An Apple A Day...



Source: http://www.wrensoft.com/

Keeps the doctor away, right? I trust you are familiar with this old saying, though more appropriate advice might be “Eat plenty of fruits and vegetables and you will be less likely to get sick.” But let’s just pretend that originators of this adage were convinced that eating apples alone would lead to fewer doctors’ visits. If we find that eating apples, while making no other changes, decreases healthcare utilization, what would this mean for the healthcare system in general? Health care costs are increasing at exorbitant rates. We are facing a shortage of primary care physicians. As our population ages, the demand for physicians will only increase. There are no quick solutions to this problem, but what we do know is that healthy people visit doctors less often. A magic solution to this public health dilemma is one that prevents illness and thus decreases healthcare utilization overall. This is the concept behind most prevention programs.

The Apple-a-day prevention program is based on the hypothesis that individuals who eat one apple, every day, will have fewer physician encounters. There are several ways to test this theory, though some are better than others. Basically what we need to examine is the exposure, in this case apples, and the outcome, which will be the number doctor appointments. What we want to determine here is causation. Do apples cause doctors appointments? Well, hopefully not. More accurately, what we want to test is whether or not apple eating is protective again doctor appointments. Simply put, does the risk of a seeing the doctor go down as apple eating behaviors increase? We could study this concept using several different approaches, with the most practical being as follows:

• Cross-sectional study - outcome and exposure are determined at the same time. To accomplish this study, we could survey patient’s, post-doctors appointment, to determine their exposure to apples. The benefit of this survey type is that data could be collected quickly at a single point in time. The downside is that we would probably see a participation bias towards people that already see the doctor more regularly, independent of their apple eating habits.

• Case-Control Study – Here we would identify two groups; one group of cases comprised of people known to have visited the doctor recently, and a control group of people who have not seen the doctor lately. We would then determine which of the cases and controls have been exposed to apples and the frequency of that exposure.

• Randomized control trial – this type of study applies a treatment or intervention and then measures the effect of the treatment. Study participants are divided into two groups and one group receives the treatment/intervention while the other group serves as a control, receiving no treatment.

Taking the approach that apple eating is potentially beneficial, I believe the best approach would be to conduct an intervention. Using a randomized control trial, we would conduct an experiment whereby one group would receive the treatment, and a control group would receive no treatment. The treatment group will be supplied with, and required to eat, a minimum of one apple per day. The control group will not be supplied with apples and will receive no instructions regarding apple intake. They will instead be supplied with a pamphlet on healthy habits.

The survey method will be identical for both the treatment and control group. To measure both exposure and disease, study participants will conduct a web-based survey, once per month, for the duration of one year. The survey tool to be used is a questionnaire developed by scientists Gladys Block, Christina Gillespie, Ernest Rosenbaum and Christopher Jenson, known as the Food Screener. The Food Screener is a validated one-page, self-report survey that provides a quick and reliable measure of nutrient intake¹. This survey will measure fruit and vegetable intake for both study groups. This method is being used to collect accurate data of comprehensive fruit and vegetable consumption, rather than just apple consumption. Focusing solely on apple consumption might have the unintended effect of swaying the control group to eat apples. It may also conceal the potential benefits to be gained from eating a wide variety of produce. An additional survey will ask respondents to report how many times they have visited a physician during the previous one-month period. They will be prompted to select which type of facility was visited; physician office, outpatient hospital, or emergency department. The data gained from these two surveys will tell us how many apples a participant consumed during a given month and the number of physician visits they made. A comparison will be made between the number of physician visits and the quantity of apple consumption. We will examine frequency of both measurements, as it is believe that higher apple consumption will lead to fewer physician visits.

Methods

The study will be conducted in three locations across the United States, so as to account for regional differences in produce consumption and healthcare utilization. The program sites include St. Louis, MO., Atlanta, GA., and San Francisco, CA. The recruitment goal is 1,000 participants from each city, for a total of 3,000 participants.

Inclusion criteria must be met prior to enrollment in the study. Potentials participants must be 18 years or older to participate. Participants are excluded if they report a chronic medical condition whereby routine medical care is required. Additionally, potential participants are excluded if they report an allergy to any fruit or vegetable. Participants must give written conformed consent, whereby they agree to comply with the assigned treatment. All participants are notified that intervention treatment (apples for intervention group and pamphlet from controls) will be paid for by the research study. The costs associated with receiving care from a physician are the responsibility of the participant. Each participant will receive compensation in the amount of $10 per month, to be awarded after the completion of each monthly survey.

Assignment into the control or intervention group will be accomplished using computer generated allocation. 500 participants from each city will be assigned to the treatment group, while 500 will be assigned as controls. Participants will be blinded as to which group they are assigned. The healthy eating habits pamphlet distributed to controls in intended to make these participants feel that are receiving some type of treatment. They will not be aware of the actual treatment procedures.

Confounders

Recognizing that healthcare utilization is heavily dependent upon many factors besides diet, the study will take these variables into consideration. Possible confounders include age, sex, health insurance status, and income level. The data from both groups will be stratified based upon these factors so that we can determine if there are outcome differences. For example, participants in the older age groups may, as a whole, visit the doctor more often.

