Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2025 Aug 16.
Published in final edited form as: J Sleep Res. 2025 Mar 13;34(6):e70041. doi: 10.1111/jsr.70041

The Impact of Dynamic Lighting on Sleep Timing and Duration for Hospitalised Patients

Andrew S Dunn 1, Barbara Rabin Fastman 2, Alan Weinberg 2, Lindsay Condrat 3, Allison Fraser 3, Rabia Khan 4, Marjorie P Zambrano Loor 4, Geetanjali Rajda 1, Octavio L Perez 4, Ayham Adawi 5, Korey Kam 5, Ankit Parekh 5, Andrew W Varga 5, Richard L Vincent 4
PMCID: PMC12353654  NIHMSID: NIHMS2066838  PMID: 40083068

Abstract

Poor sleep is common in hospitalised patients due to multiple factors, including disruption of the circadian rhythm. Few studies have examined programmable artificial lighting systems in hospital patient rooms, and few have achieved meaningful improvement in sleep. We sought to determine how novel dynamic lighting affects sleep timing and duration compared to standard hospital lighting. Patients were admitted to rooms on a cardiology unit with customised intervention or standard lighting. The lighting system delivered blue-enriched light during the day, a melanopic stimulus twice daily and blue-depleted light in the evening. Sleep/wake probability was measured in 30-s epochs using mattress sensors to capture sleep timing and nocturnal sleep duration. Subjective sleep duration and alertness were assessed with sleep diaries and the Karolinska Sleepiness Scale (KSS), respectively. A total of 87 patients were enrolled. Subjects experiencing customised lighting demonstrated significantly advanced rest/wake activity phase by 160 min and overall greater sleep probability. Overnight sleep duration (11 p.m.–7 a.m.) was 66 min greater in the lighting condition (266 vs. 200 min, p < 0.05). Patients in the intervention group reported higher levels of alertness during the morning (KSS score 3.8 vs. 4.9, p = 0.01) and evening (5.4 vs. 7.1, p = 0.01). A lighting system programmed to entrain the circadian rhythm and provide a daytime melanopic stimulus on a hospital unit was associated with advanced circadian phase, increased nocturnal sleep duration and increased perceived morning and evening alertness. These results suggest that dynamic lighting systems have the potential to improve sleep for hospitalised patients.

Keywords: circadian rhythm, dynamic lighting, Karolinska sleepiness scale, mattress sensor, sleep duration

1 |. Introduction

Hospitalisation is associated with disrupted sleep for myriad reasons, including noise, being awoken for vital signs or laboratory draws, toileting and lighting (Dobing et al. 2016; Wesselius et al. 2018; Khalid et al. 2022), leading to a decrease in overnight sleep duration of approximately 1–2 h compared to the non-hospital setting (Dobing et al. 2016; Wesselius et al. 2018; Arora et al. 2011). Poor sleep for hospitalised patients has been associated with decreased melatonin and cortisol levels, increased blood pressure and decreased physical activity (Arora et al. 2011; Kuzmik et al. 2022). The impact of impaired sleep may be more serious for patients at risk of complications. For example, decreased rapid eye movement (REM) sleep has been associated with an increased risk of hospital-acquired delirium for patients admitted to the intensive care unit (Sun et al. 2021; Wilcox et al. 2024).

Sleep in the hospital is likely misaligned with the subjective night due to a combination of insufficient light exposure during the day and excessive light exposure at night (Bano et al. 2014). Greater exposure to natural light for hospitalised patients has been associated with improved outcomes, including increased survival in acute myocardial infarction and decreased length of stay (LOS) for patients admitted with bipolar disorder (Benedetti et al. 2001; Beauchemin and Hays 1998), as well as other hospitalised patients with general medical conditions (Park et al. 2018; Jafarifiroozabadi et al. 2023). Though the disruption of sleep is routine and causes have been identified, many hospitals have not taken steps to mitigate the sleep impairment. Lighting has been proposed as a key environmental factor in the systems engineering of the hospital ecosystem and can be harnessed to improve circadian rhythmicity and potentially better align sleep with the subjective night in hospitalised patients (Hughes and Clancy 2005; Ulrich et al. 2008). A study of hospitalised patients comparing standard lighting with lighting programmed to provide gradual changes in luminance across the day and 2 h of bright blue-enriched light late in the morning found that the study lighting system was associated with an increase in sleep of 6 min per day (Gimenez et al. 2017). Relatedly, combining exposure to natural daylight with glasses that either emitted or blocked short-wavelength light in the morning and evening respectively resulted in circadian phase advancement in hospitalised patients (Mangini et al. 2024). Factors that could influence the degree of circadian phase change and modulate sleep in the hospital include the timing, intensity, and spectra of light, including the inclusion of a melanopic stimulus.

We developed a dynamic lighting system that delivers a programmable spectrum and intensity of light designed to better entrain the circadian rhythm for hospitalised patients. We installed the lighting system in a two-bedded hospital room and compared objective sleep timing and duration, as measured by mattress sensors; subjective sleep duration with a sleep diary; subjective sleepiness with the Karolinska sleepiness score (KSS); and subjective lighting perception and preference with surveys, with a concurrent control group in an adjacent room with standard hospital lighting. We additionally evaluated LOS between conditions. Our primary hypothesis was that the customised lighting intervention would entrain rest/activity rhythms to the subjective night, thereby increasing nocturnal sleep and decreasing levels of subjective sleepiness.

2 |. Materials and Methods

2.1 |. Participants

The study was conducted at a 1141 bed tertiary care academic teaching hospital in New York City. From June 7, 2021, to June 30, 2022, patients were enrolled from the lighting intervention and control rooms on the hospital’s general cardiology unit, which admits patients who require telemetry monitoring for acute cardiac conditions. Patients who met criteria for admission to the cardiology unit were assigned to the intervention or control room through standard hospital procedures based on bed availability, rather than on clinical features. Patient comorbidities were gleaned from the electronic medical record.

