Avoidable visits to the paediatric emergency department: associated factors and lessons learned from the pandemic

Abstract

Purpose

The main goal of this study is to identify the associated factors with avoidable admissions in ED, comparing pre-COVID and COVID periods.

Methods

This was retrospective study that took place in a Paediatric Emergency Department of a metropolitan, university-affiliated hospital in Portugal. All visits to paediatric emergency department between 2014 and 2020 were considered.

Results

There was a decrease of 7.2% points in avoidable visits between pre-COVID and COVID periods. Considering both periods, this study identifies older ages, being admitted to the paediatric emergency department between 4 and 7 a.m., referral and having visited the emergency department previously within 72 h as major factors associated with a reduced likelihood for avoidable visits. On the other hand, it identifies an increased likelihood of avoidable visits in the 3 to 5 years old age group, visits that occurred during the Summer and visits that occurred between 8 and 11 p.m. When considering what changed between pre-COVID and COVID periods, while having visited the paediatric emergency department 72 h prior made it less likely for the patient to be an avoidable visit during the pandemic period, this tendency has inverted, making it more likely for return visits to be avoidable.

Conclusions

The relatively low decrease in avoidable visits’ ratios between pre-COVID and COVID periods, associated with the similar distribution of attendance during the day and lower odds ratio of avoidable visits during periods when primary care is available, suggests that avoidable visits are a chronical problem of the National Health system’s structure and its usage, not having a single factor nor a combination of factors as a driving force. Nevertheless, this study identified several factors associated with avoidable visits to the emergency department. Therefore, it can aid policy makers to create targeted interventions to mitigate this problem.

Significance

AbstractSection What is Known

- Excess of avoidable visits have been a problem in the Emergency Department for several decades.

- The fear experienced during COVID-19 pandemic, influenced the visits to the emergency departments.

AbstractSection What is New

- The change in demand introduced by the COVID-19 was used to provide insights regarding the factors associated with avoidable visits.

- A triage system’s agnostic definition of avoidable visit was chosen, to better reflect the visit’s lack of necessity of hospital resources, including hospital admission.

Introduction

Background

Excess of avoidable visits have been a problem in the Emergency Department (ED) for several decades as it leads to overcrowding, unnecessary exposure to the hospital environment and unnecessary costs to the National Health System (NHS). Crowding in ED is well described, leading to poor outcomes and generally worse quality of care [1,2,3,4], and despite the impact of avoidable visits (AV) on crowding, patients subject themselves to the hazardous hospital environment [5,6,7]. Being most of them self-referral patients, there might be a lack of literacy to weigh the risks of exposing themselves or their children to this environment [8]. This is a structural NHS and knowledge asymmetry problem that clashes with patients’ perspective, as the type of care division exists only on the caregiver side, and the patient is most of the time unaware of the primary, secondary, or even tertiary type of attention most appropriate for the condition [9]. Avoidable attenders are also an unnecessary burden to the NHS since the cost is significantly greater compared to a visit to a primary care provider (PCP) [10]. Most studies regarding this problem attempt to describe the visits focusing on chronic illnesses or socio-economic aspects to characterize avoidable visits to the ED. In the United States avoidable ED visits were found to vary from 13.7 to 27.1% [11]. However, the variability is high and heavily dependent on the method used to classify an avoidable visit [12,13,14].

In Portugal, the National Health System (NHS) has universal coverage. Nevertheless, there are cost associated to attendance and use of health care services [15]. Primary health care (PHC) is organized into 55 groups of primary healthcare centres (ACES) and, in conditions that are not severe and do not require urgent care, should be the first contact that the patient has with the health system. However, a study conducted in two portuguese hospitals identified that 40% of ED visits were classified as inappropriate. The study considered inappropriate the 2 less urgent levels of the Manchester triage system [16].

The fear experienced during the pandemic due to coronavirus (COVID-19 disease) and the recommendations by the government and health authorities for people to stay home could have influenced the visits to the emergency departments. Many studies all over the world reported that ED visits were declined during the lockdown, states of emergencies or just during the several months that the pandemic lasted [17, 18] and in particular, in paediatric ages [19, 20] and in avoidable ED visits [21, 22]. So, it is essential to characterize the factors that can be involved in this possible decline in Portugal, particularly regarding avoidable admissions, due to the portuguese Health Service (HS) paradigm and concerning Paediatric Emergency Department (PED), the factors involved might be different.

