Improving Hospitalized Children's Sleep by Reducing Excessive Overnight Blood Pressure Monitoring

This prospective study assessed the impact that reducing unnecessary overnight BP monitoring could have on improving sleep in the general medicine ward of Comer Children’s Hospital at University of Chicago Medicine. A year-long needs assessment from June 2017 to June 2018 involved surveying patient caregivers and staff about sleep and disruptions in the hospital and at home. Survey inclusion criteria required that the patient’s caregiver had stayed with the hospitalized child overnight and that the patient was awake, cognitively intact, between the ages of 30 days and 18 years, and admitted to the general pediatrics, neurology, or gastroenterology services.

After the completion of a preliminary needs assessment, during which families identified nurse and/or physician interruptions and VS monitoring as the most disruptive to patients, the team focused on reducing overnight BP monitoring through a combination of nurse and physician education and EHR order set updates. Key primary and secondary drivers that impact the project’s aims were developed (Fig 1).

To better define “unnecessary” BP monitoring, exclusion criteria were developed with nursing and physician leadership, which determined that the following diagnoses warranted continuation of overnight BP monitoring: sepsis and/or serious bacterial infections, dehydration, cardiac disease, pre-existing diagnosis of hypertension, documented hypertension or hypotension during admission, medications that alter VSs (opioids, intravenous immunoglobulin, magnesium, etc), anaphylaxis, status asthmaticus, or kidney disease. These exclusion criteria were intended not as an exhaustive list but rather a supplement to clinician judgment.

Behavioral “nudges” were first defined by behavioral economists Thaler and Sunstein¹⁸  as small systems changes that can predictably encourage desired behaviors, thus nudging consumers toward one decision over another. The concept of nudges has since been used in health care QI work, and many studies have successfully used EHR nudges to alter clinician behavior.^19–23 

Pursuing VS monitoring as the top sleep disruptor, in May 2018, we incorporated nudges into the pediatric general medicine VS order set of the EHR system (Epic, Verona, WI). The default was modified to require physicians to select whether patients needed overnight BP checks in addition to other VS monitoring (Fig 2A). “Overnight” was defined as the checks routinely performed at 12 am and 4 am. This modification allowed physicians to continue overnight VS monitoring without BP checks if clinically indicated. A reference column was added to providers’ EHR patient list to indicate which patients had overnight BP monitoring ordered (Fig 2B).

Intern education sessions occurred during intern orientation in June 2018. Although all residents were invited, interns were specifically targeted because they are primarily responsible for EHR ordering. A short presentation discussed the intervention, which was also summarized and distributed on badge cards for quick reference (Fig 3).

The intervention was discussed with both the day and night nursing teams during their daily huddles throughout the month of June 2018. Nurses were empowered to discuss with physicians when their patients could forgo overnight BP monitoring. A review of patients’ overnight BP orders was incorporated into night rounds, which occurred nightly between the charge nurse and the resident overnight team. This check-in encouraged interprofessional communication and created a regular opportunity to de-escalate patients’ BP orders.

After the active intervention period, communication occurred periodically through e-mail and visiting physician and nursing rounds. Using continuous improvement strategies, these touchpoints reinforced the intervention, allowed the team to relay real-time data regarding important process measures, and created space for stakeholders to provide feedback.²⁴ 

The primary outcome measures of the intervention were reported sleep duration and physician and/or nurse disruptions collected via patient-caregiver surveys throughout the duration of the study. Most weekdays from February 2017 to May 2019, research assistants surveyed a convenience sample of patients to inquire about sleep habits in the hospital and at home. All surveys included the Potential Hospital Sleep Disruptions and Noises Questionnaire, which has been validated in adult patients and asks about specific nighttime disruptions in the hospital with a Likert-type scale.^25,26  This survey has been shown to have concurrent validity with objective room entries for pediatric patients.^16,17  Surveys for patients <24 months old included an adaptation of the validated expanded Brief Infant Sleep Questionnaire (Supplemental Information), and surveys for patients >24 months old used the validated Pediatric Sleep Questionnaire (Supplemental Information).^27,28  All survey data were stored in the Research Electronic Data Capture database (version 7.6.0; Vanderbilt University, Nashville, TN).

EHR order set use data regarding overnight BP monitoring were considered our primary process measure. Monthly reports from the Epic Clarity database between March 2018 and May 2019 were used to track adoption of the new overnight BP order options. These reports detailed VS orders for all patients in the pediatric general medicine service, and therefore, the proportion of patients forgoing overnight BP monitoring could be examined over time.

Another process measure involved cross-sectional surveys assessing the effectiveness of physician and nurse education sessions. A survey asked learners to agree or disagree with statements regarding their knowledge, attitudes, and behaviors surrounding overnight BP monitoring and the importance of high-quality inpatient sleep using a Likert-type scale (Supplemental Information). A historical pre-post design was used to measure the immediate knowledge transfer of this education session and also obtain a commitment to change.^29,30 

As a balancing measure, the team collaborated with pediatric intensivists and the pediatric emergency team (PET) to monitor patient care escalations. The method our hospital uses to track PET responses and outcomes does not allow PET activations to be linked to individual patient data, such as who had overnight BP monitoring orders. Therefore, to create a balancing measure, we partnered with our pediatric intensivists and the PET to develop a notification plan if they were activated to the bedside of a patient with orders for no overnight BP monitoring. We also asked nurse managers and staff nurses to proactively report any adverse patient care events related to the intervention and followed up monthly to address any concerns these teams raised.