Research coordinators reviewed the list of patients admitted to the study rooms daily and obtained patient consent for the study. Exclusion criteria included inability to speak English or Spanish, due to limitations in translation services, and patients admitted and discharged over the same weekend were not approached due to the lack of availability of a research coordinator. Patients in the intervention and control lighting rooms were asked to consent to use recordings of their sleep data from mattress sensors (Moyen et al. 2024) (Eight Sleep, New York, NY), a daily sleep diary, assessments of subjective sleepiness using the nine-point Karolinska sleepiness scale twice daily, and surveys of lighting perception and preference. Patients were not told about the study hypothesis. Patients could decline participation or consent to one or more of the study components. A total of 110 patients were approached, and 87 consented to one or more study procedures.

Patients were not asked to consent to being admitted to the intervention lighting room as the spectrum of light across the programme falls within the range of hospital standards. A patient education card with a graphic describing the lighting programme was provided to patients admitted to the intervention room, which states, “This hospital room uses advanced lighting to provide you with a natural light experience. This room provides automatic changes in lighting that look like natural daylight. The lighting begins at 7 a.m. and ends at 11 p.m. Please check with the nursing staff if you have a question or request about the lighting.” Data collection from the electronic medical record included patient demographics, comorbidities, home medications, hospital orders for and administration of as-needed medications for sleep, LOS on the intervention unit, total hospital LOS, performance of any procedures and escalation of care to an intensive care unit. This study was approved by the Institutional Review Board of the Icahn School of Medicine at Mount Sinai protocol number (19–00376).

2.2 |. Surveys and Questionnaires

The lighting perception/preference surveys, sleep diary, and KSS surveys were offered to patients daily on weekdays by a research coordinator to gather patients’ perception of sleep quality (Akerstedt and Gillberg 1990). Participants were approached for the KSS typically between 9 a.m. and 10 a.m. At that time, they were provided 2 questionnaires regarding sleepiness, the first worded “Please choose the answer choice that best matches how sleepy or awake you felt this morning,” and the second worded, “Please choose the answer choice that best matches how sleepy or awake you felt last night.” The surveys and diary were not completed on days patients were not available, such as being off the unit for a test, or on weekends. We adapted a survey on the perception of lighting from a previously published study and administered the survey to patients in the study rooms (Topf 1985; Eklund and Boyce 1996). A modified version of this survey was administered to nursing staff that contained questions regarding the impact of the custom lighting on work/tasks the nurses do in the patient rooms.

2.3 |. Customised Lighting

The study rooms had two beds each, were adjacent, and were in an inner tower with windows facing a hospital atrium. In the lighting intervention room, we developed a novel multichannel integrative lighting system that delivers controlled light spectral power distributions (SPD) across the visual and non-visual spectra. The custom lighting system was developed by modifying commercial 2′ × 2′ ceiling light fixtures (Simon, Barcelona, Spain) with six (6) channels of LEDs (Seoul SunLike 6500, 4000, 2700 K and PC-Amber, Korea and 480 and 505 nm Philips Lumileds, Schiphol, The Netherlands). These were installed in the tile spacing used for standard hospital lighting over each bed in the lighting intervention room. Additionally, wall lights (Archer, Acuity Brands, Conyers, GA) above each patient bed-head were re-engineered to provide programmed uplight and downlight/reading (four channels, two up and two down/reading from Seoul SunLike 6500, 2700 K and PC-Amber, Korea) (Figure 1). The bathroom light (Mercury Lighting L65, Fairfield, NJ) was modified from fluorescent to three (3) LED channels to follow the programmed circadian lighting present in the room. Colour temperature across the period of programmed lighting changed in the following pattern: 7:00–7:15 a.m.: PC-Amber (1732 K), 7:15–8 a.m.: 2700 K, 8–10 a.m.: 4000 K, 10–10:30 a.m.: 6500 K, 10:30 a.m.–12:30 p.m.: 6500 K + (cold white LED enhanced with “bluish” light (480 and 505 nm)), 12:30–1 p.m.: 4000 K, 1–2 p.m.: 2700 K, 2–3 p.m.: 6500 K+, 3–5:30 p.m.: 4000 K, 5:30–10:30 p.m.: 2700 K, 10:30–11 p.m.: 2200 K. Light was programmed across these lighting fixtures (Casambi, Espoo, Finland) to deliver a changing spectrum designed to entrain the circadian rhythm.

FIGURE 1 |.

FIGURE 1 |

Open in a new tab

Example of the customised lighting in the lighting intervention room during a time of high-melanopic light stimulus (6500 K) (A) and during a time of low melanopic light stimulus (2700 K) (B).

Programmed lighting began at 7:00 a.m. The daily cycle included two periods (10:30 a.m. to 1:00 p.m. and 2:00 p.m. to 5:30 p.m. for a total of 6 h) of blue-enriched light delivered for maximum melanopic stimulus, based on the melanopic-equivalent daylight illuminance (m-EDI) as defined by the International Commission on Illumination (CIE S 026/E:2018) (Figure 2). Levels of blue light were depleted from 6:00 p.m. to 10:00 p.m. Starting at 11:00 p.m., there was no programmed light in the room until the morning cycle began at 7:00 a.m. The bathroom night lighting delivered a blue-depleted phosphor-converted amber-based lighting. During the daytime, the bathroom light SPD tracked the programmed circadian lighting present in the room. Intervention light cycled automatically. The standard wall switch provided the intervention lighting when in the “off” position and turned on standard lighting in the “on” position. This option for standard lighting was provided in the event full light was needed (e.g., during venipuncture) whilst intervention lighting was cycled to low intensity. The patient or a staff member could temporarily turn off all lighting through the use of a specialised switch mounted near the bed. If turned off, the customised lighting automatically returned at the next programmed interval. SPD measurements were taken to calculate the delivered light in lux for the visual component and m-EDI lux for the non-visual component using a calibrated Konica-Minolta CL-500a spectrometer (Konica Minolta, Ramsey, NJ) at the position of the patient’s head in an assumed horizontal position. Example SPD measurements during the duration of customised lighting between 7:00 a.m. and 11:00 p.m. are shown in Figure 2. The control room utilised standard hospital lighting (LED ceiling panels and fluorescent wall lights and bathroom lighting), consisting of white light with a fixed spectrum at 4000 K that was always on between 7 a.m. and 11 p.m.

