To make progress in linking personalized diet counselling process for CMR conditions to intermediate outputs, we need to know first what interventions are being offered and then if they are associated with relevant outputs, like short-term changes in diet quality. The literature was searched for any comparable studies and none were found that holistically addressed both the content or food aspects of diet counselling, as well as the behaviour change techniques being used. Within the context of implementation studies, it is also relevant to consider the feasibility and value of data collection methods. A major challenge is to document process in sufficient detail to detect important differences, while remaining feasible in the typical practice setting. Health behaviour counselling is a complex activity and researchers in past studies have been considering care process at many levels, from basic paper review of study methods in published papers to video review of observed clinical encounters. This analysis represents an attempt to use RD self-report as a possible approach to document key aspects of process that impact diet quality.
Use of diet quality as a short-term output is still relatively uncommon, but is increasing, as more researchers are publishing personalized nutrition studies. The value of such tools lies in assessing both baseline diet quality relative to the population as well as assessing change in groups after intervention. For example, the large European Food4Me study (n=1269 completers) tested the hypothesis that subjects given phenotype and genotype data would improve diet quality more in a 6-month internet counselling study, compared to control subjects (68). The 2010 US HEI was calculated, and all intervention groups combined demonstrated a 4-point mean change over 6 months, compared to a 2.3 point change in the control group, for a between group difference, adjusted for baseline of 1.27 points (95% CI, 0.30, 2.25, p=0.010). Baseline HEI values were 49.1 to 49.5, slightly lower than the mean of 55.4 (SD = 8.32) found in a German household study (69). The few studies using versions of the HEI in MetS have shown results similar to ours, in terms of increases in fruits and vegetables and reductions in sugar intake as examples (70, 71). More work on the development and measurement properties of diet quality tools to assess dietary change in counselling studies is warranted and may yield important insights in interpreting the results of different studies.
Documentation of encounter time and channel in diet counselling is common (72), but there is little agreement on best practices for number or types of contact. In Mitchell et al.’s review of clinical trials of dietetic consultations in primary care there was a range of 19 consultations (mean 5.6); and 25-600 minutes of contact time (73). In our study, the decision to fix the number of encounters was intended to ensure sufficient dose of intervention and potential for clinical indicators to change, for all participants. In considering our results with other diet studies, multiple contacts over time seem to be important for people to practice new skills and to help develop new food habits. Most changes in our study had occurred by 3 months. More work is needed to better define ranges for maximal effectiveness in most patients. In this regard, interventions that focus on weight loss vs changes in diet quality should be considered separately, as diets for weight loss involve restriction of caloric intake and increased hunger, suggesting a possible need for different types and length of encounter time, to maximize potential for adherence. Changes in diet quality persisted to 12 months in this study that did not have a strong focus on weight loss. Evidence for longer term maintenance of changes to diet quality are limited.
Multiple FBG were documented by the RDs throughout the study; on average >2 FBGs/person/contact. There was a strong emphasis on the Balanced meals and the ‘Healthy plate’ concept as measured by frequency of mention and use of these patient resources, in line with Canada’s Food Guide in current use during the study (51). Surprisingly, however, the more times the concept was documented, the more adverse the 3-month HEI-C, compared to when it was mentioned fewer times or not at all. In post-study interviews, a few RDs mentioned starting with the Balanced meals concept before moving on to specific food advice but indicated that if subjects did not make progress on eating more balanced meals, the concepts would be reviewed again (Stevens, undergraduate thesis). Further work is needed to unravel cause and effect, as well as the utility of the balanced meals concept, for counselling different patients. It was not surprising that counselling to decrease alcohol would be associated with lower 3-month HEI-C, although no specific intervention study on this topic was found.
Goals of consuming poultry more often than red meats, increasing plant protein, increasing fish and eating breakfast were less commonly promoted, yet were associated with potentially clinically relevant 6-point higher 3-month HEI-C (71, 74). From population studies, it is known that the standard deviation of HEI scores tend to be in the range of 12 points (similar in our study); therefore a 6-point difference in HEI-C would represent an effect size of ~0.5; which many people would consider a “medium” effect (74).
