Making the Most of Mealtimes (M3): protocol of a multi-centre cross-sectional study of food intake and its determinants in older adults living in long term care homes

A significant barrier to LTC research is receiving consent from homes, staff and residents to participate in the data collection. The best way to ensure participation is to build a trusting relationship [83]. We recruited 30 of 32 homes purposively at the time of submitting our proposal for funding; each provincial investigator leading the project in their province conducted this recruitment based on their prior relationships with these sites. All of these initially recruited homes participated in the project. Confidentiality of home identity was also assured to promote feelings of trust [83] and a memorandum of understanding between the university and the individual LTC homes was completed. This agreement was reviewed and signed by home management in a face-to-face meeting with the researchers and/or the provincial coordinators in each province approximately two months prior to starting recruitment of participants in that home. Facility management, director of care (nursing), food service management and the clinical dietitian were involved in these initial meetings and all signed this agreement to ensure good communication about the study. A primary contact person for each home was determined and this person was involved in all communications that supported the data collection. In most LTC homes, this role was assigned to the director of food service, but in other homes, it was the clinical dietitian. Post data collection, a summary document described nutrition risk, risk for dysphagia, oral health, and any other concerns for each participant. This summary was provided to the home’s dietitian and primary contact person to support the resolution of these potential problems. As well, most homes were interested in their menu nutrient analysis and this information was provided to them within a month of the data collection. When the data collection was finished across the country, home representatives were invited to a stakeholder meeting, which included government representatives, decision makers/knowledge users and other local LTC homes to present site-specific anonymous data on key variables for the province. This helped to build awareness of the study, to start work with respect to improving provincial policy, and to begin intervention development. In separate individual face-to-face meetings, each home was provided with their specific results, benchmarked with other homes in their province. All homes were receptive to this engagement and expressed appreciation for being involved in the study and receiving their own data to support quality improvement initiatives. As a result of these relationship building efforts, many homes have expressed interest in being part of further research focused on developing and testing interventions to improve food intake of residents.

As the researchers were in the home for approximately 20 days in a month, a clear understanding of their purpose and the expectations for staff around resident data collection were needed. A variety of strategies were used to promote this communication and build positive relationships. Management discussed the project with staff on the study units, using a script provided by the researchers. A similar script was provided for communication books. The provincial coordinator was the first researcher to enter the unit and started with training the seconded home staff to determine resident eligibility. They were typically introduced to key staff members by the home contact person and procedures for reviewing resident charts were discussed. DEAP was conducted on the first or second day by the coordinator as a way of initiating data collection for the units. All researchers wore nametags and were trained to be as unobtrusive as possible. The provincial coordinator was on site the first day that the RAs entered the home; it was found to be beneficial if the RAs could be introduced to key staff members and shown around the home and the study units the day before they began mealtime observations. The researchers worked as a team to identify residents and ensure a complete data collection. However, staff cooperation was needed for help in initially identifying and finding participating residents as they were often not in their rooms and arm bracelets for identification were not typically used. Where staff was needed for resident data collection (e.g., risk factors for poor food intake as reported by care staff for the PG-SGA), care staff were asked when a convenient time would be for this brief data collection. Where more intensive time was required (e.g., interviewing care staff for completion of interRAI LTCF), the home provided an additional staff member for coverage and this cost was billed to the project. All of these activities helped to build a positive relationship with the staff and supported a quality data collection from residents.

Staff participation was promoted by providing back-fill for staffing on selected days when the staff PDC questionnaire was completed. Staff also appreciated the $5 gift certificate. As discussed, most of the staff members were hesitant to use the provided laptops for their data entry and hard copy versions of the questionnaire were used. They required additional assurance that their individual results would remain anonymous and in no way would home management know who participated. When results were presented to management, staffing classification of ‘nursing’ or ‘other’ was only used to promote anonymity of staff that was limited in numbers (e.g., recreation staff).

Homes were chosen to have a sufficient number of residents (~50) to ensure that 20 eligible residents could be recruited. In some sites, random selection of the unit to recruit resident and care staff participants was not possible due to the size of the home. A care staff member, typically a regulated nurse (RPN/LPN), was shown how to: a) develop the list of eligible residents from the randomly selected study units, b) randomly select residents and/or the alternative decision maker (ADM) to be approached using a random number table, c) use a script to speak with the resident/ADM, and d) note reasons for lack of interest in the study. This training took a significant amount of time on the part of the provincial coordinator and specialized forms, tables and written instructions were provided to the home to make it as error-free as possible. The staff member randomly recruited 40 participants, as we had determined from a pilot that approximately 1 in 2 residents/ADM will refuse participation. This took a significant amount of time; the home invoiced the research project for the hours required by the seconded home staff member who completed this work.