FIGURE 2 |.

FIGURE 2 |

Open in a new tab

Control vs. intervention lighting illuminances (photopic & melanopic): Photopic (lux) and Melanopic (melanopic equivalent daylight illuminance (m-EDI) (lux) illuminance during the 24-h cycle in both the intervention lighting condition (“lighting”) and the control lighting condition (“control”).”

2.4 |. Mattress Sensor Data Collection

Mattress sensors from the company Eight Sleep were pre-installed in the 4 study beds. Vigilance states in hospitalised patients were estimated using mattress sensors, which used a proprietary combination of movement, heart rate, respiratory rate and temperature to output sleep or wake vigilance states in 30-s epochs. The resulting hypnograms were extracted using the mattress application programming interface (API) and converted to binary 0/1 to denote the sleep or wake state. In addition to binary sleep and wake outputs, the sensor could output “no data,” which occurred primarily when the patient was not in the bed or the sensor hub was not powered or connected to the internet. Epochs of “no data” are referred to as “indeterminate time” in the Results below. A total of 60 participants (30 in the control condition and 30 in the lighting condition) had at least some measurable mattress sensor data.

2.5 |. Sleep Assessments

Wake probability across the 24-h cycle was assessed per individual by summating all instances of wake in any 30-s epoch and dividing by all instances of sleep or wake within that specific epoch across all days during the subject’s full LOS within the designated hospital room, ignoring epochs in which data were absent. Thus, for each 30 s epoch, probability values ranged from 0 to 1, with a value of 1 denoting that the subject was awake for that particular epoch across all days of their stay. Using cosinor analysis (R version 3.6.3 with the “cosinor” package), wake probability was fitted with a sinusoid (Y ~ sin(X) + cos(X)) per subject and the following parameters were calculated: (a) MESOR (mean of the fitted sinusoid), (b) BATHYPHASE (trough of the sinusoid representing the time at which the maximum probability of sleep occurred or, minimum probability of wake). Wake probability data at the group level were also fitted with a sinusoid.

We additionally evaluated total overnight sleep time, defined as sleep between the hours of 11:00 p.m. and 7:00 a.m., averaged per subject across all nights of stay. Individual patient nights in which greater than 4 h of mattress sensor data were absent were excluded from the analysis (there were 4 subjects in the control condition and 0 subjects in the lighting condition for whom all nights during the entire LOS were excluded based on this criterion, resulting in 26 control subjects and 30 custom lighting subjects). Sleep duration between the hours of 11:00 p.m. and 7:00 a.m. was evaluated in two ways. First, total raw minutes of sleep were summated and compared between conditions. Second, because the absolute number of minutes of sleep may be biassed by differing amounts of missing data per night, we report the number of minutes of sleep per hour of recorded data.

Of note, sleep/wake as estimated by the mattress sensor was not adjudicated/curated by sleep diary information because this could not be done in any consistent manner. Sleep diaries were never collected during weekends (even though mattress data may have been collected), and even during the weekdays, hospitalised patients showed variable consistency in completing sleep diaries.

2.6 |. Data Analysis

Chi-squared tests were used to compare proportions of demographic and health comorbidity data between conditions. The student t-test for continuous variables and Fisher’s exact test for discrete data were utilised to determine whether the treatment and control arms were comparable. Lighting perception/preference surveys utilised a Likert scale. The nonparametric Wilcoxon Rank Sum test was employed for the comparison between two groups for the surveys.

All sleep timing and duration variables, including the mean difference in sinusoid-estimated BATHYPHASE and MESOR, and the mean difference in nocturnal duration of sleep, wake, indeterminate time and sleep duration normalised to hour of measured sleep/wake for subjects in the lighting group and the control group were evaluated and compared using the Wilcoxon rank-sum test. Evaluation of average nocturnal duration of sleep as a function of lighting condition, diabetes mellitus type 2 (DM2) status, chronic kidney disease (CKD) status and length of lighting exposure was conducted using a general linear model (the glm_model function in R version 3.6.3) according to the model: average nocturnal sleep duration ~ CKD + DM2 + length of lighting exposure + lighting condition. Length of lighting exposure represents the number of nights up to and including the last night of hospitalisation admission in which a “valid” sleep recording occurred, where “valid” is defined as > 50% of the overnight period (e.g., > 4 h) with an actual sleep or wake output (not indeterminate).

The number of KSS surveys and sleep diaries completed per patient varied, depending on the overall length of admission and the number of surveys completed per admission. Reported average KSS values reflect the last morning and evening KSS survey completed by each patient who completed one or more of these instruments. The last response was chosen to maximise the impact of lighting exposure on the patient experience. We also report change in KSS scores across mornings and evenings (last KSS score—first KSS score) in those individuals who completed more than one morning or evening KSS. For sleep diary data, we report average self-reported sleep time per 24 h, summating both nocturnal sleep and any daytime napping.

