Introduction

Poor nutrition among older people in hospital is well recognised in many countries, and a pooled analysis of data based on the Mini Nutritional Assessment tool recently reported an estimated prevalence of malnutrition of 39% among older inpatients in Belgium, Switzerland, Germany, Italy and Sweden (1). In England nutritional screening surveys using the Malnutrition Universal Screening Tool (MUST) similarly found a prevalence of malnutrition of 38% among patients in Care of the Elderly wards (2). Malnutrition often predates admission and is associated with poor outcomes of hospital care including increased mortality (3) and longer lengths of stay (4). Factors recognised widely to contribute to poor dietary intake include acute illness, co-morbidities, cognitive impairment, low mood and medication (5). The hospital environment has also been scrutinised, with reports from Australia and other countries of food being placed out of reach or going cold because of clinical care at mealtimes, and a lack of assistance from ward staff with eating (6). In the UK the Care Quality Commission reported that 12% of hospitals surveyed in 2012 were not meeting the minimum legal standards for mealtime care of older people (7).

Initiatives to address this issue include the use of coloured trays to identify patients at risk of poor nutrition, and protected mealtimes when non-urgent clinical care stops to allow patients and staff to focus on patients’ dietary intake. However the few studies of protected mealtimes, conducted in the UK and Australia, have typically not demonstrated an increase in patients’ dietary intake (8, 9). The few trials of additional mealtime assistance have shown mixed results. Increased dietary intake has been shown in a study employing dietetic assistants to help older women with hip fracture (10) and when using undergraduate nutrition students to target feeding assistance for older dysphagic patients (11). However among older general medical patients additional assistance from healthcare assistants has not been demonstrated to increase patients’ dietary intake (12), nor among patients with dementia (13).

The use of volunteers to help older people in hospital at mealtimes has attracted recent interest but little evaluation (14). An Australian study of 23 patients reported increased energy and protein intakes at two lunchtimes when volunteers were present, compared to two days when they were helped by ward staff as routine care (15) and an American study reported increased dietary intake among 34 patients helped by volunteers (16). Other small studies have reported volunteer activity (17) or anecdotal benefits (18) but did not measure dietary intake. Volunteers are likely to be increasingly important in an era when healthcare systems are generally limited in both financial resources and the ability to recruit sufficient nursing staff. In order to address this gap in the literature through a larger controlled study, the Southampton Mealtime Assistance Study aimed to evaluate the use of trained volunteer mealtime assistants to help older inpatients. The feasibility and acceptability of recruiting and training volunteer mealtime assistants in this study has previously been reported (19). This paper reports the impact of the introduction of mealtime assistants on the intake of energy and protein among older female medical patients.

Methods

Study design

This was a quasi-experimental study with a pre-post-test design. Patients on the intervention ward were compared during the observational (non-intervention) year and the intervention year with parallel comparison with patients on a control ward in the same department during both years. The period of data collection of one year pre and post intervention was chosen to account for issues related to seasonality such as patients’ medical conditions and changes in menu and food choice. During the intervention year trained volunteers provided mealtime assistance to patients in the intervention ward on weekday lunchtimes while those on the control ward received usual care from nursing staff. The protocol has been previously described (20). Data collection was repeated in both years on both wards: the primary outcome was patients’ dietary intake, collected as individual patient records over 24 hour periods. Demographic data was abstracted from the medical records and the observed levels of confusion among patients and mealtime assistance requirements were recorded.

Participants and setting

All participants aged 70 years and more admitted between February 2010 and January 2012 to two female acute medical wards within the same department at a university hospital in England were eligible for inclusion. The selection of female wards reflects the hospital policy of single sex wards, the resources available for the study and the recognition that the majority of older people are female. Only patients who were nil by mouth, tube fed, on the Liverpool Care Pathway for the dying, or being nursed in a side room were excluded. The patient case-mix, medical and nursing care was very similar on both wards. At the end of the observational year the hospital coincidentally introduced the red tray system to heighten awareness of patients at risk of malnutrition, and also reinforced protected mealtimes whereby clinical activity was suspended during mealtimes to allow staff to concentrate on helping patients to eat. These changes took place in both wards.

Intervention

Volunteers attended a half day training session with a speech and language therapist and dietitian on aspects of assisting older patients at mealtimes including practice in safe feeding strategies (19). They then assisted the nursing staff on the intervention ward during weekday lunchtimes during the intervention year. After cleaning patients’ hands and trays prior to lunch, specific assistance included encouragement to eat, opening up packages, cutting up food and feeding patients. The nursing staff indicated which patients needed feeding and the volunteers offered additional help to other patients, recording the help that was actually given at the end of each mealtime. The volunteers were not available at other mealtimes or weekends. In total 29 volunteers assisted patients on the intervention ward during 229 lunchtimes during the year; patients on the control ward received usual mealtime care from the nursing staff.

