The protocol for the RCT follows the SPIRIT statement guidelines (Standard Protocol Items for Randomized Trials) (22) and is registered with the Clinical Trials website (ClinicalTrials.gov identifier: NCT03788993). The SPIRIT checklist and other appropriate details are included as an online supplement (Appendices 1 and 2).
Trial design, setting and interventions
This is a single-centre, unblinded, two-arm, parallel-group, pragmatic effectiveness RCT of differences in the mean duration of acute psychiatric hospitalization in days for individuals exposed to experimental lighting compared with normal lighting conditions.
The study is located at a newly built acute psychiatric unit at St. Olavs Hospital, Østmarka (Trondheim, Norway) which serves a catchment area of 300,000 people. The unit has 40 patient rooms divided equally between two wards that are built around two atriums. Each hospital ward has the same layout and facilities with five rooms (25% of the total) designated as ‘psychiatric intensive care beds’ (targeted at the most severely ill patients and offering a higher staff-to-patient ratio than the rest of the ward). The light intensity (photopic lux) is similar in both wards, but individuals are exposed to a different spectrum of light in each ward (e.g. melanopic lux is kept below 25 in the blue-depleted unit).
During the recruitment phase, patients admitted to the unit are randomized 1:1 to the experimental or control (normal) lighting conditions.
- Experimental condition: a 20-bedded ward with tunable light emitting diode (LED) lamps. At 18:00 hours (h) the lighting undergoes a 30-minute transition during which the green and blue LEDs are dimmed to produce blue-depleted amber coloured lighting. At 06:50h a 10-minute transition programme changes the light colour to normal indoor lighting (3000 Kelvins of colour temperature) which then continues until 18:00h. The light intensity is dimmed to 20% (of the maximum) from 23:00h to 6:50h.
As well as the LED system, blue-blocking window filters are deployed in the evening. All television sets have permanent blue-blocking filters and the outdoor area has external lights that block blue light. Use of electronic media is not restricted (unless access is limited in accordance with an agreed treatment plan), but patients are provided with blue-blocking screens that can be attached to the front of all electronic devices. If a patient leaves the blue-depleted unit after 18:30 they are offered blue-blocking glasses to wear. The amount of blue light in the ward was assessed prior to commencing the RCT. This demonstrated that the light exposure is well-matched to the amount shown in laboratory settings to minimally suppress melatonin (23, 24) and this was tested in an onsite pilot study with healthy adult volunteers (ISRCTN12419665).
- Control condition: a 20-bedded ward with normal indoor lighting installed. The light intensity is dimmed to 20% (of the maximum) during the night (from 23:00h to 06:50h).
Participants and procedure
The flowchart for the RCT is shown in Figure 1.
Admissions occurring at any time during each 24-hour period and on any day of the week throughout the 6-month recruitment period are considered eligible for study inclusion.
No specific recruitment strategy is employed, but should recruitment fail to reach the required sample size (for any reason) within an allocated 6-month time period, we will recommence recruitment in late 2019 i.e. all recruitment will be undertaken during months with similar daytime light exposure (contact HK for further details).
A frequent barrier to research into acute psychiatric admissions is that, at the time of hospitalization, many individuals lack mental capacity to give informed consent (they may be severely ill, suicidal, lack insight or be hospitalized involuntarily) and/or clinicians and researchers may regard a discussion about research participation inappropriate or unethical. However, as noted in acute medicine research, excluding a large proportion of patients from an RCT purporting to investigate acute admissions can bias the included sample so that it is unrepresentative of the inpatient population and undermines the generalizability and real world impact of any study findings (25). For this reason, we did not employ any patient exclusion criteria but, with ethical approval, used a post-randomization, deferred (also known as delayed) consent procedure (25-27). The RCT eligibility criteria are as follows:
Inclusion criteria: All individuals aged > 18 who are admitted to the acute inpatient unit during the study period are eligible for randomization. Any patients who are re-admitted during the study period are eligible for re-randomization.
Exclusion criteria: There are no pre-randomization exclusion criteria, but individuals may be withdrawn from the study immediately post-randomization (see Withdrawal criteria).
Randomization is based on an allocation concealment procedure.
As soon as the decision to admit the patient is confirmed, the individual is randomly allocated to one of the two arms of the RCT using a web-based randomization program with a variable block design. This occurs without any consultation with wards (regarding bed availability, case-mix or staffing levels, etc).