3 |. Results

3.1 |. Demographics

A total of 110 patients were approached, and 87 consented to enrollment, including 42 in the intervention group and 45 in the control group. Of the patients who consented for one or more of the study components, 84 consented to the use of the mattress data, 73 consented to the lighting perception/preference surveys, and 57 to the sleep diary. Data were collected for 60 patients who consented to the mattress sensor, 63 patients who consented to the lighting perception/preference survey, and 50 patients for the sleep diary. Patient characteristics are shown in Table 1. The mean age was 66.3 years, and 50.6% were female. A total of 16 patients had been taking an as-needed sleep medication at home (8 participants in the lighting intervention group and 8 participants in the control group). Overall, the two groups were similar for most baseline variables, with the exception that more patients in the intervention arm had diabetes (59.5% vs. 31.1%, p = 0.01) and chronic kidney disease (47.6% vs. 24.4%, p = 0.02). There were no significant differences in the prevalence of sleep apnea or insomnia diagnoses between conditions. The number of days of sleep data collection via the mattress sensors varied from 3 days to 23 days. Each individual subject included in the analysis in either condition contributed at least 1 day of validly collected mattress sensor data.

TABLE 1 |.

Patient characteristics.

Control group (n = 45) Intervention group (n = 42) p
Race/ethnicity, n (%)
 White 20 (44.4) 15 (35.7) 0.41
 Black or African American 11 (24.4) 14 (33.3) 0.36
 Asian 1 (2.2) 0 (0.00) 1.0
 Native American 1 (2.2) 0 (0.00) 1.0
 Hispanic 12 (26.7) 11 (27.5) 0.93
 Other identity 13 (28.9) 14 (33.3) 0.65
Age, years ±SD 65.3 ± 16.4 67.2 ± 13.8 0.56
Female, n (%) 26 (57.8) 18 (42.9) 0.16
Comorbidities, n (%)
 Diabetes 14 (31.1) 25 (59.5) 0.01*
 Hypertension 35 (77.8) 31 (73.8) 0.67
 Coronary artery disease 28 (62.2) 24 (57.1) 0.63
 Congestive heart failure 24 (53.3) 24, (57.1) 0.72
 Cardiac arrhythmia 25 (55.6) 17 (40.5) 0.16
 Valvular heart disease 21 (46.7) 13, (31.0) 0.13
 Chronic kidney disease 11 (24.4) 20 (47.6) 0.02*
 COPD 5 (11.1) 6 (14.3) 0.66
 Asthma 9 (20.0) 7 (17.1) 0.73
 Prior cancer 4 (8.9) 7 (16.7) 0.28
 Current cancer 2 (4.4) 1 (2.4) 0.60
 Depression 7 (15.6) 7 (16.7) 0.89
 Anxiety 4 (8.9) 3 (7.1) 0.76
 HIV 1 (2.2) 1 (2.4) 0.96
 Arthritis 14 (31.1) 14 (33.3) 0.82
 OSA 6 (13.3) 7 (16.7) 0.66
 Insomnia 1 (2.2) 3 (7.1) 0.35
 Home sedative/hypnotic medication 8 (17.8) 8 (19.0) 0.88
Open in a new tab

3.2 |. Sleep/Wake Timing and Duration

The wake probability distribution double-plotted across two 24-h cycles for all participants in the lighting and control conditions is shown in Figure 3. Goodness of fit of the cosinor at the group level (adjusted R2) was 0.34 for the lighting condition (p < 0.001) and 0.28 for the control condition (p < 0.001). The bathyphase (trough) of the sinusoid (reflecting minimum wake probability/maximum sleep probability) occurred at 6:32 a.m. in the lighting condition and 9:10 a.m. in the control condition (mean bathyphase difference between control and lighting group: 160 ± 18.1 min, p < 0.01, Z = 2.64). Furthermore, we observed a significant difference in the average mesor value of each participant’s sinusoid, reflecting overall reduced wake probability (increased sleep probability) in the lighting condition (lighting mesor 0.52 ± 0.11, control mesor 0.58 ± 0.06, p < 0.01, Z = 2.21). Wake probability data at the group level were also fitted with a sinusoid yielding similar mesor values shown in Figure 3 (lighting mesor 0.53, control mesor 0.61).

FIGURE 3 |.

FIGURE 3 |

Open in a new tab

Wake probability distribution double-plotted across two 24-h cycles for all participants in the lighting (red) and control (blue) conditions. The bathyphase (trough) of the sinusoid (reflecting minimum wake probability/maximum sleep probability) occurred 160 min earlier in the lighting condition. The average mesor value for the lighting condition sinusoid was significantly lower than the control condition sinusoid, reflecting overall reduced wake probability (increased sleep probability) in the lighting condition.

