Dynamics of healthcare inequalities in type 2 diabetes mellitus across the COVID-19 pandemic: a real-world population-based study

Abstract

Background

healthcare inequalities have been widely documented, yet the COVID-19 pandemic may have altered their magnitude and direction. This study aimed to analyse the dynamics of healthcare inequalities in patients with type 2 diabetes mellitus (T2DM) before and after the pandemic using a real-world data approach.

Methods

we conducted a real-world observational study including T2DM patients aged ≥ 45 from the CARhES cohort, a population-based dataset integrating clinical and administrative information from Aragón (Spain), between 2017 and 2022. Healthcare utilisation was assessed at two levels: Primary Care and specialist care. Socioeconomic, clinical, and healthcare variables were retrieved from electronic health records. We described healthcare utilisation patterns and trends across the study period and estimated adjusted prevalence ratios (PRs) for inequality axes (age, gender, migrant status, socioeconomic level, and rurality) using Poisson regression models at three time points (2017, 2020, 2022), adjusting for multimorbidity.

Results

A total of 86,407 T2DM patients were included. Almost all patients consulted a general practitioner (GP) each year and increased over time, while specialist consultations declined during the pandemic and had not recovered by 2022. Socioeconomic inequalities persisted or widened across most axes. After the pandemic, patients aged ≥ 80 were less likely to visit GP than younger patients, reversing pre-pandemic trends. Women continued to show higher GP use than men but fewer consulted with specialist, although the magnitude of the observed difference was small. Immigrants’ access to specialists decreased relative to natives in 2022. active people with low socioeconomic status showed the lowest specialist consultations, while mutualists had the highest during the pandemic. Rural residents maintained greater reliance on Primary Care but fewer specialist visits post-pandemic. The explanatory power of socioeconomic variables declined in GP models, whereas multimorbidity gained influence during and after COVID-19.

Conclusions

Healthcare inequalities among T2DM patients persisted and some patterns shifted after 2020. Although GP utilisation increased, nursing and specialist follow-up did not recover to pre-pandemic levels. These patterns underscore the need for targeted interventions to optimise care for older adults, women, migrants, rural residents, and low-income populations, with a focus on promoting digital inclusion and developing tailored healthcare pathways to advance health equity.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12913-026-15132-7.

Background

Type 2 diabetes mellitus (T2DM) is a chronic disease characterised by high blood glucose levels, which pose serious health risks due to macrovascular and microvascular complications [1]. Effective clinical management is essential for controlling this disease and reducing its impact on health. Therefore, health systems must ensure that these patients have timely and equitable access to preventive care, diagnostic services and evidence-based treatments.

Despite the importance of healthcare, inequalities in access to and utilization of healthcare based on gender, race, socioeconomic status, or place of residence are well known [2]. For example, women have difficulty accessing specialist care and are less likely than men to be hospitalised for the same health problem [3, 4]. People with a low socio-economic status experience similar difficulties, including lower adherence to medication [5, 6]. Immigrants report barriers at all levels of healthcare, and there are also differences in healthcare service use and access by place of residence [7, 8]. These factors lead to poorer management of cardiovascular risk factors in people in vulnerable groups, worsening prognosis and health outcomes [9].

The COVID-19 pandemic was a turning point in the healthcare attention of chronically ill patients, including those with DM. Across Europe, many chronic patients experienced disruptions in Primary Care attention and other levels of care [10] due to lockdowns and the reallocation of healthcare resources. To mitigate these lockdowns in the attention of patients with DM, national governments and public healthcare systems implemented alternative care strategies, such as the large-scale and rapid adoption of telehealth and remote consultations [11]. In parallel, European health authorities and patient organisations promote a renewed focus on patient empowerment and self-management [12], in order to maintain continuity of care under constrained healthcare conditions.

Nonetheless, these strategies had an unequal impact on the population, which could have increased pre-existing inequalities [13, 14]. Evidence on how this differential impact reshaped healthcare utilisation within chronic disease populations remains scarce, and it is still unclear whether care patterns among patients with T2DM, a group whose continuity of care is particularly sensitive to disruption, have returned to pre-pandemic levels [15, 16].Therefore, the objective of this study was to describe socioeconomic inequalities in healthcare utilisation among a population-based cohort of patients with type 2 DM using Real-World Data. We examined disparities across five traditional axes of social stratification (age, sex, country of birth, socioeconomic status, and place of residence) and the dynamics of utilisation at three key points: pre-pandemic (2017), during the pandemic (2020), and post-pandemic (2022).

Methods

Study design, setting and population

This population-based observational study was conducted in the CARhES cohort (CArdiovascular Risk factors for hEalth Services research), a dynamic open cohort based on Real-World Data from the Aragón Health Service. It includes individuals aged 16 and older diagnosed with hypertension, diabetes, or dyslipidemia in Aragón from 2017. Aragón is an autonomous community in northeastern Spain with a population of around 1.3 million. It has an aging population, with a low birth rate and a high proportion of elderly residents, particularly in rural areas [17]. The study protocol was approved by the Ethics Committee for Clinical Research in Aragon (CEICA PI21/148), and informed consent was waived due to the study’s retrospective and population-based nature, with no direct participant interaction. More information about the cohort can be consulted elsewhere [18].

Eligibility for this study was defined as participation in the CARhES cohort with a diagnosis of type 2 DM or at least one prescription for an oral antidiabetic medication in 2017 (the start of follow-up). Women diagnosed with polycystic ovary syndrome and prescribed metformin were excluded from the study, provided they did not have a concurrent diagnosis of Type 2 DM. To improve the specificity of type 2 DM identification, the cohort was restricted to individuals aged ≥ 45 years. In younger populations, antidiabetic prescriptions are frequently related to non–type 2 DM conditions, such as polycystic ovary syndrome, increasing misclassification risk, while type 2 DM prevalence rises from mid-adulthood onward. This age threshold reduces diagnostic heterogeneity and enhances internal validity.

