Development and feasibility testing of the Pediatric Emergency Discharge Interaction Coding Scheme

Abstract Background Discharge communication is an important aspect of high‐quality emergency care. This study addresses the gap in knowledge on how to describe discharge communication in a paediatric emergency department (ED). Objective The objective of this feasibility study was to develop and test a coding scheme to characterize discharge communication between health‐care providers (HCPs) and caregivers who visit the ED with their children. Design The Pediatric Emergency Discharge Interaction Coding Scheme (PEDICS) and coding manual were developed following a review of the literature and an iterative refinement process involving HCP observations, inter‐rater assessments and team consensus. Setting and participants The coding scheme was pilot‐tested through observations of HCPs across a range of shifts in one urban paediatric ED. Main variables studied Overall, 329 patient observations were carried out across 50 observational shifts. Inter‐rater reliability was evaluated in 16% of the observations. The final version of the PEDICS contained 41 communication elements. Results Kappa scores were greater than .60 for the majority of communication elements. The most frequently observed communication elements were under the Introduction node and the least frequently observed were under the Social Concerns node. HCPs initiated the majority of the communication. Conclusion Pediatric Emergency Discharge Interaction Coding Scheme addresses an important gap in the discharge communication literature. The tool is useful for mapping patterns of discharge communication between HCPs and caregivers. Results from our pilot test identified deficits in specific areas of discharge communication that could impact adherence to discharge instructions. The PEDICS would benefit from further testing with a different sample of HCPs.


| Background
Emergency departments (EDs) are the leading providers of unscheduled care. 1,2 Comprehensive discharge communication is a key component in the provision of quality care in these settings. 2 With over 85% of patients discharged home from the ED, ensuring that they have the necessary information to manage their care at home after leaving the ED is vital. 2,3 However, discharge communication is often hindered by the chaotic and fast-paced nature of the ED, which can result in frequent interruptions for health-care providers (HCP). 4 Other environmental barriers that impact discharge communication in the ED include overcrowding, noise, patient and caregiver stress, and time constraints. 4,5

| Importance
Inadequate discharge communication can have undesirable consequences for the patient and family, such as underutilization of follow-up services, adverse drug events and parental uncertainty. 3 The effectiveness of standardized instructions to enhance discharge communication in the ED is equivocal. 6,7 The content of discharge instructions in an ED setting has been shown to vary, and there is currently no consensus on the optimal content and delivery format across different emergency practice settings and illness presentations. [8][9][10][11] Further, there is a lack of policy in place to support discharge communication practice in an ED context. 6,12 Patient/caregiver comprehension of discharge communication has been found to be an important factor to improve care at home and prevent unnecessary return visits. 13 As such, greater understanding of the patterns and characteristics of discharge communication in a paediatric ED is needed to inform the design of discharge communication strategies and policies and improve outcomes for children and families. At present, there are no tools available, which could be used to characterize and study discharge communication in a paediatric ED.

| Goals of this investigation
The aim of our pilot project was twofold. First, we sought to develop a discharge communication coding scheme and coding manual that could be used to accurately and reliably code discharge communication between HCPs and parents in a paediatric ED. Second, we conducted a pilot study to test the reliability of the coding scheme and begin to describe the content and patterns of discharge communication between paediatric emergency department HCPs and parent caregivers.

| Development of the PEDICS
An interdisciplinary research team consisting of an ED physician, a registered nurse, a psychologist, a pharmacist and a knowledge translation researcher was established to assist with development and revision of the Pediatric Discharge Interaction Coding Scheme (PEDICS) and coding manual. Following a review of the literature, an initial list of 34 discharge communication behaviours was developed to populate the coding scheme. 3,4,6,8,[13][14][15][16] A research assistant then shadowed three staff physicians and nurses during shifts in a paediatric ED to adjust the sequence of the discharge communication behaviours on the list to mirror how they occur in practice. The research team also grouped similar codes under parent nodes to improve the flow of the tool. We adopted a broad definition of discharge communication to include the exchange of information to inform caregivers about the diagnosis of their child, the treatments received in the ED and plan for follow-up after discharge. Operational definitions of the codes (ie definitions based on observable characteristics) were developed for each discharge communication behaviour, including examples of what would and would not reflect the behaviour. Definitions were written using general descriptive terms that could be applied by interprofessional coders.

