What Each Design Can Prove: Judging the Evidence That a Health System's Diabetes Prevention Program Works
Student Name
American College of Education
HLTH5623: Epidemiology and Public Health for Healthcare Administrators
Module 3 Assignment
Instructor Name
May 15, 2028
The Question and What Is Already Established
A regional health system, invented for this paper, launched a year-long lifestyle change program for adults with prediabetes three years ago, modeled on the national Diabetes Prevention Program. About 1,400 patients have enrolled. The chief medical officer wants to know whether the program works well enough to expand to all 22 primary care clinics, which would add roughly $1.1 million to annual operating costs. Before approving that expense, the finance committee has asked for evidence from the system's own patients rather than from studies elsewhere. The question, stated precisely, is whether enrolling in the program reduces the proportion of patients with prediabetes who progress to type 2 diabetes within three years.
Whether the underlying intervention can work is already established by strong evidence. The original randomized trial showed that intensive lifestyle change cut new cases of diabetes by more than half, 58 percent, relative to placebo across roughly three years (Diabetes Prevention Program Research Group, 2002). Long-term follow-up found that the reduction persisted but narrowed, with incidence 27 percent lower in the lifestyle group than in the placebo group over a mean of 15 years (Diabetes Prevention Program Research Group, 2015). An evaluation of the national program delivered in community settings found that participants attended a median of 14 sessions, lost an average of 4.2 percent of body weight and that 35.5 percent reached the 5 percent weight loss goal, with more attendance associated with more weight loss (Ely et al., 2017). The system does not need to prove that lifestyle change can prevent diabetes; it needs to know whether its own program delivers enough of it.
Design One: An Ecological Comparison
The quickest design would compare rates of new diabetes diagnoses in the counties or clinic catchment areas where the program is offered with rates in areas where it is not. The data exist in the system's electronic records and in public county data. An ecological study compares groups rather than individuals, so it can show whether areas with the program have lower diabetes incidence, but it cannot show whether the individuals who enrolled were the ones who avoided diabetes. The program reaches only a small fraction of each area's adults with prediabetes, so any area-level difference is likely to be driven by other factors such as income, age and access to care. This design can describe patterns and generate hypotheses; it cannot establish the program's effect.
Design Two: A Cross-Sectional Survey
A second option is a survey of current patients with prediabetes asking about program participation, diet, activity and weight, analyzed alongside their most recent A1c. Celentano and Szklo (2019) describe cross-sectional studies as measuring exposure and outcome at the same point in time, which makes them useful for estimating prevalence and associations but unable to establish which came first. Patients whose A1c improved might be more likely to remember and report participation, and patients who were already more active might be more likely to enroll. The survey could show that participants have lower A1c values on average, but not that the program caused the difference. It establishes association at a moment, not effect over time.
Design Three: A Retrospective Cohort
A stronger option uses the electronic record to build a cohort of all patients diagnosed with prediabetes over the past three years, comparing those who enrolled in the program with those who were eligible but did not, and following both groups for new diabetes diagnoses. Because exposure is recorded before the outcome, a cohort design establishes the time order that a cross-sectional survey cannot, and it allows calculation of incidence and relative risk. Its limitation is selection. Patients who choose to enroll in a year-long program are likely to be more motivated, more health conscious and better able to attend sessions than those who do not, all of which may independently lower their risk. This is sometimes called healthy volunteer bias, and it tends to make a program look more effective than it is.
Statistical methods can reduce but not eliminate that bias. Matching participants with nonparticipants on baseline A1c, body mass index, age, insurance and number of primary care visits, or using propensity scores to balance those characteristics, makes the comparison fairer. Motivation and health habits are rarely recorded in the chart, however, so the most important differences may remain. A cohort study can show that participants developed diabetes less often; it cannot fully show that participation was the reason.
Design Four: A Staged Rollout
The system's plan to expand to all 22 clinics offers an opportunity for a much stronger design. Instead of expanding everywhere at once, the system could randomize the order in which clinics begin offering the program, with a new group of clinics starting every four months over about eighteen months. At any point, some clinics would offer the program and others would not yet, and the comparison of diabetes incidence among eligible patients in each group would reflect the program's availability rather than patients' decisions to enroll. This is a cluster randomized design with a stepped rollout, and it answers a slightly different question: whether offering the program to a clinic's patients reduces diabetes incidence among all eligible patients. That population-level question is the one that matters for a decision about expansion, because the system cannot choose who enrolls.
