Improving performance on night shift: a study of resident sleep strategies

Abstract

Study Objectives:

To identify sleep strategies of internal medicine residents transitioning to night shift and report their effect on performance.

Methods:

Residents logged hours of sleep and work starting 3 days prior to the first night shift and continuing through the next 8 days. Cohorts were defined by sleep logs and compared separately by transition strategy, total hours of sleep, amount of sleep occurring at work, weekend sleep schedule, and residency training year. Data from logs were entered into the Fatigue Avoidance Scheduling Tool to measure predicted Performance Effectiveness (PE) during each night shift.

Results:

Twenty-three residents were evaluated. The Sleep Banking transition strategy (n = 2) had higher PE (mean = 88.6%) than all other sleep strategies combined (n = 21, mean = 80.9%; P = .016). Additionally, residents who slept an average of 8–9 hours daily during their week of night shifts had a higher mean PE compared to those who slept < 6 hours (86.8% vs 78.6%; P = .014).

Conclusions:

Residents who engaged in Sleep Banking prior to the first night shift had higher PE and spent less time above a 0.05% blood alcohol concentration equivalent compared to all other strategies. Similarly, PE and time spent above a 0.05% blood alcohol concentration equivalent improved with increased average hours slept per day during the week of night shifts. Optimizing performance on night shift through the adoption of efficacious sleep strategies is imperative to mitigate patient safety issues that may result from poor alertness and cognitive abilities.

Citation:

Cushman P, Scheuller HS, Cushman J, Markert RJ. Improving performance on night shift: a study of resident sleep strategies. J Clin Sleep Med. 2023;19(5):935–940.

BRIEF SUMMARY

Current Knowledge/Study Rationale: Studies evaluating optimal sleep schedule strategies for transitioning to night shift have not been reported. Thus, there was a need to assess which strategies employed to mitigate the adverse effects of night shift are more effective or potentially maladaptive. This study analyzed the impact various sleep transition strategies had on performance.

Study Impact: Identifying effective methods for transitioning to night shift is a worthwhile endeavor that requires further investigation.

INTRODUCTION

Approximately 16% of wage and salary workers in the United States work an evening, night, or rotating shift.¹ Numerous health consequences have been linked to shift work, including accidents, type 2 diabetes, weight gain, coronary heart disease, stroke, and cancer.² These outcomes are likely related to the myriad of metabolic ramifications found to be associated with circadian rhythm disruption, including increased blood pressure and inflammatory markers, reversed cortisol rhythms, reduced heart rate variability, and decreased insulin sensitivity.³ In addition to these health consequences, shift work also has a negative impact on alertness, performance, and cognitive abilities.^4–7 Maltese et al⁵ evaluated 51 intensivists from 3 intensive care units and found that the cognitive abilities of the intensivists were significantly altered following a night shift in the intensive care unit. Harrison et al⁶ evaluated 31 emergency medicine residents over 4 weeks using sleep logs, actigraphy, real-time reported sleepiness, and performance on a vigilance task. They found that alertness and performance were lowest during night shifts. McHill and Wright⁷ evaluated 15 adults over a 6-day period of simulated night shift work and found that they had decreased performance on multiple standardized cognitive performance tests compared to day shift.

Despite the impact on health and performance, shift work is an unavoidable aspect of certain career fields. Workers in industries such as hospitality, leisure, transportation, utilities, and the wholesale and retail trades are more likely to work a nondaytime schedule than workers in other industries.¹ Likewise, most internal medicine residency programs use a night float rotation, on average comprising 2.4 months of training, though some programs have a notably longer time spent on nights.⁸ Thus, there is a need to assess which strategies employed to mitigate the adverse effects of night shift are more effective or potentially maladaptive.

Studies evaluating optimal sleep schedule strategies for transitioning to night shift have not been reported, though 1 study examined sleep schedules for days off between shifts. Petrov and colleagues⁹ surveyed 213 night-shift nurses regarding their sleep-wake patterns during their days off. The participants were grouped into 5 cohorts based on their reported sleep schedules on days off. Questionnaires were used to gather self-assessments of adaptation to night shift. Participants who chose to nap during their previously normal work hours (“Nap Proxy” group) and those who chose to stay awake for a 24-hour period when switching from nocturnal to diurnal sleep (“No Sleep” group) reported poorer adaptation to night shift work.