Looking over the rest of the FBG, it can be noted that goals defined in terms of specific food changes were associated with statistically increased HEI-C in the expected direction, but of lesser magnitude, while FBG that were focused on reducing nutrients (sodium, total fats, calories) or aimed at affecting the whole diet (glycemic index, mindful eating) were not associated with higher HEI-C, compared to not using that FBG. Our null results for mindful eating are consistent with results of a recent meta-analysis (75).
With respect to use of 16 specific BCT, Goal setting is an essential task in counselling and accounted for 15% of all BCTs recorded, yet more goal setting (>3times/3 months) was associated with a statistically poorer HEI-C at 3 months (-5 points), compared to fewer or no mentions, which did not differ from each other. Reasons for more goal setting being associated with poorer HEI-C may have reflected challenges in the joint goal setting process. Goal setting is well documented as an essential BCT; however adverse effects of more goal setting, over multiple visits has not been previously reported to our knowledge, as most reviews to date have only reviewed written descriptions of study methods (76). Some have suggested that more goal setting may be a symptom of issues with achieving realistic goals within the constraints of daily life, which vary tremendously from person to person. Further work to explain these associations is needed.
More Self-monitoring (> 4 mentions) was associated with improved HEI-C, and the HEI-C of the no mention and 1-4 mention groups did not differ from each other, which is consistent with the literature (32, 77–80) and RD experience (56). Focus on past success was associated with a statistically significant 4-point higher HEI-C at 3 months, as would be expected. Feedback on past performance, however, found that HEI-C scores were 4 points higher in the intermediate category, and overall association F test was relatively weak (p<0.037). Further work is needed to see if the two approaches differ in effectiveness. None of the rest of the BCTs were statistically associated with improved HEI-C.
Several meta-analyses have used the CALO-RE taxonomy to complete paper reviews of the BCT used in studies of adults where at least some studies related to dietary change outcomes (81–86). Results were mixed. Four of the six studies attempted to relate use of BCT to an outcome. Goodwin et al. did not find an association with mortality (81), nor did McDermott et al. find an association with physical activity or eating behaviours (83). In weight loss studies, Hartman et al. found that “comparing behaviour with others” was the only BCT associated with greater weight loss (84). We did not record this BCT. Lara et al. found that several BCT were associated with increased intake of fruits and vegetables, specifically “planning social support”, “goal setting”, “follow-up prompts” and “providing feedback on performance” (85). This was most comparable to our analysis as we used three of these BCT, excluding follow-up prompts as study contacts were fixed. One relevant study used the CALO-RE taxonomy to analyse BCT observed in video-recordings of initial consultations to promote exercise among older adults (87). “Providing information about where and when to perform the behavior” (86%) and “setting outcome goals” (82%) were used most often. Self-monitoring was rarely mentioned. In this study, a key issue may have been that it was an initial consultation
More recent relevant results have used the later Behaviour Change Technique Taxonomy version1, BCTTv1 (62), which greatly expanded the number of BCT from 40 in the CALO-RE taxonomy to 93 items. Alageel et al. conducted a relevant review of CVD prevention in primary care (88) that found highly heterogenous results and limited use of BCT. Cradock et al. reviewed diet and physical activity interventions in diabetes for changes in HbA1c and weight (89). Four of 46 BCT identified were associated with >0.3% reduction in HbA1c: 'instruction on how to perform a behaviour', 'behavioural practice/rehearsal', 'demonstration of the behaviour' and 'action planning' Samdal et al. reviewed physical activity and healthy eating in overweight and obese adults, but focused on assessing BCT used only with intervention groups and not control groups, and combining diet and exercise (80). Goal-setting and feedback on outcomes of performance were significant in meta-regression in the long-term. Technology driven diabetes prevention studies were reviewed by Van Rhoon et al. (90). The techniques of social support (unspecified), goal setting (outcome/behaviour), feedback on behaviour, and self-monitoring of outcome(s) of behaviour were identified in over 90% of interventions, effective for weight loss (3% at 6 months; ≥ 5% at 12 months).
