A change in provision of food and drink was developed and piloted in one Norfolk County Council run care home, with substantial investment in staff time, décor and equipment. The changes in this first home were judged very successful, endorsed by both staff and residents, and there was a suggestion of a reduction in numbers of residents' falls (30% fewer falls in the 3 months after the change compared with the 3 months before), reduction in falls needing assessment at the local accident and emergency department, better physical and mental state for several residents who had struggled to eat enough previously, increased interest in food, more socialising during meal times, and more relatives eating meals at the home.
For this reason it was decided to explore how we might address the effects of such changes more rigorously. We aimed to develop and assess a pragmatic methodology based on planned service development and maximising use of routine data collection while employing a controlled design and some additional measured outcomes. We set out to trial these changes in food and drink provision more formally in 3 further homes. As the 3 homes had already been chosen, the study design incorporated an additional 3 matched homes (matched by resident population and home size) as controls. We operated a pragmatic assessment using a mixture of pre-change and post-change data collected through questionnaires, interviews, routine data collected by the homes and blood tests. One intervention home and the matched control home specialised in dementia care as care homes in Norfolk were on the cusp of becoming mixed homes with some residents with dementia, enabling the appropriateness of changes across populations with differing levels of cognition to be addressed. Ethical approval for the study was obtained from the University of East Anglia, Faculty of Health, Ethics Committee. The trial was registered as ISRCTN86057119 (see http://www.controlled-trials.com/ISRCTN86057119
For a pictorial description of the study chronological flow see figure . During the spring and summer of 2006, a few weeks before the change in food and drink was due to take place in the intervention home, before baseline data collection or requests that residents take part in the study, residents, relatives and staff in each intervention home and its paired control were informed about the changes at open meetings (relatives were sent written invitations). All were told that the aim of the changes was to improve residents experience of eating and drinking in their homes, that there was a suggestion that this may impact positively on residents health (but that this was not clear) and that the changes were going to occur in some homes imminently, and in others a year later (with specific information provided on the home that was being addressed). There was concern among relatives that the changes were going to result in higher charges or were a cover for reduced meal quality or reduced care, and these concerns were directly addressed by Norfolk County Council managers at each home.
Study time flow. Study flow for intervention and control homes.
All residents, staff and relatives in control and intervention homes were given questionnaires (relatives were sent them in the post with stamped addressed envelopes, residents and staff were provided a box in which to post their completed forms) about current food and drink. During this same period residents were informed about the study and asked whether they would like to participate in any or all of the following: allowing researchers access to their routinely collected data (from a variety of sources, collectively referred to as 'care notes'); participating in an interview (to collect demographic data, mid-upper arm circumference, grip strength, hydration status, depression, anxiety and cognitive status); and/or providing a fasting blood sample.
Where interested, residents were asked to provide written informed consent. Home managers were approached to check whether any residents would be unable to provide informed consent due to impaired cognition - for such residents relatives were asked for informed assent (blood tests were not requested of such residents or their relatives). Where a relative provided informed assent an interview was only undertaken where the resident appeared happy and relaxed during the process - where this was not the case the interview was not attempted or immediately terminated. Ethically this system provided a double check from relatives that participants were likely to have approved of participation in the study when at full cognitive ability, and also appeared happy to participate on a moment-by-moment basis. As some residents were forgetful the study, its purpose and process were re-described at each meeting to every participant and all were offered chances to participate or not at each planned contact. While most residents who had provided consent were happy to participate in the interviews, several declined blood tests on the mornings that these were scheduled, due to feeling tired or unwell.
Care note (routine) data were collected in the intervention and control home pair for the year preceding the change at the intervention home (first year, much of this time prior to enrolment in the study), and for the year following the changeover (second year), see figure . This included whether the resident had fallen, number of falls, medications used, mobility, weights, height, major illnesses, contact with health care professionals and hospital admissions. Interviews were held just before the change over (first year), and one year later (second year).
