Sleep is vital to recovery from illness, yet it is frequently interrupted in the hospital setting. Existing literature relying on survey data identifies vitals, medications, and pulse oximetry as major disruptors to sleep. This study was designed to assess the degree to which these candidate sleep disruptors are associated with objective room entries.
Room entry sensors were placed on doors to 18 rooms on acute medical–surgical units at a tertiary academic center. The number of entries into rooms between 10 Pm and 6 Am were logged on patients admitted to hospital medicine services from March 2021 through February 2022. Medical records were reviewed to extract orders for vital sign frequency, medication timing, continuous pulse oximetry, and intravenous fluid use overnight. Negative binomial regression was used to evaluate associations.
Room entry data were collected for 112 admissions and 192 patient-nights. There was an average of 7.8 room entries per patient-night. After adjustments for the other variables and for patients represented in multiple nights, vitals ordered every 4 hours were associated with a 1.3-fold increase in room entries (95% confidence interval 1.0–1.5; P = .013), as were medications scheduled during overnight hours (1.3; 95% confidence interval 1.0–1.5; P = .016). There was no association between room entries and continuous pulse oximetry use. After adjustment, there was also no association with administration of intravenous fluids.
Vitals ordered every 4 hours and medications scheduled during sleep hours are independently associated with increased room entries and may be reasonable initial targets for quality improvement interventions designed to minimize nighttime disruptions.
Adequate sleep is critical to recovery from illness,1,2 yet hospitalized children have their sleep interrupted frequently and receive significantly fewer hours of sleep than recommended.3,4 When surveyed, caregivers consistently report that vitals, pulse oximetry, medications, and alarms are most disruptive to their children’s sleep.5–8 A quality improvement (QI) intervention designed to decrease overnight blood pressure measurement frequency demonstrated improvement in caregiver-reported disruptions and reported quantity of patient sleep.6 Although the number of caregiver-reported nighttime awakenings has correlated with objective room entries,4 it is unclear whether factors that cause awakening correlate with caregiver reports.
Our study was designed to assess the degree to which common inpatient practices are associated with objectively measured nighttime room entries. These data are intended to inform QI interventions designed to minimize overnight disruptions.
Methods
Wireless entry sensors (YoLink) were placed on 18 doors on the acute medical–surgical units at a tertiary academic center consisting of 80 total general medical–surgical beds. These sensors communicated via hub devices to a smartphone application. Within the smartphone application, date/time stamps of each door opening and closure were logged. Doors were selected for sensor placement on the basis of strength of wireless signal achieved. Sensors were small and discreet. Front-line providers were neither educated on the purpose of the sensors, nor were they specifically blinded to the study.
Patients were assigned to hospital rooms on the basis of bed availability. Patients admitted to the hospital medicine service to rooms with room entry sensors were included in this study. The number of room entries between 10 Pm and 6 Am was the primary outcome. Sensor data were accessed and logged Monday through Friday from March 2021 to February 2022. The first night of admission/transfer to the floor and any nights involving rapid response team activation or codes were excluded to minimize outlier nights.
Definitions of events were established a priori. Sensors were able to distinguish between a door in an open position and a closed position; thus, we defined a room entry to be sensor activation for entry, followed by sensor activation for closure. There were 2 exceptions to this:
if a door opening and closure occurred within 30 minutes of another door opening and closure, and each individual opening and closure were separated by <30 seconds, this was counted as a single room entry (as opposed to 2 entries) to account for the likelihood of the door being closed while the provider remained in the room; and
if a door was left open (no door closure after a door opening) for >2 hours, the night was excluded to prevent systematic underestimation of room entries, because multiple entries/exits were likely to have occurred while the door was left open.
Sensors that lost wireless signal were similarly excluded.
Medical records of patients meeting inclusion criteria were reviewed to record demographic data and extract orders for vital sign frequency (every 4 hours [Q4H] versus less frequently), scheduled medications given between 10 Pm and 6 Am (confirmed given in the medication administration record [MAR] and recorded as yes/no), continuous pulse oximetry use between 10 Pm and 6 Am (confirmed used on the basis of documentation of continuous pulse oximetry device data points, which are automatically uploaded in the vitals flowsheet), and intravenous fluids (IVF) use between 10 Pm and 6 Am (as documented in the MAR and recorded as yes/no). No patients in this study had vital signs collected more frequently than Q4H.
Negative binomial regression was used to evaluate correlation of the 4 predictor variables with the number of room entries. We evaluated each predictor individually in simple regressions and then included all 4 predictors in 1 multiple regression model. Data from some patients were captured on multiple nights. Because of this, we adjusted for clustering by individual for each model.
