A prospective cross-sectional study was conducted over a six-month period (May to October 2013) in 15 clinical areas: six acute medical and six surgical wards, an intensive and high dependency care unit, and one aged care and one rehabilitation ward. These clinical areas were considered representative of services provided in this 672-bed tertiary referral teaching hospital in Australia. The study involved technical monitoring and concurrent direct observation of the nocturnal environment, patient interviews and nursing reports for each clinical unit for two consecutive nights between 22:00 and 07:00 h.
Patients and nursing staff were recruited for this study. Non-probability convenience sampling was used to recruit patients to participate in a semi-structured interview to provide their individual perspective on sleep in the clinical setting. Patients admitted to any one of the 15 clinical wards on the two consecutive nights were eligible to participate in the interviews, unless they were intubated; were receiving end-of-life care; had a medical diagnosis of florid psychosis, dementia, confusion or expressive disorder; or the medical or nursing staff considered them clinically unsuitable.
Nursing staff working the night shift during the monitoring period were invited to complete a survey regarding their perceptions of the quality of their patients’ sleep and what they considered to be sleep-disturbing factors. Nursing staff were not required to provide direct patient care to the patients who participated in the semi-structured interview. Rather, nursing staff provided an overview of the clinical environment at the time of monitoring.
Data were collected using three methods: subjective data were obtained from nursing staff and patients, objective clinical data were derived from environmental monitoring and observational data were documented by the study research assistants. These data were concurrently obtained over two consecutive nights on each ward. The data collection tools were adapted, with approval, from documents designed by the Imperial College Healthcare (United Kingdom National Health Service) to more accurately reflect the Australian hospital’s clinical environment in terms of service delivery and design.
The researchers made no attempt to blind clinical staff or patients to the study purpose. Staff were informed that technical and observational information were being collected about the clinical environment in general, and not specifically about them or the patients. No alterations were made to the clinical environment, such as reminding staff to turn off lights or alerting them to alarms.
Four instruments were used to measure the study outcomes which had been developed by researchers at the Imperial College Healthcare (UK) based on existing sleep questionnaires and review of the literature: the Patient Sleep Interview Form , the Nurses’ Self-report Form , the Environmental Sleep Observation Form and environmental monitoring devices .
Patient sleep interview form
The patient interviews comprised open-ended questions concerning the patients’ perceptions of their sleep experience in the hospital. The interview questions addressed sleep quality; sleep hygiene behaviours at home and if and how their sleep differed in the hospital; and which factors they believed inhibited, prohibited or aided their sleep in the hospital. The patients were interviewed once the environmental monitoring of the clinical ward was completed, and their responses were recorded verbatim in writing by the research assistant conducting the interview.
Nurses’ self-report form
This questionnaire elicited nurses’ perceptions of their patients’ quality of sleep in the clinical ward, including the factors that they perceived inhibited the patients’ ability to sleep inclusive of the clinical environment. Nurses’ suggestions on how to improve patients’ sleep were also requested. Comments provided by clinical staff pertained generally to their clinical environment and were not matched to the patients who participated in the semi-structured interviews.
Environmental sleep observation form
The Environmental Sleep Observational Form was used to log the frequency of predetermined categorised noise sources which had the potential to inhibit or disturb patients’ sleep, such as conversations, alarms, telephone calls and patient buzzers. The documentation of observational data was completed by two research assistants (who were not registered nurses) during each night of clinical monitoring for one hour at four time points: 23:00, 02:00, 04:00 and 06:00 h. These timeframes were selected to reflect the spread of activity throughout the night in the clinical setting.
Environmental monitoring devices were positioned in the clinical environment to monitor noise, luminance and temperature. A total of 12 noise, light and temperature monitors were positioned in a range of clinical areas in each ward, including the nursing station, corridors, and single and shared patient rooms in order to capture the activity in the clinical areas. The devices were positioned directly behind the patients’ beds, permitting the monitoring of stimuli from the patients’ perspective, while minimising their effect on the work of the clinical staff. In clinical areas, such as nursing stations and corridors, sound monitoring devices were mounted on the wall in the centre of the room. This occurred concurrently with the patient interviews and nursing surveys.
Noise and luminance monitoring
Noise and luminance levels were recording using Extech sound level meters (Model SDL600, frequency range 31.5 Hz to 8 KHz) and Extech light meters (Model SDL400), which both conformed to American National Standards Institute and International Electro-technical Commission Standards.
Noise levels were recorded in A-weighted decibels (dB), with a fast response time (125 ms) to permit the capture of peak noises (within a decibel Min-Max from 30 to 180 dB), and noises that occurred rapidly. The A-weighted filter was used because it attenuates the curve that describes loudness frequency for the human ear. Whilst luminance levels were recorded in Lux, with a recording Min-Max of 0 to 1999 Lux accuracy: +/− 4% + 2 disability glare threshold.
Noise and light meters were located within close proximity to each other and were programmed to log data at five-second epochs, and were calibrated, as per the manufacturer’s recommendations, before each recording. All logged data from the sound and luminance meters were saved to a secure digital 2 GB memory card in a spreadsheet format.
Temperature and humidity monitoring
Temperature and humidity were recorded at 30-s epochs using ThermoLoggers (Thermodata Pty Ltd) (accuracy +/− 1 °C and =/− 0.6% humidity, range temperature − 40 °C to + 85 °C), as temperature and humidity is less variable than the other environmental measures. These devices were calibrated prior to being positioned within the clinical environment as per manufacturers’ recommendation and were positioned within close proximity to the sound and luminance monitors. Data was then downloaded directly from the thermologgers into Excel (Microsoft 2010) spreadsheets.
All data were collated, coded and entered into the Statistical Packages for Social Science (IBM SPSS, version 20) for quantitative analysis. Descriptive statistics of mean and standard deviation (SD) were applied. The chi-square statistic was used to determine any differences in sleep quality reported by patients and staff. Missing data related to environmental monitoring (noise, light, temperature and humidity) was managed via a data reduction method which deleted cases which involved missing data. To replace missing data an additional night (9 h) of environmental monitoring was undertaken to replace missing data. This method was applied in three cases. The interview responses from the patients and nurses were transcribed from the interview form, and underwent content analysis using NVivo (10 QSR). The coding of the responses was continuous and iterative, with responses coded to emerging themes that were then synthesised into smaller theme matrices. The codes identified were descriptive (rather than interpretative) of the responses in an effort to reduce researcher bias.