Objectives

The main goal of this study is to identify the associated factors with avoidable admissions in the ED, comparing pre-COVID and COVID periods.

Methods

This retrospective study compares avoidable with non-avoidable PED visits. This comparison is done considering several factors: some known at the moment of admission and others to measure the impact of avoidable visits’ on the PED itself. In order to compare the influence of a pandemic on the associated factors, this analysis was split into two, pre-COVID and COVID period on the 13th of March 2020, when the portuguese government set the first state of emergency.

This paper is structured according to the RECORD statement, i.e., The REporting of studies Conducted using Observational Routinely-collected health Data [23].

Setting

Urgent and emergency care in Portugal is provided by hospitals with ED and healthcare centres. Healthcare centres provide primary care, each serving the population of a geographically defined catchment area, mainly during office hours. Nevertheless, centres provide primary care until 11p.m. on weekdays, and during business hours on weekends, for urgent situations. There are appointment slots called “Open medical appointment” used to treat acute illness and injury, scheduled for the same day, attributed to walk-in patients. However, the available slots are limited.

Hospitals serve a much larger geographically defined catchment area located on top of the portuguese NHS referral network. However, hospitals in Portugal cannot redirect walk-in patients from the ED regardless of their medical condition.

The NHS in Portugal is funded mainly by taxes, nevertheless, there is a patient’s co-payment that was set at 20 € per visit to the ED in 2011 [24]. However, there are situations where patients get free medical care, e.g., during pregnancy, patients under 12 years old and patients with low income [25]. From January of 2015, the age limit to be entitled to free medical care was raised to include all minors, i.e., those less than 19 years old [15].

This study took place in the PED of a metropolitan, university-affiliated hospital with a catchment area of approximately 800,000 inhabitants, receiving approximately 75,000 visits per year from an estimated population of 137,016 children or adolescents [0.17 years] (data supplied on request) [26]. The hospital receives both adults and children; however, these departments are organised and managed separately.

At any moment, in the PED, there are always two senior physicians trained in Paediatric Emergency Medicine, two to three medical residents (depending on the workload), eight nurses, and three auxiliary staff per shift. Different teams work in 12-hour shifts providing 24 h per day coverage. The PED triages visitors from Level 1 through 5 according to the Paediatric Canadian Triage and Acuity Scale PaedCTAS [27]. To each level is assigned a different degree of urgency i.e. Level 1 - Resuscitation, Level 2 - Emergent, Level 3 - Urgent, Level 4 - Less Urgent and Level 5 - Non Urgent. In Portugal this type of triage is used in 3 out of 14 Paediatric emergency departments.

All patients who attended the hospital’s PED (i.e., from 0 to 17 years old) in a 7-year period (between 01/Jan/2014 and 31/Dec/2020) were considered. However, patients who left the PED without being seen by a physician or left against medical advice were excluded from the analysis.

Variables

The dependent variable was defined by the conjunction of several PED markers, resulting in a restrictive definition of avoidable. A PED visit was considered avoidable if the patient was discharged home and no diagnostic tests (i.e., blood tests and radiology exams), procedures, or medications are required during the stay. And if the patient is not asked to stay further observation, i.e. it was not necessary the physician to better assess the condition’s evolution. N.b. COVID tests were not included in the diagnostic tests’ variable.

Definition of Avoidable visit: - Not medicated; - No radiologic exams; - No analysis; - Discharged home. - Not asked to stay for observation;

As regression predictors, i.e., factors.

• Front door information.

  • Age (years) – patient’s age group;

  • Sex – patient’s sex;

  • Season – Season of the year of patient’s PED admission;

  • Weekend – if the day of the patient’s PED admission was a weekend or weekday;

  • Time of day – Time of day of patient’s PED admission, grouped in 4-hour slots;

  • Residence municipality – patient’s residence, this variable was grouped;

  • Referral – A referral patient was defined as a ‘not walk-in’ patient, e.g. referral from PCP, private clinic, or other hospitals;

  • Made return visit 72 h prior- Patient had visited the PED 72 h prior to current visit, i.e., the current visit is a return visit;

  • Visit by frequent attender – Visit made by frequent attender, i.e., > four visits per year.