Our analysis appears to be among the first to use self-report of BCT by practitioners to document process and to consider changes in BCT over the course of an intervention. A benefit of having practitioners document their BCT is that we are likely to be obtain more accurate and detailed information compared to a review of study descriptions as reported in published papers. A key challenge, however, is feasibility. We were aware of this from the outset and completed RD reviews to limit the number of BCT, used a training video and created drop-down menus for quicker data entry. The finding that only 2.0 BCT/person/contact to 3-months were documented, suggests substantial under-reporting was occurring. BCT have been defined as the smallest identifiable components that in themselves have the potential to change behaviour (63). Clearly, in any counselling or group program many BCT are being used. For example, MacPherson et al. used the BCT taxonomy v1 to document BCT they used in group classes for diabetes prevention and identified 30 BCT per session (91). Therefore, the BCT reported by the RDs were their perceptions of the important or noteworthy BCT in the session. It may be worthwhile in future work to only focus on a few BCT if self-report methods will be used. For example, we found two older analyses that were relevant. One study was found of coaching methods used in the Diabetes Prevention Program study, where coaches documented their coaching techniques (92). Lifestyle coaches used problem-solving with most participants and regularly reviewed self-monitoring skills. In the other study, participants in a long-term clinical trial (ADDITION-Plus) reported on their own use of eight BCT related to diet and physical activity at one year. “Goal setting”, “goal review” and “preparing for setbacks” was related to decreased % fat in diet, but not to physical activity (93). Given our experience in the challenges RDs had in documenting BCT and the current very long lists of BCT that have been developed, there is a need to find some middle ground for practical assessment of groups of BCT that can be used in future effectiveness studies of diet counselling (63). Further methodological work is needed to identify the BCT that capture differences in techniques that are feasible to measure and can be associated with outputs.
RDs are often educating on food skills, and this is explicit in the NCP terminology (72). This focus is well justified by evidence of poor food skills among many Canadians (94).
The CALO-RE taxonomy was not very explicit on this aspect, as also noted by Hollywood et al. in their review of food skills interventions (82). The RDs in this study made variable use of the option to add food skills to their documentation. We believe skills training was under-reported in this study and needs to be addressed in future studies. The more recent version of the 93 item Behavioural Change Taxonomy v.1 has addressed this gap, through 2 of 19 categories (‘Shaping Knowledge’ and ‘Repetition and Substitution’) (62).
Other work in health behaviour change may have application to future diet counselling studies. There has been increasing emphasis on the concept of fidelity in delivery of complex behavioural interventions, with use of video recording and analysis to document process and content. Two examples for RD counselling in cancer patients (95) and weight loss (96) were located, but they did not use the BCT taxonomies to describe process. Given our state of knowledge, we believe a strong focus on fidelity may be premature within personalized diet counselling. For example, use of motivational interviewing has been widely promoted in the profession and was very commonly reported in this study (10% of all BCTs), yet there was no association with change in diet quality. We believe a more directed approach is now needed to identify the key BCT, FBG and skills training that can be explicitly associated with relevant diet change.
Participant engagement, the degree that participants will use tools from the intervention in their lives, is a final area requiring more work. While we conducted an overall patient satisfaction study as part of the CHANGE project (41), and asked about confidence to maintain diet change, we did not assess patient engagement in detail. Walton et al. (97) completed an interesting systematic review of 66 studies for complex health interventions assessing both fidelity and engagement, only 6 of which included a diet intervention. Members of the group have published a five-step process for developing engagement measures for implementation studies that may be useful (98).
Strengths and Limitations
This is one of the first studies to attempt to link dietetic care process assessed by practitioners themselves to short-term counselling outcomes. We have succeeded in demonstrating significant associations for several FBG to changes in diet quality in a well conducted pre-post study. We were less successful in eliciting BCT. The self-report format was feasible for a primary care-based study, and efforts were made to ensure reliable documentation of counselling over the 4-year study. Use of a generic diet quality measure, the HEI-C, was a strength in that it was possible to quantify several promoted food intake changes in a single measure.
A self-report format using a check-all-that-apply list of possible FBG and BCT was employed as is common in clinical studies, but the wide variability in overall reporting suggests that RDs varied in the extent of recording. Further work to assess reliability and validity of self-report methods is now justified. For example, it is known that forced choice responses will elicit more responses, but both methods result in similar rank ordering (99). Our analysis approach was very basic, being limited to descriptive statistics and univariate associations and ignored the complexity and likely interactions among different elements of the counselling process.