In assessing nocturnal sleep duration, we observed that total overnight sleep time based on data from the mattress sensors during the 11:00 p.m.—7:00 a.m. period was 66 min greater for patients in the lighting group than the control group (266 ± 24 min vs. 200 ± 21 min, p = 0.035), without significant differences in the duration of wake time or indeterminate time during the same interval (Figure 4), though nocturnal wake time trended towards lower values in the lighting condition (147 ± 16 min vs. 192 ± 13 min, p = 0.053). To account for potential differences in the duration of indeterminate output during the 11:00 p.m.—7:00 a.m. period, we additionally quantified the number of minutes of sleep during this nocturnal period normalised to the total number of minutes of binary sleep or wake output (i.e., excluding epochs of indeterminate time). We observed significantly more sleep per hour during the nocturnal period in the lighting condition than in the control condition (36.9 ± 2.6 min sleep per hour vs. 28.4 ± 2.4 min per hour, p = 0.022) (Figure 4). Finally, in order to evaluate the potential maximal effect of the custom lighting condition vs. the control condition, we evaluated sleep, wake and indeterminate time durations during the single overnight period (11 p.m. to 7 a.m.) representing the last night of hospitalisation admission in which a “valid” recording occurred, where “valid” is defined as > 50% of the overnight period (e.g., > 4 h) with an actual sleep or wake output (not indeterminate). Using such an approach, we observed that the custom lighting condition was associated with a significant decrease in overnight wake time vs. the control condition (145 ± 36 min vs. 200 ± 35 min, p = 0.03) without significant differences in sleep duration (265 ± 55 min vs. 206 ± 53 min, p = 0.13) or indeterminate time duration (71 ± 34 min vs. 74 + 33 min, p = 0.87). The night representing this calculation did not differ significantly between conditions (5.8 ± 0.8 nights in the lighting condition vs. 4.5 ± 0.5 nights in the control condition, p = 0.19). To further assess average nocturnal sleep duration as a function of a potential contribution of duration of lighting exposure whilst also assessing for contributions of DM2 and CKD (the 2 medical disorders in which significant differences were observed between lighting conditions), we ran a general linear model defined by: average nocturnal sleep duration ~ CKD + DM2 + length of lighting exposure + lighting condition. We observed a significant effect of lighting condition on nocturnal sleep duration after controlling for CKD, DM2 and length of lighting exposure (p = 0.043) (Table 2). Differences in sleep duration were unlikely to be influenced by sedative medication use as we observed no statistically significant difference in the use of as-needed medications for sleep (1.8 doses per 10 patient days in the lighting condition vs. 1.3 doses per 10 patient days in the control condition, p = 0.11). Subjective sleep estimates from patient sleep diaries were not different between conditions (6.4 ± 1.9 h in the lighting condition vs. 6.4 ± 1.6 h in the control condition, p = 0.99). Mean LOS was not different between conditions (7.4 ± 7.8 days in the lighting condition vs. 6.7 ± 7.9 days in the control condition, p = 0.68).

FIGURE 4 |.

FIGURE 4 |

Open in a new tab

Absolute nocturnal sleep duration (11 p.m. to 7 a.m.) was significantly increased in the lighting condition (orange) in comparison to the control condition (blue) (A), whilst absolute wake duration trended towards a decrease (p = 0.053) (B). There was no significant difference in the duration of indeterminate time (C) as measured with the mattress sensor. Normalised sleep duration (minutes of sleep per hour of measured sleep or wake) was also significantly increased in the lighting condition vs. the control condition (D). *p < 0.05. n.s. = not significant.

TABLE 2 |.

General linear model evaluating the relationship between lighting condition, CKD, DM2, and length of lighting exposure on average nocturnal sleep duration.

Coefficients Estimate Standard error t value p
Intercept 197.6 33.7 5.9 < 0.001
Lighting condition 77.2 37.2 2.1 0.043*
CKD 7.6 36.6 0.21 0.84
DM2 −41.1 39.3 −1.0 0.30
Length of lighting exposure 2.5 4.8 0.52 0.60
Open in a new tab

3.3 |. Subjective Alertness

Patients in the intervention group reported higher levels of alertness (i.e., lower KSS) during the morning (KSS 3.8 vs. 4.9, p = 0.01) and evening (5.4 vs. 7.1, p = 0.01). The change in either the morning or the evening KSS was assessed for patients who completed more than one morning or evening KSS during the study period, irrespective of hospital admission duration. On average, alertness increased in both the morning and evening for the intervention group but not the control group, though we failed to observe that this was statistically significant (−0.94 vs. 0.25 change in morning KSS, p = 0.13 and −0.81 vs. 0.53 change in evening KSS, p = 0.19). Less than half as many patients in the intervention group reported taking a nap compared to the control group, though this difference did not reach statistical significance (16.7% vs. 40.0%, p = 0.19).

3.4 |. Patient and Nurse Subjective Lighting Assessments

Patients in the intervention group were more likely than patients in the control group to agree with a statement that “the lighting fixtures are too bright,” (60.7% vs. 39.5%, p = 0.03), and that they would want “more control over the room lighting on the ceiling” (78.6% vs. 50.0% p = 0.01), and were less likely to agree that it is “easy to adjust the lighting pattern” (53.6% vs. 96.6%, p = 0.04). No differences in perception of lighting between rooms were noted for all other questions, including whether the lighting is “uncomfortably bright,” whether the lighting is “comfortable during the day,” or “comfortable during the night,” or for questions on the bathroom lighting or light from the adjacent bed. In the survey of nurses on the intervention room, 29% agreed that light was comfortable during the day, 12% agreed that light was comfortable at night, 12% agreed that the lighting was easy to use, 65% agreed that they would want more control of the lighting, 82% agreed that lighting was worse compared to other hospital rooms, and 88% reported that all or most patients complained about the study lighting.

4 |. Discussion

Our study found that dynamic lighting programmed to provide a more natural lighting spectrum that cycles over the course of the day, plus melanopic stimuli, advanced circadian rest/activity phase by 160 min compared to standard hospital lighting, resulting in the observed 66 min increase in overnight sleep duration between 11:00 p.m. and 7:00 a.m. The shift in phase likely aligned subjects’ sleep more closely with the subjective night; but in addition, subjects in the lighting condition showed a lower mesor of wake probability, reflecting increased sleep probability across the full 24-h continuum. Whilst the full reasons for the observed overall increase in sleep probability in the lighting condition are unknown, it is established that sleep that is aligned to the subjective night is generally longer in duration and less disrupted (Czeisler et al. 1980; Pilcher et al. 2000; Luckhaupt et al. 2010). Additionally, the study population comes from the largest urban city in America and likely experiences greater 24-h light exposure and delayed circadian phase compared to rural populations (Carvalho et al. 2014). Thus, it stands to reason that in such a population, an intervention that maximally advances phase and aligns sleep with the subjective night could also result in overall increased sleep duration. Our finding that patients in the intervention group reported higher levels of alertness during the morning and evening provides further support for a causal effect of the lighting system, though we acknowledge that because alertness questionnaire timing was consistently between 9 and 10 a.m., alertness responses may be influenced by the interval between wake and questionnaire completion.

