Keywords
choice architecture - clinical decision support - electronic health record - inpatient sleep - nudges
Background and Significance
Background and Significance
Computerized provider order entry (CPOE) and passive clinical decision support (CDS) systems are integral components of electronic health records (EHRs). These systems are designed to streamline workflows and promote evidence-based care.[1] To ensure the best patient outcomes, it is essential to understand how the design of CPOE environments, also known as choice architecture, influences clinician behavior.[2]
Computer-generated defaults, while offering benefits, can also lead to automation biases that may pose risks to patient care.[3]
[4]
[5]
[6]
[7]
[8] One concern arises from default around-the-clock frequencies, leading to the overuse of hospital resources and decreased sleep opportunities.[3]
[9] Overnight interventions, such as medication administrations or vital sign checks, have been identified by patients and staff as significant sleep disruptions.[10]
[11]
[12]
[13]
[14]
[15]
To address these challenges, we implemented a bundled approach using nudges, which includes various strategies to influence decision-making such as EHR order modifications, EHR-integrated messaging, and educational interventions.
Objectives
The objectives of our study were three-fold: (1) to identify EHR defaults that unnecessarily lead to around-the-clock interventions; (2) to assess the effectiveness of “nudge interventions,” informed by the principles of choice architecture, in altering clinician behavior toward selecting sleep-friendly orders, and (3) to investigate whether the adoption of sleep-friendly ordering, influenced by modifications of the EHR choice architecture, would effectively decrease overnight interventions in different hospital settings.
Methods
This observational study was conducted at the University of Rochester Medical Center, where millions of orders are placed yearly through its EHR system (Epic Systems Corporation, Verona, Wisconsin, United States).
Development of Intervention
A multi-step process was undertaken to develop the intervention. An interdisciplinary team from a 25-bed inpatient neurology unit (INU) at an academic medical center (AMC), including unit leadership, nursing, clinicians, pharmacists, and informaticists, conducted a thorough needs assessment involving EHR data analysis, staff surveys, and focus group discussions between October 2020 and February 2021. We prioritized the neurology unit due to their desire to improve sleep for neurological patients' recovery and operational support.
Staff identified vital signs, neurological (neuro) checks, and noncritical medications ordered every 4, 6, 8, and 12 hours as common sleep disruptions in surveys and focus groups.[12] In contrast, medications ordered twice, three times, and four times daily were considered sleep-friendly because they default to daytime-only schedules in the EHR. Other relevant literature has highlighted similar observations regarding the impact of orders associated with flexible versus inflexible medication schedules.[16]
[17]
Subsequently, we reviewed EHR medication administration data from 11 p.m. to 7 a.m. We found that acetaminophen and heparin deep vein thrombosis (DVT) prophylaxis accounted for 30% of overnight medications administered in the INU. We also found that the volumes of scheduled three times daily (TID) and every 8-hour (q8h) medication orders across the AMC were similar, but q8h orders were disproportionately more likely to be administered overnight (34% for q8h vs. 15% for TID orders).
We organized two 1-hour presentations in August 2021. The first session was conducted with the medical/surgical pharmacy team, while the second was with inpatient neurology residents and staff. Led by our pharmacist and neurology resident champions, these presentations emphasized the importance of selecting daytime schedules over around-the-clock schedules. During these meetings, we confirmed a preference for heparin over enoxaparin on the INU due to its shorter half-life and reduced risk of intracranial bleeding.
Challenges related to variability in heparin TID schedules between hospitals were identified. The community hospital affiliate (CHA) defaulted to an 0800, 1400, and 2000 schedule, while the AMC defaulted to 0900, 1700, and 2300. We advocated for CHA and AMC schedule alignment; however, stakeholders at the AMC did not approve this request due to concerns about nurse shift changes in the emergency department.
Next, we proposed EHR choice architecture modifications based on this medication administration data to decrease overnight heparin and acetaminophen administrations. Acetaminophen orders did not include a TID quick button, and heparin DVT prophylaxis defaulted to q8h. We added a TID quick button to all acetaminophen liquid and tablet orders and changed the heparin preselected default from q8h to TID. Order sets and panels with acetaminophen and heparin were updated accordingly.
Based on survey input, we suggested EHR modifications to all vital signs and neuro check orders. Before the intervention, orders featured four quick buttons for frequency selections: one time, every 4 hours, every shift (qshift), and daily. We added two quick buttons: every 4 hours while awake (q4WA) and TID while awake (TIDWA), with corresponding time ranges outlined (see [Fig. 1]). These selections also drove changes in the worklists of nurses and patient care technicians.