Data collection

Date of birth, height, weight, body mass index (BMI), Malnutrition Universal Screening Tool (‘MUST’) score (21), and primary diagnosis were abstracted from patients’ clinical records. The level of mealtime assistance given to each patient at lunchtime on the days when dietary intake was measured was observed and recorded as none; cutting up food and preparation of the tray eg opening packages; encouragement to eat; feeding patients; patient refused to eat . The presence on that day of any confusion (a deliberate term to capture delirium, dementia and medication side-effects) among the patients was rated by each patient’s nurse as none, mild, moderate or severe.

The primary outcome was the dietary intake of energy and protein by individual patients, which was recorded for a 24 hour period on 14 weekdays in the observational year (7 days on both the intervention and control wards approximately monthly between April and December) and 12 weekdays in the intervention year (6 days on each ward in a similar manner between April and November). The dietary assessment days included a range of weekdays and there was no systematic difference in the availability of volunteers or ward nurses on these days. Typically 16-20 patients were assessed in each day and were evaluated once during their admission. Each individual patient record included all food and drink consumed by the patient at three mealtimes, seven drinks rounds, sip feeds, and additional food and drinks provided by relatives or consumed at other times including during the night. Research nurses recorded all food items delivered to patients and collected the trays at the end of each meal service or drinks round. All leftover food items, drinks and sip feeds were weighed individually. To derive weights of food consumed, portion size information was obtained from the hospital caterer whose portion control measures ensured that each meal component was within 10% of the stated weight; for other foods, standard portion sizes or manufacturers’ information was used as described in the study protocol (20). The nutrient intake of individual patients was derived from the weight of food consumed and its nutritional content. The latter was obtained from the hospital caterer, UK food composition tables and manufacturers’ data.

A sample size of 100 subjects per ward in each year was estimated to be sufficient to detect a difference of 218 kcal/ day in dietary energy intake with 80% power at the 5% level, assuming a mean (SD) dietary intake of 1300 (550) kcal/day (20).

Statistical Analysis

Data were double entered, cleaned and analysed using Stata release 13 (StataCorp, Texas, 2010). Normality was assessed and variables transformed by taking the natural log or deriving Fisher-Yates z-scores where necessary. Summary statistics - mean (SD), median (IQR), number (%) - were used to describe the participants and their dietary intake. Differences in these between the wards and years were assessed using t-tests, χ2 tests, Fisher’s exact tests and Mann-Whitney tests as appropriate. Planned sub-group analysis of clinical factors used to identify those at high risk of poor nutrition (defined a priori as patients who either had a MUST score of 1 or more, were confused, were prescribed a soft diet or additional sip feeds) were conducted. Univariate and mutually adjusted generalised estimating equations (GEE) analyses were used to assess the association between these factors, as well as the level of mealtime assistance, on the dietary intake of energy and protein.

Results

In total 221 patients (intervention ward) and 186 patients (control ward) had their dietary intake assessed over a 24 hour period, representing 77% of the total ward population available on those days. In both years a small number of discharges or deaths led to incomplete records which were excluded, as were a few patients with long admissions who had been previously studied (Figure 1). The patients’ mean age was very similar in both wards and both years, and overall there was little difference in participants’ characteristics, MUST scores, or primary diagnosis between wards in each year (Table 1). Patient characteristics did not differ significantly between year one and year two on either ward (data not shown). In both years, the majority of patients had a BMI within the normal range and a low MUST score indicating a low risk of malnutrition; however confusion was a common problem on both wards.

Figure 1
figure 1

Participant recruitment

Full size image

As expected, there was little difference in the amount of mealtime assistance given when comparing the control and intervention wards during the observational year. Typically each bay of six patients had one nurse helping at each lunch time. However, during the intervention year, the level of mealtime assistance on both wards remained similar. On the control ward this continued to be provided by nursing staff; on the intervention ward two or three volunteers replaced nursing staff on weekday lunchtimes, enabling them to focus on patients with swallowing difficulties and other duties. This meant that for both wards there was no change in the overall level of feeding assistance provided from the observational to the intervention year.

Patients’ daily energy and protein intakes were very low, with median (IQR) values of 1039kcal (709, 1414) and 38.9g (26.6, 54.0) respectively (Table 2). There was no difference in patients’ daily or lunchtime energy and protein dietary intake between the intervention and control wards in either the observational year or the intervention year, nor when comparing intakes in the observational and intervention years following the introduction of volunteer feeding assistants on the intervention ward (data not shown). Lunch was the main meal on both wards with a median of 32% energy and 37% protein intake, while intake between meals (drinks, sip feeds and extra foods) accounted for 19% and 17% respectively of overall total daily energy and protein intake (data not shown). Planned subgroup analysis considered differences between wards according to patients’ risk of malnutrition, current confusion, use of soft diets and sip feeds. With the exception of a difference in energy and protein intakes, that were higher among confused patients on the control ward during the intervention year, there were no other differences in patients’ intakes of energy and protein observed in these subgroups, when comparing the wards in either the observational or intervention years.

