In this quasi-experimental study, we introduced vegan menus and educational material on sustainable diets in a cafeteria of a large German hospital over a period of three months. We compared the changes in adherence to the EAT-Lancet PHDI and in GHGE between hospital employees consuming their meals in the intervention and the control cafeterias. Individuals with self-assigned plant-based dietary practices showed higher PHDI score points and lower GHGE than those consuming mixed diets. While the intervention effects using a DID analysis were not significant, the PHDI tended to increase stronger in the intervention group than in the control group. A similar trend was seen for GHGE. Individuals consuming the vegan menu more frequently showed greater, but not significant, increases in PHDI than those never or rarely consuming the vegan menu.

The mean PHDI of 25 points in this study was about 7 points higher than the mean PHDI in a population-based cohort study with 43–73 year-olds in Sweden; which applied the same approach for PHDI construction [21]. The mean GHGE was 3·3 kg CO2eq./d, comparable to Greenpeace’s estimation for consumers of meat 3–4 times a week [29]. However, in order to stay within planetary boundaries, our participants would have to achieve full points in all PHDI red meat and dairy categories and halve their GHGE [6].

The correlation between PHDI and GHGE in our study is imperfect; participants with high GHGE can attain high score points on the PHDI. This is because both outcomes operationalize different aspects of sustainability. PHDI combines healthfulness with adherence to six planetary boundaries in all relevant food categories, while GHGE characterize only one of those boundaries. The PHDI scoring system rewards high intakes of recommended foods and punishes high intakes of unrecommended foods. Further, the system is based on discrete steps from 0 to 3 points, and only accommodates changes within defined cut-offs for each food group. Reducing beef consumption from three times to once a week does not increase the score, as a points change is only awarded for consuming beef less than once a week. GHGE, on the other hand, is a continuous variable and is very sensitive to dietary changes regarding food groups with high GHGE, emphasizing the effects of diet on climate change.

In this sample of 190 hospital employees, we saw trends for improved adherence to the PHDI and reductions in GHGE after the 3-months intervention period, but no significant effects. Descriptive analyses showed a slightly higher increase in PHDI among participants with self-assigned vegetarian diet than individuals on vegan, pescatarian, flexitarian, and mixed diets. Participants with high PHDI at baseline might have had lower capacity to improve than individuals with low baseline values.

The significant increase in consumption of legumes suggests beneficial effects attributable to our intervention. Since there were no significant differences in the consumption of dairy and meat products, we infer that participants may have compensated for reduced consumption in these categories at home or that reported intakes for these categories were imprecise. This may be due to the detailed breakdown of individual components within dairy and meat categories (dairy: 10, meat: 13) compared to the legume category (4 components), making it easier for participants to lose track of total intake. Ultimately, our GHGE calculation highlighted that an increase in legume consumption without a concurrent reduction in meat and dairy production does not result in decreased emissions.

The study sample size was small as it was not calculated for the outcomes of this paper and the proportion of individuals lost to follow-up was large (72%). Even though the numbers were similar between groups, arguing against selection bias, we run into the possibility of type II error with insufficient power to detect an effect.

Furthermore, previous facility-based diet interventions (FBDI) have had longer durations (6 months to 2 years) or included individual counselling [18]. We did not consider counselling to be an appropriate means for the present intervention programme due to its high demand for human resources. Nevertheless, it appears plausible that participants with high mitigation capacity, namely consumers of vegetarian, flexitarian, and mixed diets will develop stronger PHDI adherence over a longer intervention period. In our intervention group, 18–33% of flexitarians and mixed diet consumers felt motivated by the cafeteria menu to adopt more sustainable dietary practices at home. However, according to the transtheoretical model multiple months to years are required to transition from the stage of contemplation into the stage of action [11].

Information materials and practical tips, as provided in the second component of our intervention, are valuable resources for promoting the transition to action because willingness and capacity to change are associated with knowledge [14]. To transfer to more sustainable diets, consumers need to know that the percentages of animal and plant-based foods in a diet are the most important determinants of sustainable diets. In particular, due to the short duration of our intervention, our educational component did not unfold its full potential. Future studies with longer durations should invest in a more strategic implementation of this component.

Finally, we speculate that the intervention programme was not fundamental enough. In fact, we still offered meat-and-fish menus in the intervention cafeteria. Consequently, participants in the intervention group may not have been subjected to large dietary changes. The effect of the intervention in the group who never consumed the vegan menu was equivalent to control conditions. Indeed, we found higher increases in mean PHDI at higher consumption levels of the vegan menu. To maximise the intervention effects, future studies should concentrate on actively engaging participants who do not initially choose vegan options.

To date, intervention studies targeting vegan or sustainable consumption in a collective meal context have been scarce, often lack a robust study design, have focused on labelling, prompting, and education [30], and have not increased the availability of sustainable menus. While a recent study reported a 45% increase in consumption of vegetarian dishes after a FBDI for more plant-based diets, there have been no studies reporting effects on individual diets using a priori indices [31]. Also, previous studies lack a common definition of sustainable diets and measure diverse outcomes [32]. This is the first report on the effects of an FBDI on the sustainability of consumers’ diets measured by the PHDI and GHGE.

Using semi-quantitative FFQ data allowed us to rank participants according to their adherence to an a priori index for sustainable diets. Of the currently available diet indices, the Stubbendorff index shows the best performance when estimating health and environmental impacts of diets [33]. We acknowledge that self-reported dietary data from FFQs can be subject to recall bias and underreporting. In addition, this FFQ has not been validated yet. Still, this does not affect the results of our DID analysis, given that all participants carry the same measurement error.

Food choices are influenced by many different factors in addition to the food environment [8]. We have controlled for confounders, observed and unobserved, by means of the DID analysis. Yet, in contrast to randomized controlled trials, DID analyses rely on the common trends assumption [25]. While we could not statistically validate this assumption in our study, conceptual arguments, similar distributions of covariates and between-group distributions of the outcome variables at baseline favour this assumption.

Lastly, there was the possibility of spill-over effects through switching between cafeterias and exchanging information between employees. This might have diluted the actual intervention effects, as could be seen in ITT and LATE. Also, ITT analysis and LATE provided stronger estimates for outcomes. This could be due to a high proportion of those participants already consuming sustainable diets switching to the intervention group.

Future research in this area should focus on the sustained effects of FBDIs on both health and environmental parameters. When performing short-term FBDIs, we propose limiting research to an accurate estimation of changes in food categories with high environmental impact. We believe that the PHDI is only a sensitive evaluation tool for evaluating FBDIs affecting all daily meals and with long durations because quite radical diet changes in multiple food categories are necessary to substantially increase PHDI. A separate analysis of changes in high-impact food categories is indispensable. To increase the accuracy of food reporting, we propose interviewer administered FFQs and a re-evaluation question on the total intake of each food category at the end of questioning.

While the promotion of plant-based and sustainable diets is an emerging field, it can draw on long-standing experiences from related fields, such as weight-loss and healthy diet interventions [30]. Future studies can use our intervention design as a starting point. However, to animate consumers not initially adopting sustainable food options, interventions should have longer durations and be progressively intensified. This could be accomplished by limiting unsustainable food options, increasing vegan options for all meals and vigorously escalating the educational component. Future studies could also make further investigations into differences in effects depending on gender. Including multiple study sites could provide a larger sample size and greater diversity, enhancing the generalizability of findings.