vs. models that did not. Results of these procedures showed that both the random effect and the contextual parameters significantly increased the log-likelyhood only for mortality and for some cardiovascular complications. However, to illustrate the results in a homogeneous way, all RRs refer to multilevel random effect models, which also include variables measured at contextual level. To compare the RRs between men and women within the same model (i.e. same outcome), we used the linear combination estimators technique (STATA command lincom).


We identified 331 923 individuals aged ≥35 years alive and resident in RMD on 1 January 2008. A total of 27 642 (8.3%) residents had a diagnosis of diabetes, whereas 304 281 (91.7%) were defined as people without diabetes. A total of 24% of diabetic patients were identified through the Hospital Discharge Registry, 82% through the Drug Prescription Registry, 66% through the Exemption Registry and 44% through the Outpatient Specialty Care Registry. Capture–recapture models showed that the four sources identified >90% of the total number of estimated cases. The demographic characteristics of the two populations are shown in Table 1.

Prevalence of diabetes was 8.3% (9.2% in men and 7.6% in women). Results from multiple logistic regression (Table 1) indicate that diabetes prevalence is higher in elderly population, in Italian born citizens and in men; however, the effect of the deprivation of the area of residence is stronger among women (ORs ranging from 1.30 to 1.78 with increasing deprivation level) than among men (OR ranging from 1.21 to 1.45).

The mean annual crude mortality rate is 12.9 for 100 person-years among diabetic patients, and 3.8 per 100 person-years among people without diabetes. Table 2 shows the estimated adjusted RRs of death and of hospital admissions (overall and for several groups of diagnosis) obtained by Poisson models, for diabetic vs. non-diabetic patients, by deprivation level.

The effect of the variables measured at contextual level followed the same direction as that illustrated for diabetes prevalence in Table 1 for the individual variables (i.e. RRs increased for male vs. female subjects, Italians vs. non-Italians), except for the variable ‘mean age’ for which a protective effect was observed at the census tract level (these RRs are not reported in tables).

Diabetes increased the risk of mortality and of almost all causes of hospitalization both in men and in women (Table 2). Increase in RRs was observed with increasing deprivation for most outcomes, except for cancer hospitalization in women. Moreover, social differences were significantly higher among women compared with men (P-values < 0.05) for ischaemic diseases (medium and high deprivation), myocardial infarction (high deprivation) and renal diseases (high deprivation), whereas deprived men showed higher increases in RRs for all causes hospital admissions (all deprivation categories), all cancers (all deprivation categories) and lower limb amputations (high deprivation category). When mortality without previous hospitalization was considered, results showed the same pattern of mortality, including previous hospitalization (not reported in tables).

Table 3 shows the RRs of mortality and of hospitalization by deprivation levels among people with and without diabetes. Mortality risk increased with increasing deprivation both in men and in women, without differences between diabetic and non-diabetic populations. As for hospitalization, diabetic and non-diabetic women showed a similar social pattern: almost no differences for all causes of hospitalization, but social differences for cardiovascular diseases. The social pattern of men was rather similar to that of women, the main difference being myocardial infarction, showing no social differences among persons with diabetes, stroke and renal diseases showing milder social differences among diabetic. Again, results from analyses of mortality restricted to never hospitalized patients confirmed the results of the entire cohort.


In this population-based study of >27 000 residents of Rome, aged ≥35 years, with a diagnosis of diabetes, the prevalence of the disease was similar to that reported in literature for Italy.2 Prevalence was higher among men, Italian born citizens and in more deprived census tract areas; moreover, social differences were larger in women than in men, suggesting gender differences in life style or in health behaviours across different social strata; these findings are consistent with other studies both in Italy4,5 and in other European countries,3 which ascribed the social and gender gradient in the prevalence of diabetes mainly to the similar social and gender distribution of obesity and sedentary lifestyle. However, even in studies where it was possible to adjust for these two factors, the social differences in prevalence were still observed,3 suggesting the role of other factors, such as stress and cigarette smoking.