Evaluation of sleep/wake patterns on the single last night of hospitalisation was rooted in the possibility that longer exposure to custom lighting might lead to stronger entrainment of the rest/activity patterns, with longer nocturnal sleep duration and shorter nocturnal wake duration; however, this analysis suffers from the fact that it reflects what is happening on a single night (e.g., an individual may be randomly having a “good” or “bad” sleep night on that one night that may not be reflective of the average sleep experience). Nonetheless, within these limitations, we observed significantly reduced nocturnal wake time on the last night between conditions and, on average, longer sleep duration (265 ± 55 min in the Lighting condition vs. 206 ± 53 min in the control condition); however, not significantly so (p = 0.13). We feel overall these observations are consistent with our observations regarding sleep/wake duration reported across the full hospitalisation in which, on average, nocturnal sleep is increased, and nocturnal wake is decreased. We further demonstrated in a general linear model that lighting condition continued to serve as a significant predictor of average nocturnal sleep duration when controlling for length of lighting exposure as well as DM2 and CKD status. This does not rule out the possibility that longer lighting exposure could better entrain rest/activity rhythms, only that the duration of lighting exposure experienced in our population did not contribute significant variance to the nocturnal sleep durations captured.

Our findings are consistent in direction with other studies that employed customised lighting in inpatient settings. Canazei and colleagues reported that their lighting intervention consisting of an “artificial dawn and dusk” and blue-depleted nighttime light in hospitalised patients with major depression advanced circadian phase and increased sleep by 11 to 27 min per week (Canazei et al. 2022). Gimenez and colleagues compared sleep for patients admitted to a standard room to a room with lighting programmed to provide gradual changes in luminance across the day, bright blue-enriched light 10:30 a.m.–12:30 p.m., and low nocturnal light, and found that sleep increased by 5.9 min in the intervention group (Gimenez et al. 2017). A cross-over trial of 12 patients on a psychiatry unit compared sleep for patients in rooms with blue-depleted lighting from 6:30 p.m. to 7:00 a.m. with rooms with standard hospital lighting and found an increase in sleep of 8 min per night (Vethe et al. 2021). In contrast, a small study of patients hospitalised with cirrhosis found no difference in sleep for patients receiving programmed lighting compared to a control group, possibly related to the significant level of decompensation in these patients (De Rui et al. 2015). Okkels and colleagues found no difference in sleep quality assessed via the Pittsburgh Sleep Quality Index (PSQI) or in subjective depressive symptoms in response to customised lighting, but did not measure sleep duration subjectively or objectively (Okkels et al. 2020).

Our study found an increase in sleep duration of substantially greater magnitude than noted in prior studies. The difference in the degree of benefit may be due to the light recipe utilised. Specifically, in addition to programmed lighting from 7:00 a.m. to 11:00 p.m., our protocol included delivery of a maximal melanopic stimulus twice daily, which has been proposed to increase daytime alertness. A brief melanopic stimulus was found to increase activation in the prefrontal cortex on functional MRI testing and improve performance on a task of memory for healthy volunteers (Alkozei et al. 2016). The Second International Workshop on Circadian and Neurophysiological Photometry brought together experts in the field to develop consensus recommendations on healthy daytime and evening lighting environments (Brown et al. 2022). Their panel recommended a minimum daytime EDI of 250 lx at eye level when seated and 10 lx starting 3 h prior to bedtime. In addition, the group recommended that a “stable and regular daily light-dark cycle is also likely to reinforce good alignment of circadian rhythms, which may further benefit sleep, cognition and health.”

The consequences of utilising lighting to increase sleep and align circadian rhythm with the natural light/dark cycle may have benefits associated with improved patient outcomes. Benedetti and colleagues found that patients with bipolar disorder admitted with a major depressive episode had a LOS that was 3.7 days shorter when admitted to rooms exposed to direct sunlight in the morning than patients in rooms without morning sunlight (Benedetti et al. 2001). Beauchemin and Hays reported decreased mortality for patients admitted to a cardiac intensive care unit with acute myocardial infarction assigned to southern-facing rooms defined as “bright” compared to northerly-facing rooms defined as “dark” (Beauchemin and Hays 1998). Artificial light designed to deliver blue-enriched morning light has been shown to improve subjective well-being, mood, and cognitive performance in non-hospital settings (Gabel et al. 2015). Whilst improvement in sleep timing and duration is an important clinical outcome on its own, understanding whether or not there are additional clinical benefits (e.g., on immune function) will be important to test for in the future.

Despite finding increased objective nocturnal sleep duration, the intervention room lighting was more likely to be perceived by patients and nurses as too bright. Similarly, few nurses agreed that light from the study room was comfortable during the day or night; most felt the room lighting was worse compared to other hospital rooms, and most reported that all or most patients complained about the room lighting. We acknowledge that it is possible that nursing impressions may have influenced patient impressions in terms of the satisfaction surveys if there was discussion between nurses and patients on this topic. Smart design of hospitals and lighting systems aiming to improve sleep duration will need to maintain high patient and staff satisfaction. Future versions of the dynamic lighting system can consider decreasing daytime light intensity or enhancing the indirect lighting component whilst maintaining a similar spectrum across the programmed period. Similarly, our finding that patients and nurses prefer more control over room lighting than was provided can be addressed through allowing patients and staff to adjust dynamic lighting using user-friendly wall switches. Patient or staff-initiated adjustments to the amount of lighting received, however, will have the potential to decrease the beneficial effects on sleep. Light interventions (such as a skeletal photoperiod approach with brief light exposure at dawn and dusk) (Olde Engberink et al. 2020) that also significantly optimise circadian phase and improve sleep but may also be more acceptable to patients and nurses would be ideal, but such an approach would have to be tested empirically. Future studies of dynamic lighting will need to assess the impact of changes to the intensity and spectrum of light and the degree of patient control of the system.