Fig. 1 (A) Preintervention vital signs order with 1 Time, Q4H, Q Shift, and Daily quick buttons. Scheduled times are typically collapsed and hidden. (B) Postintervention vital signs order to encourage sleep-friendly ordering. Q4H (6 a.m.–10 p.m.) and TID (8 a.m.–10 p.m.) quick buttons were added, Q shift quick button was removed. Scheduled times only show the first occurrence. To view future occurrences, the end user must click the calendar icon. Screenshots shared with permission from © 2023 Epic Systems Corporation. Q4H, every 4 hours; Q Shift, every shift; Q4H (6 a.m.–10 p.m.), every 4 hours while awake; TID (8 a.m.–10 p.m.), three times daily while awake.
We removed the qshift quick button from all neuro checks and vital sign orders, although it remained available as a searchable frequency. Unlike the other quick buttons, the EHR did not display the default q8h schedule for qshift to the ordering clinicians, causing confusion.
Specifically, when clinicians selected the qshift frequency, it automatically scheduled tasks q8h on nursing and patient care technicians' worklists. Clinicians were unaware of this q8h schedule on nursing worklists because the EHR did not show them this information. This lack of visibility led to confusion about the meaning of qshift in terms of scheduling. We did not change the every 4-hour default.
The final development step for choice architecture modifications was to assemble stakeholder groups consisting of nursing informatics, clinical informatics, pharmacy informatics, and operational leadership for the orders, who then reviewed and approved the above-proposed interventions for the AMC and CHA.
Implementation of Intervention
The EHR modifications were implemented in stages based on approval and build. The acetaminophen TID quick button was made available in August 2021, the heparin default change in October 2021, and the vital signs/neuro checks quick button changes in November 2021.
We used Secure Chat messaging, facilitated by the secure messaging feature within our Epic EHR system, to promote the intervention. In October 2021, pharmacists began messaging clinicians, asking them to consider modifying q8h frequencies to TID for all noncritical medications. In November 2021, resident champions sent messages to clinicians assigned to neurology inpatients at the start of each 2-week rotation, requesting them to review all around-the-clock orders and modify them if appropriate. We also promoted the initiative through educational means, including a 1-hour presentation to internal medicine residents and hospitalists in October 2021, our EHR educational newsletter in November 2021, and signage in the INU. [Fig. 2] provides a timeline of intervention development and implementation.
Fig. 2 Intervention timeline. Needs Assessment: October 2020–February 2021. INU Staff, Clinician, and Pharmacy Training and Engagement: August 2021–November 2021. Acetaminophen TID Quick Button Added: August 2021. Heparin Default Frequency Changed to TID: October 2021. Vital Signs/Neuro Check Quick Buttons Changes Made: November 2021. Resident Champions Secure Chat Neurology Clinicians: began November 2021. Education for Non-neuro Residents Rotating on Neurology Service: began January 2022. Secure Chat Expanded to All Covering Clinicians on the INU: April 2022. INU, inpatient neurology unit; TID, three times daily.
Data Analysis
We defined overnight as 11 p.m. to 7 a.m. based on average patient-reported sleep onset and offset on the INU. We used a single interrupted time-series analysis to estimate the proportion of sleep-friendly vital signs, neuro checks, and medication orders. We also evaluated changes in vital sign and neuro check assessments using EHR flowsheet data and medications administered using medication administration data. We analyzed the level (y-intercept) and the trend (slope) for pre- and postintervention periods, which provided information about the baseline status and the rate of change in the outcome variable over time, respectively, helping to evaluate the impact of interventions on the observed time-series data. We focused on two parameters: (1) immediate effect, defined as the change in mean levels between the last preintervention month and the first postintervention month following a 1-month washout period, and (2) change in slope between the preintervention period and the postintervention period following washout.
We analyzed data from February 2021 to June 2022 for all interventions. For vital signs, neuro checks, heparin DVT prophylaxis, and all medication administrations ordered TID or q8h, the pre-intervention phase spanned 8 months, followed by an 8-month postintervention period, separated by a 1-month washout when the nudge went live. Acetaminophen administrations had a 6-month preintervention period because the acetaminophen nudge went live 2 months earlier than the other nudges and a 10-month postintervention period, also separated by a 1-month washout. This varied duration of pre- and postintervention phases was important for capturing each intervention's impact over time and facilitating a comprehensive analysis across the same timeframe.
We included TID and q8h frequencies to analyze all medication administration data as the implemented nudges influenced them. Sleep-friendly vital signs and neuro check orders included two times daily, q4WA, TIDWA, and daily. Order frequencies and overnight interventions were extracted from the EHR and analyzed using R, version 4.1.2.
