Study design

This was a multicenter study, where each center was randomized into either treatment or control: three intervention arms and one control group. We compared the results obtained in the intervention group with the control group, which received no intervention and continued to provide their standard care. To ensure homogeneous group representation, nursing homes were stratified based on the facility size and the availability of kitchen implements, with no exclusion criteria required. Inclusion criteria was to be a nursing home placed in Catalonia (Spain) and the randomized unit was the nursing home using a simple randomization method (all nursing homes were numbered and, using Stata, random numbers were assigned to each group).

The intervention program consisted of educational sessions. These were undertaken from 15–30 September 2019, with the attendance of up to two staff participants per center, often including the nursing home managers or other senior nursing home staff. Educational sessions focused on guiding nursing and caregiving staff, resident cooks, and the directors of nursing homes. Each session consisted of small groups of 15–25 staff participants from multiple nursing homes receiving a three-hour educational training session by several expert nutritionists from the Alicia Foundation (https://www.alicia.cat/en/), a research center devoted to technological innovation in cuisine, improving eating habits, and evaluating food and cooking heritage. After a one-hour theoretical lecture, two hours were allocated to practical lessons on aspects including meal planning, ingredients, food texture and presentation and other related issues (food temperature, colors, quality, and quantity). The sessions on presentation provided more emphasis on colors and how to present meals. The texture sessions reinforced how to provide the most suitable texture-modified diets and techniques of dysphagia assessment. It should be noted that the sessions were conducted at the Alicia Foundation, located 62 km from Barcelona city, which challenged the participation of some centers. To improve the involvement of centers in the intervention groups that could not attend physical sessions, video recordings of their specific sessions were sent to them. We also provided participating centers with hard copy supplementary materials from the Health Department to help nursing home staff improve their managerial decisions as supporting material after the sessions had taken place. A follow-up survey to assess the intervention impact was planned four months after the baseline survey, but because of the SARS-COV-2 outbreak was postponed until April 2021.

The control group received no material or educational sessions. Thus, all nursing homes allocated to the control group made their own decisions on how to produce and present meals as they had done previously. This trial was not an open-label trial, so none of the participants knew whether they were in the intervention or control group. We asked participants to avoid communicating the contents of the educational session and held each group session on a different date.

The study adhered to the Declaration of Helsinki and was approved by the Ethics Committee of Universitat Internacional de Catalunya (FCES-2018-01).

Data collection

The list of all nursing homes in Catalonia was obtained from the Public Health Department of the regional government of Catalonia (Generalitat de Catalonia) by sending a letter to the Department of Social Welfare and the Ministry of Health in Spain. Each center was contacted to invite them to participate, and those nursing home managers who agreed provided written informed consent.

A purpose-developed survey was administered online (a pdf version was available) between February and June 2019 to all participating centers. Facilities were classified based on the participants’ responses regarding facility size (number of residents in quartiles) and available kitchen tools (kitchen robots and mixers). The baseline survey contained 31 questions related to four main aspects, mostly by means of closed questions: (i) center characteristics, including the number of residents, the number of chefs and staff responsible for providing food and supplements, and the presence/absence of dietitian support; (ii) the kitchen size and availability of cooking tools; (iii) the availability of different daily menu options, the degree of food modification (categorized into high (own kitchen), moderate (both catering and preparation of garnishments or other dishes), or low (full catering)) and the staff’s prior participation in nutrition courses on texture-modified diets. Moreover, some characteristics of the residents were collected, such as the dependence level (based on the Department of Social Services criteria, the dependence degree was classified as I (mild), II (moderate), or III (high)); nutritional status, classified following the Mini Nutritional Assessment scale as normal nutritional status or at risk of malnutrition or malnutrition [17] and body mass index [BMI]) categories (underweight, normal weight, overweight and obesity) [18], need of help (full, partial) when feeding themselves, the number of meals taken per day at the nursing home (options: breakfast, lunch, snack, dinner) and the incidence of diseases among current residents during the previous thirty days (acute respiratory diseases or psychological stress; pressure ulcers; vascular ulcers; ulcers due to moisture; chronic diarrhea; constipation and vomiting). This survey was developed considering existing literature, personal interviews, and focus groups with participating nursing homes and day-care centers.

The follow-up survey included seven questions regarding the effects of the intervention. Questions were asked about the availability of chefs and dieticians, the staff responsible for making nutritional decisions, the degree of food handling, the availability of texture-modification tools, and the level of satisfaction with food textures and appearance. Closed questions were asked (yes/no). We used this information to compare the baseline levels and assess the intervention’s effect on decisions concerning food management. Managerial decisions around food could refer to: i) whether the right staff were responsible for providing food or supplements, ii) specifically whether there were in-house dietitians, and ii) whether there was a change in the number of implements available for texture modification. The interaction could impact firstly, by changing some or all of the three outcomes mentioned above; or secondly, between baseline and follow-up i) by improving/reducing the degree to which appropriate people led the administration of food and supplements; ii) whether the number of a new in-house dietitians increased/declined; or iii) whether the number of kitchen tools (robot, mixer or both) increased/declined. We considered the intervention led to an improvement (worsening) if the metrics mentioned above increased (declined) between baseline and follow-up.

Statistical analyses

Nursing home characteristics were analyzed descriptively at baseline, assessing whether there were statistically significant differences between the conditional means in the intervention and control groups (that is, controlling for the mean difference of all the variables simultaneously) using a logistic regression.

