In recent years, there has been a growing interest in how our dietary habits affect cognitive performance and mood( Reference Hammersley, Reid and Atkin 1 , Reference Jager, Dye and Bruin 2 ). A healthy diet, including a high consumption of fruit and vegetables, fatty fish and nuts, and a low intake of SFA, has been associated with cognitive benefits in children, old adults and in midlife( Reference Lamport, Saunders and Butler 3 – Reference Pribis, Bailey and Russell 6 ). Conversely, skipping meals or eating sugary snacks can result in low blood glucose levels, causing slowed reaction time, lowered concentration and mood( Reference Hoyland, Lawton and Dye 7 – Reference Muff, Reinhardt and Erbel 9 ).
A previous study demonstrated that healthcare staff often eat unhealthily or skip meals because of limited time to take a break, as well as limited access to healthy meals and snacks( Reference Lemaire, Wallace and Dinsmore 10 ). This was supported by an employee survey at the Copenhagen University Hospital Herlev-Gentofte in Denmark, where stated reasons for non-users of the staff canteen were both limited canteen opening hours and lack of time and opportunity to leave the ward( 11 ). A high proportion of the staff participating in the survey responded that they would make use of vending machines with healthy snacks, if available( 11 ). It is generally acknowledged that there is an abundance of chocolate on hospital wards, rather than access to healthy meals or snacks, as chocolate is often given to hospital staff by patients as a token of their gratitude( Reference Cheung 12 ). Moreover, a prospective cohort study by El-Sharkawy et al.( Reference El-Sharkawy, Bragg and Watson 13 ) among nurses and physicians highlighted that a significant proportion of nurses and physicians were dehydrated at the start and end of medical and surgical shifts. Dehydration was associated with some impairment of cognitive function.
The observed lack of access to proper meals and beverages at work among healthcare personnel may have implication not only in a health promotion perspective but also in terms of reaction time and, ultimately, in patient security. Night-shift workers typically have the most limited opportunities to leave the ward and most limited access to canteen services or other meal options. Nabe-Nielsen et al.( Reference Nabe-Nielsen, Jorgensen and Garde 14 ) showed in a cross-sectional questionnaire study that working environment interventions in the eldercare sector may reach shift workers to a lower extent compared with fixed day workers. This group of employees has also been associated with more unfavourable dietary habits compared with day workers( Reference Hemio, Puttonen and Viitasalo 15 ). Also, there is some preliminary evidence that shift work may adversely influence cognitive functions. Titova et al.( Reference Titova, Lindberg and Elmstahl 16 ) observed that current and recent former shift workers performed worse on a trailmaking test, which measures executive cognitive function, compared with non-shift workers.
As far as we know, research examining the effect of optimal workday nutrition on cognitive function in healthcare staff is limited to a study by Lemaire et al.( Reference Lemaire, Wallace and Dinsmore 17 ). They demonstrated a positive effect of offering healthy food on physicians’ cognitive score, reaction time and blood glucose levels. The study was relatively small (twenty participants) and the intervention only lasted 1 d. For the majority of the participants (seventeen of twenty) the study took place during the day.
A review by Schröer et al.( Reference Schröer, Haupt and Pieper 18 ) concluded that employees’ dietary behaviour can be influenced by workplace interventions based solely on nutritional education or in combination with environmental modifications. They point out that maintaining employees’ health and work performance, including introduction of workplace health promotion and prevention programmes, is of great importance. However, there is still a lack of sufficient evidence through randomised controlled trials on effective and feasible strategies to promote healthy eating among healthcare staff, including workers on shift schedules.
The objective of the present study was to examine the effectiveness of increased availability of healthy meals at work, including a cold lunch meal, bottled water and a snack, on reaction time, mood and total dietary intake among hospital healthcare staff. We further aimed to conduct a subgroup analysis examining data separately for the participants working on shift schedules.
Methods
Study design
The study used an 8-week randomised cross-over design, initiated with a 1-week run-in period (Fig. 1). Participants were divided into fifteen groups, with four participants in each, according to the sequence of recruitment. All participants entered the study simultaneously. The number of participants allocated to each group was because of practical reasons, as there was only capacity to test four participants a day. The groups were block-randomised to start either with the intervention period or the control period. The randomisation process was carried out by an independent research assistant drawing lots from several small closed labels with ‘control’ or ‘intervention’ written on them. The subjects were informed about their allocation before the start of the study.
The diagram of the study design. 
, Intervention diet; 
, habitual diet; 
, assessment of reaction time, mood, dietary intake.