Rather than using self-assessments of adaptation, this study aimed to use a biomathematical fatigue and performance model to address the following knowledge gaps: (1) specific sleep strategies used by internal medicine residents transitioning to night shift and their effect on performance, (2) whether the use of caffeine or sleep aids had an impact on predicted performance, (3) whether sleeping at work had an impact on predicted performance, and (4) whether switching back to daytime wakefulness on days off impacts performance when returning to night shift.

METHODS

Study design

The study design was a prospective cohort study, without randomization. Over the course of 1 year, all internal medicine residents rotating at Miami Valley Hospital for a 2-week night shift were asked to participate in the study. The principal investigator recruited participants via email. A notification of the initiation of the study was sent to all internal medicine residents at the beginning of the study, and residents were reminded of the study 1 week before their 2-week night float blocks were scheduled to begin. Participants were asked to complete a log that documented hours of sleep and work starting 3 days prior to their first night shift and continuing through the next 8 days (total of 11 days of logs). The logs had no personal identifiers, and completed surveys were emailed to the principal investigator via a dedicated study email address.

The sleep logs were used to define 7 transition strategy cohorts: Sleep Banking (sleeping an average of ≥ 9 continuous hours in the nights preceding the first night shift), Early Transition (≥ 2 hours forward transition in sleep schedule), Day Nap (≥ 1 hour daytime nap prior to the first shift), No Sleep and Nap (staying awake the night prior to the first night shift and a day nap), Early Transition and Day Nap (combination of the 2 aforementioned strategies), and No Change (maintaining prior sleep schedule). Residents who employed certain combinations of strategies (eg, Sleep Banking with Early Transition) were grouped into the Other Combination cohort, as their numbers were too low to form individual cohorts but had schedules unique from the other cohorts.

Sleep logs were also used to define 5 cohorts based on mean hours of daily sleep prior to the first night shift: < 7 hours, 7–8 hours, 8–9 hours, 9–10 hours, and > 10 hours. Cohorts were similarly defined for mean hours of daily sleep during the first week of night shifts. Separately, 4 cohorts were defined based on number of hours of sleep occurring at work: no sleep, < 1 hour, 1–3 hours, > 3 hours.

Additionally, 3 cohorts were defined based on maintenance of sleep schedule on the weekend days: Diurnal Sleep (< 33% of sleep during nighttime hours), Nocturnal Sleep (> 50% of sleep during nighttime hours), and Partial Transition (33%–50% of sleep during nighttime hours). Separately, resident training year (PGY) and use of caffeine and sleep aids were also compared.

The data from logs were entered into the Fatigue Avoidance Scheduling Tool (FAST) software to measure predicted Performance Effectiveness during each night shift. The FAST software uses the Sleep, Activity, Fatigue, and Task Effectiveness Model (SAFTE) to evaluate fatigue impairment and performance.¹⁰ The SAFTE is a scientifically validated fatigue and performance model approved for use in multiple industries (eg, aviation, transportation) to accurately measure and predict the effect of fatigue on physical and cognitive performance.^10–12 SAFTE predicts changes in cognitive performance based on circadian rhythms in metabolic rate, cognitive performance recovery rates associated with sleep, cognitive performance decay rates associated with wakefulness, and cognitive performance effects associated with sleep inertia. Sleep inertia refers to the transitional state between sleep and wake marked by drowsiness and cognitive impairment. The primary output of the SAFTE model is Performance Effectiveness, expressed as a percentage, which is a summation of the homeostatic processes and the circadian processes over time, and includes transient adjustments for sleep inertia.¹²

Mean Performance Effectiveness during night shift work hours for the first week was calculated using the FAST software. The FAST software also provided the percentage of time on night shift spent below a 77% Performance Effectiveness threshold for the week. For comparative interpretation, the FAST software includes blood alcohol concentration (BAC) equivalence to Performance Effectiveness. The BAC equivalence is based on studies comparing the effects of sleep deprivation and alcohol on performance on a driving simulator¹³ and cognitive test performance.¹⁴ The percentage criterion of 77% (0.05 BAC equivalent) is chosen for multiple reasons. The current BAC limit in the United States is 0.05 for commercial drivers. The current BAC limit in the United States for noncommercial drivers is 0.08. The Centers for Disease Control and Prevention lists impaired judgment, lowered alertness, reduced response to emergency driving situations, and other adverse effects at 0.05.¹⁵ Dry and others¹⁶ found that significant cognitive decrements in speed of information processing, reductions in working memory, and increases in errors of commission are seen at 0.048 BAC. The National Transportation Safety Board has proposed lowering the legal limit to 0.05 in the United States since BAC levels as low as 0.05 have been associated with significantly increased risk of fatal crashes.¹⁷ More than 100 countries have already established BAC limits at or below 0.05.