The procedure was developed and is managed by the Unit for Clinical Research (at the Department of Medicine and Health, NTNU) and can be implemented at any time of the day or night. Randomization is undertaken by the nurse coordinating the intake process (who participated in pre-trial instruction and training), but they cannot influence the process in any way. When the intake nurse logs in to a web-based programme, the randomization program issues an authentication code that is sent to the unit (a telephone back-up system is available should there be any problems with the web-based programme).
As randomization occurs at the point of admission, all exclusions are de facto post-randomization. Immediately post-randomization, there are two potential reasons for withdrawal from the RCT:
1) Lack of availability of rooms (as allocated at randomization): acute wards operate at high levels of bed occupancy, so sometimes there will be no rooms available in the ward to which the individual is allocated (i.e. the randomization process cannot be completed).
2) Clinical imperative: sometimes a senior clinician may decide that it is inappropriate to admit an individual to the room to which they are randomized. The most frequent reasons for this decision are clinical concerns about (a) how this admission would affect the case mix within the ward (e.g. it may be inappropriate to co-locate several patients with acute mania) and/or (b) following through with the randomization process may compromise the safety, care or treatment of existing inpatients or of the individual being admitted (e.g. it may not be possible to provide the appropriate staff-to-patient ratio required if all individuals with higher levels of need are co-located).
During the admission withdrawal may occur because:
1) The individual is unwilling to give written informed consent at any time during their admission (when approached according to the deferred consent procedure) or is unable to give informed consent for the duration of the study (i.e. they remain persistently and severely ill and/or lack mental capacity).
2) The consent procedure was incomplete: an individual may be discharged early or have an unplanned discharge (so they may not be approached about study participation or have only given verbal, but not written consent).
3) A patient will be withdrawn from the study if they are absent for >24 hours from the ward to which they randomized (e.g. they may be transferred to a medical ward for several days; a patient may request or clinicians may instigate transfer to another ward; medical or nursing staff may decide a patient should be transferred to the other ward participating in the RCT because of clinical, case mix or staffing issues, etc.).
4) An individual can decline to participate at any stage of the study and/or a mental health professional can recommend withdrawal of an inpatient from the RCT if they have any clinical concerns regarding an individuals’ participation (e.g. if they believe a patient has experienced an adverse event associated with exposure to the blue-depleted light).
A record will be kept of all reasons for withdrawal.
Descriptions of all measures are provided below, and the timing of study assessments is summarized in Table 1.
A key consideration in the selection of assessment tools was that they were already used routinely or could easily be incorporated into ward procedures, and that clinical staff were familiar with or had received training in the use of the instruments. The design and setting of the study mean it is unfeasible to blind patients, clinicians or investigators. However, we have used electronic hospital records and advanced technology to collect objective data on sleep-wake cycles.
Baseline demographic and clinical information
The intake assessment records detailed information on the following:
- age, sex, ethnicity, marital status, living situation, years of education, employment status,
- current diagnosis or diagnoses (according to the International Classification of Diseases 10) (28). As in our previous research, consensus expert opinion is employed to review all diagnoses at discharge. If the intake and discharge diagnoses differ, the latter is employed in analyses (as acute intake diagnoses may be less reliable).
- type of admission (voluntary or involuntary), number of psychiatric admissions and total number of inpatient bed-days in the two years prior to the index admission and current prescribed medication.
- details of current presentation including evidence of sleep problems (during the preceding month), level of functioning, alcohol and substance misuse, risk of or actual harm to self or others, current physical health.
- past psychiatric and forensic history, and history of comorbid medical illnesses.
The primary outcome measure is mean duration of admission in days per individual. Admission is defined as the date and time of commencement of the intake assessment (recorded electronically when the randomization code is generated); discharge is defined as the date and time the patient left the lighting condition to which they were randomized for >24h.
Secondary outcomes are focused on clinical changes over time and treatments or other interventions reported during admission.
- Clinical Assessments
- Objective assessments:
Sleep-wake cycle: Individual sleep and activity patterns will be assessed using de-identified data collected via radar (Xethru sensors) installed in each room. The sensor is a low-powered ultra-wideband radar that allows contact-free assessment of sleep patterns and wakefulness with high accuracy compared to polysomnographic (PSG) recordings (29).