We performed an ordered probit regression based on the nature of the endogenous variables (worsened, no change, or improved managerial decisions concerning food-related issues) to estimate the effects attributable to the intervention, accounting for the characteristics at the baseline year. The following empirical model was used: (1) where = 1 if nursing home i during time t (t = 0 before the intervention and t = 1 after the intervention) experienced any improvement in the considered outcomes related to dietary decisions (including the availability of texture-modification tools and the dieticians’ role in the nursing homes). E represents a dummy variable that equals 1 for the treatment group; x_i represents the changes in the characteristics of the nursing home (facility size and available kitchen tools); α_i represents the unobserved fixed effects for nursing homes, and ε_i is the random noise, which is assumed to be independent of the explanatory variables. The main parameter of interest is δ—the effects of the nutrition workshops. Standard errors were corrected. Results were assessed on the intention-to-treat basis. Concerning the nursing home characteristics, although our interest lies in the impact of the treatment, we also included the following variables that might impact managerial decisions on nutrition-related choices and might also affect basal differential characteristics when randomizing: (i) variables identifying the percentage of users with the list of diseases mentioned in the “data collection” section above; (ii) change in the number of residents between the baseline and follow-up periods; (iii) previous percentage of residents who required full help to be fed; (iv) previous nutritional training received by the staff; (v) records of patients’ food intake; (vi) time passed between the two surveys; (vii) the percentage of residents affected by SARS-COV-2. These variables would help to correct possible biases due to systematic differences in nursing homes. Moreover, the predicted probability of falling into one of the arms (treatment or control) was used as an instrumental variable. We also computed the marginal effects for those nursing homes that participated in on-site workshops and those who received online training by including the distance from the nursing home to the workshop location as a second instrument, as distance may have conditioned participation. We measured the effect of each type of workshop by using simultaneous equations to jointly estimate all treatments. Statistical analyses were performed using STATA software, version 17.0 (STATA Corporation, College Station, TX, USA).

Limitations and implications

This study has some limitations that are worth considering. It was impossible to assess the nutrition status of residents and related problems due to pandemic restrictions, as well as to collect any other information at the residents’ level. The initial idea was to perform the study at the individual level, the resident, but after the COVID-19 outbreak, the only possibility to run the study, as agreed with the corresponding institutional body, was to perform it at the nursing home level. The promising result identified in the study was observed with the intervention despite the constraints imposed by the SARS-COV-2 pandemic, which had a huge impact on nursing homes in all countries, particularly in Spain [25, 26]. Based on official data [27], around 20,200 residents– 47–50% of Spanish nursing home residents–died during the initial outbreak of the SARS-COV-2 pandemic. Although the intervention was implemented in October 2019, the pandemic led us to postpone the follow-up survey until the end of April 2021. Therefore, in the statistical analysis, we included the time passed between baseline and follow-up as a way of identifying possible differences between facilities. Additionally, day-care centers were forbidden to operate during the first wave of the pandemic (March 2020 –May 2020), which prevented us from following up with the centers in that category that participated during the baseline year. This led to the removal of day-care centers from the study and the results reported in this article. Although the data collected on nursing homes and day-care centers are similar at baseline, adding post-intervention information from day-care centers could have affected the applicability of results. Nevertheless, delays caused by the pandemic allowed us to observe the effects of the intervention over a more extended period and in a particularly complicated situation. In this regard, the 60.1% response rate of the follow-up survey constitutes a success considering the situation. However, attrition issues still had an impact on the study, although inverse probability weights were used to account for this issue. The 24% increase in the availability of texture-modification cooking tools observed after controlling for the time difference between the assessments is promising and unexpected. The pandemic’s impact likely worsened the previously observed characteristics and resources available to nursing homes. This is because nursing homes worst affected by the pandemic would have had less flexibility when making decisions on aspects such as nutritional choices. There are other factors outside the scope of this study that could have been relevant to improving residents’ food intake or related to it. These include changes in residents’ nutritional status, which was only assessed and recorded during the baseline data collection. While this study assessed the availability of texture-modification tools, results on the improvement of food presentation were not discussed, although this variable might be measured with bias. Actually, for future studies, we may consider including a more objective measure to assess the effect of modified food textures, such as the International Dysphagia Diet Standardization Initiative (IDDSI). The IDDSI has a standard classification system for the classification of texture modified foods to address dysphagia [28]. Finally, we conducted a not open-label study clustering participants based on their location to avoid information crowding out and switched their initial assignment if the facility belonged to the same institution. These considerations should have helped to minimize the presence of bias.

Despite these limitations, this study is one of the largest assessing a nutritional education intervention targeting nursing home managers in Spain. The efficacy of this intervention encourages us in post-pandemic times to propose nutrition education as a new policy program of the health department to improve the service provided by nursing homes.

Those making managerial decisions can benefit from the current evidence reported in this article. The evidence can help decision-makers in nursing homes to improve food presentation and textures which should lead these facilities to reductions in malnutrition prevalence, avoidance of adverse events, and higher customer satisfaction, as the existing evidence has already demonstrated. The educational sessions assessed in the present study have demonstrated that increasing the availability of texture-modification tools is one of the mechanisms for better food presentation. Moreover, helping decision-makers with non-clinical issues, such as food presentation, encouraged them to increase the use of dietitians. Incorporating such health professionals in collaboration with nurses might lead to the formation of multidisciplinary teams who can provide facility residents with better assistance. Further studies that analyze this effect in the medium and long term are needed.