This study was approved as nonhuman subjects research through the university’s institutional review board.
Results
There were 4680 eligible patient-nights. Patient-nights were excluded if the patient was not on the hospital medicine service (61%), the night was the first night of admission/transfer (22%), the door was open for >2 hours or the sensor lost wireless signal (16%), or there was a rapid response or code (<1%). Ultimately, 192 patient-nights met inclusion criteria, including data from 112 unique admissions and 109 unique patients. Table 1 details patient demographics.
. | n . | % . |
---|---|---|
Age, y | ||
Infant (<1) | 27 | 24.1 |
Early childhood (1–5) | 36 | 32.1 |
Late childhood (6–12) | 18 | 16.1 |
Adolescence (13 and older) | 31 | 27.7 |
Primary diagnosis | ||
Gastrointestinal | 30 | 26.8 |
Respiratory | 28 | 25.0 |
Neurologic | 15 | 13.4 |
Rheumatologic | 10 | 8.9 |
Ophthalmologic and/or ENT | 9 | 8.0 |
Psychiatric | 6 | 5.4 |
Musculoskeletal and/or skin | 4 | 3.6 |
Renal and/or GU | 4 | 3.6 |
Hematology | 2 | 1.8 |
Endocrine | 2 | 1.8 |
Other | 2 | 1.8 |
Technology dependencea | 42 | 37.5 |
. | n . | % . |
---|---|---|
Age, y | ||
Infant (<1) | 27 | 24.1 |
Early childhood (1–5) | 36 | 32.1 |
Late childhood (6–12) | 18 | 16.1 |
Adolescence (13 and older) | 31 | 27.7 |
Primary diagnosis | ||
Gastrointestinal | 30 | 26.8 |
Respiratory | 28 | 25.0 |
Neurologic | 15 | 13.4 |
Rheumatologic | 10 | 8.9 |
Ophthalmologic and/or ENT | 9 | 8.0 |
Psychiatric | 6 | 5.4 |
Musculoskeletal and/or skin | 4 | 3.6 |
Renal and/or GU | 4 | 3.6 |
Hematology | 2 | 1.8 |
Endocrine | 2 | 1.8 |
Other | 2 | 1.8 |
Technology dependencea | 42 | 37.5 |
ENT, otolaryngology; GU, genitourinary.
Includes enteral feeding pumps, positive- pressure ventilation, tracheostomy and/or ventilator, intracranial shunts, and central lines.
In the entire cohort, there was an average of 7.8 room entries (SD 4.0) per patient-night. The minimum number of room entries was 1 and the maximum was 23. Of the 192 patient-nights, 149 (78%) included vitals ordered Q4H, 74 (39%) included the utilization of continuous pulse oximetry, 46 (24%) included administration of IVF, and 98 (51%) included the MAR-documentation of administration of at least 1 scheduled medication. Q4H vital signs and the administration of scheduled medications overnight were associated with the largest difference in mean room entries compared with when those were not done (2.5 and 2.5 additional room entries, respectively) (Table 2). Table 3 includes the association of each nighttime intervention with room entries, evaluating each metric individually and then adjusted for patients who had data collected on multiple nights. In this analysis, adjusted by patient, Q4H vitals were associated with a 1.4-fold increase in room entries (95% confidence interval [CI] 1.2–1.7; P < .001). Scheduled medications and IVF administration were associated with a 1.4-fold (95% CI 1.1–1.7; P = .001) and 1.3-fold (95% CI 1.0–1.6; P = .029) increase, respectively.
. | Not Present . | Present . | ||||||||
---|---|---|---|---|---|---|---|---|---|---|
. | n . | Mean . | SD . | Min . | Max . | n . | Mean . | SD . | Min . | Max . |
Q4H versus ordered | 43 | 5.9 | 3.0 | 1 | 13 | 149 | 8.4 | 4.1 | 1 | 23 |
Pulse ox used | 118 | 7.5 | 4.0 | 1 | 23 | 74 | 8.4 | 3.9 | 1 | 22 |
IVF administered | 146 | 7.4 | 3.7 | 1 | 19 | 46 | 9.4 | 4.7 | 4 | 23 |
Scheduled med administered | 94 | 6.6 | 3.2 | 1 | 17 | 98 | 9.1 | 4.4 | 1 | 23 |
. | Not Present . | Present . | ||||||||
---|---|---|---|---|---|---|---|---|---|---|
. | n . | Mean . | SD . | Min . | Max . | n . | Mean . | SD . | Min . | Max . |
Q4H versus ordered | 43 | 5.9 | 3.0 | 1 | 13 | 149 | 8.4 | 4.1 | 1 | 23 |
Pulse ox used | 118 | 7.5 | 4.0 | 1 | 23 | 74 | 8.4 | 3.9 | 1 | 22 |
IVF administered | 146 | 7.4 | 3.7 | 1 | 19 | 46 | 9.4 | 4.7 | 4 | 23 |
Scheduled med administered | 94 | 6.6 | 3.2 | 1 | 17 | 98 | 9.1 | 4.4 | 1 | 23 |
Max, maximum; med, medication; Min, minimum; ox, oximetry.