Cohort studies of diabetic patients have been conducted in many countries including Italy,4,810,12,1825 investigating either mortality10,12,19,2324 and/or hospitalization.10,12,19,24 However, few Italian studies, all in Northern Italy, have evaluated the role of socio-economic factors on health outcomes of persons with diabetes.4,10

We confirm that diabetes increases the risk of mortality, specifically a 50% increase in men and 59% in women; these results are similar to those reported in other areas of Northern Italy,4,2527 but less marked than what reported in other areas of Europe and North America.8,28 We also show that diabetes increases the risk of hospitalization4 for most of the causes considered an that, for some cardiovascular complications, women are more at risk than men.29,30

As for deprivation, we observed a direct relation of the deprivation index of the area of residence both with mortality and with most causes of hospitalization. This direct relation affects in an almost similar way both diabetic and non-diabetic population. However, for some causes, as cardiovascular diseases and ischaemic heart diseases in men, the effect of social differences was present only among non-diabetic patients.

The effect of social gradients with respect to mortality and hospitalization observed among diabetic subjects is modest compared with those without diabetes, both for some acute (myocardial infarction, stroke) and chronic (ischemic heart disease, nephropathy, retinopathy, amputation) complications, thus suggesting that the same level of care is guaranteed to all persons with diabetes equally. Indeed, in some cases, such as myocardial infarction in male subjects, the presence of diabetes seems to play a protective role on the effect of inequality.

Published results on the relationship between social gradients and health outcomes are controversial across Europe and North America,8,3134 but our results are consistent with a recent multicenter study conducted in two large cities in Northern Italy, that reported a lower social gradient for the risk of death among diabetic compared with non-diabetic patients.10

As far as the risk of hospitalization is concerned, the low effect of deprivation observed among diabetic patients is consistent with recent findings reported elsewhere in Italy10 and in Europe.3 This might indicate that patients are cared regardless of their social status as recently shown in a study from Northern Italy,35 and that being in charge of a diabetologist and/or a general practitioner helps to eliminate the social disadvantages concerning acute cardiovascular complications that are present among non-diabetic patients. This is likely to happen in our study population, as hospitalization and all other form of care in Italy are free of charge and guaranteed to anyone. However, we were only able to explore social differences in access to hospital care; the difference between low social gradient in hospital admissions and the relatively higher social differences in mortality could be partially explained by differences in the quality of outpatient care. The only residual socio-economic difference in hospitalization rates in our cohort was observed among deprived women for cardiovascular complications. This result could reflect some underlying unhealthy lifestyle risk factors that are more common among socially disadvantaged populations, especially for women.3

This study extends to an area of central Italy results that were limited to the Northern cities (Turin, Venice), confirming the low impact of socio-economic position on diabetes care and the peculiarities of the Italian situation with respect to what is known in Europe and in the USA. Compared with previous works, we also added the analysis for a selection of causes of hospitalization, showing some consistency with what has been shown with respect to mortality.