Employing best available scoring algorithms, raw data from daily recordings will be used to estimate total sleep time (TST), sleep onset latency (SOL), number of nocturnal awakenings, wake after sleep onset (WASO), and final wake time for each participant, along with day-to-day variability in sleep onset, sleep offset and TST.
- Observer assessments
- Clinical Global Impression scale (CGI) is rated on a 1-7 scale (high scores indicate worse clinical or functional status) and is a well-established outcome measure in trans-diagnostic studies (30). In this RCT, psychiatrists, psychologists and nursing staff trained in the use of the CGI, will record a consensus score for each study participant during the daily ward meetings. During the weekends, the psychiatrist on duty and nursing staff will provide the CGI consensus ratings.
The CGI ratings will be based on all available information (nurse observations, clinical assessments, hospital medical records, etc.) and will be used in two ways in this RCT: (a) to monitor day-to-day changes in mental state and functioning, and (b) to record overall change from admission to discharge.
- Clinical Global Impression, Severity subscale (CGI-S) is a likert scale ranging from 1 to 7 (from ‘Normal, not at all ill’ to ‘Among the most extremely ill patients’) (30). The CGI-S ratings are benchmarked relative to the total inpatient population and are scored on two occasions only: the morning after admission to the unit and at discharge (based on the preceding 24 hours).
- Clinical Global Impression, Improvement subscale (improved version: iCGI-I) captures change over time with ratings ranging from -6 (maximum deterioration) to +6 (ideal improvement) (31). The iCGI-I is used (a) to monitor day-to-day changes in mental state and functioning, and (b) to record overall change from admission to discharge.
- Risk of Harm to Self or Others: Suicide risk is assessed daily (rated according to level of risk and/or need of continuous observation) and risk of aggressive behaviour is assessed three times per 24 hours using the Brøset Violence Checklist (BVC) (32, 33). The BVC is a 6-item scale with good psychometric properties and the sum score indicates risk of violence (low=0). Incidents of aggressive behaviour will be systematically recorded using the Staff Observation Aggression Scale-Revised (SOAS-R) (34) and interventions employed will be recorded.
- Treatments and Interventions
- Medications: Daily doses and classes of medications or other treatments or interventions prescribed per individual during admission will be recorded.
- Change in admission status: if a patient is admitted involuntarily, we will record the time until their status is reclassified as voluntary (as a marker of improved insight and mental capacity). Similarly, for some individuals time to change from voluntary to involuntary status will need to be recorded.
Patient-related experiences and other outcomes:
For everyone involved with the new unit, a key aspiration is to try to capture information regarding patient experiences of the experimental condition and to evaluate benefits and harms. In this RCT, we will assess acceptability (adherence, perceived satisfaction and benefits) or harms (side-effects and adverse effects) using observer and self-rated assessments.
For individuals allocated to the blue-depleted conditions, adherence with the intervention is assessed using an item checklist to record any exposure to normal lighting (duration and reasons), whether blue-blocking glasses were worn as appropriate (e.g. when exiting the unit) and whether blue-blocking filters were employed on media devices.
- Satisfaction and Benefits
Mean levels of patient satisfaction with an admission are routinely assessed using the standard patient satisfaction questionnaire completed at discharge. The questionnaire was developed by Norwegian Institute of Public Health, is used throughout the Norwegian Health Care system and consists of 10 items scored on a 5-point Likert scale (1=low satisfaction). Some items are relevant to examining experiences of the different lighting conditions, side effects (see below) and perceived benefit of the admission (1 = no benefit).
- Side effects and adverse events
The frequency of any side effects or adverse events experienced by individuals admitted to each lighting condition will be recorded using the 8-item Headache and Eye Strain Scale (HES), which has been shown to be sensitive to exposure to different lighting conditions (35). This is supplemented by eight items that may reflect side effects of acute psychiatric treatments (e.g. dizziness, gastro-intestinal disturbances, daytime sleepiness, poor sleep quality, and restlessness, etc.). Each symptom is rated on a 4-point scale (ranging from absent to severe).
To capture any putative adverse events experienced during the admission we will record the occurrence of any serious or untoward incidents in each ward (such as non-accidental and accidental deaths, near fatal events, severe violence, etc.). Also, we will note if any patients are transferred out of the blue-depleted light environment because of clinical opinion suggests that it is having detrimental effect on the individual.