. | No Adjustments . | Adjustment of Variance for Cluster by Individual Patient . | ||||||
---|---|---|---|---|---|---|---|---|
. | IRR . | Lower 95% CI . | Upper 95% CI . | P . | IRR . | Lower 95% CI . | Upper 95% CI . | P . |
Q4H versus ordered | 1.4 | 1.2 | 1.7 | .000 | 1.4 | 1.2 | 1.7 | .000 |
Pulse ox used | 1.1 | 1.0 | 1.3 | .150 | 1.1 | 0.9 | 1.4 | .327 |
IVF administered | 1.3 | 1.1 | 1.5 | .002 | 1.3 | 1.0 | 1.6 | .029 |
Scheduled med administered | 1.4 | 1.2 | 1.6 | .000 | 1.4 | 1.1 | 1.7 | .001 |
. | No Adjustments . | Adjustment of Variance for Cluster by Individual Patient . | ||||||
---|---|---|---|---|---|---|---|---|
. | IRR . | Lower 95% CI . | Upper 95% CI . | P . | IRR . | Lower 95% CI . | Upper 95% CI . | P . |
Q4H versus ordered | 1.4 | 1.2 | 1.7 | .000 | 1.4 | 1.2 | 1.7 | .000 |
Pulse ox used | 1.1 | 1.0 | 1.3 | .150 | 1.1 | 0.9 | 1.4 | .327 |
IVF administered | 1.3 | 1.1 | 1.5 | .002 | 1.3 | 1.0 | 1.6 | .029 |
Scheduled med administered | 1.4 | 1.2 | 1.6 | .000 | 1.4 | 1.1 | 1.7 | .001 |
IRR, incidence rate ratio; med, medication; ox, oximetry.
Table 4 presents multiple negative binomial regression models including all 4 predictor variables. Q4H vitals were associated with a 1.3-fold increase in room entries (95% CI 1.0–1.5; P = .013). Scheduled medications were also associated with 1.3-fold increase in room entries (95% CI 1.0–1.5; P = .016).
. | No Adjustments . | Adjustment Variance for Cluster by Individual Patient . | ||||||
---|---|---|---|---|---|---|---|---|
. | IRR . | Lower 95% CI . | Upper 95% CI . | P . | IRR . | Lower 95% CI . | Upper 95% CI . | P . |
Q4H versus ordered | 1.3 | 1.0 | 1.5 | .013 | 1.3 | 1.0 | 1.5 | .013 |
Pulse ox used | 1.1 | 0.9 | 1.2 | .399 | 1.1 | 0.9 | 1.3 | .517 |
IVF administered | 1.2 | 1.0 | 1.4 | .054 | 1.2 | 0.9 | 1.4 | .157 |
Scheduled med administered | 1.3 | 1.1 | 1.5 | .001 | 1.3 | 1.0 | 1.5 | .016 |
. | No Adjustments . | Adjustment Variance for Cluster by Individual Patient . | ||||||
---|---|---|---|---|---|---|---|---|
. | IRR . | Lower 95% CI . | Upper 95% CI . | P . | IRR . | Lower 95% CI . | Upper 95% CI . | P . |
Q4H versus ordered | 1.3 | 1.0 | 1.5 | .013 | 1.3 | 1.0 | 1.5 | .013 |
Pulse ox used | 1.1 | 0.9 | 1.2 | .399 | 1.1 | 0.9 | 1.3 | .517 |
IVF administered | 1.2 | 1.0 | 1.4 | .054 | 1.2 | 0.9 | 1.4 | .157 |
Scheduled med administered | 1.3 | 1.1 | 1.5 | .001 | 1.3 | 1.0 | 1.5 | .016 |
IRR, incidence rate ratio; med, medication; ox, oximetry.
Discussion
Vital signs ordered Q4H and the administration of scheduled medications were associated with increased room entries between 10 Pm and 6 Am in general medicine patients admitted to a hospitalist service, when examined individually and with other predictor variables. Q4H vital signs and overnight scheduled medications are promising targets for QI work designed to minimize nighttime disruptions.