Analyses were conducted separately for three key time points: pre-pandemic (2017), during the pandemic (2020), and post-pandemic (2022). Eligible participants were followed from 2017 until 2022 or until censoring due to death, loss to follow-up, or migration to another autonomous community.

Data sources and variables

Data was obtained from the CARhES cohort. CARhES was constructed through BIGAN, a data lake that integrates all the information available in the healthcare system of the Aragón Health Service. For this study we obtained information of three different categories: socioeconomic, clinical and healthcare utilisation.

Regarding socioeconomic variables, we included information about the five axes previously mentioned. Age was categorized into three groups: 45 to 64 years old, 65 to 79 and ≥ 80. Gender was divided into male and female. Country of birth was categorized into Spain or other. Socioeconomic level was classified based on income and employment status as follows: employed individuals earning < €18,000 per year, employed earning ≥ €18,000 per year, unemployed individuals, pensioners earning < €18,000 per year with free pharmacy access, pensioners earning ≥ €18,000 per year, mutualists and others (including those with special conditions or uninsured status). This classification uses two complementary proxies to capture socioeconomic status: employment situation and pharmaceutical co-payment level. In the Spanish public health system, co-payment rates are directly determined by income thresholds, meaning that co-payment level is a reliable indicator of income. Mutualists are public employees, mainly civil servants, affiliated to the special administrative mutualism schemes, which constitute an insurance regime distinct from the general scheme that covers most of the population. Unlike the remaining categories, mutualists may choose annually between public and private healthcare provision [19]. This distinct coverage arrangement, rather than income level alone, is what differentiates them from the other groups and justifies their separate consideration. This classification, combining employment status with the income-determined pharmaceutical co-payment rates, has been previously applied to characterise socioeconomic position using administrative data in the Spanish setting [20], and was operationalised for the present cohort as described in its profile [18]. This approach provides a coherent, single-dimensional classification of socioeconomic status based on administratively verified data. Using administrative data minimises misclassification, since patient status is formally registered and verified when they join the health system. Finally, place of residence was categorized as urban or rural based on the individual’s basic healthcare area [21].

Information on clinical conditions other than type 2 DM was obtained from the Morbidity Adjusted Groups (GMA) system. GMA integrates all recorded medical diagnoses from primary care and hospital discharge databases. From this data source, we extracted summary indicators for major comorbidities, including heart failure, chronic obstructive pulmonary disease (COPD), ischaemic heart disease, stroke, chronic kidney disease, cirrhosis, dementia, and obesity. In addition, GMA provides an overall morbidity burden measure for each patient. This variable quantifies the cumulative impact of an individual’s health conditions on functional status and healthcare needs, accounting for disease severity, the presence of comorbidities, and healthcare resource utilisation. Further details on the GMA methodology are available elsewhere [22]. Additionally, to characterize the study population, we classified individuals based on their residence in a long-term care facility (institutionalized patients), using information available from the Aragón Health Service healthcare system.

Finally, we obtained data on overall healthcare utilization for these patients, including all healthcare contacts regardless of whether they were diabetes-related. We focused on routine, scheduled healthcare relevant to the ongoing management of type 2 diabetes (Primary Care and specialist care) as our aim was to characterise the continuity of chronic disease management. Emergency department visits were not included, as they capture acute and unplanned demand governed by different determinants Information on healthcare utilisation was obtained from the BIGAN data platform within the framework of the CARhES cohort. BIGAN integrates routinely collected healthcare data for secondary use in research under strict legal, ethical, and governance standards. All data provided to researchers are pseudonymised at source, with direct personal identifiers removed and replaced by unique, non-reversible codes, ensuring that individuals cannot be identified at any stage of the analysis. Data access and use were authorised by the competent Research Ethics Committee (CEICA PI21/148). No identifiable personal information was accessed by the research team.

Statistical analyses

We calculated the proportion of patients who had at least one contact with primary care services, distinguishing between visits to general practitioners and to primary care nurses, as well as the proportion of patients who had at least one consultation with specialist care. Specialist care utilisation was analysed both in aggregate and separately for each medical specialty.

Missing data were minimal for the variables used in stratification and adjustment. Sex, age group, and socioeconomic level were complete in all annual cohorts from 2017 to 2022, while the variable country of birth had only one missing value per year (~ 0.001%). Given this negligible volume of missing data, analyses were conducted using complete-case analysis without imputation. We further verified that missingness in country of birth did not vary meaningfully by socioeconomic level, and that socioeconomic level was complete across all country of birth strata, minimizing the risk of bias due to differential missingness. The eligible number of individuals with type 2 DM analysed for each year can be consulted in Table 1 annex. Also, a close cohort was analysed in order to evaluate population follow-up throughout the study. The results are available in Table 2 annex.

Descriptive statistics were used to summarize the characteristics of the study population. Continuous variables were reported as means with standard deviations (SD). Categorical variables were presented as absolute (N) and relative frequencies (%). Group comparisons (by gender) were conducted using Poisson regression models with a log link function to estimate the corresponding p-values. A significance level of p < 0.05 was considered statistically significant for all analyses. Given the observed gender disparities in healthcare utilisation and comorbidities, stratified analyses by gender were also performed to better capture differential patterns.

To assess changes in healthcare utilisation over time (2017–2022), trend analyses were performed using generalized logistic regression models (GLM) with a logit link function. McFadden’s pseudo R² coefficients were calculated to assess model fit. Multivariable regression models were used to examine the associations between socioeconomic factors (age, gender, socioeconomic level, etc.) and healthcare use outcomes. Poisson regression models with a log link function were employed.

Therefore, for trend analysis of healthcare utilisation over time, generalized linear models with a binomial distribution and logit link were used. For cross-sectional analyses examining inequalities in healthcare use, Poisson regression models with a log link were applied.