Each subject was required to attend three test sessions: on the 1st day of run-in period, on the last day of the control period and on the last day of the intervention period, respectively. The tests were completed under standardised conditions – that is in the same location and at the same time in the morning after the subjects ate their habitual breakfast. Consumption of coffee and other drinks high in caffeine, for example, black tea, was not allowed 10 h before the tests.
Participant recruitment
Physicians, nurses and nursing assistants employed at the Copenhagen University Hospital in Herlev-Gentofte in Denmark were recruited for the study through posters displayed around the hospital, and through advertisements on the hospital Intranet in August 2012. Approximately 130 persons attended information meetings, ninety-six were selected for screening and sixty were enrolled as study participants. Screening and enrolment was performed by one of three research assistants. Exclusion criteria were BMI below 18·5 or above 35 kg/m2, use of staff canteen more than once a week, weekly working hours below 30, no direct contact with patients, food allergy or intolerance, pregnancy, extreme level of physical activity and vacation of more than 1 week during the study period. The study period ran from September until December 2012.
Participants were classified as shift workers if the majority of their working hours were in the evening, at night and/or at the weekend.
Ethical statement
All subjects signed a written consent after receiving oral and written information about the study. Because of the design of the study, no approval was needed from the Research Ethics Committee according to the Danish law. The study was registered at ClinicalTrials.gov, identifier: NCT01739569.
Intervention content
During the intervention period the participants were offered a cold meal, a bottle of water and a snack on all workdays/shifts. Two cold meals, two snacks and two bottles of water were delivered for participants working a double shift. No educational activities or recommendations were made regarding dietary intake. In the control period the participants were instructed to continue with their habitual dietary intake. Both groups were asked to maintain their body weight and their habitual level of physical activity throughout the study period.
The cold meals were pre-packed lunch meals selected from the ordinary Keyhole menu at the hospital staff canteen. The Keyhole is a Nordic nutrition label designed to make it easier for customers to choose healthier food alternatives. Meals eligible to carry the Keyhole label must fulfil certain conditions, including appropriate portion sizes, a maximum amount of fat, salt and sugar and a minimum amount of dietary fibre, whole grain, fruit and vegetables( Reference Lassen, Beck and Leedo 19 ). Each weekday the participants could choose between six different Keyhole cold meals – that is three salads, two closed sandwiches and one open rye sandwich. The selection of cold lunch meals changed according to availability at the weekend. There was a total of seventy-six different keyhole-labelled cold lunch meals in the selection during the whole study period.
The snacks were designed especially for the present study based on results from a focus group session with a total of nine healthcare employees conducted at the hospital before the beginning of the study. They were further designed so that they could be eaten ‘on the go’, and to have an energy content of 800–1000 kJ corresponding to about 10 % of the recommended total daily energy intake for adults( 20 ). The snacks consisted of three elements: a fresh in-house-baked bun or an in-house-produced muesli bar, a piece of fresh fruit and a bottle of water (500 ml). There were seven different buns or muesli bars in rotation: a rye bun with dark chocolate or dates on Monday and Thursday, a wheat bun with spinach/feta cheese or potato/pesto on Tuesday and Friday and a muesli bar on Wednesday. At weekends the same bun as delivered on Friday was generally supplied. The type of fresh fruit provided changed according to availability.
All subjects had to provide a working plan for the intervention period in order for the study to provide a proper service. Moreover, participants had to order the cold meals by 12.00 hours on Friday the week before delivery at the latest. A salad meal and the days’ snack were delivered to the participant as a default, if they failed to place an order on schedule. The food was delivered to the respective wards between 11.00 and 12.00 hours on weekdays and between 14.00 and 16.00 hours the Friday before for weekend shifts.
The nutrient composition of the intervention meals is described in Table 1. The average serving size of the cold meals was 366 (sd 85·7) g and that of the offered snacks was 63 (sd 19·3) g.
Nutrient composition of the meals offered in the intervention period (Mean values and standard deviations)
E%, energy percentage.
* The values are means of all seventy-six keyhole-labelled lunch meals served during the whole study period.
† The values are means of all seven buns/muesli bars and eight different fruits served during the whole study period.
Anthropometrics
Body weight was measured using the same body composition analyser TBF-300 (TANITA), in the same location, and standardised to the nearest 0·1 kg, ensuring that participants voided their bladder immediately before weighing, and that clothing was the same each time (i.e. no coat, no shoes, empty pockets, phones removed). BMI was calculated using the measured weight and self-reported height.
Reaction time test
Reaction time was measured by Go/No-Go test from the Test-battery for Attentional Performance version 2.1( Reference Zimmermann and Fimm 21 ). This task tested the participants’ ability to perform under time pressure while suppressing inappropriate behaviour responses( Reference Sorensen 22 ).