In our study population, there was an average of 7.8 room entries per patient-night, which is slightly <2 other pediatric studies using objective methods to record room entries. Erondu et al,4 using data from hand hygiene devices, recorded an average of 10 room entries per night. Their study was conducted within a unit that included patients admitted to pediatric hospital medicine, gastroenterology, and neurology services. Hinds et al9 asked caregivers and front-line providers in 2 pediatric cancer centers to manually log room entries and noted an average of 11.3 entries with a similar range as our study (3–22 entries). A lower average number of room entries in our study may be related to different severity of illness in the included populations or the different methods of measuring room entries. Further, previous studies, although strengthened by their use of objective metrics, did not specifically evaluate the presence of modifiable factors associated with increased room entries.
Both Peirce et al5 and Cook et al6 used the validated Potential Hospital Sleep Disruptions and Noises Questionnaire to survey pediatric caregivers for their subjective assessments of the nighttime interventions most disruptive to sleep. In each of these studies, vitals, medications, and continuous pulse oximetry were identified by caregivers as the most significant disruptions. Both relied on subjective measures of sleep disruption and did not specifically correlate these caregiver-identified disruptions with objective room entries.
These subjective reports have subsequently guided QI interventions targeting interruptions to sleep in the hospital setting. Published QI work on pediatric sleep in the hospital has focused on metrics representing proxies for sleep disruption. Cook et al6 used caregiver-reported sleep duration and disruptions, as well as the proportion of patients who did not have a blood pressure measurement ordered as their metrics. Lee et al10 analyzed both whether active vital signs (blood pressure and temperature) were ordered and obtained at 4 Am, whereas Lin et al11 tracked the percentage of patients admitted with failure to thrive or hyperbilirubinemia who had vital signs ordered and measured overnight (midnight and 6 Am). Mozer et al12 measured the percentage of orders with sleep-friendly administration times.
Our study fills a gap in the literature that is relevant to both past and future QI work seeking to improve patient sleep by establishing which modifiable factors are associated with increased room entries. Our work confirms that vital sign frequency and the timing of medication administration are reasonable proxies for room entries. Our work also suggests that these 2 interventions (ie, vital signs and scheduled medication administration) may be higher-yield targets for QI leaders seeking to decrease unnecessary disruptions overnight. Because each of these modifiable factors appears to be independently associated with door entries, targeting both could have an additive effect.
Pulse oximetry use was not associated with increased room entries in our study, possibly because of nurses having the ability to silence alarms without entering patient rooms in our hospital. In the unadjusted model, IVF use was associated with increased room entries, but when adjusted for the other variables, this increase was no longer significant. This may reflect the grouping of tasks involving maintenance and monitoring of the IV line with other cares, such as vital signs and/or medication administration.
Our study has several limitations. It was conducted at a single-center, university-based children’s hospital, and the results may not be generalizable to other sites. Although the number of patient room entries does not necessarily reflect actual sleep disruption, others have demonstrated with video observations that the most common cause for awakening after sleep initiation was a staff member entering or exiting a room.13 Additionally, our institution utilizes monitors that can be centrally silenced outside of patient rooms, so room entries may not fully capture sleep disturbances attributed to alarms. Room entry sensors, although small and inconspicuous, are visible to providers looking for them. Although this fact theoretically had the potential to influence provider behavior, most front-line providers were not aware of their purpose or our study. Although we were unable to differentiate between room entries attributable to hospital staff versus caregivers, all room entries have the potential to disturb patient sleep. Finally, we were unable to adjust our results for individual patient illness severity; however, our intention was to identify broadly applicable targets for QI intervention in a patient population appropriate for admission to general acute care units.
Conclusions
Vital signs ordered Q4H and medications scheduled during sleep hours are independently associated with increased room entries and may be reasonable initial targets for QI work designed to minimize nighttime disruptions. Targeting both could have an additive effect.
FUNDING: No external funding.
CONFLICT OF INTEREST DISCLAIMER: The authors have indicated they have no conflicts of interest relevant to this article to disclose.
Dr McDaniel conceptualized and designed the study, led data collection, analysis, and interpretation, drafted the initial manuscript, and reviewed and revised the manuscript; Dr Seshadri contributed to the design of the study, participated in data collection, and critically reviewed and revised the manuscript; Dr Tackett contributed to the design of the study, conducted analysis and interpretation of data, and critically reviewed and revised the manuscript; Dr Ralston supervised the conceptualization and design of the study, supervised data collection, analysis, and interpretation, and critically reviewed and revised the manuscript; and all authors approved the final manuscript as submitted and agree to be accountable for all aspects of the work.
Comments