The models were estimated separately for three predefined time points: pre-pandemic (2017), during the pandemic (2020) and post-pandemic (2022). The unit of analysis was the annual proportion aggregated at the study year level (2017, 2020 and 2022) rather than a single longitudinal model incorporating repeated individual-level observations. This approach was chosen because the research question focused on describing and comparing population-level patterns at specific time points, rather than modelling individual trajectories over time. The three selected years represent distinct epidemiological periods, and a cross-sectional comparative approach was employed to capture period-specific estimates and examine how these exceptional contextual shifts may have influenced utilisation dynamics”.

Three sequential modelling strategies were implemented: (1) univariate models evaluating each socioeconomic factor independently; (2) multivariable models adjusted for socioeconomic factors; and (3) fully adjusted models additionally accounting for morbidity burden. These analyses were intended to describe adjusted associations and contextualise observed inequalities rather than to infer causal effects.

Poisson regression models with a log link were used to directly estimate prevalence ratios (PRs) and their 95% confidence intervals (95%CIs) for healthcare utilisation outcomes. The 95% CIs and p values from the Poisson models were calculated using robust sandwich-type variance estimators to obtain valid measures of uncertainty for binary outcomes. Finally, to characterise temporal disruptions associated with the COVID-19 pandemic, segmented regression analyses were conducted within an interrupted time series framework, with 2020 defined a priori as the interruption point.

All statistical analyses were conducted in R (version 4.4.2) using the package glm for all regression models.

Results

In this study, a total of 86,407 patients diagnosed with type 2 DM were identified, comprising 54.8% of men. The mean age was 70.4 years and the population was predominantly Spanish. The most frequent socioeconomic level was low-income retired people (57.5%) and living in urban areas (68.9%). Patients with type 2 DM had a high morbidity burden and a high prevalence of comorbidities, especially chronic kidney disease and obesity.

There were statistically significant differences between men and women in our study. Majority of male participants fell within the 45 to 79 age range, while there were more women in the 80 and above age group. Furthermore, women participants also had a lower socioeconomic level and a higher morbidity burden than men. Women had a higher frequency of heart failure, depression, chronic kidney disease, dementia and obesity than men (p < 0.001). The description of the total sample and stratified by gender can be consulted in Table 1.

Table 1.

Characteristics of the study population, categorized by gender

	Total	Men	Women	p
	N = 86,407	N = 47,398	N = 39,009
Age Mean (SD)	70.39 (11.74)	68.41 (11.32)	72.81 (11.80)	< 0.001
Groups of age N(%)				< 0.001
45–64	27,908 (32.30)	18,160 (38.31)	9,748 (24.99)
65–79	36,021 (41.69)	19,973 (42.14)	16,048 (41.14)
≥ 80	22,478 (26.01)	9,265 (19.55)	13,213 (33.87)
Country of birth N(%)				0.027
Spanish	81,616 (94.46)	44,845 (94.61)	36,771 (94.27)
Other	4,790 (5.54)	2,553 (5.39)	2,237 (5.73)
Socioeconomic level N(%)				< 0.001
Actives < 18 K € per year	8,510 (9.85)	5,328 (11.24)	3,182 (8.16)
Actives ≥ 18 K € per year	5,993 (6.93)	4,747 (10.02)	1,246 (3.19)
Unemployed	3,260 (3.77)	1,821 (3.84)	1,439 (3.69)
Mutualist	448 (0.52)	260 (0.55)	188 (0.48)
Pensioners < 18 K € per year	49,663 (57.48)	23,145 (48.83)	26,518 (67.98)
Pensioners ≥ 18 K € per year	16,665 (19.29)	11,337 (23.92)	5,328 (13.66)
Other	1,868 (2.16)	760 (1.60)	1,108 (2.84)
Place of residence N(%)				< 0.001
Rural	26,881 (31.11)	15,224 (32.12)	11,657 (29.88)
Urban	59,526 (68.89)	32,174 (67.88)	27,352 (70.12)
Institutionalized N(%)	929 (1.08)	381 (0.80)	548 (1.40)	< 0.001
Morbidity burden Mean (SD)	10.52 (5.38)	10.01 (5.36)	11.15 (5.33)	< 0.001
Comorbidities N(%)
Heart failure	4,668 (6.68)	2,254 (5.85)	2,414 (7.69)	< 0.001
COPD	5,930 (8.48)	4,363 (11.33)	1,567 (4.99)	< 0.001
Depression	12,676 (18.14)	4,052 (10.52)	8,624 (27.48)	< 0.001
heart disease	8,586 (12.28)	6,107 (15.86)	2,479 (7.90)	< 0.001
Stroke	4,239 (6.07)	2,424 (6.29)	1,815 (5.78)	0.005
Chronic kidney disease	15,390 (22.02)	7,964 (20.68)	7,426 (23.67)	0.001
Cirrhosis	4,116 (5.89)	2,417 (6.28)	1,699 (5.41)	0.001
Dementia	3,640 (5.21)	1,220 (3.17)	2,420 (7.71)	0.001
Obesity	17,481 (25.01)	8,361 (21.71)	9,120 (29.06)	0.001

N: number; p: statistical significance; SD: standard deviation; % percentage; COPD: chronic obstructive pulmonary disease

We studied the dynamics of health care use by gender for the period 2017–2022. Figure 1 shows that nearly all patients had at least one general practitioner visit during the study period, with consistently higher utilisation among women, although sex differences narrowed over time. Overall use of general practitioner services increased significantly (p < 0.001). In contrast, primary care nursing visits were more frequent among women but showed an increase up to the onset of the COVID-19 pandemic and a marked decline thereafter, resulting in an overall decreasing trend for both sexes (p < 0.001). Specialist care utilisation also declined significantly between 2017 and 2022 (p < 0.001), with a clear inflection point in 2020 and levels in 2022 remaining below those observed before the pandemic.

Fig. 1.