• Impact on the PED.

  • Triage level - PED’s nurses triage visitors according to the Canadian Triage and Acuity Scale Paediatric (PaedCTAS) [27], and a level from 1 to 5 is assigned to the patient during the stay.

    • Level 1 – resuscitation.

    • Level 2 – Emergent.

    • Level 3 – Urgent.

    • Level 4 – Less urgent.

    • Level 5 – Non urgent.

  • Discharge physician’s specialty – Describes the discharging physician’ clinical specialty;

  • Returns within 72 h – if the patient will return to the PED within 72 h, i.e., the current visit is the index visit of later readmission.

Statistical analysis

The dataset provided for this analysis was exported from the hospital’s information system.

Descriptive statistics were used to analyse the characteristics of the PED visits. Logistic regressions were performed, and odds ratio with 95% confidence intervals and p-values were calculated for each variable. When considering the impact of avoidable visits in the PED, the variables known at the moment of admission and interactions were also included as adjustments to the model.

All the data analysis was performed in R version 4.0.2 (2020-10-10). The integrated development environment (IDE) used was RStudio Version 1.4.1103.

In Fig. 1 it is presented a diagram to provide an overview of the study.

Fig. 1

Study’s diagram representing the period split and comparisons within each period

Results

There was a total of 516,851 visits to the PED within the study period. From these, patients who left without being seen (n = 11661) and who left against medical advice (n = 303) were removed from the analysis. Furthermore, there were 3,552 (0.69%) missing observations in the variable “Avoidable ED visit” and 4553 (0.88%) in the variable “Discharge destination”. There were negligible (< 0.5%) or no missing data in the other variables of interest. Triage level 1 was also removed from the regression analysis for its low frequency. There were only 22 considered avoidable visits from 2014 to 2019 and no avoidable visits triaged level 1 in 2020. Hence, after removing 16,507 visits due to the exclusion criteria and missing data, 500,344 visits remained for analysis.

Descriptive data

Table 1 shows the distribution of PED visits’ characteristics over the study period. Up until and including 2019 attender’s characteristics remained stable. Walk in patients (i.e., non-referrals) in the PED steadily decreased. However, in 2020 there was a considerable decrease in attendance and an increase in referrals to the PED.

Table 1 Distribution of PED visits’ characteristics from 01/jan/2014 to 31/dec/2020, stratified by year

Outcome data

There was a total of 192,104 (41.1%) and 11,071 (33.9%) PED avoidable visits in the pre-COVID and COVID periods, respectively. A complete breakdown per variable can be seen in Table 2.

Table 2 Number and percentage of PED visits for all study variables by time period (i.e., pre-COVID and COVID periods) and main outcome (i.e., avoidable and not avoidable)

Main results

Table 3 shows the logistic regression results considering the factors known about the patient when entering the PED as the independent variables and being AV as the dependent variable. There has been a steady decline in the likelihood of AV through the years. Regarding age, there is an increased likelihood of AV at younger ages. Older ages remain less likely to be AV both in pre-COVID and COVID periods. Furthermore, males are more likely to be AV, however, the effect size is small. AVs are more likely to happen in the summer. Nevertheless, in the COVID period there was an increase in likelihood in all seasons compared to the Fall. AV are more likely to happen during the weekend, however, the effect is small and not statistically significant in the COVID period. When considering the time of day, AVs are less likely after midnight, during the morning and more likely in the afternoon and towards midnight. Only Porto (where the hospital is located) and Gondomar have a similar likelihood of AV. AV are less likely to reside in other referral municipalities. Patients who do not reside in the catchment area are more likely to be avoidable. Referrals are less likely to be AV. Nevertheless, there was a slight increase in the COVID period. Patients who were in the PED 72 h prior to the current visit were less likely to be AV, this inverted in the COVID period. Frequent attenders were more likely to be AV, with a negligible magnitude of effect. However, the effect inverted, and the magnitude increased in the COVID period.