| Study setting and population
A convenience sample of eight physicians and nine registered nurses working in an academic paediatric ED, with an annual census of 27 000 patients, agreed to participate in the pilot study. Data were captured during a 5-month period of observation from June 2013 to October 2013. Trainees, such as nursing students, medical students and residents, were excluded from observations. Written informed consent was obtained from each participant. During an observation block, HCPs obtained verbal assent and consent from the patient and caregivers, respectively, before the coder entered the room. The location where the communication occurred and the number of interruptions during each interval were recorded. It was also recorded whether the interval was the first interaction with the patient/ caregiver and whether the patient was discharged during the observation block. This was performed to track the number of complete data sets of observed communication behaviours between the HCP and caregiver during a patient visit. The individual initiating the communication was also recorded. For example, when a HCP discussed a communication element with the caregiver in response to a question (RTQ), this was noted on the coding sheet. In addition to coding communication behaviour, we also collected the following demographic data for each unique patient observation: time of triage, chief complaint and diagnosis at time of discharge.

| Pilot testing and refinement of PEDICS
Following the initial pilot testing, a second coder was trained. The two coders contributed to refining the tool and the coding manual using an iterative process and under the supervision of the team. A second coder was present for 16% (n=8) of the observation blocks. Following every two inter-rater observation blocks, the coders met with the research team to review any discrepancies and to make modifications to the coding manual and the coding sheet as needed. The final version of the coding scheme included 41 distinct codes, which were organized and grouped into 11 common nodes through a consensus process by the team. Each code was limited to one to three words to facilitate formatting on a single coding sheet. The communication elements were grouped into eleven categories based on the usual flow of care processes in the ED: (i) introduction, (ii) tests, (iii) medications given in the ED, (iv) discharge, (v) diagnosis, (vi) treatment plans, (vii) medications for home, (viii) social concerns, (ix) follow-up, (x) clarification and (xi) conclusion (Table 3).

| Analysis
Cohen's Kappa score was calculated for each of the 41 codes in the PEDICS. 17

| Ethics approval
The study was approved by the research ethics board at the institution where data were collected (approval #: 1014414).

| Demographics
Overall 329 patient observations were carried out across 50 observational shifts (24 physician shifts, 168 physician-caregiver observations; 26 nurse shifts, 161 nurse-caregiver observations). There were a total of 148 complete observations, meaning the observation included HCP and caregiver/patient interactions from admission to discharge within the 4-hour observation block. The majority of children seen during the observation shifts were categorized as less urgent. The distribution of Canadian Triage Acuity Scale (CTAS) scores was as follows: triage level four (less urgent) (n=170, 56.1%) followed by levels three (urgent) (n=89, 29.4%) then two (emergent) (n=38, 12.5%), and the least number of children were triage levels five (non-urgent) (n=3, 1.0%) and one (resuscitation) (n=3, 1.0%). Common presenting complaints included fever, head injury, upper and lower extremity injury, vomiting and diarrhoea, cough and abdominal pain. The triage score was not captured for 26 (8%) of the observations. As shown in Table 1, children under the age of six comprised the largest patient population observed. The age of the patient was not recorded for 15 patients. The mother accompanied the child in 54.1% (n=85) of the nurse observations and 45% (n=75) of the physician observations ( Table 2). Both parents were present in 26% (n=41) of nurse observations and 35% (n=58) of physician observations. As shown in Table 3, the most common communication elements observed for both nurses and physicians related to introduction. The least common communication elements observed for both nurses and physician were communication elements related to "social concerns." The discharge communication behaviour of "asking whether the caregiver or patient needed clarification" was observed during 5.6% (n=9) of the nurses' observation, compared with physicians who performed this communication behaviour in 60.7% (n=102) of the patient encounters.
Thirty of 41 codes resulted in Kappa scores between .61 and 1.0 (Table 3). In general, those communication behaviours that were less frequently observed had lower Kappa scores.