Go/No-Go is a complex reaction time test of 2 min 45 s, where five different patterned squares appear randomly one-by-one on the computer screen. Two of these squares are defined as target stimuli, upon the appearance of which the subject should react as quickly as possible with a key press; no reaction is required to the other squares. The trial number is sixty, of which twenty-four are targets and thirty-six distractors. Average reaction time (ms) for correct responses and number of errors and omissions were recorded. In this study, the sum of errors and omissions is presented as ‘Errors’.
Mood-related measures
The Profile of Mood States (POMS) questionnaire was used to assess participants’ subjective mood state. The questionnaire consists of sixty-five mood-related adjectives, of which the subjects are asked to state their experience over the past week on a five-point scale from 0 (not at all) to 4 (extremely)( Reference McNair and Heuchert 23 ). Items are summarised in six factors: Tension-Anxiety, Confusion-Bewilderment, Anger-Hostility, Depression-Dejection, Fatigue-Inertia and Vigour-Activity. A total mood disturbance score is the sum of the first five factors minus Vigour-Activity Score. High scores indicate greater mood disturbance on all scales except Vigour-Activity.
Dietary intake record
Subjects completed 4-d self-reported dietary records to provide information about their dietary intake during the study. This was done during the run-in week, and during the last week of both the intervention and the control periods. The subjects were instructed in dietary recording during the run-in meeting, and were asked to record quantities and items of all foods and drinks consumed. The dietary recordings had to be performed in concordance with participants’ working schedule, so the recordings reflected intake during workdays. The participants were given feedback and advice after delivering the first dietary record on how they could complete it even more comprehensively. Dietary intake on days when only a partial recording was completed was not included in the analyses. A research assistant blinded to the result of the randomisation manually entered the dietary records using MADLOG VITA® software (based on data from www.frida.fooddata.dk).
Power calculations
The power calculations of the total group of participants were based on the primary endpoint, reaction time and were based on a study by Jakobsen et al.( Reference Jakobsen, Kondrup and Zellner 24 ), in which a positive effect of a high-protein diet on reaction time was reported. With a power of 0·9, an α=0·05, a δ=−25·5 and an sd 40, a sample size of fifty-four was required. A total of sixty participants were included in the study in order to allow for possible dropouts.
Statistical analysis
Statistical analyses were performed using IBM SPSS Statistics version 19.0. Initially a possible existence of a carryover effect of treatment sequence on all outcomes was examined. Mean difference between absolute values at the end of each period was tested for normality and compared with each other. When data were normally distributed, a Students’ t test was used to compare values between the periods. For Depression-Dejection Score, number of errors in reaction time test and intake of SFA, PUFA and water, data were not normally distributed and the non-parametric equivalent (Mann–Whitney U test) was used to compare the absolute values at the end of the periods. Normally distributed data are presented as means and standard deviations, whereas data not normally distributed are presented as medians and ranges. All statistical tests and P values were two-sided and a cut-off level of P<0·05 was used for assessing statistical significance. Data analysis was made both on the total group and on a subgroup of the participants working shift schedules (n 16); however, these groups were not compared with each other.
Results
Sixty subjects entered the study. One subject dropped out during the first period because of long-term sick leave and was not included in the analysis. The run-in characteristics of the fifty-nine study participants are listed in Table 2.
Run-in characteristics of the study participants (n 59) (Numbers and percentages; mean values and standard deviations; medians and ranges)
E%, energy percentage.
The mean values of all variables after the two periods, and the mean differences between the periods, are presented in Table 3. The participants were weight-stable throughout the whole period.
Reaction time, mood-related scores and dietary variables and differences between the mean values after the intervention and control period for the group as a whole (n 59) (Mean values and standard deviations; mean values and 95 % confidence intervals)
E%, energy percentage.
No significant difference in reaction time or number of errors in the reaction time test was observed between the intervention and the control period. Likewise, the subgroup analysis in the participating shift workers (n 16) showed no differences in either reaction time or in number of errors in the reaction time test between the intervention and the control period (Table 4).
Reaction time, mood-related scores and dietary variables and differences between the mean values after the intervention and control period for the shift workers (n 16) (Mean values and standard deviations; mean values and 95 % confidence intervals)
E%, energy percentage.