Results
No significant differences in demographics (age, sex, race, diagnosis) were observed among the patients admitted to the INU pre- and postintervention, although patient turnover decreased during the postintervention phase. See [Table 1] for additional details on demographics.
Table 1
Patient demographics on the inpatient neurology unit, academic medical center, and community hospital affiliate
Demographic
|
INU pre
N = 1,059
|
INU post
N = 664
|
AMC pre
(N = 1,992,372)
|
AMC post
(N = 2,291,340)
|
CHA pre
(N = 391,389)
|
CHA post
(N = 422,346)
|
Age (SD)
|
61.4 (19.1)
|
61.6 (20.1)
|
59 (23)
|
58 (23)
|
62 (20)
|
62 (19)
|
Gender (% female)
|
49%
|
52%
|
37.4%
|
37.3%
|
45.7%
|
43.8%
|
Length of stay (d), median (IQR)
|
3 (1–6)
|
3 (2–7)
|
3 (2–8)
|
4 (2–9)
|
2 (1–3)
|
3 (2–7)
|
Race (% Caucasian)
|
74.5%
|
75.5%
|
73.6%
|
74.1%
|
74.4%
|
74.5%
|
Primary diagnosis (% acute stroke)
|
36.8%
|
38.7%
|
0.011%
|
0.011%
|
0.336%
|
0.349%
|
Abbreviations: AMC, academic medical center; CHA, community hospital affiliate; INU, inpatient neurology unit; IQR, interquartile range; SD, standard deviation.
Note: Preintervention is defined as April 2021–October 2021, and postintervention is defined as November 2021–June 2022.
Uptake—Medications
Between February 2021 and June 2022, the EHR linked 298,152 overnight medication administrations to TID and q8h orders at the AMC (10.7%) and 75,998 at the CHA (11.1%) out of a total of 2,796,354 and 685,669 overnight medication administrations, respectively. Within the INU at the AMC, 4,640 overnight medication administrations were linked to TID and q8h orders out of 19,159 total medication administrations (24.2%). Furthermore, we examined overnight TID and q8 acetaminophen administrations, revealing 64,637 out of 117,657 at AMC (54.9%), 16,572 out of 40,922 at CHA (40.5%), and 536 out of 1,821 at the INU (29.4%). Similarly, we reviewed overnight heparin DVT prophylaxis administrations, finding 17,374 out of 19,685 at AMC (88.3%), 3,760 out of 4,638 at the CHA (81.1%), and 849 out of 854 (99.4%) at INU linked to TID and q8h orders.
Following the EHR modifications in heparin and acetaminophen choice architecture, coupled with education and Secure Chat reminders from pharmacists on q8h medications, there was a significant increase in sleep-friendly medication orders. The most considerable and immediate increase was noted for heparin across all locations, with mean (95% confidence interval [CI]) increases of 63% [40, 86], 72% [67, 78], and 89% [83, 94] for the INU, AMC, and CHA, respectively. Acetaminophen also showed immediate and significant increases, albeit generally smaller than heparin. Notably, the INU at the AMC experienced the highest percentage increase in acetaminophen TID selection with a mean (95% CI) increase of 34% [6, 74]. [Table 2] provides a detailed breakdown of the immediate effect and the change in slope following the interventions.
Table 2
An interrupted time series analysis to examine the percentage of sleep-friendly medication order frequencies selected
Order type
|
Location
|
Change in slope
|
95% CI
|
p-Value
|
Immediate effect %
|
95% CI
|
p-Value
|
All medications
|
AMC
|
−0.44
|
[−1.10, 0.18]
|
0.15
|
4.3
|
[1.2, 7.5]
|
0.01
|
CHA
|
0.68
|
[−0.11, 1.5]
|
0.08
|
5.6
|
[1.6, 9.6]
|
0.01
|
INU
|
0.57
|
[−3.6, 4.8]
|
0.77
|
−17
|
[−38, 4.3]
|
0.11
|
Acetaminophen
|
AMC
|
−0.27
|
[−1.6, 1.0]
|
0.66
|
8.1
|
[2.5, 14]
|
0.01
|
CHA
|
−0.83
|
[−3.4, 1.7]
|
0.50
|
8.3
|
[−3.0, 20]
|
0.13
|
INU
|
3.7
|
[−5.4, 13.0]
|
0.39
|
34
|
[−6.0, 74]
|
0.09
|
Heparin
|
AMC
|
−0.53
|
[−1.7, 0.65]
|
0.35
|
72
|
[67, 78]
|
<0.0001
|
CHA
|
−0.26
|
[−1.4, 8.6]
|
0.63
|
89
|
[83, 94]
|
<0.0001
|
INU
|
−3.1
|
[−7.7, 1.5]
|
0.16
|
63
|
[40, 86]
|
<0.0001
|
Abbreviations: AMC, academic medical center; CHA, community hospital affiliate; CI, confidence interval; INU, inpatient neurology unit; q8h, every 8 hours; TID, three times daily.