Table 1
figure 2

Description of participants’ characteristics

Full size image

The intervention did not result in greater numbers of patients receiving assistance with feeding, and dietary intakes were comparable across wards. The data were therefore combined in order to explore influences on patients’ daily intake of energy and protein. The clinical factors identified a priori as likely to impact on dietary intake were a MUST score of 1 or more, confusion, being prescribed a soft diet or receipt of additional sip feeds and also the level of mealtime assistance (Table 3). Univariate analysis demonstrated that lower energy intakes were associated with increasing levels of mealtime assistance, with more severe confusion, and with having a soft diet. In the mutually adjusted model only the increasing level of mealtime assistance remained associated with lower intake. Protein intake was similarly negatively associated with increasing mealtime assistance and confusion, but demonstrated a positive association with use of sip feed supplements. The associations with need for feeding assistance and use of sip feeds were robust to mutual adjustment.

Table 2
figure 3

Dietary intake of participants

Full size image

Discussion

There is a pressing need to improve the nutrition of older hospital inpatients in a cost-effective manner. The Southampton Mealtime Assistance study is the first large controlled trial of the use of volunteers to help at mealtimes on an acute female medical ward for older people, where a large proportion of patients were confused and required help at mealtimes. The volunteers were able to help patients with feeding and the intervention was sustained over one year (19). However, the introduction of trained volunteers did not increase patients’ dietary intake.

The lack of difference in energy and protein intake when comparing wards or years of study may reflect the acuity of illness and frailty of the participants. By pooling the data we were able to explore the role of clinical factors as predictors of intake in this large population of older patients who were acutely unwell. The need for and level of mealtime assistance given was the factor most strongly associated with patients’ low intake of energy and protein (independently of other factors such as confusion and the need for a soft diet) and may be a useful simple clinical indicator of patients at risk of poor nutrition.

Table 3
figure 4

Association of energy and protein intake with participants’ characteristics

Full size image

Similar levels of feeding assistance were observed on the two wards in both the observational and the intervention years. In the intervention year this was achieved by nursing staff on the control ward, but by volunteers and nursing staff on the intervention ward. The trained volunteers were able to successfully provide mealtime assistance and help patients to eat to the same extent as the nurses, and interview data suggests that they provided a high quality service (22). Importantly, trained volunteer mealtime assistants have the potential to free up nursing time to complete more clinical tasks.

Our results are similar to a recent Australian study of mealtime volunteers which also found no significant increase in energy and protein intake (23). Indeed previous studies of mealtime assistance have reported mixed results (10-16). However, a recent meta-analysis by Tassone and colleagues has demonstrated a significant increase in mean dietary energy and protein intake among older inpatients with mealtime assistance (24), combining studies using volunteers and paid staff acting as mealtime assistants.

Most previous studies comparing additional mealtime assistance with usual care have not reported the levels of mealtime assistance achieved with either the assistance or usual care, and a difference in level of assistance provided may be one of the factors behind the mixed results reported. One study (of protected mealtimes and/or additional nursing feeding assistance to male and female medical inpatients, mean age 81 years) which did document the level of mealtime assistance reported an increase in mealtime assistance from 30% to 76-80% patients, with a reduction in other tasks by nursing staff during meals, although this study also found no significant difference in total energy or protein intake (25). The authors suggested that patients with feeding dependency or cognitive impairment gained most benefit from mealtime assistance, which may be consistent with the findings of this study.

