Overnight consult message volume and patterns to pediatric subspecialty fellows

Abstract

In children's hospitals, pediatric subspecialists provide crucial diagnostic and therapeutic guidance for hospital‐based clinicians around‐the‐clock. Consultation commonly occurs through secure text message systems (STMS) with potential important impacts on trainee sleep and wellbeing. We sought to quantify STMS metadata to on call subspeciality fellows overnight in a single children's hospital over a 6‐month period (March to August 2024). We analyzed STMS metadata for message counts and timing patterns. We found significant differences in call shift message volume across subspecialists (p < .001). Fellows in the hospital received 32 more messages per shift compared to fellows at home (53 to 21; p < .001). We found significant differences in the median longest time between messages across subspecialties (p < .001), ranging from 1.6 to 10.0 h. Guidance on clinical workload duty hours for trainees was developed prior to technological advances in clinical communication and requires further evaluation considering overnight messaging burden and impact on sleep.

INTRODUCTION

Pediatric subspecialty fellows provide crucial diagnostic and therapeutic guidance to pediatric residents and hospitalists caring for acutely ill patients in children's hospitals. Overnight consultation can add value to patient care and optimize resources, ¹ and is an important component of fellowship training, but it carries substantial risk of sleep‐related impairment and burnout. ² , ³ Over the past years, clinical communication has evolved to occur through secure text messaging systems (STMS), including for subspecialty consultation. ⁴ These systems enable asynchronous communication for consulting fellows and in‐hospital care teams and have fundamentally altered the way clinical work is performed. STMS have increased the ease of communication, but messages are often frequent and superficial which may lead to unintended consequences for clinical workflows, patient safety, and clinician well‐being. ⁵ , ⁶ , ⁷ Accreditation on Graduate Medical Education (ACGME) work hour guidance does not address these technological changes. The impact of these evolving communication technologies on subspecialty trainee workload—particularly during overnight call—remains largely unmeasured, limiting the ability to identify necessary policy, workflow, or educational changes. We sought to characterize message volume and timing for pediatric subspecialty fellows at a single institution and examine the association of message patterns and trainee call shift location. We hypothesized that trainees in the hospital would receive more messages during the overnight shift as compared to those at home.

METHODS

Study setting

We performed a cross‐sectional study analyzing 6 months of Epic Secure Chat STMS data (March to August 2024) for nine different subspecialties at a children's hospital. Epic Secure Chat use started in March 2021 for clinical communication at the institution. Epic Secure Chat messages are grouped by conversation and can be sent and received on both mobile and desktop devices. Our institution provided general guidance on phone calls versus secure messaging, but no specific guidance around overnight consult messages. The protocol was approved by our Institutional Review Board with waivers of consent and HIPAA authorization.

Data collection

We extracted Epic Secure Chat metadata from the Epic Clarity Database. Metadata fields included senders, recipients, sent and read timestamps, conversation ID, and patient ID. Conversations are groups of messages threaded together between users and have a unique identifier. Conversations may also be linked to the patient chart by users. All secure chat messages sent to the on‐call fellow during their call shift from 5 p.m. to 7 a.m. were included in the analysis. We used a convenience sample of 9 pediatric subspecialties based on schedule availability. We extracted on‐call schedules for each subspecialty. Three subspecialties are in‐hospital for call shifts (Cardiology, General Surgery, and Neurology) while six are at home (Endocrinology, Gastroenterology, Infectious Diseases, Pulmonology, Rheumatology, and Urology). Schedules were confirmed by a fellow from each subspecialty. Each subspecialty only had one trainee on call each evening, apart from Neurology, which had two trainees on call as primary and secondary coverage for 37% of the dates analyzed. Clinical responsibilities across subspecialties include consults from the emergency department, inpatient hospital teams, and outside hospitals. For those fellows in the hospital, responsibilities include direct patient care and procedures. The trainees were listed as first contact for all subspecialities except for General Surgery and Urology, where they are listed as second contact.

Data processing and analysis

For each subspecialty, we calculated the total messages received by the trainee on call for each call shift and by hour, and count of unique conversations per call shift. For neurology call shifts with two trainees, we counted the same message sent to both trainees simultaneously as one message read. Median messages per shift were determined for each subspecialty. We compared message totals before and after midnight. We calculated the longest time on shift without reading a chat (maximum of 14 h). Median longest time between chats was determined for each subspecialty. Descriptive statistics were calculated as medians and interquartile range (IQR) for continuous variables and frequencies and percentages for categorical variables. Differences in continuous variables were assessed using Kruskal–Wallis rank sum test. Differences in categorical variables were assessed using Pearson's χ ² test. Data processing and statistical analysis were conducted using R Statistical Software (v 4.4.1; R Core Team 2021). For all tests, statistical significance was defined as a p‐value < .05.

RESULTS

We analyzed 6 months of STMS data across nine different subspecialties for fellows on overnight clinical call shifts. A total of 123 fellows received 70,855 messages during the study period. Ninety three percent (19,943 of 21,463) of STMS conversations were linked to a chart, for a total of 8,101 unique patients. We found a significant difference in median total messages per shift across each subspecialty (p < .001, Table 1). The highest median total messages were in Neurology (N = 123, IQR: 88, 168) and lowest in Rheumatology (N = 6, IQR: 1, 14). We found significant differences in median message volume when comparing fellows taking in‐hospital call compared with those at home (53 vs. 21; p < .001, Table S1). Fellows on call in the hospital received a median difference of 32 more messages a night than fellows on home call. Neurology had the highest number of conversations per shift (N = 33, IQR: 27, 40) and Rheumatology had the lowest (N = 3, IQR: 1, 5). In house fellows had a median difference of 10 more conversations per night compared to at home fellows (18 vs. 8; p < .001, Table S1).