Sample size

The Wright State University Internal Medicine Residency Program is a 3-year educational program with 25 residents in each class. Two residents (1 intern and 1 senior) are each on a 2-week period of night float for Miami Valley Hospital. For 1 academic year, a total of 25 night float blocks were included in this study, thus the planned sample size was the maximum possible for the study period: 50 participants.

Statistical analysis

The primary outcome was Performance Effectiveness, a continuous variable measured by the FAST program. The independent samples Mann-Whitney Test was used for comparisons involving 2 groups (sleep strategies), and the independent samples Kruskal-Wallis Test was used for comparisons involving 3 or more groups (hours of sleep at work, hours of sleep prior to the first night shift, hours of sleep during the first week of night shifts, sleep schedule on weekends, and PGY level). Inferences were made at the 0.05 level of significance with no need to correct for multiple comparisons. Our study used a study-wide alpha of 0.05 rather than an alpha of 0.05 for each statistical test. Analyses were conducted with IBM SPSS Statistics 25.0 (IBM, Armonk, NY).

RESULTS

Twenty-five of the 50 eligible residents completed logs. Two residents either incorrectly filled out the logs or were on alternate shift schedules, and their data were unusable. Therefore, 23 residents were evaluated. One of the 23 residents completed the log through the first week of night shift but excluded his/her weekend schedule. Consequently, this person was included in all cohort comparisons except the maintenance of sleep schedule on weekends. The mean predicted Performance Effectiveness (PE) for all residents was 81.6% (95% confidence interval = 79.7%–83.5%) during the first week of night shifts, with residents spending a mean of 34.7% (95% confidence interval = 28.0%–41.5%) of their shifts above the 0.05% BAC equivalent. Table 1 reports the results for all comparisons.

Table 1.

Comparisons for 23 residents in an internal medicine residency on sleep strategies and related characteristics.

Characteristic	Mean Predicted Performance Effectiveness	P	Percent > 0.05 BAC Equivalent	P
Transition strategy		.016*		.047*
Sleep Banking	88.58		13.45
Day Nap	83.78		21.60
Early Transition and Day Nap	81.98		37.18
Early Transition	80.30		42.28
No Sleep and Nap	79.41		45.35
No Change	78.99		35.75
Other Combination	81.58		26.53
Hours of sleep prior to the first night shift		.019*		.027*
Less than 7 hours	75.81		54.00
7–8 hours	80.20		40.30
8–9 hours	82.22		33.95
9–10 hours	85.63		17.90
More than 10 hours	82.46		35.65
Hours of sleep during the first week of night shift		.014*		.011*
Less than 6 hours	78.60		45.53
6–7 hours	82.14		34.00
7–8 hours	80.22		37.38
8–9 hours	86.81		16.08
Hours of sleep during night shift		.09†		.008†
One hour or more	83.41		24.69
Less than 1 hour	80.19		42.46
Caffeine use on night shift		.40†		.20†
Yes	81.18		37.17
No	82.22		30.94
Sleep aid use on night shift		.71†		.92†
Yes	81.14		35.27
No	81.75		34.55
Residency year–performance effectiveness		.26†		.06†
PGY1	80.56		40.93
PGY2/3	82.84		29.06
Weekend sleep schedule		.09*		–
Diurnal	89.11		–
Nocturnal	82.70		–
Partially shifted	81.02		–

^*Kruskal-Wallis Test. †Mann-Whitney Test. PGY = residency training year.

Transition strategy

The Sleep Banking cohort had a mean predicted PE of 88.58%, followed by Day Nap 83.78%, Early Transition and Day Nap 81.980%, Early Transition 80.30%, No Sleep and Nap 79.41%, and No Change 78.99%. Other Combination was 81.58%. Sleep Banking (n = 2, mean = 88.58%) was higher than all other sleep strategies combined (n = 21, mean = 81.01%) [P = .016].