Many homes struggled with this first step of data collection and we did not get complete records on participation rate as a result. However this two-step process for attaining consent resulted in only minimal problems with withdrawal post recruitment. Based on records available from homes, 84% of eligible residents/ADM reached by home staff agreed to be contacted by researchers. Not all eligible residents were contacted by the researchers; if they reached the 20 patient quota per home, the remaining list of eligible participants who had agreed to be approached by researchers was not exhausted. As resident participation rates for Canadian LTC research ranges from 65–72% [83‐85], M3 had a successful participation rate which is attributed to the relationship building, training of home and research staff, as well as the non-invasive protocol. Participants were representative of the units where they lived. Specifically mean age (86.9 SD 7.81 vs. 87.0 SD 8.2 respectively), and proportions of males (male 32% v. 31.5% respectively) or those requiring an ADM for consent (74.6% vs. 72.8% respectively) did not differ between participants and eligible non-participants. Attrition due to death was minimized by recruiting and collecting all data within one-month. Ten eligible residents died between identification by the home staff (screening) and recruitment by the researchers, while six participants died during data collection. These latter participants were retained however, as they each had a minimum of two days of observed food intake, and thus provided key data for analysis. Where prudent, sensitivity analyses will be used to determine if these patients unduly affect interpretation of findings.

We opted for tools and scales that had a minimal burden for residents to promote a complete data collection and chose where possible tools that were familiar to homes and staff (e.g., interRAI LTCF). As a result, we achieved high completion rates for most resident-level measures: PG-SGA (99.8%), calf circumference (98%), STAND (69.7%; 398/517 not on thickened fluids), oral health exam (89.8%; 574/639, 559 (87.5%) with complete data on exam), food satisfaction (52.3%), and smell test (47.3%). The latter two were expected to be only completed in those residents with sufficient cognition to do so, which limited the sample for these measures. STAND was not completed for residents that consumed thickened fluids (10.6%), as we did not want to expose these residents to the risk of aspiration; the remainder of resident participants where STAND was not completed either refused participation or were unable to follow the simple commands required. The proportion that completed STAND demonstrated the importance of an a priori decision to create a composite variable to identify dysphagia risk, which would ensure 100% identification of this important covariate. Dysphagia risk was defined as: thickened fluid prescription or failed STAND or exhibited any coughing or choking during three observed meals.

As noted, data collection on food intake was rigorous with protocols for weighing and estimating food clearly outlined and followed by the RAs. There were some challenges in attaining non-consecutive days or timing. Specifically, where travel from the provincial research centre required an overnight stay, the nonconsecutive day requirement for meal observation could not always be adhered to. We were able to observe 6+ meals for almost all participants (99%) with only 6% of participants having one or more meals that were not observed as they were consumed outside of the home. With a minimum of two days of intake for all residents, and little missing data for other days, we consider this data sufficiently thorough to address the primary research questions. The poorest collection on food intake was the evening snack as estimated by care staff. Other than using reminders with evening staff as well as clear instructions and daily follow up, no further suggestions can be provided to support complete and accurate data collection when RAs are not in the home to observe food intake. However, as 80–90% of calories are consumed during meals in LTC [86], we anticipate minimal bias in our primary outcomes due to inaccuracy in collection of consumption of food in the evening (after supper) or overnight. Nutrient analysis protocols promoted rigor and consistency. If a home recipe was unavailable, the RAs were trained to use the following steps: 1) use another home’s recipe for a similar food item in their province; if this was unavailable, 2) use a recipe from another province; and if this was not available, 3) consult Food Processor™ standard recipes and other sources as required, checking ingredients with the home’s cooks and/or food service management.

RAs were trained to be as unobtrusive as possible. As care staff in the dining room was not aware of the specifics of data collection and only observed the RAs weighing residents’ food, we anticipate that care staff’s behaviours towards residents were unchanged. Multiple measures for mealtime scans and meal behaviour observations were averaged with standard deviations reviewed to determine any potential outliers.

A significant amount of time was spent by the lead provincial coordinator and a research analyst in cleaning and checking the data that was submitted on RedCAP™. Specifically, PG-SGA was checked by having the lead investigator review randomly selected participants from all provinces to ensure consistency in malnutrition classification based on the charted components of this measure, as this instrument does require clinical judgment. Some sites also found it challenging to report care staffing levels on the home survey. Checking was done in the data cleaning phase by confirming values with a home informant.

Pilot data were collected prior to the main study to confirm feasibility of all measures and the time necessary to complete the data collection in home. In each home, RAs were present for approximately 10–11 h per day for 15–20 days over a one-month period to collect residents’ food intake data. This was the most laborious and expensive portion of the data collection. They typically worked 3–4 days per week and had up to three, one-week breaks during the 1 year of data collection. Two provinces were able to complete data collection within ten months as it was the preference of the research staff to take fewer breaks. Two provinces did require overnight stays for research staff due to the commuting distance to the provincial research centre; the research budget accounted for these additional costs. RAs did have time between meals to conduct nutrient analysis of resident intake and menus; however, some ‘administrative’ days in each province were required when RAs were not in a LTC home to catch up on key aspects of data entry and cleaning. Cohesiveness of the provincial teams ensured data collection with minimal challenges, considering the scope and the breadth of this study.