Differences between mean values of the mood-related scores did not reach the level of significance in the group as a whole (Table 3). In contrast, several significant beneficial differences in the mood-related scores were seen among the participating shift workers (Table 4, Fig. 2). The intervention period resulted in a 31·1 % lower Fatigue-Inertia Score (P=0·003), a 15·3 % higher Vigour-Activity Score (P=0·041) and a 42·7 % lower Total Mood Disturbance Score (P=0·017) compared with the control period. Moreover, a trend for shift-working participants to feel less confused (P=0·070) and angry (P=0·084) was observed.
Scores of Fatigue-Inertia, Vigour-Activity and Total Mood Disturbance in shift workers after intervention (
) and control (
) period. Values are means, and standard deviations represented by vertical bars. Statistical differences are based on Students’ t test. Significantly different from the control period: * P<0·05; ** P<0·005.
Several differences in daily dietary intake were seen between the study periods in the group as a whole. The intake of fat (P=0·030) and PUFA (P=0·003) was lower, and the intake of carbohydrate (P=0·008), dietary fibre (P=0·031) and water (P<0·001) was greater in the intervention period than in the control period. No differences were observed in energy, SFA or protein intake (Table 3).
The intervention period resulted in a significantly higher intake of water compared with the control period among shift workers (Table 4). None of the other dietary components reached a significant difference between the periods, although the same tendencies were observed as in the whole study group.
Discussion
In this randomised, cross-over design study the effectiveness of increased availability of healthy cold meals, water and snacks at work on participants’ reaction time, mood and dietary intake was examined in a group of sixty healthcare staff. A subgroup analysis was made on the sixteen participating employees working shifts. The study found a beneficial effect on dietary intake for the group as a whole (Table 3), and a beneficial effect on several mood-related parameters for participating shift workers (Table 4, Fig. 2). However, no effect on reaction time was found in either the whole group or in the shift-work subgroup.
Effect on dietary intake
This study providing healthy foods and snacks, as well as bottled water, at work was an effective way of both improving employees’ dietary intake (e.g. a lower total fat and a higher fibre intake) and increasing their water intake on working days.
Similar effects were seen in a short-term 2-d cross-over study( Reference Lemaire, Wallace and Dinsmore 17 ) in physicians, which found that higher intake of nutrients and improved hydration could be reached by providing healthy meals and beverages during work. The authors also concluded that implementing a nutritional intervention in a healthcare system is feasible on a practical level. Sharma et al.( Reference Sharma, Paolicelli and Jyothi 25 ) conducted a cross-sectional analysis describing current policies and practices targeting the nutrition and physical environment in five large hospitals. They emphasised the need of environmental and policy-based strategies to successfully improve healthy nutrition and/or physical activity among employees, including making healthy snacks readily available (e.g. in vending machine and break rooms) especially for the night-shift workers who often do not have access to canteens( Reference Sharma, Paolicelli and Jyothi 25 ).
Torquati et al.( Reference Torquati, Pavey and Kolbe-Alexander 26 ) conducted a systematic review on the effectiveness of intervention studies promoting diet and physical activity in nurses. They found that dietary outcomes were generally positive, although studies in this area were very limited. The positive effect found in this study on food intake is supported by other ‘real-life’ intervention studies among other employee groups using environmental strategies for improving employees’ dietary intake – that is canteen takeaway( Reference Lassen 27 ), fruit schemes( Reference Alinia, Lassen and Krogsholm 28 ) and offering healthy canteen food( Reference Lassen, Beck and Leedo 19 , Reference Lassen, Thorsen and Sommer 29 ).
Reaction time
The lack of observed statistically significant effect of the intervention on reaction time is in discordance with the findings of a previous intervention study, in which the effect of increased availability of healthy meals was examined in physicians( Reference Lemaire, Wallace and Dinsmore 17 ). Results from this smaller and shorter term study suggested a significant improvement in reaction time when healthy meals were provided for 1 d. Although the study by Lemaire et al.( Reference Lemaire, Wallace and Dinsmore 17 ) also used a cross-over design, the study used a different method to assess reaction time than the present study. In a similar study in young males, in which all foods were provided for 3 weeks and reaction time was measured using the same method as in the present study, a significant positive effect of the diet on reaction time was seen( Reference Jakobsen, Kondrup and Zellner 24 ). However, this study investigated the effect of a high-protein diet, which was not the focus of the current study. Moreover, the results from the study by Jacobsen and co-authors might not be fully comparable to ours, as 88 % of the subjects in the current study were females, who have shown to have slightly longer reaction time compared with males( Reference Lassen, Lehmann and Andersen 30 ), and because our baseline reaction time was approximately 100 ms higher compared with the baseline reaction time in their study. One of the explanations for the somewhat surprising lack of effect on reaction time and mood could be the fact that the study population was relatively healthy at the beginning of the study. The intake of total fat, carbohydrates, protein and dietary fibre, except for SFA, all met Nordic Nutrition Recommendation 2012 reference ranges for intake( 20 ). Another explanation could be that a large part of our study population was made up of nurses who worked at outpatient clinics. The hospital outpatient clinics usually have regular opening hours and planned breaks, giving a good opportunity to take time off for both lunch and coffee breaks. In addition, 88 % of this study population comprised women, who are generally known to be more health conscious and have healthier eating habits than men( Reference Lassen, Lehmann and Andersen 30 ).