Dynamics of healthcare attention by gender. 2017–2022

When analysing the different specialists associated with the attention of patients with type 2 DM for the period 2017–2022, we observed different trends by specialist and gender. In men, the proportion of patients who visited cardiologists and nephrologists at least once a year increased. On the contrary, visits to vascular surgeons and ophthalmologists decreased. For women, none of the specialists analysed increased over this period, and visits to endocrinologists, vascular surgeons and ophthalmologists decreased in a statistically significant way between 2017 and 2022 (p < 0.001) (annex Table 3).

The number of patients analysed for each year and the absolute magnitude of healthcare use and its associated uncertainty, can be consulted in annex Table 4. Also, the intensity of visits was calculated among users (≥ 1 visit) with the aim of characterizing the frequency of follow-up once contact with the healthcare system exists. This information can be consulted in annex Table 5.

With regard to the determinants of having visited a general practitioner at least once (Table 2), the likelihood was higher among older people, women, natives, employed individuals, low-income pensioners and those living in rural areas. After adjusting for patients’ comorbidities, older people were less likely to consult a general practitioner than younger people, and these differences persisted and even increased slightly after the pandemic (PR for people aged 80 and over: 0.95; 95%CI: 0.95–0.96). Differences in general practitioner use between pensioners and the active population increased, showing lower use in pensioners after adjustment for comorbidities. Mutualists changed from a higher use than the active population in the pre-pandemic period to a lower use post-pandemic (PR: 0.75; 95%CI: 0.73–0.78). Differences by gender and place of residence remained stable over the period. In terms of model fit, it decreased during the pandemic period. Regarding the post-pandemic period, the explanatory power of socioeconomic variables was lower than in the pre-pandemic period, but on the contrary, the inclusion of comorbidities increased the explanatory power more than in the pre-pandemic period.

Table 2.

Social determinants of general practitioner use in pre-pandemic, pandemic and post-pandemic. Univariate and multivariate models

	2017			2020			2022
	Model 1 PR (95%CI)	Model 2 PR (95%CI)	Model 3 PR (95%CI)	Model 1 PR (95%CI)	Model 2 PR (95%CI)	Model 3 PR (95%CI)	Model 1 PR (95%CI)	Model 2 PR (95%CI)	Model 3 PR (95%CI)
Age
65–79	1.12(1.12–1.13)	1.03(1.02–1.04)	0.96(0.95–0.96)	1.03(1.02–1.03)	1.02(1.01–1.02)	0.94(0.93–0.95)	1.09(1.09–1.10)	1.07(1.06–1.08)	0.97(0.96–0.98)
≥ 80	1.36(1.35–1.37)	1.19(1.18–1.20)	1.03(1.02–1.03)	1.21(1.21–1.22)	1.14(1.14–1.15)	0.97(0.96–0.98)	1.27(1.26–1.28)	1.19(1.18–1.20)	0.95(0.95–0.96)
Sex
Women Country of birth	1.19(1.19–1.20)	1.14(1.13–1.14)	1.13(1.13–1.14)	1.18(1.18–1.19)	1.15(1.15–1.16)	1.14(1.14–1.15)	1.18(1.18–1.19)	1.14(1.14–1.15)	1.13(1.13–1.14)
Foreigner Socioeconomic level	0.84(0.83–0.85)	0.93 (0.92–0.94)	0.96(0.95–0.97)	0.93(0.92–0.94)	0.96(0.95–0.97)	0.99(0.98-1.00)	0.91(0.90–0.92)	0.96(0.95–0.97)	0.99(0.98-1.00)
Active ≥ 18 K	0.88(0.87–0.89)	0.91(0.90–0.92)	0.93(0.92–0.94)	0.87(0.86–0.88)	0.90(0.89–0.91)	0.92(0.91–0.93)	0.83(0.82–0.84)	0.87(0.86–0.88)	0.90(0.89–0.91)
Unemployed	1.01(1.00-1.03)	1.03(1.01–1.04)	0.97(0.96–0.98)	0.91(0.90–0.92)	0.92(0.91–0.93)	0.87(0.85–0.88)	0.96(0.95–0.98)	0.98(0.96–0.99)	0.90(0.89–0.91)
Mutualist	1.25(1.21–1.29)	1.22(1.18–1.25)	1.17(1.13–1.20)	0.95(0.92–0.98)	0.94(0.91–0.97)	0.91(0.88–0.94)	0.79(0.77–0.82)	0.78(0.75–0.80)	0.75(0.73–0.78)
Pensioner < 18 K	1.26(1.25–1.27)	1.12(1.11–1.13)	0.99(0.98–0.99)	1.05(1.04–1.06)	0.96(0.96–0.97)	0.85(0.84–0.85)	1.10(1.09–1.11)	0.98(0.97–0.99)	0.83(0.82–0.84)
Pensioner ≥ 18 K	1.01(1.00-1.01)	0.99(0.98-1.0)	0.91(0.90–0.92)	0.86(0.85–0.86)	0.86(0.85–0.87)	0.79(0.78–0.80)	0.91(0.90–0.92)	0.89(0.88–0.90)	0.80(0.79–0.81)
Other	1.04(1.02–1.06)	1.02(1.01–1.04)	0.98(0.97-1.00)	0.87(0.86–0.89)	0.85(0.84–0.87)	0.82(0.80–0.83)	0.91(0.90–0.93)	0.89(0.88–0.91)	0.84(0.83–0.86)
Residence
Urban	0.76(0.76–0.77)	0.78 (0.77–0.78)	0.77(0.77–0.77)	0.79(0.78–0.79)	0.80(0.80–0.80)	0.79(0.79–0.80)	0.76(0.75–0.76)	0.77(0.76–0.77)	0.76(0.76–0.76)
R2	-	0.0713	0.2082	-	0.0421	0.1552	-	0.0613	0.2681

Model1: univariate; Model2: adjusted by age, sex, country of birth, socioeconomic level and residence; model3: adjusted by variables of model 2 and morbidity burden; PR: Prevalence ratio; 95%CI: 95%Confidence interval; Reference categories: 45–64, men, Spanish, active<18 K, rural residence; R2: model fit

Analysis of the dynamics of primary care nurse utilisation (Table 3) showed higher use in the pre-pandemic period among older individuals, native-born patients, pensioners, and those residing in rural areas. The highest prevalence ratios were observed in patients aged ≥ 80 years, although this association weakened over time. After adjustment for comorbidities, women exhibited lower utilisation than men during and after the pandemic. Socioeconomic gradients were also observed, with pensioners, particularly those with incomes < 18 K, showing the highest likelihood of at least one annual nurse visit, although this excess risk diminished across the study. Urban residence was consistently associated with lower use compared with rural areas. As observed in the general practitioner models, the explanatory contribution of socioeconomic variables declined over time (from 0.0808 to 0.0549), whereas the explanatory power of comorbidity-adjusted models increased in the post-pandemic period.