Caregiver survey responses comparing in-hospital sleep duration preintervention and postintervention were analyzed by using an X-bar R statistical process control (SPC) chart, which measures both mean and moving range over time. Rules outlined by the American Society for Quality were used to identify special cause variation.³¹  The prevention mean was used as a baseline and a process change was implemented when special cause variation was identified.

Responses to the Potential Hospital Sleep Disruptions and Noises Questionnaire were dichotomized by defining a score of 1 as “no sleep disruption” and a 2 or higher as “sleep disruption.” The postintervention cutoff was defined as July 1, 2018, and the proportion of patients who were disrupted pre- and postintervention was compared via χ² test. All data were analyzed by using Stata 15.0 (Stata Corp, College Station, TX).

The proportion of EHR orders forgoing overnight BP monitoring was analyzed and plotted by using a single-group interrupted time series analysis (ITSA). Although final data are reported at one year postintervention, interim data were analyzed monthly to track the intervention and provide real-time feedback to stakeholders.

Physician and nursing survey responses from pre- and posteducation sessions were calculated and dichotomized at the median value of 4, with scores of 4 or 5 considered “agree” and scores of <4 considered “disagree.” The dichotomized data were compared by using a 2-sample test for proportions.

Survey methods were approved by The University of Chicago Institutional Review Board and the intervention was approved via QI determination. Institutional leaders at The University of Chicago sponsored this intervention via the Center for Healthcare Delivery Sciences and Innovation’s Choosing Wisely challenge.

In the preintervention period, 106 and 157 surveys were completed by caregivers of patients <24 months old and >24 months old, respectively, and in the postintervention period, 58 and 73 surveys were completed by caregivers of patients <24 months old and >24 months old, respectively. There were no statistically significant differences in age, sex, race, or primary diagnosis category between the two survey populations (Table 1). Survey results are shown in Figs 4 through 6. Caregivers reported a statistically significant increase in sleep duration of 82.4 minutes for patients >24 months old (411.1 minutes before versus 493.5 minutes after; 95% confidence interval [CI] 39.6 to 125.1; P < .001; Fig 4). The SPC chart showed special cause variation beginning in July 2018 based on a baseline mean of 411; a process change was introduced at that time, and the final SPC chart confirmed an increase in sleep duration from 411 minutes to 494 minutes (Fig 5). No statistically significant change in sleep length was seen for patients <24 months old (414.9 minutes before versus 393.0 minutes after; 95% CI −77.9 to 34.3; P = .44). Caregivers of patients >24 months old also reported a 25% decrease in nocturnal awakenings, although this difference was not statistically significant (2.0 awakenings before versus 1.5 awakenings after; 95% CI −1.0 to 0.03; P = .06).

Caregivers reported a 19% decrease in overnight disruption by nurses and physicians (42% before versus 23% after; P < .001), a 12% decrease by temperature monitoring (22% before versus 10% after; P = .004), and a 9% decrease by pulse oximetry monitoring (32% before versus 23% after; P = .05; Fig 6). The frequency of other commonly reported disruptions, such as medication administration, did not significantly change (30% before versus 30% after; P = .95).

There were 458 unique patients in the preintervention period and a total of 493 VS order records, signifying that some patients’ VS orders were modified throughout their hospitalization. There were 1214 unique patients and 1472 VS records in the postintervention period. Of these patients, 510 had orders forgoing overnight BP, which represents a gross average of 42% of all pediatric general medicine patients during the study period. ITSA showed that the proportion of patients postintervention without overnight BP orders increased in intercept by 50.7% after the active intervention period (95% CI 41.2% to 60.3%; P < .001) with a subsequent −0.16% decrease in slope each week (95% CI −0.32% to −0.01%; P = .037), as shown in Fig 7.

Results from education session surveys showed that nurses have a higher level of baseline knowledge and more positive baseline attitudes regarding forgoing overnight BP monitoring than did interns. Postsession surveys also show that education sessions were effective in improving intern comfort with forgoing overnight BP monitoring. Results of physician and nurse education sessions are detailed in the Supplemental Information.

Throughout the intervention period, the pediatric intensivists and PET did not report any instance of a patient care escalation event occurring in patients with orders to forgo overnight BP monitoring. Additionally, nursing staff reported no patient safety concerns related to patients without overnight BP monitoring.

Cost analysis of our intervention was estimated by using the principle of time-driven, activity-based costing.³²  Nurses at our hospital estimated spending 5 minutes performing each BP check, which occurs twice overnight. Our pediatric hospital medicine service manages an average of 15 patients each night. After the intervention, an average of 6 patients every night forwent overnight BP checks. Using the average staff nurse salary in Chicago of $75 000 per year and assuming nurses work 36 hours per week, 48 weeks per year, the average annual cost savings of our intervention due to recovered nursing time can be approximated at $15 842.01.