However, our study has some limitations that could affect the results. First, 10% of diabetic patients, probably at a less severe stage of the disease, were not identified, and, consequently, were included in the reference (diabetes-free) population; as these cases represent a minimal percentage of the non-diabetic population, the impact of this misclassification should be considered negligible. On the other hand, as all the subjects included in the diabetic population have a diagnosis made by a physician (i.e. not self-reported), the chance of false-positive cases is low. Second, even if we excluded patients aged <35 years, we could not distinguish between the types of diabetes; as >95% of cases are type 2 diabetes, our results mainly reflect the outcomes of non–insulin-dependent diabetes. Third, lack of information on time of diagnosis and of results of laboratory tests did not allow us to consider the severity and the duration of the disease in the analyses; even though use of anti-diabetic drugs was included in all analyses of diabetic patients as a proxy of severity, some residual confounding could affect our results. Fourth, as mortality follow-up was short, it is possible that the severity of the disease could have affected early mortality, especially among hospitalized patients. However, analyses of mortality restricted to never hospitalized patients confirm the finding of higher RRs compared with non-diabetic population. Fifth, although the literature suggests that locally based deprivation indexes should be used when exploring socio-economic variability in small areas,36 we used the regional-based index in place of the municipal-based one for ∼10% of the population; however, the occurrence of a bias in the estimates should be excluded, as results of the sensitivity analyses, where models included the regional-based deprivation index for the entire data set, yielded to similar results (data not shown). Moreover, we are aware that further socio-economic information based on income level, not available for this data set, could provide more insight on the role of economic disadvantage vs. social disadvantage in diabetes care. Sixth, all analyses are based on claims reimbursed by the National Health Service. Although health care is virtually free of charge for any citizen with a chronic condition, it cannot be excluded that some individuals performed laboratory exams, visits, took drugs or were hospitalized with costs paid by themselves or reimbursed by private health insurances, and thus were not recorded in the data sources considered in this study. As this is more likely to happen in well-off citizens, social differences would be underestimated. Finally, the deprivation index is a group variable, thus it was computed as the average of the deprivation index of each individual living in the census tract. Because of the homogeneity/heterogeneity of the deprivation index of individuals living in the same census tract, the association between deprivation and outcomes could be attenuated.

In conclusion, we confirm that diabetes increases the risk of mortality and hospitalization for cardiovascular complications. We also found that social differences in health outcomes do exist both in diabetic and in non-diabetic populations. However, the intensity of the social gradient does not differ substantially between these two populations, suggesting that the care to persons with diabetes living in Rome, at least at hospital level, is provided without social disparities, and in some cases, it protects from the adverse effects of social inequalities. Furthermore, these results extend to a metropolitan area of central Italy previous results from Northern Italy showing that diabetes reduces the established inverse association of low social position with adverse health outcomes and corroborate the suggestion that the model of diabetes care run in Italy by the networks of diabetes centres and general practitioners contributes to the levelling off of disparities in health care.37 The Italian model of diabetes care deserves to be further investigated, as it could represent a model for the care of other chronic conditions and for contrasting social inequities in health.


The study was partially funded by a project aimed at building disease management models for diabetes, coordinated by the Istituto Superiore di Sanità, and a project aimed at investigating disequalities in health, coordinated by the Istituto Nazionale per la Promozione della Salute delle Popolazioni Migranti ed il Contrasto delle Malattie della Povertà.

Conflicts of interest: None declared.

Prior publications

Preliminary results of this work were presented as oral communications at the XXXIII National Conferences of the Italian Epidemiology Association (22–24 October 2009, Modena, Italy), at the XXXIV National Conferences of the Italian Epidemiology Association (9 November 2010, Florence, Italy) and at the VI Congress ‘Preventing diabetes complications from basic research to patient care’ (5–6 March 2012, Rome, Italy).

Key points

  • This is the first study conducted on a large unselected population of diabetic patients resident in Rome (Italy), to assess the role of socio-economic conditions of the area of residence on prevalence and health outcomes of diabetes.

  • There is an inverse relationship between socio-economic position and diabetes prevalence, hospitalization and mortality.

  • The social gradients of health outcomes observed among diabetic patients are similar, or even lower, compared with those in non-diabetic patients, suggesting that the access to hospital care of diabetic patients is not affected by social disparities.


The authors wish to thank Margaret Becker of the Agency for Public Health of the Lazio Region for her editorial help. V.F., K.B. and M.M. contributed to the data collection. V.F. and P.P. carried out and wrote the statistical methods. V.F., K.B., P.P. and M.L.G. carried out the data analyses. All authors contributed to the data interpretation and to the final version of the manuscript, which was written by V.F., P.P. and R.G.


View Abstract