Table 3.

Social determinants of primary care nurse use in pre-pandemic, pandemic and post-pandemic. Univariate and multivariate models

	2017			2020			2022
	Model 1 PR(95%CI)	Model 2 PR(95%CI)	Model 3 PR(95%CI)	Model 1 PR(95%CI)	Model 2 PR(95%CI)	Model 3 PR(95%CI)	Model 1 PR(95%CI)	Model 2 PR(95%CI)	Model 3 PR(95%CI)
Age
65–79	1.38(1.37–1.39)	1.10(1.09–1.11)	1.03(1.03–1.04)	1.34(1.33–1.35)	1.11(1.10–1.12)	1.04(1.03–1.05)	1.32(1.32–1.33)	1.13(1.12–1.14)	1.03(1.02–1.04)
≥ 80	1.81(1.80–1.82)	1.39(1.37–1.40)	1.22(1.21–1.23)	1.71(1.69–1.72)	1.36(1.35–1.37)	1.17(1.16–1.18)	1.61(1.60–1.62)	1.33(1.32–1.34)	1.09(1.08–1.10)
Sex
Women Country of birth	1.09(1.09–1.10)	1.00(1.00-1.01)	1.00(1.00- 1.01)	1.08(1.08–1.09)	0.99(0.99-1.00)	0.99(0.98–0.99)	1.05(1.05–1.06)	0.97(0.96–0.97)	0.97(0.96–0.97)
Foreigner Socioeconomic level	0.64(0.63–0.64)	0.84(0.83–0.85)	0.86(0.85–0.88)	0.74(0.73–0.75)	0.94(0.93–0.95)	0.96(0.94–0.97)	0.73(0.72–0.74)	0.88(0.87–0.89)	0.90(0.88–0.91)
Active ≥ 18 K	0.89(0.87–0.90)	0.88(0.86–0.89)	0.90(0.88–0.91)	0.85(0.83–0.86)	0.85(0.84–0.87)	0.87(0.86–0.89)	0.88(0.87–0.89)	0.88(0.86–0.89)	0.90(0.89–0.92)
Unemployed	1.19(1.17–1.22)	1.23(1.20–1.25)	1.16(1.14–1.18)	1.05(1.03–1.07)	1.07(1.05–1.09)	1.01(0.99–1.03)	1.10(1.08–1.12)	1.13(1.11–1.15)	1.05(1.03–1.07)
Mutualist	1.33(1.28–1.38)	1.18(1.14–1.23)	1.13(1.09–1.18)	1.22(1.17–1.27)	1.11(1.06–1.15)	1.08(1.04–1.12)	1.09(1.04–1.13)	0.99(0.94–1.03)	0.96(0.92-1.00)
Pensioner < 18 K	1.89(1.87–1.92)	1.54(1.52–1.56)	1.36(1.34–1.38)	1.63(1.62–1.65)	1.38(1.36–1.36)	1.22(1.20–1.24)	1.56(1.55–1.58)	1.32(1.30–1.34)	1.14(1.12–1.15)
Pensioner ≥ 18 K	1.46(1.45–1.48)	1.30(1.28–1.31)	1.19(1.18–1.21)	1.28(1.27–1.30)	1.16(1.15–1.18)	1.08(1.06–1.09)	1.24(1.23–1.25)	1.11(1.09–1.12)	1.01(0.99–1.02)
Other	1.21(1.18–1.23)	1.22(1.19–1.24)	1.16(1.13–1.19)	1.05(1.02–1.07)	0.84(0.83–0.84)	0.99(0.97–1.02)	1.09(1.06–1.11)	1.10(1.07–1.12)	1.04(1.02–1.06)
Residence
Urban	0.78(0.78–0.79)	0.81(0.80–0.81)	0.80(0.80–0.80)	0.81(0.81–0.82)	0.84(0.83–0.84)	0.83(0.83–0.84)	0.81(0.81–0.82)	0.84(0.83–0.84)	0.83(0.82–0.83)
R2	-	0.0808	0.1556	-	0.0611	0.1293	-	0.0549	0.1605

Model1: univariate; Model2: adjusted by age, sex, country of birth, socioeconomic level and residence; model3: adjusted by variables of model 2 and morbidity burden; PR: Prevalence ratio; 95%CI: 95%Confidence interval; Reference categories: 45–64, men, Spanish, active<18 K, rural residence; R2:model fit

Regarding the evolution of specialist consultation (Table 4), we observed that this was more frequent in patients with type 2 DM from 45 to 64 years-old, while people from 65 to 79 showed the lowest risk of attendance. Men had a statistically significant higher risk of visiting a specialist than women, remaining stable for the period analysed. There were no differences by country of birth during the pre and pandemic period, but we observed a lower risk of specialist use in foreigners after the pandemic (PR: 0.97). Workers of low socioeconomic level showed the lower risk of specialist attendance, while mutualists showed the highest use during the pandemic. Regarding the place of residence, a higher risk of visiting a specialist was observed in patients living in urban areas after the pandemic.

Table 4.