The staged design has costs. It delays access for some clinics by up to eighteen months, it requires careful tracking of which patients were eligible in which clinic at which time and three-year diabetes outcomes will take several years to observe. Process measures, such as enrollment, session attendance and weight change at twelve months, can be read much sooner and are known from the national evaluation to predict the program's effect.
Recommendation
The recommendation is to proceed with expansion using a randomized staged rollout, and in the meantime to conduct the retrospective cohort analysis with propensity matching to give leadership an early, if imperfect, estimate. The cohort result should be reported with an explicit statement that it likely overstates the program's effect because of self-selection. The ecological comparison and the survey should not be used to judge effectiveness. This combination gives the chief medical officer a usable answer within months and a strong answer within a few years, without delaying the program for the clinics that start first.
Conclusion
The published trials establish that lifestyle intervention can prevent diabetes; the system's own question is whether its program delivers that benefit to its patients. An ecological comparison can only describe patterns, a cross-sectional survey only associations, and a retrospective cohort can establish time order but remains vulnerable to the self-selection of motivated participants. A staged rollout with clinics randomized to their start dates can answer the question that matters for expansion. Matching the design to what it can prove is what keeps a health system from spending $1.1 million a year on a conclusion its evidence cannot support.
References
Celentano, D. D., & Szklo, M. (2019). Gordis epidemiology (6th ed.). Elsevier.
Diabetes Prevention Program Research Group. (2002). Reduction in the incidence of type 2 diabetes with lifestyle intervention or metformin. New England Journal of Medicine, 346(6), 393-403. https://doi.org/10.1056/NEJMoa012512
Diabetes Prevention Program Research Group. (2015). Long-term effects of lifestyle intervention or metformin on diabetes development and microvascular complications over 15-year follow-up: The Diabetes Prevention Program Outcomes Study. The Lancet Diabetes & Endocrinology, 3(11), 866-875. https://doi.org/10.1016/S2213-8587(15)00291-0
Ely, E. K., Gruss, S. M., Luman, E. T., Gregg, E. W., Ali, M. K., Nhim, K., Rolka, D. B., & Albright, A. L. (2017). A national effort to prevent type 2 diabetes: Participant-level evaluation of CDC's National Diabetes Prevention Program. Diabetes Care, 40(10), 1331-1341. https://doi.org/10.2337/dc16-2099
How this HLTH 5623 Module 3 example is structured
HLTH 5623 Module 3 in many sections reads study designs for what they can actually establish; your classroom's instructions decide the designs and the example. This example states the question, summarizes what the landmark trial and a national program evaluation established, and then examines four designs the system could use with its own data, from weakest to strongest, naming for each the specific bias that limits it. The recommendation matches a design to what the system actually needs to know. The paper treats designs as tools with different capacities rather than as a ranking to memorize.
HLTH5623 Module 3 questions, answered
What does HLTH5623 Module 3 usually ask for?
HLTH5623 Module 3 in many sections asks students to examine epidemiologic study designs and explain what each can and cannot establish, often by applying them to a healthcare question. Common designs include ecological, cross-sectional, case-control, cohort and randomized studies. Your classroom's instructions decide which designs and which example to use.
What is healthy volunteer bias?
It is a form of selection bias in which people who choose to take part in a program or study are healthier or more motivated than those who do not, so their better outcomes may reflect who they are rather than the program itself. It often makes voluntary programs look more effective than they are.
What is a stepped rollout or stepped wedge design?
It is a design in which groups such as clinics begin an intervention at different times, in a randomly assigned order, until all have it. At each point some groups have the intervention and others do not, allowing comparison while every group eventually receives it. It suits programs a system plans to expand anyway.
Write yours, or have the desk draft it
This paper is an original model document written by our desk, not a submitted student paper and not an official American College of Education document. Read it for the moves, then write your own to the instructions in your classroom. If you want one built to your exact prompt and rubric, the first custom sample is free and arrives in 24 to 48 hours.