The validation of a questionnaire to assess barriers to enteral feeding in critically ill patients: a multicenter international survey

In many areas of healthcare there is a gap between what research evidence indicate ought to be done and what actually happens in clinical practice [1]. The recognition of this problem, together with a heightened focus on quality improvement and evidence-informed practice has stimulated interest in research examining the causes of this ‘knowledge-practice gap’ [2]. A growing body of literature supports the need to identify factors that limit or restrict implementation of best practices [2‐6], so that interventions can be selected to address these barriers and improve care delivery.

For instance, when we consider nutrition therapy in critically ill patients, on the one hand, several Clinical Practice Guidelines (CPGs) have been published summarizing evidence from over 200 randomized controlled trials (RCTs) [7‐12]; while on the other, observational studies of nutrition practice consistently report large variation in practices across Intensive Care Units (ICUs) [13‐16]. Overall, the provision of nutrition therapy is suboptimal, with patients, on average receiving less than 60% of their prescribed calories and protein [13].

To gain a better understanding of the reasons for this knowledge-practice gap in critical care nutrition, we conducted multiple case studies in 4 ICUs in Canada [17]. This qualitative analysis was guided by one of the most often cited theoretical frameworks regarding barriers to knowledge use, the knowledge-attitudes-behaviour framework, by Cabana et al.[6]. The analysis led to the development of an extended and revised framework which provided a comprehensive description of factors impeding adherence to critical care nutrition guidelines [18]. Although useful in illuminating potential barriers, this framework did not enable the identification and measurement of these barriers. The ability to assess and quantify barriers is necessary to be successful at tailoring interventions to overcome them and improve practice [5]. Consequently, we developed the ‘Barriers to Enterally Feeding Critically Ill Patients’ Questionnaire [19]. We focused on the provision of enteral nutrition (EN) (i.e. nutrition delivered via a tube placed into the gastrointestinal tract), rather than other aspects of nutrition therapy such as parenteral nutrition (PN) (i.e. intravenous feeding) or nutrient supplementation, because it is the preferred type of feeding [13], practice recommendations related to EN are uniformly endorsed across published guidelines [7‐11], and critical care providers generally agree with the recommendations [20].

In total, 55 ICUs were included in the analysis, and 1153 patients were accrued across these sites. Figure 1 shows how the study sample was determined. Tables 2 and 3 report the ICU and patient characteristics. The majority of ICUs were closed units (78%) in teaching hospitals (75%) located in Australia and New Zealand (40%) or North America (33%). Included patients had a mean age of 61 years (standard deviation (SD)) 17, were predominantly admitted with a medical condition (65%) and a mean APACHE II score of 22 (SD 8). Twenty three percent died within 60 days of their ICU admission.

The majority of patients received EN either alone (n = 819 (71%)) or in combination with PN (n = 189 (16.4%)). Sixty-four patients (5.6%) received PN alone and 81 patients did not receive any artificial nutrition therapy. The mean adequacy of calories from EN and total nutrition were 48% (SD 17) and 60% (SD 16) respectively (Table 4). Figure 2 illustrates the adequacy of calories from EN for all sites and by the 5 geographic regions across the 12 days of observation. A total of 1439 completed barriers questionnaires were included in the analysis. On average the response rate was 30% (range 6 to 62%), equating to a mean of 23 completed questionnaires per ICU (site range 1 to 65). The mean overall barrier score was 23 (SD 11). Table 4 describes the overall and subscale barriers scores by geographic region.

None of the models evaluating potential effect modification were significant at a p-value of <0.1. Table 5 reports the results of the bivariate analysis of the association between the individual covariates and adequacy of EN. Sex, patient admission type, patient admission diagnosis, proportion of nurse respondents working in the ICU for greater than 5 years, and geographic region were significant at p < 0.25 and were selected, together with hospital and ICU type, to be evaluated as potential confounders using the change-in-estimate criterion. Admission diagnosis was highly collinear with admission type, and as the latter contributed less degrees of freedom than the former, admission diagnosis was not considered further. The estimates changed by greater than 10% between the unadjusted and adjusted models for 4 of the evaluated variables, namely; geographic region (236%), hospital type (13%), patient type (17%), and proportion of respondents working in the ICU > 5 years (51%). Consequently all adjusted analyses controlled for evaluable days, APACHE II score, geographic region, hospital type, patient admission type, and proportion of respondents working in the ICU > 5 years.

Table 6 shows the results of the unadjusted and adjusted regression models of the association between overall and subscale barrier scores and adequacy of enteral and total nutrition. A significant inverse association was observed, indicating that a 10 point increase in overall barrier score has a negative impact on nutrition practice, resulting in a 3.5 (Standard error (SE) 1.3) and 4.9 (SE 1.3)% decrease in adequacy of calories from enteral and total nutrition respectively. Although a significant association was observed for each of the 5 subscale barrier scores and adequacy of total nutrition, the association was not significant for subscales 1 and 2 with the primary outcome of adequacy of calories from EN. The effect size observed in the sensitivity analysis excluding ICUs contributing less than 10 questionnaires (N = 49) was similar (-3.0 (SE1.3)% and -4.9 (SE 1.3)% for enteral and total nutrition adequacy respectively (p-values <0.05).

The results of our analysis are corroborated by the results of our recently completed prospective study evaluating the feasibility of a guideline implementation intervention tailored to overcome barriers to feeding critically ill patients (i.e. PERFormance Enhancement of the Canadian nutrition guidelines by a Tailored Implementation Strategy: The PERFECTIS Study) [27, 28]. In that study, which also utilized the Barriers to Enterally Feeding Critically Ill Patients questionnaire, we observed a 10 point decrease in overall barriers score and a 5% increase in total nutrition adequacy following implementation of the tailored intervention. This magnitude of change is equivalent to the change in estimate seen in our regression analysis providing further evidence that barriers are an important factor leading to poor adherence to guideline recommendations in clinical practice.

