Introduction

Primary care practices are expected to be the first level of contact for paediatric patients with mild or non-urgent presentations. Referral to the emergency department (ED) may be necessary when diagnostic or therapeutic resources are required but unavailable in the primary care setting.1 However, up to 30% of all ED visits in high-income countries are by low-acuity ambulatory patients.2 This proportion is as high as 40% for the paediatric population and is mainly related to respiratory conditions.3 4 Indeed, young children are more likely to present to the ED for low-acuity conditions because of parents’ misperception of the severity of their child’s illness.5 6

Some strategies have been proposed to reduce low-acuity ED visits including redirecting patients to care alternatives perceived to be more efficient, such as walk-in clinics or primary care practices.7 Indeed, some studies have suggested that walk-in clinics and primary care facilities offer the population less costly, more accessible and equivalent quality of care than overcrowded EDs.8–10 However, data supporting this assumption are scarce and contradictory, especially in the paediatric population.11–13 We aimed to compare the value of care (ie, health outcomes and costs) given in EDs and walk-in clinics for ambulatory children presenting with acute respiratory conditions.

Methods

A parallel study following the same methodology was performed concurrently for the adult population and was previously published.14

Study setting and design

A retrospective cohort study was conducted for the fiscal year 1 April 2016 to 31 March 2017 in one ED and one walk-in clinic in Québec city (Canada). The participating ED is a paediatric tertiary care centre where an average of 34 000 children are seen every year. The walk-in clinic, managed by a team of family physicians, attends to over 30 000 patients of all ages. Similar to the ED, this walk-in clinic uses the Canadian Triage and Acuity Scale (CTAS) on patient arrival and can conduct X-rays. Additional diagnostic tests, such as laboratory tests, are available, but these are forwarded to a nearby hospital for analysis.

To estimate the cost of each patient visit at the two sites from the public payer’s perspective, a time-driven activity-based costing (TDABC) methodology was applied. This method requires estimates of two parameters: (1) the unit cost ($/min) of human resources, equipment, consumables and overheads and (2) the time required to carry out healthcare processes administered to patients.15 The cost of a care process is proportional to its duration, as measured in the field by a time-movement study. For example, the cost of ED triage can be estimated by summing the unit costs associated with the nurse, consumables and overheads allocated to the ED and multiplying by the triage duration as follows:

Cost of ED triage=(unit cost of nurse+unit cost of consumables+unit cost of overhead)×triage duration=($C0.78/min+$C0.07/min+$C0.17/min)×7.1 min=$C7.24

The cost of the index visit from the public payer’s perspective was thus estimated by summing the costs of the care processes administered from arrival to discharge, and the cost of publicly covered prescription drugs dispensed within 24 hours of the visit. The cost associated with return visits happening 3 and 7 days later was assigned to the healthcare facility where the initial visit occurred. This was calculated by combining: (1) the average cost of index visit in the relevant healthcare setting, serving as an estimate for the return visit cost and (2) costs related to medical services and prescription drugs obtained from provincial databases, which were not initially factored into the cost of the index visit. The total cost of care on day 3 after the index visit included expenditures on medical services (eg, consultation, laboratory, imagery) from day 1 to day 3 and on prescription drugs on days 2 and 3. The total cost of care on the day after the index visit included expenditures on medical services and prescription drugs from the index visit up to that point.

The TDABC estimates were calculated from the perspective of the Québec Ministry of Health, which represents the public payer. Details of the methodology have been described in open-access publications.16 17 The unit cost of cost elements, a summary of overhead expenses and estimated cost of important care processes are reported in online supplemental table S1, S2 and S3.

Participants

Consecutive patients were included if they were children: (1) aged between 2 and 17 years old; (2) ambulatory during their entire visit or consultation and (3) discharged home with a diagnosis of upper respiratory tract infection (URTI), pneumonia or acute asthma. Children were excluded if they had sought medical attention for a similar issue within the preceding 30 days or were not covered by the provincial health insurance plan.

Patient and public involvement

None.

Data source

Research assistants collected data from electronic medical records at both locations. They compiled various information including age, biological sex, presenting complaint(s), comorbidities (Charlson index score), number of regular medications (serving as an indicator of comorbidity burden18), arrival and discharge times, CTAS score, vital signs on arrival (heart rate, saturation), laboratory tests, imaging and interventions conducted during the index visit, as well as discharge diagnosis and prescribed medications. Health outcomes such as return visits, hospitalisations and mortality rates, use of publicly covered pharmaceutical services and physician billing were sourced from provincial administrative databases (Régie de l’assurance maladie du Québec and l’Institut de la statistique du Québec).