Note: The percentage of sleep-friendly medication orders was calculated as (TID)/(TID + q8h) × 100 for all medications, acetaminophen, and heparin. Nudges were implemented as follows: all medications: Secure Chat messaging, October 2021. Pre-period: February 2021–September 2021; post-period: November 2021–June 2022; acetaminophen: TID quick button added, August 2021. Pre-period: February 2021–July 2021; post-period: September 2021–June 2022; Heparin: default changed from q8h to TID, October 2021. Pre-period: February 2021–September 2021; post-period: November 2021–June 2022. The immediate effect is the mean change in the percentage of sleep-friendly medication order frequencies between the last preintervention month and the first postintervention month following wash-out.
Furthermore, the proportion of overnight medication administrations linked to TID and q8h orders significantly decreased in some locations. Heparin exhibited the largest immediate decrease in overnight administrations with mean (95% CI) decreases of 9.5% [5.3, 14.0], 9.8% [7.8, 12.0], and 18% [14.0, 23.0] for the INU, AMC, and CHA, respectively. For all medications ordered TID and q8h, the CHA had an immediate decrease in overnight administrations by 2.5% [1.3, 3.6]. [Table 3] provides a detailed breakdown of the immediate effect and change in slope.
Table 3
An interrupted time-series analysis to examine the percentage of overnight medication administrations after choice architecture modifications
Order type
|
Location
|
Change in slope
|
95% CI
|
p-Value
|
Immediate effect
|
95% CI
|
p-Value
|
All medications
|
AMC
|
0.11
|
[−0.095, 0.31]
|
0.27
|
−0.63
|
[−1.6, 0.38]
|
0.20
|
CHA
|
−0.44
|
[−0.67, −0.21]
|
0.0013
|
−2.5
|
[−3.6, −1.3]
|
0.0006
|
INU
|
−0.90
|
[−2.3, 0.51]
|
0.19
|
6.7
|
[−0.38, 14]
|
0.062
|
Acetaminophen
|
AMC
|
−0.096
|
[−0.43, 0.24]
|
0.54
|
−1.7
|
[−3.2, −0.27]
|
0.024
|
CHA
|
−0.53
|
[−1.0, −0.017]
|
0.044
|
−2.0
|
[−4.3, 0.21]
|
0.072
|
INU
|
−0.65
|
[−3.6, 2.3]
|
0.64
|
4.8
|
[−8.0, 18]
|
0.43
|
Heparin
|
AMC
|
−0.17
|
[−0.56, 0.22]
|
0.36
|
−9.8
|
[−12, −7.8]
|
<0.0001
|
CHA
|
−0.62
|
[−1.6, 0.34]
|
0.18
|
−18
|
[−23, −14]
|
<0.0001
|
INU
|
−1.0
|
[−1.8, −0.19]
|
0.020
|
−9.5
|
[−14, −5.3]
|
<0.0001
|
Abbreviations: AMC, academic medical center; CHA, community hospital affiliate; CI, confidence interval; INU, inpatient neurology unit; q8h, every 8 hours; TID, three times daily.
Note: Overnight was defined as 11 p.m.–7 a.m. The percentage of overnight medications was calculated as: (overnight TID + overnight q8h)/(total TID + total q8h) × 100. Nudges were implemented as follows: all medications: Secure Chat messaging, October 2021. Pre-period: February 2021–September 2021; post-period: November 2021–June 2022; acetaminophen: TID quick button added, August 2021. Pre-period: February 2021–July 2021, post-period: September 2021–June 2022; heparin: default changed from q8h to TID, October 2021. Pre-period: February 2021–September 2021; post-period: November 2021–June 2022. The immediate effect is the mean change in the percentage of sleep-friendly medication order frequencies between the last preintervention month and the first postintervention month following wash-out.
The observed increases in sleep-friendly medication orders and overnight medication administrations were sustained over time, as depicted in [Figs. 3] and [4].
Fig. 3 Trend (interrupted time series analysis) in overnight medication administrations linked to TID and q8h frequencies. % of administrations at night was calculated as (overnight TID + overnight q8h)/(total TID + total q8h) * 100. The dashed vertical lines represent the 1-month washout period, where the nudge was introduced in all areas. Acetaminophen order nudge began August 2021, heparin order nudge and secure messaging nudges for all medications began October 2021. AMC, academic medical center; CHA, community hospital affiliate; INU, inpatient neurology unit; q8h, every 8 hours; TID, three times daily.