There are a number of possible reasons why provision of additional assistance at lunchtimes did not result in an increase in dietary intake. It is possible that mealtime assistance was already adequate. However, the interview and focus group data previously reported confirmed that nurses of all grades of seniority perceived a lack of sufficient help with mealtimes in the observational year which improved with the introduction of volunteers on the intervention ward (22). The second possibility is that there was inadequate identification of individuals who were likely to benefit from the assistance. In this pragmatic study the nurses decided who needed assistance in both wards and in both years, and while this may be subjective it is representative of clinical practice. It is possible that there were misjudgments and/or failure to identify patients who would have benefitted from additional help at mealtimes – although the lack of change in assistance given between the observational and intervention years in the intervention ward makes this less likely. It is also possible that the type of mealtime assistance was not appropriate or sufficient to increase dietary intake. However the volunteers were trained and their competency to deliver mealtime assistance was checked. Importantly the volunteers had more time than ward nurses to help individual patients (22) so this seems unlikely. It is possible that there were not enough volunteers present at mealtimes but during the study the field notes indicated that three was the optimal number and when there were more volunteers they were underemployed. It may be that the volunteers replaced some nursing staff freeing them up for other tasks. The study demonstrated that the volunteers provided the same quantity of support as the nurses but it is possible that there were differences in the quality of mealtime support. However the interviews suggest that the volunteers, who had more time, delivered a high quality service (22). The two study wards were adjoining and it is possible that the Hawthorne effect operated during the study. However the study lasted for two years and there was little transfer of staff between wards; additionally the lack of change from the observational year to the intervention year implies that any effect operated during the both years. Finally in this study we had the power to detect a difference of 218 kcal which would have represented an increase of around 20% energy intake which would be unlikely at one meal. Although measurable changes in intake were not shown in this study this is not consistent with the interview data from staff and volunteers who reported that volunteers did achieve an increase in intake for individual patients but often only by a small amount (22).

The low measured intakes of energy and protein in this study are a concern, and were found despite provision of mealtime assistance, alongside local hospital initiatives to reinforce protected mealtimes, the use of coloured trays and routine screening for malnutrition. Other authors have reported similarly low energy and protein intake among older general medical inpatients (15, 24). These are unlikely to reflect habitual intakes, and are below estimated energy requirements of sick older people (26). We do not think that these low intakes can be explained by underreporting or by limitations in the methods we used to record intake. Research staff were present on the wards from 7am until 10pm and any additional foods consumed during the day or night were recorded. Although a 24-hour period may be insufficient to capture individual energy intake with accuracy, it is sufficient to characterise group intakes (27). The most likely explanation for such low intakes is that our study included acutely unwell patients as well as frail patients with dementia, many of whom were observed to eat very little despite assistance. These results suggest that patient factors in an acute medical setting have a profound effect on food consumption. Nausea, acute infections, higher BMI and cancer have been identified as predictors of poor nutrition among general medical inpatients (28). Other authors have also demonstrated that feeding dependency and delirium contribute to poor dietary intake among older in-patients, (29) and we have recently reported that poor appetite is common among older female inpatients(30).

The energy and protein content of food items offered is also important with regard to achieving adequate dietary intake. In our study protein intake was significantly higher among sip feed users, which may indicate a role for these supplements among patients requiring mealtime assistance. The use of fortified or energy dense foods for older inpatients also needs further evaluation as interviews with patients and relatives previously reported (22) have indicated that many patients found the portion sizes too large; smaller energy dense meals suitable for older people may be needed.

This study had many strengths. It assessed the dietary intake of a large population of older female in-patients, representing the majority of ward patients and, importantly, included those who were confused. Thus it should be representative of older female patients in acute medical wards in other settings. The detailed characterisation of the dietary intake of 407 acutely unwell older inpatients over a two year period produced individual food records which included supplements, drinks and snacks: few studies report weighing individual food components in such detail. However the study also had limitations. The participants were all female and the study took place in two wards in one hospital. The pre-post-test study design is a limitation as it cannot exclude a bias due to differences in the participants or external circumstances impacting the study, although participants’ characteristics were similar across both wards and in both study years. Further research is needed, and in particular randomised controlled trials, to assess the impact of volunteer mealtime assistants on the dietary intake of older men and patients in different settings.

Conclusion

This is the first large controlled trial of volunteer mealtime assistance. We have shown that trained volunteers were able to give mealtime assistance to many acutely unwell older female patients and deliver quality mealtime care on a large scale in an effective and sustainable manner, with the potential to release time for nursing staff to complete other clinical tasks. The study participants had a low median intake of energy and protein despite provision of mealtime assistance, reinforcement of protected mealtimes, the use of coloured trays and routine screening for malnutrition. This highlights the importance of patient factors associated with acute illness such as confusion, and a stratified approach including oral and parenteral nutritional supplementation may be required for some acutely unwell patients. The level of mealtime assistance required was the factor most strongly associated with patients’ poor intake of energy and protein and may be a useful simple indicator of patients at risk of poor nutrition.

Acknowledgements: The authors are grateful to Gemma Rood, Norma Diaper and Anne Marie Lewis (research nurses), Sanet de Wet (speech and language therapist), Kirsty Porter (dietitian), Kim Sutton (voluntary services manager), the volunteers, ward and catering staff, and the study participants.

Conflict of interest statement: None declared

Ethical standards: The study received full approval from the local research ethics committee and was conducted according to the Declaration of Helsinki and Good Clinical Practice requirements. It was registered with ClinicalTrials.gov NCTO1647204.