Sleep Banking cohort participants spent a mean of 13.45% of their shifts above the 0.05% BAC equivalent threshold, followed by Day Nap 21.60%, No Change 35.75%, Early Transition and Nap 37.18%, Early Transition 42.28%, and No Sleep and Nap 45.35%. Other Combination was 26.53%. Sleep Banking (n = 2, mean = 13.45%) spent less time above the 0.05% BAC threshold than all other sleep strategies combined (n = 21, mean = 34.78%) [P = .047].

Hours of sleep prior to the first night shift

Participants who slept an average of 9–10 hours daily prior to their first shift had a mean predicted PE of 85.63%, followed by > 10 hours (82.46%), 8–9 hours (82.22%), 7–8 hours (80.20%), and < 7 hours (75.81%) [P = .019].

Participants who slept an average of 9–10 hours daily prior to their first shift spent a mean of 17.90% of their shifts above a 0.05% BAC equivalent threshold, followed by > 10 hours (35.65%), 8–9 hours (33.95%), 7–8 hours (40.30%), and < 7 hours (54%) [P = .027].

Hours of sleep during the first week of night shift

Participants who slept an average of 8–9 hours daily during their week of night shifts had a mean PE of 86.81%, followed by 6–7 hours (82.14%), 7–8 hours (80.22%), and < 6 hours (78.60%) [P = .014].

Participants who slept an average of 8–9 hours daily during their week of night shifts spent a mean of 16.08% of their shifts above a 0.05% BAC equivalent, followed by 6–7 hours (34.00%), 7–8 hours (37.38%), and < 6 hours (45.53%) [P = .011].

Caffeine use on night shift

Participants who used caffeine during night shifts (n = 14) had a mean PE of 81.18% compared to participants not using caffeine (n = 9), who had a mean PE of 82.22% [P = .40].

Sleep aid use on night shift

Participants who used sleep aids to adjust to their circadian rhythm to night shift (n = 6) had a mean PE of 81.14% compared to participants not utilizing sleep aids (n = 17), who had a mean PE of 81.75% [P = .71].

Hours of sleep at work during night shift

Participants who slept 1 or more hours on night shifts (n = 10) had a mean PE of 83.41%, whereas participants who slept less than 1 hour on night shifts (n = 13) had a mean PE of 80.19% [P = .09].

Weekend sleep schedule

Participants who maintained diurnal sleep on the weekend had a higher mean PE when returning to night shift compared to those who switched to nocturnal sleep and those who partially shifted (89.11% vs 82.70% vs 81.02%) [P = .09].

PGY1 vs senior residents (PGY2 and PGY3)

PGY1 residents (n = 11) had a mean PE of 80.56% during their first week of night shifts compared to senior residents (n = 12) who had a mean PE of 82.84% [P = .26]. PGY1 residents (n = 11) also spent a mean of 40.93% of their shifts above a 0.05 BAC equivalent, while senior residents (n = 12) spent a mean of 29.06% of their shifts above a 0.05 BAC equivalent [P = .06].

DISCUSSION

Night shift work is an unavoidable aspect of an internal medicine residency. Given the gravity of responsibility an internal medicine resident often has on night shift (eg, admitting patients, running codes, covering the intensive care unit), optimizing performance on night shift is imperative to mitigate patient safety issues that may result from poor alertness and reduced cognitive abilities. The primary goal of this study was to identify which sleep strategies employed by residents were most beneficial to performance.

In our primary outcome, we found that Sleep Banking had the highest Performance Effectiveness and less time spent above a 0.05% BAC equivalent threshold compared to all other strategies. This finding may be a product of total time slept, as Sleep Banking had an average daily sleep of 9.92 hours compared to an average of 8.28 hours for all other residents. Interestingly, this advantage appears to reach a zenith with the 9–10-hour group, as Performance Effectiveness and less time spent above a 0.05% BAC equivalent did not improve further for participants who slept > 10 hours. This could potentially have been due to multiple factors. First, only 2 participants slept a mean of > 10 hours daily prior to their first shift, which limits the power of this cohort. These 2 participants appear to have highly variable sleep in subsequent days after the first night shift, ranging from 4 to 12 hours per day, but it is difficult to draw conclusions on this given the limited power. Second, it is possible that this amount of sleep may signal a sleep-deprived state or an undiagnosed sleep disorder. Future studies may consider including medical and psychiatric comorbidities as a variable, or may consider evaluation for a previously undiagnosed sleep disorder. Additionally, we found that Performance Effectiveness and less time spent above a 0.05% BAC equivalent improved with more sleep during the first week of nights. No resident was able to sleep > 9 hours during the first week of night shifts (mean daily sleep for all residents was 6.78 hours), likely due to the time constraints of working 12-hour shifts. With over a third of total resident night shift time above 0.05% BAC, studies evaluating alternative shift duration (eg, 10-hour shifts or implementation of swing shifts) might determine if allowing residents more time for sleep improves performance.