Fig. 4 Trend (interrupted time series analysis) in sleep-friendly medication orders. % of sleep-friendly orders was calculated as TID orders/(TID orders + q8h orders) * 100. The dashed vertical lines represent the 1-month washout period, where the nudge was introduced in all areas.
Uptake—Neuro Checks and Vital Signs
After implementing choice architecture changes, along with education and Secure Chat reminders from resident champions in the INU, we observed a notable increase in sleep-friendly vital sign and neuro check orders in all areas. This is depicted in [Figs. 5] and [6]. The most significant immediate increase occurred in the INU. Sleep-friendly vital sign orders exhibited mean (95% CI) increases of 14% [0.5, 28], 6.7% [5.2, 8.2], and 4.3% [3.2, 5.5] for the INU, AMC, and CHA, respectively. Sleep-friendly neuro check orders also saw increases of 15% [0.2, 30], 8.1% [1.9, 14.0], and 1.2% [0.15, 2.6] for the INU, AMC, and CHA, respectively. See [Table 4] for the immediate effect and change in slope.
Table 4
An interrupted time series analysis to examine the percentage of sleep-friendly vital signs and neurological check order frequencies selected
Order type
|
Location
|
Change in slope
|
95% CI
|
p-Value
|
Immediate effect
|
95% CI
|
p-Value
|
Neurological checks
|
AMC
|
−0.69
|
[−1.9, 0.54]
|
0.25
|
8.1
|
[1.9, 14.0]
|
0.015
|
CHA
|
−0.022
|
[−0.30, 0.25]
|
0.87
|
1.2
|
[−0.15, 2.6]
|
0.075
|
INU
|
−1.4
|
[−4.4, 1.6]
|
0.32
|
15.0
|
[−0.2, 30.0]
|
0.053
|
Vital signs
|
AMC
|
0.0086
|
[−0.30, 0.31]
|
0.95
|
6.7
|
[5.2, 8.2]
|
<0.0001
|
CHA
|
−0.069
|
[−0.29, 0.15]
|
0.51
|
4.3
|
[3.2, 5.5]
|
<0.0001
|
INU
|
−0.26
|
[−3.0, 2.4]
|
0.84
|
14.0
|
[0.5, 28]
|
0.043
|
Abbreviations: AMC, academic medical center; CHA, community hospital affiliate; CI, confidence interval; INU, inpatient neurology unit; TIDWA, three times daily while awake; Q4WA, every 4 hours while awake.
Note: The percentage of sleep-friendly orders was calculated as (TID + BID + Q4HWA + TIDWA + daily)/(total orders) × 100 for neurological checks and vital signs. Raw numbers of all overnight neurological checks and vital signs documented in EHR flowsheets were analyzed pre- and postintervention. Nudges were implemented as follows: vital signs and neurological checks: TIDWA and Q4WA quick buttons were added, and the qshift quick button was removed in November 2021. Pre-period: March 2021–October 2021; post-period: December 2021–July 2022. The immediate effect is the mean change in the percentage of sleep-friendly medication order frequencies between the last preintervention month and the first postintervention month following wash-out.
Fig. 5 (A) Trend (slope) in overnight neurological checks performed. (B) Trend (slope) in sleep-friendly neuro check orders. % of sleep friendly orders is calculated as sleep friendly orders/total orders * 100%. The dashed vertical lines represent the 1-month washout period, where the nudge was introduced in all areas.
Fig. 6 (A) Trend (slope) in overnight vital signs performed. (B) Trend (slope) in sleep-friendly orders. % of sleep-friendly orders is calculated as sleep-friendly orders/total orders * 100%. The dashed vertical lines represent the 1-month washout period, where the nudge was introduced in all areas.
After quick button changes, the INU observed substantial reductions in every 4-hour frequencies for vital sign checks (−56%) and neuro checks (−50%), as well as qshift vital sign checks (−91%), and neuro checks (−92%). Similar but less pronounced decreases were observed in the AMC and CHA. Overall, TIDWA was preferred over q4HWA across all locations. Additionally, despite the decline in every 4-hour and qshift frequencies, every four hours remained the most ordered frequency.
At the same time, there were decreases in overnight vital signs and neuro checks performed across the organization (see [Figs. 5] and [6] for interrupted time series [ITS] analysis graphs and [Table 5] for immediate effect and change in slope). The ITS graphs for the INU show the most substantial increases in sleep-friendly neuro check orders and decreases in neuro checks performed from January through February 2022. The most substantial decreases in overnight medication administrations occurred from April through June 2022 (see [Table 5]).