Outcomes

The primary outcome measure was the proportion of children who revisited any ED or outpatient clinic at 3 and 7 days following the index visit. The secondary outcome measures included (1) the mean cost of care provided during the index visit, as well as within 3 and 7 days after the index visit; (2) hospital or intensive care unit admissions within 30 days; (3) deaths within 30 days; (4) administration of antibiotics for URTI; (5) performance of chest X-rays for URTI and acute asthma19 and (6) compliance with provincial practice guidelines for corticosteroid prescribing for acute exacerbation of asthma.20 21

Statistical analysis

Descriptive statistical analyses are presented as means and SD or 95% CI for continuous variables, while counts and percentages are used for categorical variables.

To capture the inefficiency of potential duplicated care, referrals from the walk-in clinic to the ED were counted in the clinic group and the ED evaluation and discharge to home was recorded as a return visit. The index visit cost was estimated as the cost of the care processes administered from arrival to discharge as determined with TDABC plus the cost of publicly covered prescription drugs dispensed within 24 hours of the index visit. The return visit cost was allocated to the index visit setting as the mean cost of an index visit in that setting, plus the expenditures on medical services and prescription drugs obtained from provincial Ministry of Health databases. All costs are reported in 2017 Canadian values (CAD).

To compare the ED to the walk-in clinic, generalised linear models were employed with probability distributions adapted to the outcome under evaluation. We used an overlap weights approach to adjust for the following potential confounders: age, sex, comorbidities, age-adjusted Charlson index, number of regular medications and vital signs. With this approach, each individual received a weight proportional to the inverse of their probability to choose a particular setting.22 23 CIs were calculated using a robust variance estimator.24

Previous literature has shown that the accuracy of the CTAS score in paediatric emergency medicine is only moderate, and that the inter-rater reliability among emergency nurses is inconsistent.25 As a result, the CTAS score was not used as an independent variable for risk adjustment. However, a subgroup analysis retaining only patients with CTAS 4 and 5 was conducted to account for residual confounding from a potential difference in disease severity between the two groups.

The value of care delivered in each care setting was illustrated with an operational effectiveness graph.26 Adjusted mean costs of care for acute ambulatory respiratory conditions were plotted on the x-axis, and 3-day return visit proportions were plotted on the y-axis. Points closest to 0 on both axes represent the highest value of care; the lower the return visit proportion and cost of care, the higher the value of care.

Statistical differences for all outcomes were calculated with a significance level (α) set at 0.05. The statistical analyses were conducted by using SAS, V.9.4 (SAS Institute).

ResultsCharacteristic of participants

Overall, 532 children were included in the ED and 201 in the walk-in clinic (table 1). The mean age (SD) was 6 (4) years and 5 (3) years in the ED and the walk-in clinic, respectively. In the ED and the clinic, 95% of children had a Charlson index score of 0. Their mean vital signs were within the normal range in both settings. The mean crude length (SD) of stay was longer in the ED than in the walk-in clinic (4.5 (2.3) hours vs 4.0 (2.6) hours). More information regarding the baseline characteristics before and after overlap weighting is presented in online supplemental tables S4, S5 and figure S1.

Table 1

Characteristics of patients, for the whole cohort and per discharge diagnosis (N=733)

Outcomes

For the primary outcome, the incidence of return visits at 3 and 7 days was 20% and 27% in the ED vs 6% and 11% in the walk-in clinic, respectively (adjusted relative risk (aRR) (95% CI) 3.17 (1.77 to 5.66) and aRR 2.24 (1.46 to 3.44)) (table 2). The mean cost (95% CI) of care (CAD) for the index visit was $C96.68 ($C92.62 to $C100.74) in the ED vs $C48.82 ($C45.47 to $C52.16) in the walk-in clinic (mean difference (95% CI): $C46.15 ($C41.29 to $C51.02)). Cost estimates at 3 and 7 days after the index visit were also significantly lower in the clinic (table 3). The operational effectiveness graph suggests that the value of care at 3 days after the index visit is higher in the walk-in clinic than in the ED, for the whole cohort and per discharge diagnosis (figure 1).

Figure 1

Operational effectiveness graphic representing the 3 days return visit proportion according to the mean cost (CAD) of the visit index, for the whole cohort and per discharge diagnosis.

Table 2

Incidence proportions and adjusted relative risks of return visits and hospitalisations in the participating emergency department (n=532) and walk-in clinic (n=201) for the whole cohort and per discharge diagnosis

Table 3

Comparison of the mean costs of care (CAD)* for the whole cohort and per discharge diagnosis

In terms of secondary outcomes, the incidence of hospitalisation within 30 days of the index visit was 2.4% in the ED vs 1.0% in the walk-in clinic (aRR 1.36 (95% CI 0.29 to 6.41)) (table 2). No patients were hospitalised in intensive care or died within 30 days of the index visit.