Social determinants of specialist use in pre-pandemic, pandemic and post-pandemic. Univariate and multivariate models

	2017			2020			2022
	Model 1 PR(95%CI)	Model 2 PR(95%CI)	Model 3 PR(95%CI)	Model 1 PR(95%CI)	Model 2 PR(95%CI)	Model 3 PR(95%CI)	Model 1 PR(95%CI)	Model 2 PR(95%CI)	Model 3 PR(95%CI)
Age
65–79	1.03(1.02–1.05)	0.97(0.96–0.99)	0.96(0.95–0.98)	1.01(0.99–1.02)	0.95(0.93–0.96)	0.96(0.94–0.97)	1.04(1.02–1.05)	0.99(0.98–1.01)	0.92(0.91–0.94)
≥ 80	1.12(1.10–1.13)	1.04(1.02–1.06)	1.01(0.99–1.02)	1.08(1.06–1.09)	1.01(0.99–1.03)	1.01(0.99–1.03)	1.14(1.13–1.16)	1.09(1.07–1.11)	1.01(0.99–1.03)
Sex
Women Country of birth	1.00(0.99–1.01)	0.98(0.97–0.99)	0.98(0.97–0.99)	0.99(0.98–1.01)	0.97(0.96–0.98)	0.98(0.97–0.99)	1.00(0.99–1.01)	0.98(0.97–0.99)	0.97(0.96–0.98)
Foreigner Socioeconomic level	0.95(0.93–0.98)	1.00(0.97–1.03)	1.02(0.99–1.05)	1.03(1.00-1.05)	1.05(1.02–1.08)	1.01(0.98–1.04)	0.95(0.93–0.98)	0.97(0.95-1.00)	0.97(0.94–0.99)
Active ≥ 18 K	0.98(0.95–1.01)	0.97(0.94-1.00)	1.01(0.98–1.04)	0.93(0.90–0.96)	0.93(0.91–0.96)	0.92(0.90–0.95)	0.96(0.93–0.98)	0.95(0.92–0.97)	1.00(0.97–1.03)
Unemployed	1.03(1.00-1.07)	1.03(1.00-1.07)	1.07(1.03–1.11)	1.02(0.98–1.05)	1.02(0.98–1.05)	1.02(0.98–1.05)	1.11(1.08–1.15)	1.11(1.08–1.15)	1.08(1.04–1.11)
Mutualist	1.15(1.06–1.24)	1.15(1.06–1.24)	1.11(1.02–1.20)	1.08(1.00-1.16)	1.11(1.03–1.20)	1.18(1.09–1.27)	1.02(0.95–1.10)	1.00(0.92–1.08)	1.07(0.99–1.15)
Pensioner <18 K	1.13(1.11–1.15)	1.13(1.10–1.16)	1.12(1.09–1.15)	1.09(1.06–1.11)	1.13(1.10–1.15)	1.06(1.04–1.09)	1.12(1.10–1.14)	1.10(1.07–1.12)	1.12(1.09–1.14)
Pensioner ≥18 K	1.06(1.04–1.09)	1.07(1.04–1.10)	1.07(1.04–1.10)	0.99(0.97–1.01)	1.03(1.01–1.06)	1.02(0.99–1.04)	1.05(1.03–1.08)	1.04(1.02–1.06)	1.07(1.05–1.10)
Other	1.06(1.02–1.11)	1.07(1.02–1.12)	1.08(1.04–1.13)	1.06(1.01–1.11)	1.07(1.02–1.12)	1.05(1.00-1.10)	1.05(1.01–1.10)	1.06(1.01–1.11)	1.12(1.07–1.17)
Residence
Urban	1.01(1.00-1.02)	1.02(1.01–1.03)	1.01(1.00-1.02)	0.98(0.97-1.00)	0.99(0.98–1.01)	0.99(0.97-1.00)	1.03(1.01–1.04)	1.03(1.02–1.05)	1.03(1.02–1.04)
R2	-	0.0047	0.0088	-	0.0044	0.0065	-	0.0048	0.0170

Model1: univariate; Model2: adjusted by age, sex, country of birth, socioeconomic level and residence; model3: adjusted by variables of model 2 and morbidity burden; PR: Prevalence ratio; 95%CI: 95%Confidence interval; Reference categories: 45–64, men, Spanish, active<18 K, rural residence; R2: model fit

The explanatory power of socioeconomic variables in the specialist’s models were lower than those for primary care. The model including comorbidities for the post-pandemic period showed the highest explanatory power (R² 0.017).

Finally, the segmented regression showed an immediate decline in 2020 for some care levels, such as primary care nursing and specialist consultation, followed by a partial recovery or a change in post-2020 trends. This pattern was particularly evident in general practitioner visits, which showed a positive change in slope after 2020 (Table 6 Annex).

Discussion

Our findings show that patients with type 2 DM have frequent contacts in both primary care and specialists across the study period. We observed notable differences in how type 2 DM patients accessed the healthcare services being studied, showing social disparities in the use.

The COVID-19 pandemic substantially altered healthcare utilisation patterns. In 2020, visits to primary care nurses and medical specialists declined and had not fully recovered by 2022, whereas consultations with general practitioners continued to rise. This divergence reflects the reorganisation of care delivery during the COVID-19 pandemic. Non-urgent specialist visits were postponed, and many primary care nursing activities were reduced or reassigned to pandemic response tasks. In contrast, primary care physicians played a central role in maintaining continuity of care for chronic patients, rapidly expanding telemedicine and assuming increased administrative workload, which sustained general practitioner visits [23, 24].