Interviews included assessment of mid-upper arm circumference,[21
] grip strength (measured by a battery powered Takei TKK5401 Grip D LED-display dynamometer, best of 3 measures on each side with arm held vertically downward at side of the chair), assessment of dehydration (visual assessment of the tongue by a trained district nurse), cognitive functioning (Mini-Mental State Exam [23
], MMSE with scores from 0 to 30, higher scores representing better cognitive function), depression and anxiety (Hospital Anxiety and Depression Score [24
], HADS, scored 0 to 21 independently for depression and anxiety, with a higher score indicating greater anxiety or depression), previous occupation, and school leaving age. For those who gave informed consent a fasting blood test was taken before breakfast just before the change over and one year later. Blood was tested immediately for haemoglobin levels, and further samples stored at -80°C so that lipids [25
] from both years were analysed in a single batch.
Anonymous questionnaires including ten questions on enjoyment of food by residents, variety, sufficiency of food and drink, preferred dining area, perceived friendliness of the home and the frequency of visitors sharing meals and drinks were distributed to residents, relatives, and staff before the intervention and a year afterwards. Enjoyment of food was assessed by rating a question 'How much do you enjoy eating here?' from 1 ('I hate eating here') to 5 ('I love eating here'). Questions to relatives and staff asked about residents levels of enjoyment, rather than their own.
The primary outcome for the intervention study was number of falls, as there was a suggested decline in falls in the pilot home. Satisfaction with meals, body weight, dehydration, cognitive functioning, depression, haemoglobin and cholesterol levels were seen as important secondary outcomes.
The changes in food and drink provision, implemented between July and December 2006 in the three intervention homes and detailed in table , were intended to improve comfort during meals (making eating more like going to a restaurant than eating in a canteen), increase the level of choice available at meal times, making eating with others a pleasurable and more sociable experience and encourage fading appetites. They were also intended to widen the availability of drinks and snacks (to visitors as well as residents), encourage greater independence on the part of residents in choosing and obtaining their own snacks, and generally reduce the feeling of institutionalisation.
Description of food and drink provision, and changes
Data from care notes were limited by missing data, as many participants were not present for the whole of the first and second years. To maximise the available data it was decided to analyse each individual's fall rate as falls per month (counting only months for which the participant was living at the home). Poisson regression was used to model the relative rate of falls per person per month after the change for the intervention vs. the control homes, adjusting for the rate of falls before the change, age, use of psychotropic drugs,[27
] and stratified cluster design using Stata v.10.0 software (each stratum containing a matched pair of homes). Design-adjusted Wald tests were used for calculation of 95% confidence intervals (CI) and P-values for the rate ratios. One resident was omitted from falls analysis as he fell 57 times over nine months, this was an extreme value.
Weight data were more limited as, in addition to censored stays, many homes did not weigh participants during the initial months of the first year despite this being considered good practice, and data were too sparse to compare the change in weight slope over the first year with the second year slope as initially envisaged. For this reason, we calculated, for each individual, the difference between mean weight in year 1 and mean weight in year 2. Adjustment for pair of homes was made through pooling by meta-analysis of weighted mean differences using random effects analysis with each of the 3 pairs entered as an individual study (using RevMan 4.2 software [29
Interview data were obtained from fewer participants than care note data. Numbers of participants dehydrated at the second interview were used to calculate a relative risk of dehydration for each pair of homes, and pooled using meta-analysis and random effects methodology on RevMan 4.2 software. MMSE score in year 1 was subtracted from that in year 2 for each participant, and combined data for intervention and control participants compared using the Mann-Whitney test. Depression score in year 1 was subtracted from that in year 2 for each participant and combined data for intervention and control participants compared using the t-test (both analyses without adjustment).
Blood tests were carried out on the smallest group of participants, and only included participants able to give their own informed consent (so that residents of the two dementia care homes were not represented). Only residents with valid samples in both years were included in the analysis. Comparison of mean changes in haemoglobin and total cholesterol levels used a linear regression model (SPSS 14), adjusting for age and the effect of being in a particular pair of homes.
Residents' satisfaction with meals was assessed in questionnaires to residents. As questionnaires were anonymous mean scores for each home in year 1 were subtracted from mean scores at year 2 for that home, and the difference between intervention and control homes analysed using the independent t-test.