This QI pilot study used a combination of electronic behavioral nudges and nurse and physician education to decrease overnight BP monitoring for pediatric general medicine patients. Education session surveys successfully documented an improvement in knowledge, behaviors, and attitudes regarding the intervention among physicians and nurses while encouraging participation by obtaining a commitment to change.³⁰  ITSA shows a 50.7% increase (target 25%) in patients postintervention who were not disrupted overnight by BP checks. Additionally, this intervention increased caregiver-reported patient sleep length by an average of 82 minutes (target 30 minutes) for children >24 months old and reduced sleep disruptions by nurses and physicians by 19% (target 20%) with no reports of related patient care escalations or safety concerns. Because overnight BP monitoring can be disruptive for pediatric patients, this study shows that patient sleep can be improved by educating physicians and nurses about whether their patients need BP monitoring and creating EHR order sets that allow physicians to forgo BP monitoring in certain patients.

Results show a statistically significant increase in patient sleep duration over the course of the postintervention period. However, it must be noted that there were months in the preintervention period when patient sleep duration was above baseline, although this increase was not statistically significant. This time actually coincides with the public announcement of the Choosing Wisely challenge and the development of the intervention with nurse and physician leadership. These events may have brought increased awareness to the proposed intervention and caused early adopters to begin modifying their behavior to include sleep-friendly practices, although this increase could also be related to other unidentifiable factors. The increase in sleep duration postintervention has less variability than preintervention, which is exemplified by the special cause variation seen in the moving range chart and lends greater evidence for the intervention’s potential effect on patient sleep.

Caregiver surveys showed no statistically significant change in sleep duration for patients <24 months old. These patients could be considered a distinct population both medically and developmentally, which may carry unique impediments to adequate sleep while hospitalized. More work is needed to determine their specific barriers to sleep so future interventions can better target their individual needs.

Patient caregivers also reported a significant decrease in disruptions due to both pulse oximetry and temperature monitoring. Although not the focus of this study, education sessions may have led some clinicians to further consider forgoing all VS monitoring in patients as appropriate and may represent a spillover effect of the intervention. Additionally, education sessions encouraged nurses to consider all overnight patient disruptions, which may have led them to modify other impactful behaviors.

Results show the proportion of patients forgoing overnight BP slowly declining in the months after the intervention. The difficulty of sustaining a QI intervention after the active intervention period is widely acknowledged in the health care setting.^33,34  In this study, data derived from real-time EHR reports will continue to provide feedback to stakeholders and their teams. Additionally, residents regularly turn over in an academic hospital, and recurrent discussions of active QI initiatives will help re-establish institutional memory. Lastly, although not explored in this study, caregivers are powerful advocates for their hospitalized children, and making them aware of active QI initiatives may prompt them to discuss the need for overnight BP monitoring with their children’s care team.

Cost analysis shows annual savings of $15 842.01 due to an average of one hour of nursing time saved each night. Although significant, other ways to measure the overall impact of this time saved could consider that nursing time away from low-acuity patients could be reallocated to those who require extra care, thus increasing the impact and value of the care nurses provide.³⁵ 

Our team is expanding this project during a second Plan-Do-Study-Act cycle by targeting other top disruptors, such as nighttime medication administration. This Plan-Do-Study-Act cycle launch will coincide with new intern orientation and will include reinforcement sessions regarding overnight BP monitoring.

Because this study took place at a single institution in an urban environment, its generalizability to other settings may be limited. Our team is currently collaborating with another institution in an effort to manage this limitation. Physician and nurse surveys were performed immediately after education sessions, which limited the assessment of any extinction in knowledge or attitudes that may have occurred after the active intervention period. Additionally, EHR order set use could not be linked to individual patient rooms, so the influence of possible confounding factors such as a patient’s severity of illness, demographics, or length of stay could not be evaluated. Similarly, because systematic patient care escalation data could not be linked to a patient’s chart, we had to use a proxy balancing measure via a proactive query of nurses and pediatric teams. Lastly, objective measures of sleep, such as actigraphy, were not used because their feasibility is limited by the short length of stay of pediatric hospital medicine patients.

This study shows that a combination of EHR behavioral nudges and nurse and physician education focused on reducing unnecessary overnight BP monitoring can successfully decrease reported disruptions and improve pediatric inpatient sleep. An intervention that leverages the EHR as a tool to improve patient-centered care can have a positive impact on the pediatric patient experience.

The team thanks the leadership and staff of The University of Chicago Center for Healthcare Delivery Sciences and Innovation for sponsoring their annual Choosing Wisely challenge and the University of Chicago Medicine Data and Analytics team for their support with data requests. The team also thanks Dr David Gozal, the chair of the Department of Child Health and the Marie M. and Harry L. Smith Endowed Chair of Pediatrics at the Women and Children’s Hospital of the University of Missouri, for his mentorship and support.

BP

blood pressure

CI

confidence interval

EHR

electronic health record

ITSA

interrupted time series analysis

PET

pediatric emergency team

QI

quality improvement

SPC

statistical process control

VS

vital sign