In our study, specialist and nursing services utilisation remained below pre‑pandemic levels by 2022, indicating that a full return to pre‑COVID‑19 care patterns had not yet occurred. In addition to ongoing system pressures and the limited virtual adaptability of some specialist services, structural changes in referral pathways and the primary–specialist interface may have shifted greater case management responsibilities to primary care, reducing referrals for non-urgent conditions. Variations in referral policies and structural barriers within heathcare system have been identified as determinants of specialist access beyond pandemic restrictions. Moreover, the continued expansion of telemedicine may have partially substituted in-person specialist visits, with heterogeneous effects across specialties and conditions [25, 26]. Importantly, the shift towards remotely delivered care does not appear to be experienced uniformly by patients. Population-based evidence shows that telephone and digital appointments are associated with less favourable patient-reported experiences than in-person visits, including insufficient time allocation, fewer opportunities to ask questions, and lower perceived responsiveness to needs [27]. Such differential experiences with teleconsultations may have discouraged some patients from engaging with remotely delivered specialist care, plausibly contributing to the incomplete recovery of specialist utilisation observed in our cohort. Nonetheless, specialty-specific inequalities may exist that were not captured by this analytical approach, representing an important area for future research.

The role of social determinants in healthcare utilisation is evident in our results. Regarding age, typically, older adults have higher healthcare utilisation, which was observed in our pre-pandemic data. However, both in the pandemic and in the post-pandemic period, the oldest patients with type 2 DM showed a lower risk of using primary care than younger patients. One hypothesis is that older patients encountered new barriers to access. The rapid digitalization of healthcare, such as online appointment systems or teleconsultations, could be excluding elderly patients who are less comfortable with technology [28–30]. Moreover, early in the pandemic many older individuals were advised to stay home and may have been fearful of in-person visits, a behavior that could persist at 2022. This finding should be explored, as it could imply that a group of patients with high clinical risk are not engaging with primary care as much as before, potentially leading to unmet needs.

Gender disparities were persistent throughout the study period. Female patients consistently made more use of primary care services than males, yet women were less likely to access specialist care. In our context, women with type 2 DM visited their general practitioners at higher rates than men. These findings have already been observed for both the general population and women with diabetes [31, 32]. This has been related with greater health consciousness but also with higher needs, such as the higher presence of comorbidities in women. On the contrary, the lower rate of specialist use could imply they may not be getting equitable access to advanced diabetes care or consultations. This gender gap in referrals and specialist utilisation continued in the post-pandemic period. Our findings are concordance with prior research indicating that women often face barriers in reaching specialist care despite higher engagement with primary care [6], and could be related with the existence of systemic gender inequalities, including provider biases, differences in health-seeking behavior, or care pathways that favor men.

Prior to the pandemic, immigrants with type 2 DM already used less Primary Care than non-immigrants, even after adjusting for morbidity burden. These findings are consistent with known barriers immigrant populations face in accessing care [33, 34]. Notably, there was no significant pre-pandemic gap in specialist care use between immigrants and natives in our cohort, indicating that, once in the healthcare system, immigrant patients with diabetes were being referred to specialists similarly than native patients. This has changed in the post-pandemic period, as immigrants were less likely to visit specialists than non-immigrants with similar morbidity profiles. This finding could be explained by the exacerbating of known barriers, such as the uncertainty of navigating the healthcare system. But also, some new challenges should be considered, such as barriers related with telemedicine or with the strain of healthcare services, that may have unconsciously de-prioritized care for marginalized groups [7]. For example, longer waitlists and tighter triage criteria might disproportionately affect those with less understanding of the system.

The place of residence also played an important role in healthcare use. Patients living in rural areas had higher primary care utilisation than urban residents, yet significantly lower specialist utilisation. This was especially evident in the post-pandemic period. The higher primary care use observed in rural communities might be explained by easier accessibility or closer patient-provider relationships in smaller towns. On the other hand, reaching specialist services usually requires travel to urban areas [35], which became especially difficult during and after COVID-19 due to travel restrictions. The expansion of telemedicine was hoped to mitigate geographic barriers, but rural uptake of telehealth has been uneven. Connectivity issues, limited digital literacy, or simply a preference for in-person care in these communities can limit telehealth’s reach, as it has been previously observed [14, 36].

Socioeconomic status showed a complex relationship with healthcare utilisation in our study. Generally, being an active person with low SES was associated with higher utilisation of general practitioners, even after adjusting by comorbidities and other socioeconomic characteristics, as it has been described in the scientific literature [37]. This is possibly reflecting a greater burden of necessities, moreover than clinical ones, and fewer alternatives to access other kinds of services or private care. This pattern changed when considering primary care nurses, which were mainly used by pensioners, which could be related with the type of care needed by old people. Interestingly, after the pandemic, the differences in general practitioner utilisation between active with low SES and retired people increased. This fact could be explained by the previously mentioned barriers that face this group to access primary care. We also observed that active people less than 18k €/year were the socioeconomic group with lower utilisation of specialist care. This aspect has been previously detailed in the scientific literature [36].

The pseudo R² values measure how much of the variation in healthcare use is explained by specific factors such as socioeconomic status or the number of chronic conditions a patient has. Our results show that socioeconomic factors became less important in explaining who used primary care during the COVID-19 pandemic, and this has not fully recovered to pre-pandemic levels. This likely reflects the widespread restriction of primary care visits during the pandemic, which reduced utilisation across all socioeconomic groups. At the same time, having chronic conditions became a stronger predictor of healthcare use, which is consistent with previous evidence identifying multimorbidity as one of the main drivers of healthcare utilisation [38]. In specialist care, socioeconomic factors explained only a small part of the variation throughout the study period, possibly reflecting the adoption of prioritization protocols favoring the most clinically complex patients during the pandemic. Further research is needed to better understand these patterns.