Our study has several important limitations. We acknowledge that the sleep/wake outputs from the mattress sensor have not been validated against polysomnography; however, the sleep/wake outputs have face validity in that greater sleep was observed during the subjective night in both conditions, and sleep/wake patterns in participants exposed to the customised lighting demonstrated an expected phase advance. We did not employ additional circadian measures such as melatonin levels or core body temperature to cross-reference with the sensor-based wake/rest activity patterns, and this can be considered in future studies. Second, there was incomplete data capture for some patients due to technical issues with the mattress sensor hardware. Efforts were made throughout the study to address mattress sensor hardware difficulties as promptly as possible. Third, patients were not assigned using a study-specific randomisation procedure to the intervention and control lighting rooms. This concern is mitigated by the assignment of cardiac patients to these rooms being based solely on bed availability by a central admissions office rather than on clinical factors, which provided similar characteristics for the study groups. The two experimental hospital rooms were immediately adjacent to one another, and although it is possible that nursing collected vital signs/blood systematically in one room before the other, because of their proximity, the difference in timing is likely to be on the order of 5 min. Fourth, we did not customise the lighting per subject in a way that catered to one’s individual chronotype, an approach that has been used in the hospital setting successfully when incorporating an approach using glasses on the patient’s face (Mangini et al. 2024). Fifth, we did not capture subjective pain data, which may have influenced sleep patterns. Sixth, study size was limited as only one intervention and one control room (two beds in each) were examined, and in part due to lower than anticipated enrolment due to the COVID pandemic. Additionally, we were unable to approach all patients admitted to the study rooms based on the availability of research coordinators during weekends, which affected patients admitted and discharged over the same weekend. The sample size allowed for conclusions on sleep duration/timing, though it did not provide adequate power to examine whether the benefits were more no-table for any subgroup, such as for patients with longer LOS in the intervention room, and does not allow definitive conclusions on the impact on LOS or changes in request frequency for as-needed sleep medications. Finally, for most patients, the lighting perception/preference surveys were not completed daily due to patients being off the unit for tests or unavailable for other reasons.

Our innovative lighting system programmed to provide blue-enriched light during the day, a melanopic stimulus twice daily, and blue-depleted light in the evening was associated with a significant phase advancement and increase in overnight sleep for hospitalised patients on a cardiology unit. Programmable lighting has the potential to be a core component of the smart design of hospitals seeking to engineer environmental factors to improve sleep and enhance well-being. These findings will need to be confirmed in a larger population with a more diverse range of acute illnesses. As sleep hygiene for hospitalised patients is further explored, a crucial question will be whether enhanced sleep can translate into improved outcomes, such as decreased incidence of delirium and decreased LOS.

Funding:

This work was supported by the Jim H. McClung Lighting Research Foundation and National Institute on Aging, AG066870, AG080609.

Abbreviations:

API

application programming interface

CKD

chronic kidney disease

COPD

chronic obstructive pulmonary disease

DM2

diabetes mellitus type 2

KSS

Karolinska sleepiness scale

LED

light emitting diode

LOS

length of stay

m-EDI

melanopic-equivalent daylight illuminance (m-EDI)

OSA

obstructive sleep apnoea

PSQI

Pittsburgh sleep quality index

REM

rapid eye movement

SPD

spectral power distribution

Footnotes

Disclosure

The authors have nothing to report.

Conflicts of Interest

The authors declare no conflicts of interest.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.