Table 1.

Message characteristics by subspecialty.

Specialty:	GIa	Endoa	IDa	Pulma	Uroa	Rheuma	Neuroa	Carda	GenSurga	p valueb	Effect sizec
Number of messages per night	56 (38, 73)	37 (26, 56)	21 (14, 29)	18 (11, 28)	14 (7, 29)	6 (1, 14)	123 (88, 168)	50 (34, 65)	23 (14, 39)	<.001	0.571
Number of unique conversations per night	20 (14, 25)	14 (10, 18)	8 (6, 10)	8 (5, 11)	4 (2, 8)	3 (1, 5)	33 (27, 40)	18 (14, 22)	7 (5, 11)	<.001	0.648
Longest time between messages per night, hours	3.2 (2.3, 4.7)	3.6 (2.6, 5.2)	6.4 (4.9, 8.1)	6.5 (4.8, 8.0)	8.6 (5.7, 10.9)	10.0 (7.8, 13.7)	1.6 (1.1, 2.6)	3.0 (2.0, 4.1)	4.5 (3.3, 6.0)	<.001	0.504
Number of conversations linked to patients	3,342 (92%)	2,519 (96%)	1,340 (90%)	1,417 (94%)	918 (95%)	542 (97%)	5,721 (94%)	2,863 (89%)	1,409 (91%)	<.001	0.091

Note: Subspecialties in order: Gastroenterology, Endocrinology, Urology, Infectious Disease, Pulmonology, Rheumatology, Neurology, Cardiology, General Surgery.

^a Median (Q1, Q3).

^b Kruskal–Wallis rank sum test.

^c Ordinal epsilon‐squared as defined in Peres, F. F. (2025) or Cramer's V.

We examined the timing of messages throughout the shift (Figure 1). All subspecialties had their highest messaging burden prior to midnight, but there was a significant difference based on fellow call location (p < .001, Table S1). We calculated the longest time between chats read while on call and found a significant difference by subspecialty (p < .001, Table 1). Neurology had the shortest time between chats read (1.6 h, IQR: 1.1, 2.6); Rheumatology had the longest (10.0 h, IQR: 7.8, 13.7).

Figure 1.

Median number of read messages by hour for on call fellows.

DISCUSSION

We found a variable burden of messaging across many subspecialties overnight, not just in total messages received each night, but also timing of messages throughout the night that could impact sleep. We found significant differences in call shift message volume and timing across subspecialties and between at home versus in‐hospital call. Our study quantifies the number and timing of messages pediatric fellows may receive during a call shift and provides baseline data across multiple subspecialties.

We found a wide range of messages per shift across each subspecialty. The differences are most likely multifactorial based on the frequency of common hospitalized conditions and patient volume. Patient acuity and urgency of concern may also explain part of the differences. Although call location alone does not fully predict message volume across all subspecialties, overall fellows physically in the hospital received more messages than those on home call, perhaps because of perceived availability, or subspecialties with higher patient acuity opt to have in‐house fellow coverage overnight. While the message volume for fellows on home call was generally lower, this was not consistent across every subspecialty. More exploration is needed to understand drivers of message volume and key differences by subspecialty.

The highest message burden was prior to midnight, with significant differences in before and after midnight messaging based on fellow location. Understanding the timing of messaging could drive updates to messaging guidance. Message timing is an underrecognized dimension of messaging patterns that helps us better understand messaging impact on sleep compared to message volume. Substantial variation in the longest time between read messages suggests some fellows experience near‐continuous interruptions, while others may have an opportunity for sleep.

The rates and timing of messages raise important questions around call duties for pediatric fellows. For those in‐hospital, this data raises questions about balancing triaging messages and time needed at the bedside for clinical care. With advances in communication technologies, the designation of home vs in‐hospital call may no longer adequately capture the workload that fellows may perform during a call shift, and considerations for how the following workday should be handled may need to account for the increased communication workload burden. Our analysis is limited to a single academic center that used a convenience sample and studied only one form of clinical communication. The fellows represented also had different contact roles within the system.

ACGME duty hour guidance for subspecialty trainees was formed prior to the widespread use of these communication technologies and does not explicitly address technology‐mediated clinical communications work, creating potential gaps between policy and modern clinical workflows. As a result, subspecialty programs may experience differing interpretations of workload, responsibility, and recovery time related to overnight call. Being at home did not protect against a high messaging burden or around‐the‐clock messaging.

The optimal balance of clinical responsibilities for clinical trainees is difficult to know. More studies are needed to understand the drivers of message volume and timing, and balance it with trainee workload as these are important factors for maintaining a pediatric subspecialty workforce. Solutions to on call message patterns should consider not only reducing message volume but also optimizing timing and batching of communication. They will need to be institution and subspeciality specific as clinical needs are dynamic, and so too should our support to pediatric trainees.

CONFLICT OF INTEREST STATEMENT

Dr. Coogle receives honoraria from Elsevier, unrelated to this project. Dr. Bonafide has current and recently completed research funding support from NIH, AHRQ, PCORI, and NSF, unrelated to this project. Dr. Ruppel has current research funding from NIH and PCORI unrelated to this project. Dr. Luo has current research funding from NIH and PCORI unrelated to this project. Dr. Tomlinson has current research funding from NIH unrelated to this project. The remaining authors declare no conflicts of interest.

Supporting information

Supporting File 1

DATA AVAILABILITY STATEMENT

Research data are not shared.