Several ancillary studies are planned. For example, ethical approval has been granted to undertake an additional study (running concurrently with the RCT) to examine the experiences of clinical staff who rotate their work schedules between the two wards (ISRCTN21603406), including sub-studies of subjective effects on sleep, cognition, and well-being, and objective actigraphic recordings undertaken during exposure to each lighting condition. Data regarding any side effects may be compared across staff and patients, but other studies will be reported separately.
The power calculation and sample size were estimated for the primary outcome measure, namely mean number of days hospitalized per individual exposed to the experimental or control lighting conditions. This outcome was chosen as hospitalization represents a major life event for patients, is the largest contributor to cost of care across psychiatric diagnoses, and it offers a proxy measure of any adjunctive benefits associated with the experimental lighting conditions over and above any gains associated with usual inpatient treatment.
The hospital database showed that there were 1639 acute psychiatric admissions between May 2016 and April 2017, with a mean length of stay of 6.3 days (range: 1-158 days; total occupied patient bed-days per annum >10,000). Assuming the experimental lighting conditions lead to a reduction in the mean length of stay from about 6 to 5 days (with a standard deviation of about 3.5 days), then 194 participants in each condition will give an 80% chance at an alpha = 0.05 to detect a difference in the length of stay of one day using an intent to treat (ITT) analysis (and >85% power to detect a reduction of 1000 patient bed-days occupied over the course of the year). We have assumed that there will be about 800 admissions to the unit over six months, and that post-randomization exclusions (lack of consent, etc.) and sample attrition (e.g. due to early discharge) will amount to about 30% (n=240). Whilst the sample size required is 400 individuals (who give written informed consent), recruitment will continue for the entire six-month period, and we will continue to randomize all admissions. As such, it is possible that we will be able to include up to 500 individuals.
Data Analysis Plan
Data retrieved from electronic hospital records and from assessments will be de-identified and the files will be stored on secure data storage servers.
The primary statistical analyses will be performed by an independent statistician who is masked to the lighting conditions experienced by each group.
We will use a linear mixed model analysis to examine the difference in mean duration of admission in days per individual according to group. As this primary outcome is heavily skewed and not normally distributed, we will use bootstrapping. The main analyses will be based on the ITT population (i.e. all individuals who gave consent and were randomized). Per protocol (PP) analyses will be limited to participants who are hospitalized for at least two consecutive evenings in the inpatient unit. There will also be some individuals who were not randomized but who give informed consent for use of some data in additional analyses.
We used a block randomization with random block length (which avoids adding to the burden of the intake process or delaying the admission procedure). As such, we will adjust for a pre-specified subset of demographic and clinical covariates (selected from those listed in the baseline assessments) which are known predictors of duration of a psychiatric admission: these are age, sex, diagnosis, presence of comorbidities, status at admission (involuntary or voluntary), and number of bed days in the previous two years (36, 37). It should be noted that some patients will be re-admitted and re-randomized during the study period, so we will use a two-level model (with admissions within patient as level one and patient as level two).
For the analysis of the primary outcome, we expect complete data on duration of admission. For analyses of secondary outcomes, we will use multiple imputations to handle missing data as appropriate. We plan to use a mixed model with random intercept and random slope for secondary analyses which have one or more values per admission. Additional analyses will include subgroup (e.g. length of stay according to diagnostic groups; use of medications according to group, etc.) and exploratory analyses (e.g. examining whether improvement in sleep variables mediates duration of admission and risk of harm to self or others; prevalence of self-reported side effects in patients compared to nurses).
Patient and public involvement
Patient and public involvement has been a feature of the development of the inpatient service and the proposed research. For example, the leader of the User Group for Mental Health, St. Olavs Hospital, has reviewed the project and offered public support. Also, representatives of the advocacy group have been involved in all the processes related to research at the new unit from the start and have had several meetings with the research group that discussed the study design. Ongoing advice and support have been provided by medical and nursing colleagues working at the unit and others working at St. Olavs Hospital. This included dialogue about what assessments could be incorporated into ward routines and the training needs of staff involved in the projects. International experts were consulted regarding the procedure for deferred consent and others offered advice regarding the programme for and delivery of the chronotherapeutic interventions.
There is considerable international interest in the lighting technology and the use of this programme as an adjunct to standard treatment in an inpatient environment. As well as publication of findings regarding primary and secondary outcomes, we will publish descriptive articles, present information about the unit and the RCT findings at national and international conferences and will allow site visits by clinicians, researchers and patient advocates who wish to view the unit.