The feasibility of the study was confirmed with 207 persons assessed for eligibility between October, 2008 and January, 2011. Forty participants were randomized into the study and 34 participants are included in the analyses—16 assigned to Usual care and 18 assigned to Intervention.
The average age of participants was 81.4 ± 8.2 (SD) and ranged from 65 to 97 years old. There were 6 males and 28 females, of which equal numbers were African and European Americans. Close to two-thirds of participants (64.7%) were not consuming enough calories to maintain their current body weight and 15.2% were underweight (having a BMI < 18.5). There were no differences observed at baseline between the Usual Care and Intervention groups. The randomization schedule was successful in balancing for both gender and BMI.
presents the behavioral nutrition goals that were selected by participants in collaboration with the RD. Participants most often selected goals that involved changing the frequency, amount, and type of foods and beverages consumed with the intent to increase overall caloric intake.
Behavioral Nutritional Intervention Self-Management Goals
presents pre and post data relevant for assessing primary outcome measures. An ANCOVA revealed no main effects of the intervention for either caloric intake, F(2, 32)= .038, p=.847 or weight, F(2. 32)=<.000, p=.996. The Usual Care group reported higher caloric intake compared with the Intervention group at baseline and post-treatment, and did not increase their caloric intake as much as the intervention group did from baseline to post-treatment. Similarly, chi-square analysis revealed no effects of the intervention for either caloric intake or weight post-treatment. The percentage of participants who were meeting caloric needs to maintain body weight did not differ significantly between groups (43.8% for Usual Care versus 27.8% for the Intervention group) (χ2=.946, df=1, p=.331); and the percentage of participants who were weight stable did not differ between the groups (60.9% for the Usual Care group versus 55.6% for the Intervention group) (χ2=.066, df=1, p=.797). Of note, in sensitivity analyses we adjusted for multiple variables (including ethnicity, living arrangement, etc.) and none of these made any difference in the results.
Pre- and Post- Primary Outcome Measures
It is unclear if the null findings are a result of reduced study power as a consequence of being unable to achieve our desired sample size or the ineffectiveness of the intervention. Regardless, the findings of our study are important because they reveal that studies such as the one we conducted with the population we targeted are feasible, but not without substantial obstacles and with limited impact on the primary outcomes of caloric intake and/or weight. This discussion focuses on the limitations of the study with insights offered into how future studies might go about things differently.
First, with respect to lower than expected recruitment, we met only approximately 50% of our targeted enrollment over an approximately two-year grant period. Our recruitment estimates were based on previous work conducted by our research team with the same population. The previous study, however, was an observational study that did not involve an intervention. Our experiences here are not unique; Sahyoun and her colleagues reached exactly 50% of their targeted enrollment in the Community Connections Demonstration Project, an intervention study supported by the Administration on Aging that recruited from a similar population and from comparable sources.40, 41
There are multiple reasons why recruitment may have been less than anticipated. It is possible that recruitment for intervention studies requires more time than recruitment for observational studies. Because greater involvement may be required of potential participants in intervention studies, older adults may be more reluctant to participate because of the perceived additional effort. Lending support to this speculation, we note that Villareal and his colleagues conducted a randomized controlled intervention of weight loss in older adults and recruited 93 participants over an approximately four- year study period that overlapped with the timeframe of our study.42
Their rate of accrual was nearly identical to that observed in our study. Had we had similar resources and time, it is likely that we would have met recruitment goals.
Additionally, we earlier reported on the significant difficulty we encountered in receiving referrals from home health nurses, discharge planners, social workers, and case managers for the study.43
Reasons why and potential solutions are described in our paper, as well as in those by Sahyoun, et al.,40, 41
We additionally encountered patient resistance to enrollment in the study because of its’ ultimate goal of weight gain or maintenance during the recovery period. In previous work with older cancer patients, we found that patients interpreted weight loss as a positive outcome of the cancer and engaged in deliberate efforts to keep the weight off.44
The same was observed in this study. In future work, we will not market such a study as one to either increase or maintain caloric intake or weight, but instead one to improve nutritional intake or energy intake.
Nonetheless, we do not believe our inability to detect an effect for the intervention is entirely a consequence of inadequate power. Roughly 60% of persons in both groups either maintained or gained weight, and 44% of the control group and 28% of the intervention group were consuming enough calories post-intervention to maintain their baseline weight. There is considerable variability for all data points and no patterns observed for either group; and, in fact, the Usual Care group demonstrated higher and better caloric intake at pre- and post-treatment compared with the intervention group and lower BMIs at baseline—though, no statistically significant differences were observed. Future work might better target under or overweight participants or those who are undereating at baseline for inclusion in order to demonstrate effectiveness in a more homogenous sample.
What else might additionally account for the null findings? The study was meticulously designed with close attention paid to detail and the intervention was delivered in the home by highly motivated RDs, all with advanced degrees and all trained in behavioral self-management techniques. Furthermore, participants selected their own behavioral goals to target for change (e.g., from a range of options including: eating with family and friends, moderating therapeutic diets, participating in home-delivered meals program, etc.) and were supported in reaching goals in collaboration with the RD interventionist. It may well be that the behavioral goals identified by participants were not ones that either 1) would have the greatest impact on the study objectives or 2) were easy to implement. In fact, as illustrated in , the goals participants overwhelmingly selected were those associated with increasing the frequency, amount, and type of foods eaten.
Of note is that our previous work revealed that eating with others and having caregiver support were significantly associated with increased caloric intake among a similar group of older adults receiving Medicare home health services.2, 38
In this study, however, participants did not want to burden family and friends by seeking their involvement in activities surrounding food and meals. In other work with the same population, we also found that food choices were motivated primarily by sensory appeal (i.e., tastes good), convenience (i.e., is easy to prepare, simple to cook, etc.), and price.45
It may have been the case that implementing dietary changes, even those that were self-chosen, were too taxing for this group who may have been experiencing competing demands of dealing with medical issues associated with their recovery. Future work in which we engage to improve nutritional well-being in this population will focus specifically on either soliciting caregiver support or eliminating all but minimal effort on the part of participants (e.g., through provision of already prepared meals).
It is also the case that our inclusion criteria included history of weight loss or currently not consuming enough calories to maintain current body weight and our sample includes underweight, as well as normal and overweight/obese participants, any of whom may have experienced weight loss. These may have contributed to making differences difficult to detect. Although, our randomization scheme was stratified according to baseline BMI, and our analyses controlled for baseline caloric intake. Additionally, we note that other goals (e.g., increasing dietary protein, improving quality of life, reducing hospital readmissions) might be important endpoints to evaluate, as well.
While our work was undertaken in a Southern US location and our sample size was small, we have no reason to believe that our findings are not generalizable to older adults who are homebound and recovering from an illness in other places. Because of problems with recruitment, though, future work might benefit from multi-site participation.