| Location of communication
The majority of physician and nurse communication occurred in the patient's room, as shown in Table 4. During the nurse observations, 36.5% of the communication was observed during the triage process.

| Interruptions
A total of 117 interruptions occurred during the 168 physicianpatient observations, and 33 interruptions occurred during the 161 nurse-patient observations. The communication element "main concern" was most frequently interrupted in both physician and nurse communications, being observed 132 times and having 24

| Time of day
As shown in Table 5

| DISCUSSION
The development and pilot testing of the PEDICS in one urban paediatric ED with a convenience sample of 17 clinicians over four months  This code was often mistakenly coded as "diagnostic test" by the second coder. The differences between these codes were problematic as one might be used when the HCP was telling the caregiver the child needed an X-ray (diagnostic test) vs telling the caregiver that the child needed to be sedated to reduce a fracture based on the X-ray results (procedure needed). The second coder in our pilot study had limited clinical experience, which was useful for revealing gaps and providing clarity to the definitions in our coding manual.
We identified a number of challenges associated with in vivo coding in a busy paediatric ED. First, the limited space and fast pace of workflow in an academic ED pose challenges for including a second coder for reliability checks. These conditions make it difficult for multiple coders to view the observation at the same time, and it was not always easy for both coders to clearly hear the exchange. Flowerdew et al. 16 also found various sources of rater errors such as "missed behaviours" and "observed behaviours not recorded/judged" due to the high vol-  often be a barrier to the adherence of the discharge plan, and therefore exploring caregivers "social concerns" in the ED would allow the caregivers to voice their concerns such as limited access to primary care or being unable to afford the prescribed medication. 26,27 Ensuring that HCPs assess these types of barriers is paramount to providing tailored and relevant discharge information to all patients and caregivers.
Interestingly, this study found that HCP and patient/caregiver communication was most likely to be interrupted during two critical nodes: (i) main concern and (ii) ED follow-up. This finding is of concern due to the importance of these communication elements for both HCP and patient/caregiver comprehension of the medical issue.
Understanding "main concern" not only includes determining why the patient and caregiver came into the department, but also helps to focus the HCP on what to communicate with the patient/caregiver to ensure their expectations for the visit were met. 26 Interruptions during these critical nodes could lead to decreased willingness on the patient/caregiver's part to follow discharge instructions. 6,28 Findings from this study also suggest that caregivers are not ac-

| CONCLUSIONS
Our pilot work establishes the need and usefulness of a coding scheme to characterize discharge communication in a paediatric ED setting. Analysis of inter-rater reliability using Kappa scores found the majority of the PEDICS communication elements to have substantial inter-rater agreement. This coding scheme is beneficial in its ability to capture the location and frequency of discrete HCP and patient/ caregiver discharge communication behaviours in a paediatric ED context. Further evaluation of the PEDICS is required with a different sample of ED HCPs including learners. Findings from our feasibility study also suggest video recording as an important data collection strategy to accurately capture verbal and non-verbal communication behaviours, strengthen inter-rater reliability and map the multitude of factors that influence discharge communication.

CONFLICT OF INTEREST
The authors have no conflict to declare.

AUTHOR CONTRIBUTIONS
JAC, JC and SP conceived the study and designed the trial, and JAC obtained research funding; JAC supervised the conduct of the trial and data collection; AT completed the data collection and RH conducted inter-rater coding; JAC, SM and AT undertook recruitment of HCPs and patients and managed the data, including quality control; JAC, JC, AM and AT provided statistical advice on study design and analysed the data; JAC and AT drafted the manuscript and all authors contributed substantially to its revision; and JAC takes responsibility for the paper as a whole.