Frequent Users of Emergency Departments: Analysis of the Characteristics and Geographical Distribution in a Local Health Authority in Rome, Italy

1. Introduction

The Italian population, like the rest of the OECD Countries, has experienced an increase in chronic illness over the last few decades, with population aging being one of its primary causes [1]; despite significant improvements in the emergency care system, the increase in the number of fragile patients and their assistance needs have led to a rise in Emergency Department (ED) attendance [2,3]. In hospitals with insufficient inpatient bed availability, a rise in ED visits and the number of emergency patients requiring admission may increase the length of stay in the ED, leading to overcrowding and compromised ED performance. Reduced ED utilization has become a primary goal for new healthcare delivery models [4,5,6]. However, the predictors of ED attendance are complex and involve multiple components [7].

There is not a single FU definition, and it depends on the length of observation time. Frequent Users (FUs) are defined as patients who repeatedly attend the EDs, at least four times in a year [8,9,10,11,12,13]. Furthermore, FUs account for 4.5% to 8% of the ED population but comprise up to 28% of ED visits [10,14]. The overuse of EDs causes two major issues: the first is associated with extra ED activity expenditures, the weight of which is constantly out of the government’s budget limits in the public sector, particularly since the end of the COVID-19 pandemic emergency [15,16,17]. The second is the risk of compromising the quality of the ED services to patients with real acute issues and appropriate ED attendance. Furthermore, it increases the risk of overcrowding, raising the risk of human errors [8,9,18].

Previous research has highlighted how FUs are more likely to suffer from chronic diseases (particularly renal failure, cardiovascular disease, diabetes and chronic pulmonary disease), mental illnesses or substance use disorders [19,20]. The prevalence of chronic conditions is higher in FUs than in the rest of the population, and timely interventions in primary care could help avoid ED visits or hospitalization [21,22]. A growing body of evidence reveals that FUs have a high level of healthcare use (other than the ED), a connection with low socioeconomic level, and an association between the characteristics of their residence and health outcomes [19,21,22,23,24,25,26,27].

Like other countries, Italy has recently focused on improving care coordination and reducing ED attendance and hospital admissions by encouraging more appropriate use of community-based services [28]. Compared with other European contexts, a relatively high proportion of Italians report that they have unmet medical needs, and access to care is particularly difficult for Italians with lower socioeconomic level [29].

Analyzing the socioeconomic profile and the geographic variation in ED use may be a critical first step to drive local policies in reducing avoidable emergency department visits in Italy [30].

According to the literature, the prevalence of health problems is essential in ED attendance frequency. Nevertheless, other factors have also been associated with high levels of ED attendance [31,32,33,34], especially in the city outskirts or in inner city areas, such as in Rome, such as housing problems and employment status, loneliness and urban marginalization, low levels of social support, proximity to the ED, and poor access to primary care [35,36,37].

Shifting the frame of reference from macro areas to a more granular unit of analysis, i.e., census tracts, improves the identification of local hotspots with high healthcare use and shifts the focus to factors found among local populations and communities, providing insight into how to geographically target interventions for the improvement of healthcare delivery [38].

This study was carried out in Local Health Authority (LHA) Roma 1 (see Appendix), one of the three LHAs in the urban municipality of Rome, in the Lazio Region. It consists of 6 Local Health Districts (LHD 1, LHD 2, LHD 3, LHD 13, LHD 14, LHD 15) and 805 General Practitioners (GPs), has approximately 1 million residents and is one of the most populous areas in Italy, with 13 EDs (out of the 22 in the Rome metropolitan area), an aging index (the number of people in age group 65+ divided by the population age 0-14, per 100) of 202 in 2024 (the Lazio Region mean is 189,0) and an old age dependency ratio (the ratio of the number of people in age group 65+ divided by the population in the working age 15-64) of 38,1 in 2024 [39].

This study aims to identify the clinical and social characteristics of FUs and to quantify and compare amount of variation in the probability of being FU attributable to GPs and LHDs.

2. Materials and Methods

2.1. Study Design

This is a retrospective cohort study that included all ED visits from January 1, 2021, to December 31, 2021, in all 13 Emergency Departments within LHA Roma 1. All patients assisted by a GP from the same LHA were enrolled in the cohort. Patients younger than 18 years and those who attended single-specialism EDs (ophthalmology, pediatrics) and attendances for obstetric or gynecological problems were excluded since they represent a population with specific needs and peculiarities that is worthy of specific focus and not comparable with the general population.