Table 3 Results from the logistic regression considering the variables known about the patient when entering the PED as independent variables and “avoidable visit” as the dependent variable, the reference is “not avoidable” (the results are presented in odds ratio, confidence interval and p-value)

Table 4 shows that triage is a good predictor for defining AV. However, there was a slight increase in the likelihood of AV triaged level 2 and a drastic decrease of AV triaged level 5 when comparing pre-COVID and COVID periods. This varies with age, as it can be seen in the interactions, however the effect size is small compared to the coefficients associated with triage levels. Index events of future return visits within 72 h are more likely avoidable. Compared to the paediatrics specialty, only Orthopaedics were less likely to treat AV.

Table 4 Results from the logistic regression considering the variables that impact the PED’s management as independent variables and “avoidable visit” as the dependent variable, the reference is “not avoidable” (the results are presented in odds ratio, confidence interval and p-value). These results are adjusted for all front door variables and age*triage level interaction

Discussion

The main objective of this study was to identify the associated factors with avoidable visits to the PED and to see how or if those factors changed during pandemic periods. Most studies regarding avoidable ED visits do not evaluate the impact of a drastic change in demand that should, in theory, only allow non avoidable patients to access the ED (28). Hence, this study uses the change in demand introduced by the COVID-19 pandemic to learn more about the problem of avoidable visits. Furthermore, there is a lack of consensus on the definitions, terminology, and concept of “avoidable visits” [12, 29]. Therefore, we chose to classify visits as avoidable if: the patient was discharged home, didn’t need blood tests, radiology exams, or medication, nor was asked to stay for observation. This definition of AV was chosen among others [12] for being triage system agnostic and better reflecting the visit’s lack of necessity of hospital resources, and therefore more valuable as a policy making decision tool for referral. For these reasons, these analyses provide a clearer and useful picture of the driving forces behind avoidable visit’s demand.

Summary of main findings

ED attendance and visits’ characteristics have been relatively stable over the past years. However, in 2020 there was a sharp fall in attendance. Despite the reduction in overall attendance between the two periods, from an average of 206 daily visits to average of 111 daily visits, there was only a decrease of 7.2% points in AV. Older children are associated with a reduced likelihood of avoidable visits. There is an increased likelihood of avoidable afternoon visits until midnight and on weekends. Referrals were associated with a reduced likelihood of avoidable visits. Having visited the PED previously within 72 h was associated a reduced likelihood of avoidable visits in the pre covid period. Being triaged level 4 or 5 increases the likelihood of being avoidable.

Results contextualization

The proportion of avoidable visits is aligned with the median (34.0%) found in a systematic review by Durand et al., for retrospective studies using explicit criteria. However, comparison to other studies is difficult since there is considerable disagreement on the methodology and criteria for categorizing ED visits as avoidable, which results ranging from 10.8 to 90.0% [14].

Despite the different methodology and considerably higher percentage of avoidable PED visits, the reduction in AV was consistent with Valitutti et al. [21], though the drop in our results was less substantial. Caretaker’s fear could have been the cause for this decline since the first wave was more difficult in Italy, and the clinical severity of COVID is extremely rare in children [30]. This hypothesis is also corroborated by the decrease in likelihood in triage level 5 and visits made by frequent attenders in the pandemic period. Furthermore, the change in likelihood during the pandemic period in small children might be a result from the decrease in infectious diseases [20]. The ratio of avoidable visits has been steadily declining over the years. This decline might be related to the increasing presence of private hospitals in Portugal, which caretakers with medical insurance might prefer because of shorter waiting times [31]. Our study identified younger ages as a factor associated with AV. Similar studies in Belgium and Italy showed the same results [32, 33]. Brousseau et al. identified different possible reasons: (1) the importance that caretakers give to immediate reassurance; (2) the perceived lack of necessary tests and treatments at primary care; (3) discrepancies in communication and access between parents and physicians; and (4) the lack of concern about the impact of non-urgent emergency department visits [34]. The availability of primary care is paramount, and the increased likelihood of AV during the late evening point in this direction. Simpson et al. [35] and Alele et al. [36] show an increased likelihood of out-of-hours AV which partially contradicts the results of this study. Considering Simpson et al., this might be due to the co-location of PCP and ED services [37], and regarding the systematic review from Alele et al., one must bear in mind the aggregation of several settings from different countries with different paradigms for providing emergency care. It also needs to be clarified what “out-of-hours” means. We interpreted it concerning the availability of primary care services, and primary care in Portugal is also available in the evening and weekends, despite having its offer significantly reduced. In our study referral is associated with a lower likelihood of AV. Contradicting the results from a systematic review from Uscher-Pines et al. [28] but in line with Alele et al. [36]. These contradicting results might be related to the different study settingsor to the organisation of health care systems. However, in the portuguese NHS, PCP apparently work as a triage system. Nevertheless, the overall low percentage of referrals mitigates the impact of this factor. Frequent attenders are associated with a lower likelihood of AV during the pandemic. Considering that some of the main driving forces of frequent attendance are chronic illnesses [38], fear of infection might be a deterrent for patients who seek care in the PED. In pre COVID period, 72 h of re-attendances were less likely to be avoidable visits. However, this likelihood was inverted during the COVID period. This might be related to additional tests requested by physicians’ added concern in the pandemic times. The low urgency levels of triage have an increased likelihood of being avoidable. Despite being expected, since some studies use it to define avoidable [39], it is very concerning, for the majority of PED visits are triaged with these levels. Almost all specialities have an increased likelihood of receiving AV. This fact also did not change during the COVID period. These results strengthen the hypothesis of chronic misuse of the portuguese NHS.