References

  1. Akerstedt T, and Gillberg M. 1990. “Subjective and Objective Sleepiness in the Active Individual.” International Journal of Neuroscience 52: 29–37. [DOI] [PubMed] [Google Scholar]
  2. Alkozei A, Smith R, Pisner DA, et al. 2016. “Exposure to Blue Light Increases Subsequent Functional Activation of the Prefrontal Cortex During Performance of a Working Memory Task.” Sleep 39: 1671–1680. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Arora VM, Chang KL, Fazal AZ, et al. 2011. “Objective Sleep Duration and Quality in Hospitalized Older Adults: Associations With Blood Pressure and Mood.” Journal of the American Geriatrics Society 59: 2185–2186. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Bano M, Chiaromanni F, Corrias M, et al. 2014. “The Influence of Environmental Factors on Sleep Quality in Hospitalized Medical Patients.” Frontiers in Neurology 5: 267. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Beauchemin KM, and Hays P. 1998. “Dying in the Dark: Sunshine, Gender and Outcomes in Myocardial Infarction.” Journal of the Royal Society of Medicine 91: 352–354. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Benedetti F, Colombo C, Barbini B, Campori E, and Smeraldi E. 2001. “Morning Sunlight Reduces Length of Hospitalization in Bipolar Depression.” Journal of Affective Disorders 62: 221–223. [DOI] [PubMed] [Google Scholar]
  7. Brown TM, Brainard GC, Cajochen C, et al. 2022. “Recommendations for Daytime, Evening, and Nighttime Indoor Light Exposure to Best Support Physiology, Sleep, and Wakefulness in Healthy Adults.” PLoS Biology 20: e3001571. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Canazei M, Weninger J, Pohl W, Marksteiner J, and Weiss EM. 2022. “Effects of Dynamic Bedroom Lighting on Measures of Sleep and Circadian Rest-Activity Rhythm in Inpatients With Major Depressive Disorder.” Scientific Reports 12: 6137. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Carvalho FG, Hidalgo MP, and Levandovski R. 2014. “Differences in Circadian Patterns Between Rural and Urban Populations: An Epidemiological Study in Countryside.” Chronobiology International 31: 442–449. [DOI] [PubMed] [Google Scholar]
  10. Czeisler CA, Weitzman E, Moore-Ede MC, Zimmerman JC, and Knauer RS. 1980. “Human Sleep: Its Duration and Organization Depend on Its Circadian Phase.” Science 210: 1264–1267. [DOI] [PubMed] [Google Scholar]
  11. Dobing S, Frolova N, McAlister F, and Ringrose J. 2016. “Sleep Quality and Factors Influencing Self-Reported Sleep Duration and Quality in the General Internal Medicine Inpatient Population.” PLoS One 11: e0156735. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Eklund NH, and Boyce PR. 1996. “The Development of a Reliable, Valid, and Simple Office Lighting Survey.” Journal of the Illuminating Engineering Society 25: 25–40. [Google Scholar]
  13. Gabel V, Maire M, Reichert CF, et al. 2015. “Dawn Simulation Light Impacts on Different Cognitive Domains Under Sleep Restriction.” Behavioural Brain Research 281: 258–266. [DOI] [PubMed] [Google Scholar]
  14. Gimenez MC, Geerdinck LM, Versteylen M, et al. 2017. “Patient Room Lighting Influences on Sleep, Appraisal and Mood in Hospitalized People.” Journal of Sleep Research 26: 236–246. [DOI] [PubMed] [Google Scholar]
  15. Hughes RG, and Clancy CM. 2005. “Working Conditions That Support Patient Safety.” Journal of Nursing Care Quality 20: 289–292. [PubMed] [Google Scholar]
  16. Jafarifiroozabadi R, Joseph A, Bridges W, and Franks A. 2023. “The Impact of Daylight and Window Views on Length of Stay Among Patients With Heart Disease: A Retrospective Study in a Cardiac Intensive Care Unit.” Journal of Intensive Medicine 3: 155–164. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Khalid A, Rafique R, Arshad M, and Hamdani M. 2022. “Assessment of Self-Reported Factors Associated With Impaired Sleep in Hospitalized Adult Patients in Internal Medicine.” Cureus 14: e25947. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Kuzmik A, Boltz M, BeLue R, Galvin JE, Arendacs R, and Resnick B. 2022. “Factors Associated With Sleep Quality in Hospitalized Persons With Dementia.” Alzheimer Disease and Associated Disorders 36: 253–258. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Luckhaupt SE, Tak S, and Calvert GM. 2010. “The Prevalence of Short Sleep Duration by Industry and Occupation in the National Health Interview Survey.” Sleep 33: 149–159. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Mangini C, Zarantonello L, Formentin C, et al. 2024. “Managing Circadian Disruption due to Hospitalization: A Pilot Randomized Controlled Trial of the CircadianCare Inpatient Management System.” Journal of Biological Rhythms 39: 183–199. [DOI] [PubMed] [Google Scholar]
  21. Moyen NE, Ediger TR, Taylor KM, et al. 2024. “Sleeping for One Week on a Temperature-Controlled Mattress Cover Improves Sleep and Cardiovascular Recovery.” Bioengineering 11: 352. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Okkels N, Jensen LG, Skovshoved LC, et al. 2020. “Lighting as an Aid for Recovery in Hospitalized Psychiatric Patients: A Randomized Controlled Effectiveness Trial.” Nordic Journal of Psychiatry 74: 105–114. [DOI] [PubMed] [Google Scholar]
  23. Olde Engberink AHO, Huisman J, Michel S, and Meijer JH. 2020. “Brief Light Exposure at Dawn and Dusk Can Encode Day-Length in the Neuronal Network of the Mammalian Circadian Pacemaker.” FASEB Journal 34: 13685–13695. [DOI] [PubMed] [Google Scholar]
  24. Park MY, Chai CG, Lee HK, Moon H, and Noh JS. 2018. “The Effects of Natural Daylight on Length of Hospital Stay.” Environmental Health Insights 12: 1178630218812817. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Pilcher JJ, Lambert BJ, and Huffcutt AI. 2000. “Differential Effects of Permanent and Rotating Shifts on Self-Report Sleep Length: A Meta-Analytic Review.” Sleep 23: 155–163. [PubMed] [Google Scholar]
  26. De Rui M, Middleton B, Sticca A, et al. 2015. “Sleep and Circadian Rhythms in Hospitalized Patients With Decompensated Cirrhosis: Effect of Light Therapy.” Neurochemical Research 40: 284–292. [DOI] [PubMed] [Google Scholar]
  27. Sun T, Sun Y, Huang X, et al. 2021. “Sleep and Circadian Rhythm Disturbances in Intensive Care Unit (ICU)-acquired Delirium: A Case-Control Study.” Journal of International Medical Research 49: 300060521990502. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Topf M 1985. “Personal and Environmental Predictors of Patient Disturbance due to Hospital Noise.” Journal of Applied Psychology 70: 22–28. [PubMed] [Google Scholar]
  29. Ulrich RS, Zimring C, Zhu XM, et al. 2008. “A Review of the Research Literature on Evidence-Based Healthcare Design.” Herd-Health Environments Research & Design Journal 1: 61–125. [DOI] [PubMed] [Google Scholar]
  30. Vethe D, Scott J, Engstrom M, et al. 2021. “The Evening Light Environment in Hospitals Can Be Designed to Produce Less Disruptive Effects on the Circadian System and Improve Sleep.” Sleep 44, no. 3: zsaa194. 10.1093/sleep/zsaa194. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Wesselius HM, van den Ende ES, Alsma J, et al. 2018. “Quality and Quantity of Sleep and Factors Associated With Sleep Disturbance in Hospitalized Patients.” JAMA Internal Medicine 178: 1201–1208. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Wilcox ME, Burry L, Englesakis M, et al. 2024. “Intensive Care Unit Interventions to Promote Sleep and Circadian Biology in Reducing Incident Delirium: A Scoping Review.” Thorax 79: 988–997. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.

RESOURCES