FUs were defined using a cut-off of ≥4 visits per patient per year. Because of the absence of a universal definition of FUs, different cut-offs (visits ≥5, 7 or 10) were also considered in the descriptive analysis.

For each patient in the cohort the following risk factors potentially associated with the study outcome were considered: gender, age, socioeconomic level (high, middle-high, medium, middle-low, low) and the presence of chronic or multiple-chronic conditions [40,41]. The socioeconomic level was calculated at the census tract level, based on the methodology developed by Nicola Caranci and colleagues [42]. This index integrates multiple socio-economic indicators derived from national census data, including educational attainment, employment status, home ownership versus rental, household overcrowding, and family structure. It provides a composite measure of socio-economic disadvantage within small geographic areas. Among the multiple-chronic conditions group of patients, high clinical complexity was identified as a five-year mortality risk higher than 10%, based on the number and type of chronic diseases they suffer from [43].

Data related to ED visits were also compared to identify differences between FUs and no FUs. In particular, triage score, principal diagnosis group in the ED, and reported symptoms at the time of arrival were considered.

2.2. Data Sources

The cohort was enrolled from the Healthcare Emergency Information System, which collects all visits to emergency services, patient demographic data, visit and discharge dates and hours, ICD-9-CM diagnosis, reported symptoms on attendance, triage score (from no urgent to emergency), and status at discharge (e.g., dead, hospitalized, or discharged at home).

The cohort was linked to the automated system of databases of residents who receive NHS assistance in the Lazio Region, thus allowing researchers to obtain information related to chronic or multiple chronic diseases, GP and LHD of each patient, and socioeconomic level based on the residence address [42,44,45]. A deterministic record linkage procedure with anonymous identification codes was used to merge the data from different information systems. To preserve privacy, each individual identification code was subsequently and automatically deidentified, and the conversion table was deleted, giving researchers the access to only fully anonymized data.

2.3. Geographical Analysis

The administrative-territorial division of the LHA Roma 1 was used to analyze the association between the percentage of FUs and urban settings, as already performed elsewhere [46]. Each of the six LHDs of LHA Roma 1 is divided into Geographical Units (GUs, in Italy called “Zone Urbanistiche”), as defined by the Municipality of Rome. It is the smallest territorial level for which population data are available in Italy and many other countries [47,48].

2.4. Statistical Analysis

The response variable is binary (FU>=4 vs no FU). Frequency distributions of FUs (visits ≥4, 5, 7 and 10) by triage, principal diagnosis and declared symptoms at ED arrival were compared to no FU. A descriptive analysis was performed reporting absolute frequencies among patient groups and characteristics of ED attendances. Univariate logistic analysis was used to identify candidate predictor variables among those collected in the dataset. A logistic multilevel model (patient < GP< District) was performed to quantify the variability in ED FU behavior attributable to LHDs and primary care physicians and to identify the role of social and clinical determinants (gender, age, socioeconomic level and presence of chronic or multiple-chronic conditions).

Age was considered as a continuous variable, the Box-Tidwell test was used to check for linearity between the “predictor” and the logit. The role of individual variables was expressed as Odds Ratios (OR); the variance components - estimated by multilevel models - were expressed in terms of Median Odds Ratios (MORs). The MOR quantifies the variation between clusters by comparing two persons from two randomly chosen, different clusters. Consider two persons with the same covariates, chosen randomly from two different clusters. The MOR is the median odds ratio between the person of higher propensity and the person of lower propensity. This measure is always greater than or equal to 1.00. If the MOR is 1.00, there is no variation between clusters. If there is considerable between-cluster variation, the MOR will be large[50]. The MORs were estimated for the “empty” model, which only includes a random intercept, and the complete model, which includes all patient risk factors. SAS (SAS Institute Inc., North Carolina) software was used for statistical analyses.