Table 5
An interrupted time series analysis to examine the change in overnight neurological checks and vital signs performed after choice architecture modifications
Order type
|
Location
|
Change in slope
|
95% CI
|
p-Value
|
Immediate effect
|
95% CI
|
p-Value
|
Neuro checks
|
AMC
|
−814.1
|
[−2,467.0, 838.4]
|
0.30
|
−10,136.1
|
[−18,481.1, −1,791.0]
|
0.02
|
CHA
|
−330.5
|
[−980.1, 319.0]
|
0.29
|
763.9
|
[−2,516.15, 4,043.86]
|
0.62
|
INU
|
238.7
|
[−10.8, 488.2]
|
0.06
|
−705.0
|
[−1,964.8, 554.84]
|
0.25
|
Vital signs
|
AMC
|
−5,867.4
|
[−15,046.1, 3,311.4]
|
0.19
|
−6,407.0
|
[−52,757.4, 39,943.5]
|
0.77
|
CHA
|
−1,324.6
|
[−5,943.7, 3,294.5]
|
0.54
|
−3,293.3
|
[−26,618.5, 20,032.0]
|
0.76
|
INU
|
251.7
|
[18.04, 485.4]
|
0.04
|
−791.7
|
[−1,971.8, 388.5]
|
0.17
|
Abbreviations: AMC, academic medical center; CHA, community hospital affiliate; CI, confidence interval; INU, inpatient neurology unit; Q4WA, every 4 hours while awake; TIDWA, three times daily while awake.
Note: Overnight was defined as 11 p.m.–7 a.m. Raw numbers of all overnight neurological checks and vital signs documented in EHR flowsheets were analyzed pre- and postintervention. Nudges were implemented as follows: vital signs and neurological checks: TIDWA and Q4WA quick buttons were added, and the qshift quick button was removed in November 2021. Pre-period: March 2021–October 2021; post-period: December 2021–July 2022. The immediate effect is the mean change in the percentage of sleep-friendly medication order frequencies between the last preintervention month and the first postintervention month following wash-out.
The case-mix index was lower in the population with while awake vital sign orders (1.27) compared with around-the-clock vital sign orders (2.27) in the INU. Adverse events, including cardiopulmonary events, rapid responses, and cardiopulmonary arrests, remained stable across the AMC and the INU. From May 2021 to October 2021, there were 668 rapid responses and 117 cardiopulmonary arrests documented, compared with 674 rapid responses and 110 cardiopulmonary arrests from December 2021 to May 2022. The INU had 23 rapid responses preimplementation and 17 postimplementation.
Discussion
This study sought to address challenges associated with decision-making information and structure in CPOE systems, which can impact patient care and experience. We found that vital sign checks, neurological checks, and medication administrations frequently occurred overnight. Medications scheduled for q8h were more likely to be administered overnight than medications scheduled TIS. Heparin and acetaminophen were often administered overnight on the INU. Other studies identify vital signs, prophylactic heparin, and other noncritical medications as potential sleep disruptors.[16]
[18]
[19]
Nudges involve system changes designed to encourage desired behaviors.[4]
[20] We implemented nudges through EHR modifications, including altering defaults, introducing visible sleep-friendly options, and providing Secure Chat reminders. While the choice architecture modifications increased sleep-friendly orders across the organization, the impact was more profound in the INU, where additional bi-weekly messaging reminders and staff education occurred.
Default Frequency Settings
We changed the heparin order's default from an around-the-clock q8h frequency to a daytime-only TID frequency. This default adjustment likely accounted for the most significant, immediate, and sustainable improvement observed. Previous studies have shown how EHR default changes can influence workflows, such as increased posttransfusion hematocrit orders after changing the default from optional to preselected[4] and reduced telemetry orders after deactivating preselection in admission order sets.[5]
Reviews on the efficacy of nudges within EHRs consistently highlight the practice of guiding choices through default options as frequently employed and effective for changing clinician behavior.[6]
[21]
[22] The success hinges on the “active opt-out” model, where clinicians are more likely to accept defaults unless they actively reject them. Clinicians tend to trust that default selections are well-founded.[21] Consequently, ensuring the appropriateness of the heparin nudge across the organization was a crucial consideration before implementation. Similarly, we retained the preselected and defaulted-on status for vital signs and neuro checks at q4h to reduce the risk of guiding clinicians toward inappropriate care, recognizing the potential consequences of altering defaults in these contexts.