Antibiotic prescription for URTI was less common in the ED than in the walk-in clinic (1.5% vs 16.4%; aRR 0.10 (95% CI 0.03 to 0.32)), whereas the incidence of chest X-ray prescription for URTI or acute asthma was higher in the ED than in the walk-in clinic (35.2% vs 9.6%; aRR 3.18 (95% CI 1.94 to 5.21)). Compliance with guidelines for the use of antibiotics in pneumonia and oral corticosteroid in asthma was good and similar in both settings (table 4).

Table 4

Incidence proportions of compliance to quality of care indicators according to practice guidelines

The subgroup analysis of CTAS 4 and 5 patients did not alter the results (online supplemental tables S6–S8).

Discussion

We conducted a retrospective cohort study and applied TDABC to compare the value of care provided to ambulatory children presenting to an ED and a walk-in clinic for an acute respiratory condition. After risk adjustment for age, sex, comorbidities and vital signs, the proportion of return visits at 3 and 7 days was up to three times higher and the cost was almost twice as high in the ED. More chest X-rays were performed in the ED for URTIs, but more antibiotics were prescribed in the clinic for the same condition.

Very few studies have compared the value of care in EDs and walk-in clinics for low-acuity ambulatory conditions in the paediatric population. Most report results overall without stratifying for adult and paediatric populations. After examining insurance claims of both adult and paediatric patients with otitis media, pharyngitis and urinary tract infections, the study by Mehrotra et al 27 conducted in Minnesota (USA) suggested that (1) the mean cost of care was five times lower in retail clinics than in the ED and (2) the quality of care, as assessed by compliance with clinical guidelines for the same conditions, was significantly higher in retail clinics than in the ED.27 28 Campbell et al reported similar trends in Ontario (Canada).28 In their prospective cohort study comparing EDs, walk-in clinics and primary care practices in the management of minor acute conditions, the risk of healthcare reuse and the cost of care were 6.5 and 4 times higher, respectively, in the ED than in the other settings. In our own study, an ED visit resulted in a 2-fold increase in costs and 7-day return visits than in the walk-in clinic, for which a 10-fold increase in antibiotic use for URTI. According to electronic medical records, a large proportion of walk-in clinic patients with URTI received a delayed prescription to be filled in the absence of natural improvement of symptoms. Although a systematic review failed to find an association between delayed prescribing and decreased ED or clinic return visits due to a lack of data, the results of this study suggest that this possible association merits further investigation in future studies. Antibiotic resistance is an important public health issue and any approach that may potentially increase antibiotic use has to be carefully assessed and weighed.

Some authors have suggested that walk-in clinics may increase overall healthcare costs by duplicating care through a high level of return following an initial visit.29 30 Our results contradict this since the adjusted incidence of return visit was lower in the walk-in clinic than the ED. However, a higher incidence of return visits does not necessarily mean that the quality of care in the ED is poor. Indeed, there is some evidence to suggest that return visits do not always reflect poor quality of care as the majority of return visits to the ED would be due to patient non-compliance with ED discharge instructions or the natural progression of the disease.31 32 Moreover, some studies highlight that younger age6 33 and lower parental level of education34 35 are predisposing factors associated with low acuity ED visit. Younger children were more likely to present with non-urgent conditions in the ED due to parental-perceived severity of illness and perception of obtaining specialist services.5

Our study has potential clinical implications. In the past, some have raised concerns about the quality of care provided in walk-in clinics to the paediatric population.36 Although campaigns on antibiotic stewardship may be needed, our results and existing literature do not support this fear. To relieve the clinical and financial burden on EDs, health authorities should possibly consider increasing access to this alternative setting for paediatric ambulatory patients with acute health concerns. A more comprehensive research programme is now needed to better determine for what conditions and under what circumstances walk-in clinics may be an appropriate care alternative to EDs.

This study has a number of limitations. First, although we have adjusted our results for several important confounders using an overlap weights approach, residual confounding may still be present (eg, socioeconomic status, education level).

Second, this study is a medical record review with the limitations on data robustness that this design implies. However, key outcome measures were extracted from provincial administrative databases, which are reliable for these outcomes. Finally, although significant variations were observed in costs and quality of care between the participating ED and walk-in clinic, this study was conducted in only two sites. Therefore, a multicentre cohort study is now needed to confirm these results and is currently in preparation.

In conclusion, the incidence of return visits and cost of care for ambulatory children with acute respiratory conditions were significantly higher in the ED, while antibiotic use for URTI was more frequent in the walk-in clinic. These data may help determine which setting offers the highest value to low acuity conditions in the paediatric population.