Implications for policy and practice

The relatively low decrease in AV ratios between pre-COVID and COVID, associated with the similar distribution of attendance during the day in both periods, and decreased likelihood of AV made by frequent attenders during the pandemic, suggests that AV might be a chronical aspect of the portuguese NHS’s structure and its usage, not having a single factor nor a combination of factors as a driving force.

The greater odds for AV during primary care out-of-hours and the decreased likelihood of patients triaging level 5 indicates that the PED is being used for low acuity conditions. Despite increasing the offer of out-of-hours primary care centres might seem to be a good way to mitigate this problem, the evidence is not strong [40]. On the other hand interventions focused on reducing AV [42], such as coordination of care, e.g., informing the patient’s family physician and suggest follow-up consultation might be effective since referral is associated with a reduction in AV. This might also have the added benefit of improving the health literacy of that specific group [42]. The implementation of walk-in centres in primary care might also be a viable option [43].

The above proposals are focused on the diversion of demand to primary care, however, there is also the possibility of creating specific patient flows for AV, or creating primary care centres inside the ED [44]. The discussion is not easy since this approach, while solving the problem, subverts the structure of the portuguese NHS, possibly creating bigger problems attracting demand and draining patients from primary and secondary care.

Even considering the decrease of AV in 2020 the cost to the portuguese NHS averaged 3,252,000 € a year in this ED alone [45]. This study provides valuable information on how to make these interventions more effective by profiling patients and their usage of the PED and helping to mitigate this burden while improving the quality of care.

To the best of our knowledge, there are no similar studies done in Portugal. Nevertheless, given the portuguese healthcare system paradigm, there is no reason to believe that the panorama in other EDs is very different from what is reported in this study.

Limitations

The data gathered had the original purpose of providing healthcare to the patients in the PED, therefore subjected to the bias of any observational study based on routinely collected data, e.g. information system downtime and the inability to control how the variables are collected.

The definition of AV was chosen among others [12] for being triage system agnostic and better reflecting the visit’s lack of necessity of hospital resources and, therefore potentially better suited for primary care. Nevertheless, the multiple definitions used in the field make comparisons less accurate.

When considering the clinical specialities, one should bear in mind that some clinical specialities require more diagnostic tests or medication than others. Therefore, and due to the chosen definition of AV, some might be biased toward or against it (e.g. orthopaedics).

Conclusions

The relatively small reduction in the proportion of avoidable visits between the pre- and post-pandemic periods, combined with the consistent distribution of attendance throughout the day and the lower odds of avoidable visits during hours when primary care is available, indicates that avoidable visits are a chronic issue deeply rooted in the structure and usage of the portuguese NHS. This suggests that no single factor, but a combination of factors serves as the primary driver. However, this study has identified several factors linked to avoidable emergency department visits, providing valuable insights that can assist policymakers in designing targeted interventions to address this issue.

Avoidable ED visits are a known problem in countries with no gatekeeping policy [46]. A tighter communication between primary and tertiary care could result in a more effective resource allocation and optimized patient flow. Studies like the present one can guide policy makers to create targeted interventions (e.g., increasing the availability of PCP to the evening or integrating a primary care area into the PED) to mitigate this problem.