Changing Composition of Frequency Options
We made it easier for clinicians to choose daytime-only schedules for acetaminophen, vital signs, and neuro check orders by adding new quick buttons. Simultaneously, we removed the qshift quick button because the q8h schedule it generated on the nursing and patient care tech worklists was not visible to ordering clinicians, leading to confusion. Our changes successfully encouraged clinicians to choose options that are less likely to disrupt sleep, and led to a reduction in overnight interventions with an immediate decrease in overnight neuro checks and acetaminophen administrations across the AMC. Importantly, these modifications achieved their impact without altering the preselected settings within order entry and order sets.
These results align with prior research, such as the SIESTA study, which observed an increase in sleep-friendly orders for vital signs and heparin after increasing the visibility of sleep-friendly options in the EHR.[19] We enhanced the visibility of sleep-friendly schedules and linked those selections to nursing and patient care technician worklists. This decision was made because, before the EHR changes, clinicians typed “Do not Disturb” in the order comments section for vital signs and neuro checks. However, staff would still check regardless of the purely informational comments. We also integrated additional elements, such as Secure Chat reminders, acetaminophen orders, and neuro checks, and provide insights into variations across different health care settings. In contrast, the SIESTA study examined differences between two general medicine units in an AMC. Furthermore, our analysis included the percentage of overnight medications originating from specific orders and the frequency of total neuro/vital sign checks performed overnight, a dimension not extensively explored in the SIESTA study.
We observed that the introduction of sleep-friendly nudges, such as TIDWA and q4WA, led to significant increases in other pre-existing sleep-friendly orders, such as daily and twice daily (see [Supplementary Table S1] [available in the online version]). For example, at the AMC, orders for “daily” neuro checks increased by 183%, and “twice daily” vital sign orders increased by 47%. This suggests that the nudges not only reduced the frequency of interventions during sleep hours but also encouraged a shift to less frequent but more appropriately timed interventions, concentrating these activities during waking hours.
This redistribution of activities to less frequent intervals likely maintained or even reduced the overall workload rather than merely shifting it to non-sleep hours. By reducing the number of checks (e.g., from six times a day to once daily), the intervention may have streamlined care and optimized workflow efficiency during the day as well. However, it is important to note that this study did not analyze the specific impact on overall workload.
Visibility of Information and Reminders
In our efforts to nudge behavior, we made information more visible by adding schedules to the TIDWA and q4WA quick buttons. Previously, this information was hidden within collapsible order sections or nursing worklists. After examining patient care technician workflows on the INU, we ensured that vital sign order frequencies were visible on their patient lists. Although the impact of these visibility enhancements is unclear, future work should include usability testing and qualitative interviews before implementation. Other studies that have used this strategy of enhancing information visibility in the EHR have reported successful outcomes, such as displaying schedules on quick buttons for antibiotic orders, showing testing costs, and changing microbiology reporting.[6]
[16]
[23]
[24]
[25]
[26]
Complementing the visibility enhancements, we introduced reminders to review orders using Secure Chat messaging within the EHR. This innovative approach encouraged clinicians to review, modify, or discontinue their patients' orders if appropriate. When pharmacists recommended adjusting medication schedules, this nudge provides expert recommendations, or a green light, to the ordering clinicians.[2] While we did not conduct an analysis of messaging data, the INU, where this nudge was prioritized, experienced higher percentage increases in sleep-friendly orders. Additionally, covering providers gave feedback to re-time messages from the morning to the afternoon, when orders are most reviewed and cleaned up, demonstrating active engagement.
Uptake Differences and Local Workflows
Our study revealed notable differences in medication administration between different hospital settings. Following the intervention, the AMC had more sleep-friendly medication orders than the CHA, but the CHA had a more significant decrease in overnight medication administrations postintervention. These differences are likely from variations in default medication schedules. The TID schedule defaults to 0800 to 2200 at the CHA, whereas at the AMC, TID defaults to 0900 to 2300.
The varying uptake emphasizes the importance of understanding local workflows when designing nudges. One study effectively restructured clinical workflows by moving overnight vital sign checks and heparin injections to 6 a.m. and creating new order sets in the EHR to match these workflow changes.[18] The uptake of such interventions is intricately linked to the routines within each health care setting. Consequently, tailoring nudge designs to align with these variations becomes paramount for success.
To evaluate the impact of our interventions on workflow efficiency, we analyzed the number of manually entered medication time adjustments for heparin and acetaminophen orders in the EHR by nurses. By averaging the data over 3 months before and after the respective interventions, we found reductions in the need for these manual adjustments.