3. Results

3.1. Descriptive Analysis

In 2021, 89,036 people attended at least once in any one of the 13 EDs within the LHA Roma 1. The characteristics of patients and ED attendances are summarized in Table 1 and Table 2. A total of 72,781 patients had a GP in the LHA Roma 1. Patients cumulatively had 99.811 attendances. FUs represented a tiny fraction of the overall population (2.7%) but were responsible for a large portion of attendances (11.3%). Among no FUs, females accounted for 51.9%, while males were more frequent among FUs (between 51.9%, for FU>=5, and 57.7%, for >=10) than among no FUs (48.1%). The age class 50-59 was the most frequent among no FUs (18.4%) and FU≥4 (17.5%) and rose with the growth of FU attendances. The socioeconomic level was low among most of the FUs (from 25.3% of FU≥4 to 24.6% of FU≥10), while high and medium-high levels were found in the no FU class. Multiple chronic conditions were present in 29.4% of FU≥4 patients, while 47.6% of FU≥4 patients did not have chronic disorders registered.

Table 2 was realized after having first classified the patients based on the number of visits to the emergency room. Emergency and urgency triage codes decreased from 24.9% in the FU≥4 group to 20.9% in the FU≥10 group. In contrast, non-urgency codes increased from 8.1% in the FU≥4 group to 14.4% in the FU≥10 group. The main diagnoses among all the patients belonged to the specific symptoms and signs group. Mental disorders were registered in 2.6% of the NO FU group but increased from 7.6% in the FU≥4 group to 15.9% in the FU≥10 group. Cardiovascular diseases were similar in the no FU and FU≥4 groups (8.6% and 8.2%, respectively). Excluding nonspecific and missing diagnoses, many acute symptoms registered as the main issue (e.g., trauma, abdominal and chest pain, dyspnea) decreased from the no FU group to the FU≥10 group. In contrast, psychomotor agitation increased from 5.1% in the FU≥4 group to 11.2% in the FU≥10 group. Social issues are present in 0.4% of diagnoses in the FU≥4 group to 0.9% in the FU≥10 group, while it is not registered in the no FU group.

3.2. Geographical Distribution

The geographical classification comprises two levels: GUs and LHDs. Five classes were identified for each level according to the relative percentage of FUs in the area. The geographical distribution represented in Figure 1 shows that some areas of LHA Roma 1 have a significantly higher percentage of FUs, mostly in LHD 1, LHD 14 and LHD 15.

Legend: the colors represent the 5-cut-off percentage of FUs on the geographical unit population level. The bold black line separates the six LHDs, the narrow black line separates the GUs.

3.3. Inferential Analysis

Multilevel analysis results are reported in Table 3. Among candidate predictor variables identified in the descriptive and geographical analysis, the most prominent effect was related to the presence of multiple chronic conditions: the higher the patient’s clinical complexity, the higher the OR for being a FU. OR increases with decreasing socioeconomic level, with an effect that is more evident for low and medium-low levels and less clear for higher levels. The GP role does not consistently suggest an essential role in determining FU behavior (MOR 1.18, Wald p = 0.061), and no significant effect was found on the LHD level (MOR 1.05, Wald p = 0.127).

4. Discussion

FUs represented 2.7% of the overall population but were responsible for 11.3% of attendances registered during 2021, in line with the literature [46], mostly in the age class 50-59. The risk of being FU increases with the patient’s clinical complexity and for low and medium-low socioeconomic level. Generally, patients with higher socioeconomic levels were less likely to go to the ED. This is in line with another study in Milan, where the odds of avoidable hospital admissions were higher among low socioeconomic levels of patients compared with the other classes [49]. Prior studies suggest that patients with low socioeconomic level perceive ED assistance to be cheaper and more accessible than ambulatory care and are often more likely to use EDs for nonurgent conditions [50,51,52].

In this study, non-urgent triage codes were more frequent in FUs (from 8.1% in the FU≥4 group to 14.4% in the FU≥10 group), and mental disorders were present in a large proportion of FUs (to 15.9% in the FU≥10 group). Psychomotor agitation and social issues are important diagnoses associated with FUs, but the results of “symptoms, signs and ill-defined conditions” and “external causes of injury and supplemental classification” diagnosis groups, as well as the main issues on admission, such as fever, chest pain or dyspnea, may suffer the influence of the COVID-19 pandemic, which may have also affected the results of time-dependent illnesses such as stroke or cardiac complaints, possibly due to concerns about COVID-19 acquisition in the hospital [53,54]. The increase of physical and mental morbidities, especially with socioeconomic deprivation, is associated with the attendance rate at EDs, supporting previous evidence of increased emergency attendance among people with chronic illnesses and psychiatric disorders who are more likely to visit EDs [19,55]; this also holds true for people with low socioeconomic level [56,57]. Across the EU, admissions connected to chronic diseases and psychiatric disorders are estimated to be potentially avoidable through better prevention and disease management to decrease their dependence on the ED [58,59]. Some authors have investigated the importance of social support in the ED for older adults, even if in a systematic review there was no significant association between ED attendance and social support [60].