When we turned out attention to secondary outcomes, we did not find any significant effects of caffeine consumption on Performance Effectiveness. Over half of residents consumed caffeine during their night shifts. We found no difference in Performance Effectiveness between residents who used caffeine and those who did not. Therefore, it did not appear that use of caffeine during night shift improved performance or interfered with residents’ sleep schedules. It is important to note that the Performance Effectiveness is predicted in this model, and it is possible that actual performance on cognitive testing may improve after consuming caffeine. Additionally, residents who utilized sleep aids to adjust their sleep schedule did not have greater Performance Effectiveness than residents who did not utilize sleep aids. Type of sleep aid was not specified.

When looking at the relationship between sleep at work during night shift and Performance Effectiveness, we found that the 2 were positively related. Residents who were able to sleep 1 or more hours during night shifts had a nearly significantly higher mean Performance Effectiveness than residents who slept less than an hour.

Our final secondary objective was to evaluate whether switching back to daytime wakefulness on days off impacted performance when returning to night shift. Our results indicated that maintaining a diurnal sleep schedule on the weekend did not lead to a difference in Performance Effectiveness when returning to the first night shift after the weekend. Since logs ended on the first night shift after the weekend, we could not compare PE for the remainder of the second week of night shifts. It is possible that a change in sleep schedule over the weekend may have a notable effect in subsequent night shifts, as was seen in the Petrov et al⁹ study.

Comparing PGY1 and senior residents, we saw no difference in Performance Effectiveness during the first week of night shift between the 2. However, interns spent a marginally significantly greater percent of shift time above 0.05 BAC equivalent compared to senior residents. This difference could be related to senior residents having mean sleep of 2.6 hours during shifts compared to approximately one-half hour for interns.

Our study had limitations. First, the study was conducted at a single internal medicine residency program. Consequently, generalizability to other residency programs in internal medicine or residencies in other specialties should be done with caution. Second, the total and subgroup sample sizes were small. Thus, estimates were not as precise as with larger sample sizes. Third, small sample sizes resulted in what were perhaps meaningful differences between groups not being statistically significant. For example, the difference in night shift time above 0.05 BAC equivalent between PGY1 residents and senior residents was 41% vs 29%. This 12% difference may represent a substantive difference in practical terms, but one that we cannot claim was not due to chance. Furthermore, the combination of sample size and wide range of transition strategies impacted cohort size (range of 2–8 residents for 6 cohorts). There were also 3 residents with other unique strategies that did not clearly fall into any of the 6 cohorts, and thus were grouped as Other Combination. Future studies with a greater number of residents would result in the benefits of larger cohorts, more precise estimates, and increased statistical power.

In order to limit the impact on resident’s time, there were some other variables that were considered but ultimately not assessed in this study, including alcohol consumption and physical activity. Both of these should be considered in future studies as they could also have a significant impact on performance, including influence on sleep disruption or unrefreshing sleep. It is also important to note that individual differences (eg, environment, dependents at home, physiologic and psychologic differences) likely influence the strategy chosen by residents. These factors were not controlled in this study in order to limit the impact on residents’ home life, but future studies would likely benefit from randomization in a controlled environment.

In conclusion, this study suggests that increasing hours of sleep per night prior to the start of a night shift block and sleeping greater than 8 hours per day during the week of night shift leads to better predicted performance at work. Sleep Banking appears to be the most promising transition strategy. Identifying effective methods for transitioning to night shift is a worthwhile endeavor that requires further investigation to optimize physician performance and, subsequently, improve patient outcomes. We hope that the results of this study guide future studies on sleep strategies for performance optimization.

DISCLOSURE STATEMENT

All authors have seen and approve of this manuscript. The results of this study were presented at the 2022 American College of Physicians Internal Medicine Meeting. The authors report no conflicts of interest.

ABBREVIATIONS

BAC

blood alcohol concentration

FAST

Fatigue Avoidance Scheduling Tool

PE

performance effectiveness

PGY

residency training year

SAFTE

Sleep, Activity, Fatigue, and Task Effectiveness