Mood-related scores
As far as we know, this study is the first to examine cognitive response in healthy subjects to an increased accessibility to cold meals, water and snacks, at work over a longer period of time. Although no significant effect was observed on the group as a whole, improved mood scores, Fatigue-Inertia, Vigour-Activity and Total Mood Disturbance, and a trend towards lower Anger-Hostility and Confusion-Bewilderment were observed in shift workers, who significantly improved their intake of water during the intervention period.
Our findings are partially supported by the results of a short-term study by Montain et al.( Reference Montain, Baker-Fulco and Niro 31 ), who found that providing healthy foods as eat-on-the-move rations for shift-working firefighters resulted in less depression and confusion ratings on the POMS scale.
Strengths and limitations
A major strength of our study was the cross-over design, which reduced the influence of confounding covariates, as participants served as their own controls. Furthermore, the study had a very low dropout rate (2 %). Finally, the study was performed in free-living environment of the participants, which increases the external validity of our findings. It suggests practical solutions to meal provision to more effectively target factors promoting employees’ health and performance in the long term.
Some limitations must also be addressed. First, the subjects were not asked to perform any intervention compliance check. Thus, for all outcome measures, the lack of intervention effects can in part be attributed to the level of compliance to the programme – that is the proportion of meals actually consumed. Second, power calculation was performed on the primary outcome reaction time for the whole group. A retrospective power calculation on reaction time for the subgroup of shift workers (n 16) showed a power of only 0·4 based on the study by Jakobsen et al. ( Reference Jakobsen, Kondrup and Zellner 24 ). Consequently, the results of the subgroup analysis should be interpreted with sample size as an acknowledged limitation. A retrospective power calculation on the mood outcome Fatigue-Inertia revealed that a difference of 2·2 would be detectable for the group as a whole based on a study by Montain et al.( Reference Montain, Baker-Fulco and Niro 31 ) (P=0·05, sd 4·2, power=0·8). Another limitation was that there was no washout period between the two periods. However, a washout period is often unnecessary if the second period is long enough to allow the carryover effect to diminish.
Participation was voluntary, giving rise to a possible self-selection bias. Employees participating in health promotion studies may be more health conscious than other employees, and therefore not representative of the general employee group. This was confirmed by a healthier dietary intake profile at run-in compared with the average Danish population. A final limitation of the study is that dietary intake findings, as well as mood-related outcomes, are based on self-reported information, which could create response bias.
The study has highlighted the need for putting the work meal and drink, especially at shift work, on the agenda. The results are promising, but more research is needed on the effectiveness, changes in both intake and cognitive function, and feasibility of intervention studies under free-living conditions.
Conclusion
This study providing healthy meals, snacks and water during working hours was an effective way of improving employees’ dietary intake during working days, which may have long-term health benefits. Moreover, in the shorter term, the intervention resulted in an increased intake of water and beneficial mood changes in shift workers. In the present study, no significant differences in reaction time was found. The novel finding of potential beneficial effects of water on mood-related parameters in shift workers requires additional investigation. More well-powered randomised controlled trials including a greater proportion of men working shifts are needed to confirm these findings.
Acknowledgements
The authors thank Katrine Rask and Annette Vedelspang for the practical help during the study period, Heidi Anker for processing the dietary data, Michael Allerup Nielsen for securing funding and support in completion of the study, Christian Bitz for scientific support and Tobias Wirenfeldt Klausen for statistical advice.
The study was funded by The Nordic Kitchen at the Copenhagen University Hospital Herlev-Gentofte. The Nordic Kitchen had no influence on the design of the study, the manuscript or its conclusions.
The authors’ responsibilities were as follows – E. L.: participated in the research design, conducted the research, analysed the data, wrote the manuscript and had primary responsibility for the final content; A. M. B.: designed the research; A. M. B., A. A. and A. D. L.: collaborated on writing the manuscript; A. D. L.: participated in the research design. All authors approve the final version of the manuscript.
The authors declare that there are no conflicts of interest.