Challenges

The biggest hurdle in this study will be compliance of participants. In order to entice participants to complete each survey, compensation will be distributed on a month-to-month basis, rather than at the end. The survey method was chosen in order to boost compliance. It can be completed within 7 minutes and since it will be web-based, users can access it the location of their choice. Even when participants are willing to report their monthly data, it can not be guaranteed that people in the intervention group will comply with the treatment of eating one apple per day. Supplying participants with a continuous supply of fresh apples should encourage compliance, but it is by no means is an assurance this will happen.

Strengths and Weaknesses

Though the study will attempt to overcome the aforementioned challenges, there are certain aspects of the study design that can not be manipulated for our benefit. There will be inherent weaknesses within this design. The first, and perhaps most significant weakness is the length of the study. It is hoped that a period of one year will be able to capture the relationship between the treatment and the anticipated outcome. Certainly, a longer time period will give even more insight into this relationship, but budgetary restraints of this study do not allow for a longitudinal study. We must consider, as a limitation, that the protective benefits of apple consumption could be over exaggerated. Yet-to-be-diagnosed health conditions could develop, without participant knowledge, during the study period. A longer study might be able to reveal such health conditions that would warrant additional physician care. Conversely, participants may be diagnosed, during the study period, with health conditions that existed before the study began. Development of these conditions may not be dependent upon apple consumption. Randomization will equalize these two circumstances.

The intervention portion of this study does not attempt to control the eating regimen of participants. By not placing participants on a diet, we feel compliance will be higher. We also feel results will be more realistic as participants who state they have followed the treatment will have had fewer challenges associated with doing so. Participants are not asked to make drastic behavior changes or give up a beloved food item. On the flip side, not controlling what all participants eat opens up the door for an un-planned cross-over. Since controls will have access to apples, as well as other produce, it is possible they will also engage in high-apple eating frequency. We have allowed the possibility of a such a cross-over because we feel it reflects the eating habits of the general population. Restricting apple intake may not provide an accurate picture of what a normal person, who gives no thought to apple consumption, is likely to consume.

It is also possible the the behavior of apple eating is associated with other healthy behaviors that are not being measured, such as physical activity. The participants who comply with the treatment program may be inherently more likely to engage in healthy behaviors as a whole. The fact that the study measurements will be self-reported opens up the possibility for inaccuracy. However, it should be noted that we feel the self-reported data will be a strength, rather than a weakness. Because surveys will be completed on computer, without the prompts of a researcher, it is hoped that respondents will answer honestly. We believe responders will report accurate and truthful answers, rather than the answers they think researchers want to hear.

Study Implications
If this intervention successfully proves that eating an apple a day can reduce the frequency of physician utilization, this would have important ramifications for future public health practice. Rather than focusing on complicated interventions that involve nutritional data and physician recommendations, this solution will be easy to implement. Interventions using this tactic will likely have higher success rates, as asking participants to change one behavior, instead of several, is more practical.
Success of this intervention could also have important policy implications. USDA recommendations suggest eating a wide variety of fruits and vegetables. If apples are determined to be a super fruit, their status within the food pyramid may be upgraded. It may be recommended that individuals eat a wide variety of fruits and vegetables, but that this variety should include at least one apple per day. School lunch programs, which rely on USDA guidelines, may require that all children receive an apple every day with their lunch.
There may also be implications for the pratice of medicine. Physicians, rather than prescribing medication, may advise patients to consume a certain number of apples and to schedule an appointment only if symtoms persist. Routine physical screenings and preventive medicine may include questioning that pertains to apple consumption. Doctors will likely suggest that their patients eat as many apples as possible.

Additional Thoughts

When designing a study, it is very evident that there are many, many different ways to assess a certain exposure or disease. I can certainly attest to the fact that every design aspect will be planned and re-planned several times over. In the end, I settled on the study and methods described here largely because it feels realistic. If often read studies where I feel researchers aimed too high by trying to measure too many variables, or perhaps asked too much of study participants. In this study, the commitment from participants is minimal and the compensation is fair, but not overly coercive. I can imagine volunteers committing to this study with full intent of complying, and then doing so honestly. The data collected here could easily be collected retrospectively by asking participants to report their exposure to apples and their use of physician services, within a specified time period. Admittedly, it may be more cost and time-effective to do. The concerns of recall bias and temporality are too large in a retrospective study, and for me, not worth the savings. Within reason, I'd rather undergo the added expense and effort of a randomized trial in order to be assured the study results are legitimate and lead to a worthwhile conclusion.

If by some strange chance this study were conducted, and we did find a link between apple consumption and physician utilization, I would like to see this study developed into a long-term study, where we could dig even deeper into this link. A longitudinal study would help us determine exactly what disease outcomes are affected by apple consumption, which might give insight into why apple consumption keeps people healthy.

On a final note, I urge you to look toward the top right hand side of this page where you will see voting buttons. I have created a very unscientific poll where you can answer two questions. These questions, in a very basic way, will assess the relationship between apple consumption and physician utilization. Please take a quick second to complete the poll and we can see if a relationship exists within our own responses.

1Block, G., Gillespie, C. Rosenbaum, E.H., & Jenson, C. (2000). A rapid food screener to assess fat and fruit and vegetable intake. American Journal of Preventive Medicine, 14(4), 284-288. doi:10.1016/S0749-3797(00)00119-7