Regarding the geographical analysis shown in Figure 1, this study characterizes variation using a more granular geographic unit of analysis. This may be considered the first step to setting up a specific public health strategy aimed at improving adequate healthcare attendance for specific populations and communities, as decided in Paris by the Ile-de-France Health Regional Agency. [46] In other studies on geographic variation in healthcare, healthcare utilization and outcomes were quantified by using geographic macro areas, but it was impossible to distinguish a single neighborhood characterized by different socioeconomic and demographic characteristics [30,49,61,62]. Further studies may help to analyze the association among FUs, population density, income and medical services or GP offices in the same area.

The multilevel analysis shows that GP does not consistently suggest an important role in avoiding FU behaviors (MOR 1.18, Wald p=0.061), similar to the territorial level. In some studies, the reduction in demand for ED seems to be achieved via the availability of at-home GP visits or improved public transport [63]. Other authors have demonstrated that frequent ED users actually have high rates of general practitioner consultations and outpatient care utilization in addition to ED use. Thus, efforts to only increase primary care access or visiting hours (on weekdays and weekends) may not necessarily result in decreased ED utilization [64].

Primary prevention of the FU phenomenon is not much discussed in the literature. These subpopulations of FU may represent different targets for action plans, such as specific and individual interventions [65,66]. The individual plans on FUs are often based on the principle of secondary prevention because they are triggered after an ED visit, followed by an intervention or a treatment by a case-management team [67].

The high rates of FUs in ED attendances could thus indicate poverty in metropolitan areas and highlight social inequalities in access to health. As stated in the New Urban Agenda, The United Nations Sustainable Development Goals focus on health promotion through several interconnected health-related targets, achievable through multisectoral approaches [68]. Thus, the decrease of FU rate and the increase of general population’s health status may be realized through joint policies with other partners (such as school or transportation actors) [69,70].

Limitations and Strengths

This study is the most comprehensive examination in a single metropolitan area in Italy that considers all FU attendances in one year and their different health statuses, socioeconomic determinants and geographical data. Although these findings are not generalizable to the entire population, they will be relevant to many urban areas with similar levels of social deprivation. Thus, it seems possible for institutions or health care providers to identify territories where residents are at higher risk of developing FU patterns and to suggest primary prevention actions. Access to regional data flows allowed to avoid missing records or information. Additionally, residents who may be living in extremely disadvantaged circumstances were probably captured by our data because of the fictitious address LHA assigned to them.

This study has some limitations. The retrospective nature of the study allowed investigation of the predictors of ED attendance at one point in time and only in LHA residents, excluding homeless without residency, foreigners or people resident in LHA Roma 1 but whose health services should be delivered from the other LHAs in Rome. Only the main diagnosis was considered, and inaccuracies in the clinical dataset may have under recorded some morbidities. The proximity of the ED and general practitioner or primary care services as predictors of ED attendance was not investigated. Other variables would be necessary to quantify COVID-19-specific attendance to better understand the influence on the data.

5. Conclusions

Frequent ED use is a major challenge in health facility management. The analysis of ED attendances and the socioeconomic and geographical factors of the FUs highlights the need for new approaches to address important issues such as socioeconomic inequalities, increasing better housing conditions and employment, and structural issues such as strategic placement of primary care services and better transport.

This study allows the identification of potential risk factors predictive of disproportionate use of the ED and helps policymakers anticipate the needs of specific categories of patients. Further studies are necessary to analyse the presence of FUs and the geographical position of hospitals, residents’ income and primary care services in the entire Lazio region. It would be useful to investigate the efficacy of territorial interventions, according to the new directives of the National Recovery and Resilience Plan on chronic conditions management at the territorial level.