This study has several strengths. Firstly, it is a large, population-based cohort of type 2 DM patients from a universal health system, using real world data from healthcare services. These factors strengthen the validity and generalisability of our findings by minimising selection biases and financial barriers that could influence utilisation patterns. We were able to analyse healthcare contacts at different levels of care provision over multiple years, including the pre- and post-pandemic periods. This provides a comprehensive view of how care has evolved. Nevertheless, some limitations must be acknowledged. Firstly, our analyses did not consider the reason for each visit. Therefore, we cannot distinguish between visits directly related to type 2 DM and other healthcare visits. However, as the objective of this research was to examine the use of healthcare services, it was important to take into account all contacts with the healthcare system. Second, our study focused on a single region in north-east Spain (Aragón), and results should be generalized with caution. Aragón is part of Spain’s universal, publicly funded healthcare system with a strong primary care focus and an aging population. These characteristics are shared with many European regions. Structural challenges, such as specialist waiting lists and rural access, are also common across European systems, suggesting our findings may be relevant to similar contexts. The results observed in the oldest age group (≥ 80 years) should be interpreted with caution, as this subgroup undergoes substantial compositional changes over the study period. Such changes may partially account for the observed variations in the adjusted patterns (Table 7 annex). Considering 2020 as a single exposure year precluded the assessment of within-year variability, including distinct pandemic phases and policy measures. Future analyses using monthly or quarterly intervals could better capture the effects of these dynamics on healthcare utilisation. Also, our definition of T2DM relied on the dispensation of antidiabetic medications in 2017. Since the use of GLP-1 receptor agonists for weight management in individuals without diabetes was still marginal at that time, the misclassification of non-diabetic individuals as having T2DM on this basis is unlikely. A further limitation is that we did not have information on the modality of healthcare contacts (remote vs. in-person). Consequently, while differences in digital access and telehealth utilisation could plausibly contribute to the observed dynamics, we cannot directly assess their role in this study. Future research incorporating patient-reported experiences of digital care, would help to clarify how teleconsultation modalities shape utilisation patterns and inequalities among patients with chronic conditions. Future research could incorporate emergency department utilisation to examine whether constrained primary care access translated into a shift towards unplanned acute care, particularly among the most vulnerable groups. Rurality was an area-level attribute assigned according to each patient’s basic health area, so it reflects the residential environment rather than individual characteristics. Also, the dichotomous rural/urban classification does not capture the internal heterogeneity of rural settings, while a gradient based on municipality size or degree of urbanisation would have characterised this axis more precisely. Unfortunately, this was the only information available in the CARhES cohort. Information on educational level, which is a key aspect of diabetes management and an important determinant of SES, was not available in the datasets used. Finally, it is important to highlight that statistically significant differences in healthcare utilisation between groups may not necessarily be clinically or public health relevant. Some prevalence ratios, particularly those close to 1, reached statistical significance but represent very small absolute differences in visit rates. Such minor discrepancies are unlikely to translate into meaningful differences in health outcomes. To enhance interpretation in terms of clinical and public health relevance, complementary absolute-scale measures are provided for the primary comparisons, specifically for socioeconomic level (Table 8 annex). Therefore, caution is warranted in interpreting these findings, and results should be considered in the context of their practical relevance rather than statistical significance alone.

Conclusions

The evolution of healthcare utilisation among patients with type 2 DM from the pre-pandemic period to the post-COVID era indicates that inequities have not only persisted, but have also exhibited shifts in their underlying dynamics.

Healthcare interventions should prioritise those who have fallen behind in both primary and specialised care. Guaranteeing healthcare access for older adults is crucial, and this can be achieved through more age-friendly telehealth platforms, proactive outreach from primary care teams, and regular home visits for those with mobility limitations. The gender disparities described demonstrate that gender-based differences in care are deeply rooted and require targeted policy attention to ensure women with chronic diseases are appropriately referred to specialists when needed. This could involve implementing gender-sensitive referral protocols or decision support tools. To improve the inclusion of immigrants, interventions such as patient navigation programmes, culturally tailored health education and interpretation services in both in-person and telehealth settings could be of help. Additionally, health systems should strengthen their efforts to provide specialist access in rural areas expanding telehealth infrastructure, offering transport support or establishing mobile specialist clinics. Finally, it is important to highlight that all these social determinants interact with each other. Therefore, an intersectional framework would be beneficial for studying healthcare inequalities and developing more equitable, efficient, and effective healthcare.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

BIGAN

Aragonese Health Data Lake

CARhES

CArdiovascular Risk factors for hEalth Services research cohort

CEICA

Ethics Committee for Clinical Research in Aragón

95%CI

95% Confidence interval

COPD

Chronic obstructive pulmonary disease

COVID-19

Coronavirus disease 2019

DM

Diabetes mellitus

GLM

Generalised linear model

GMA

Morbidity Adjusted Groups (Grupos de Morbilidad Ajustada)

GP

General practitioner

PR

Prevalence ratio

R²

McFadden’s pseudo R-squared

SD

Standard deviation

SES

Socioeconomic status

T2DM

Type 2 diabetes mellitus

Author contributions

I.A-P: Conceptualization, Methodology, Writing – original draft, Writing – review & editing, Supervision, Project administration. M.J.R: Conceptualization, Methodology, Writing – review & editing, Supervision. S.M: Conceptualization, Methodology, Writing – review & editing, Supervision. A.G: Data curation, Formal analysis, Writing – review & editing. S.C-F: Data curation, Formal analysis, Writing – review & editing. I.B-U: Writing – review & editing. I.L-F: Writing – review & editing. All authors read and approved the final manuscript.

Funding

This study has been funded by the Instituto de Salud Carlos III through the project PI22/01193 andcofunded by the European Union. It was also supported by the Research Group in Health Services in Aragón (GRISSA) (B09-26R) that is part of the Department of Innovation, Research and University at the Government of Aragón (Spain). The funder had no role in study design, data collection andanalysis, decision to publish, or preparation of the manuscript.

Data availability

The datasets generated and analysed during the current study are not publicly available due to legal and ethical restrictions imposed by the data governance policies of the BIGAN platform and the conditions established by the Ethics Committee for Clinical Research in Aragón (CEICA; PI21/148). Aggregated outputs supporting the findings of this study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

All data were pseudonymized, and the research protocol was approved by the Clinical Research Ethics Committee of Aragón (CEICA; Code: PI21/148). The requirement for informed consent was waived by CEICA due to the pseudonymized and retrospective nature of the data analyzed. This study was conducted in accordance with the principles of the Declaration of Helsinki.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.