We would expect that the association between nurse reported barriers to enterally feeding patients and our primary outcome of adequacy of calories from EN to be stronger than with our secondary outcome of adequacy of calories from total nutrition (EN + PN + propofol), because the questionnaire focuses on barrier to the provision of EN and not PN. However, in our regression analyses we observed a 3.5% decrease in EN adequacy compared to 5% with total nutrition adequacy. Furthermore, we observed a significant relationship between all 5 subscales of the barriers questionnaire and total nutrition adequacy, but no association between subscales 1 and 2 with EN adequacy. Further study is required to confirm these observations, explore the reasons for them, and conclude if the association with nutrition adequacy differs by the type of barrier. This knowledge may lead to revisions to the barriers questionnaire and inform the design of interventions whereby barriers that have the greatest impact on nutrition adequacy are targeted.

Although the magnitude of our observed association was statistically significant, the clinical significance of a 3 to 5% change in nutrition adequacy is unclear. Given that on average patients in our study were prescribed 1800 Kcals, a 5% decrease in nutrition adequacy would be equivalent to providing 90 less kcals, which in layman’s terms is the same as a single glass of apple juice per day. We have previously demonstrated that an increase of 1000 kcals per day is associated with a 24% decrease in mortality in this patient group [31]. Consequently, when using this questionnaire, interventions need to target much larger changes in barriers score to ensure that the impact on nutrition outcomes is clinically relevant.

The ‘Barriers to Feeding Critically Ill Patients Questionnaire’ focused on modifiable barriers that were amenable to change at the local level; consequently, the small association observed may in part be because of the presence of other barriers hindering the provision of EN that were not part of the scope of the questionnaire such as ‘higher’ level barriers (e.g. type of health care system, clinical education) or other contextual factors (e.g. leadership support, best practice culture). In a previous analysis, we observed that hospital type (i.e. teaching vs. non-teaching), admission category of the patient (i.e. medical vs. surgical) and sex of the patient (i.e. female vs. male) were significant predictors of EN adequacy [32]; demonstrating that specific hospital, ICU, and patient characteristics can positively or negatively influence nutrition provision. Thus the success of interventions targeting local modifiable barriers may be compromised if these other factors that influence nutrition delivery are not considered.

This approach of analyzing local barriers (using a questionnaire) and tailoring solutions to the identified barriers is foundational to customized knowledge translation, although empirical data showing the success of such an approach is modest [4]. Change directed at documented barriers may not change practice fully, in part because, not all barriers may be identified. Furthermore, it is not yet clear whether customized facilitators of practice change, adapted to local barriers, are more effective than generic facilitators of practice change [4].

The geographic region in which the ICU was located was identified as an important confounding factor of the association between barriers and nutrition adequacy in our dataset. The number of ICUs in each region was small (i.e. 7 to 22), negating our ability to conduct subgroup analyses to better understand the nature of the confounding. Further study is required with more ICUs to confirm this observation. However, it is possible that this variable is a composite measure of other variables that may be associated with the presence of barriers and provision of nutrition such as the type of health care system, models of care delivery, staffing ratios, and education.

There are several limitations that should be considered when interpreting our results. First, participating ICUs were not a random sample of sites but rather a voluntary sample; consequently this sampling strategy may have introduced selection bias if participating sites have a greater interest in nutrition or desire to improve practice compared to the target population. Participating ICUs were predominantly closed units in academic centers, which are two factors that have been associated with higher performance [32]. This may limit the generalizability of our findings. There may have also been selection bias associated with the response rate of 30% for the barriers questionnaire if the perceptions of respondents differed from non-responding nurses. Second, the barriers questionnaires was distributed to critical care nurses, responses may differ by profession therefore the observed association needs to be confirmed amongst dieticians and physicians. Third, the barriers questionnaire was distributed at the same time as the chart audit of nutrition practice; however, we cannot be certain that the nurses who completed the questionnaire are the same as those who cared for the patients included in the study. However, respondents were asked to identify important barriers based on their general experience in the ICU and not with regard to a specific patient, therefore this discrepancy should not have biased the results. Fourth, as with any self-administered survey, responses to the barriers questionnaire reflect factors that the nurses’ perceive to be important barriers in their ICU, which may not be synonymous with ‘true’ barriers. Consequently, the averaged site-level responses can only approximate the true ICU average barrier score with measurement error, resulting in regression dilution. This may have attenuated our estimates of the association between nutrition adequacy and the true barriers score at the ICU. Fifth, the nutrition practice data were abstracted from the patients’ hospital chart; therefore the accuracy of these data depends on accurate chart documentation. Sixth, there is considerable controversy in the nutrition literature as to the optimum nutrition requirements during critical illness. In our study, the prescription of goal calories was determined at the local site by the dietician or physician and therefore their clinical judgment may have influenced the primary and secondary outcome of enteral and total nutrition adequacy. Finally, as in any observational study, there may be residual or unmeasured confounding not accounted for by the regression model.

In a large sample of international ICUs, we observed that barriers to enterally feeding critically ill patients (measured by a recently developed questionnaire) are inversely associated with nutrition adequacy (measured by a chart audit). Our results provide evidence to support the conceptual underpinnings of knowledge translation research that barriers impede adherence to guideline recommendations in clinical practice. Further research is required to evaluate whether the strength of the observed association differs by type of barrier, profession of the critical care provider, or geographic location of the hospital, and if identifying barriers using our questionnaire can inform interventions that optimize nutrition practice.