Before the intervention, the average number of manually adjusted heparin orders per month was 234, which decreased to 110 after the intervention, representing a 53% reduction. Similarly, the preintervention average for acetaminophen was 1,360 manually adjusted orders per month, which decreased to 1,159 postintervention, reflecting a 15% reduction. Reasons for medication time adjustments included promoting uninterrupted sleep, grouping care, and patient preference. These reasons did not change postintervention, suggesting that the reduction in time adjustments was primarily due to the improved alignment of order times with patient needs rather than a shift in the underlying reasons for making such adjustments.
This study emphasizes the need to refine nudging strategies in health care. Multiple nudging techniques can enhance patient-centered care. Introducing real-time digital nudges, like Secure Chat messaging, can engage clinicians. Further research should examine how different nudges can work synergistically. The INU trends show significant increases in sleep-friendly neuro check orders and a corresponding decrease in neuro checks performed from January through February 2022. There were also substantial reductions in overnight medication administrations from April through June 2022. In January, we started incorporating education for nonneurology residents rotating on the INU. In April, we started sending Secure Chat messages to all covering clinicians on the INU including internal medicine residents. These improvements occurred 6 months postintervention.
EHR designers and vendors should decrease display fragmentation on planned overnight interventions. For example, a visual timeline of future laboratories, nursing assessments, and medications can allow a more seamless understanding of a patient's upcoming night. CDS in CPOE systems can suggest schedule adjustments to bundle overnight interventions.
Furthermore, the study highlights the importance of considering workflows of different environments within a health care system. Future research directions may involve examining the impact of choice architecture modifications to optimize nudging strategies for diverse health care settings and determining the impact of such strategies on patient sleep outcomes and patient and staff perspectives.
Limitations
This study has several limitations. It was not a randomized controlled trial, so we cannot determine causality. The design does not allow us to determine if one intervention was more impactful than another. Other factors such as staffing shortages, decreased patient turnover postimplementation, and reduced room entries due to COVID-19 pandemic may have influenced overnight interventions. We used interrupted time-series analysis to assess the intervention's impact. Our analysis assumes no lagged effect but considers a washout period. The data may have seasonal components which we did not account for. Due to the close succession of the intervention start times (e.g., January and April), there was insufficient time between them to estimate the effect of each intervention separately. Consequently, our analysis focused on the overall period starting from the first intervention. This approach does not isolate individual impacts. Lastly, we did not collect patient-specific sleep outcomes where our modifications were implemented.
Conclusion
In conclusion, our study demonstrates the significance of tailored EHR modifications and nudging strategies in influencing clinician decision-making and patient care. The varied impact across nudge types and different settings underscores the importance of nudge design and considering local contexts. This study offers valuable insights for EHR designers, vendors, and clinicians to explore and refine nudging strategies to promote inpatient sleep. Future studies may build on this work by demonstrating the impact of such EHR modifications on patient sleep outcomes and the overall inpatient care experience.
Clinical Relevance Statement
Clinical Relevance Statement
This study's EHR modifications, informed by choice architecture principles, notably improved sleep-friendly ordering and reduced unnecessary overnight interventions. The study advocates for ongoing exploration of nudging strategies in health care settings, offering valuable insights for EHR designers, vendors, and clinicians aiming to optimize patient care and improve inpatient sleep quality.
Multiple-Choice Questions
Multiple-Choice Questions
-
What is the primary focus of the study's intervention to increase overnight sleep opportunities?
-
Modifying hospital staffing levels
-
Enhancing patient communication
-
Implementing changes in electronic health record (EHR) choice architecture
-
Introducing new medication administration protocols
Correct Answer: The correct answer is option c. The study's primary focus is on implementing comprehensive modifications to electronic health record (EHR) choice architecture to optimize the scheduling of overnight vital sign checks, neurological checks, and medication administrations.
-
How did the study enhance the visibility of information in the electronic health record (EHR) to influence clinician decision-making?
-
Introducing targeted messaging reminders
-
Adding timeframes on quick buttons
-
Adjusting staffing levels in the EHR system
-
Removing the qshift button
Correct Answer: The correct answer is option b. The study enhanced choice saliency by adding the timeframes next to the order frequency, providing a more seamless understanding of defaulted schedules.
-
What does the study emphasize regarding the uptake differences between the academic medical center and community hospital affiliate settings?
-
Similarity in nursing workflows
-
Irrelevance of local workflows in nudge interventions
-
Importance of tailoring interventions to local norms
-
Uniform impact of interventions across settings
Correct Answer: The correct answer is option c. The study highlights the significance of tailoring interventions to local norms, emphasizing the importance of understanding and adapting to variations in local workflows for successful implementation of nudge interventions.