HLTH 5013 Module 2 Study Design Comparison Example

Does the County's Lifestyle Program Keep Prediabetes From Becoming Diabetes? Comparing Five Study Designs for One Local Question

Student Name

American College of Education

HLTH5013: Epidemiology and Statistics

Module 2 Assignment

Instructor Name

October 12, 2026

The Question

The composite county described in the first module found about 1,400 new diagnoses of type 2 diabetes a year. Three years ago, the health department began referring adults whose A1c falls in the prediabetes range to a year-long lifestyle change program modeled on the Diabetes Prevention Program. About 900 people have enrolled. The board now asks whether the program works in this county. Put as a research question: among adults in the county with prediabetes, does enrollment in the lifestyle program reduce the three-year incidence of diagnosed type 2 diabetes compared with no enrollment?

The efficacy of intensive lifestyle intervention is already established. In the original randomized trial, adults coached on diet and exercise developed diabetes 58% less often than those on placebo, while metformin cut the rate by 31% (Diabetes Prevention Program Research Group, 2002). The local question is different: whether a community version, delivered by the county to its own residents, produces a real benefit. That distinction, between efficacy under trial conditions and effectiveness in practice, shapes which designs are useful.

Randomized Controlled Trial

In a randomized trial, eligible adults would be assigned by chance to the program or to usual care, and diabetes incidence would be compared after three years. Randomization is the design's great strength: it balances known and unknown factors, such as motivation, diet and family history, between groups, so that when outcomes differ, the program is the most plausible reason. Grimes and Schulz (2002a) rank randomized trials highest among designs for testing whether an intervention works, and this is why.

For this county, a standard trial has serious problems. Denying a program with established efficacy to half of those who want it would be hard to justify ethically and politically, three years of follow-up for hundreds of people is expensive, and the department has no research staff. A waitlist design, in which the control group receives the program after a delay, eases the ethical problem but shortens the comparison period to months, too short to observe diabetes incidence.

Cohort Designs

A cohort study follows people grouped by exposure, here enrollment or not, forward to see who develops the outcome. Grimes and Schulz (2002b) describe cohort studies as the best observational design for determining incidence and examining several outcomes of an exposure, while warning that who ends up in each group, and who drops out along the way, can bias the answer. The county could run a prospective cohort, enrolling people now and following them for three years, or a retrospective cohort, using the health information exchange to identify everyone with prediabetes three years ago, classify them by enrollment and look up who has since been diagnosed.

The retrospective version is feasible now, since the data exist. Its central problem is confounding. Residents who sign up probably differ from those who do not in ways that matter, such as drive, attention to health and a regular source of care, which could produce a lower diabetes incidence even if the program did nothing. The analysis would need to adjust for measured differences, such as age, baseline A1c, body mass index and insurance, and ideally match enrolled and non-enrolled people on the propensity to enroll. Unmeasured motivation would remain a threat. A cohort study can tell the board that enrolled residents did better; it cannot, by itself, prove that enrolling is why.

Case-Control Design

The case-control design works backward from the outcome. Cases would be adults with prediabetes three years ago who have since developed diabetes; controls would be similar adults who have not; and the study would compare how many in each group had enrolled in the program. Schulz and Grimes (2002) describe case-control studies as efficient for rare outcomes and long latency periods but prone to selection and information bias, and they stress that controls have to be drawn from the very population whose members became cases.

Here the design offers little advantage. Diabetes is not rare among people with prediabetes, the county already has records for the whole population at risk, and a cohort analysis of the same records gives incidence directly. The case-control approach would also produce an odds ratio, which overstates the relative risk when the outcome is common. It would make sense only if records on enrollment were incomplete and had to be gathered by interview, which is not the case.

Cross-Sectional Design

A cross-sectional survey would measure, at one point in time, whether adults had ever enrolled and whether they now have diabetes. It is cheap and quick, and it could describe participation by neighborhood, language or insurance. It cannot answer the causal question, because it cannot establish which came first: some people may have enrolled after being diagnosed, and people who developed diabetes early may have left the program. For this question, the cross-sectional design is useful only for describing who the program reaches.

A Stepped Rollout

One further design deserves attention because it fits how public health programs actually expand. The department plans to add the program at four more clinics over the next two years. If the order in which clinics start is chosen at random, and data are collected at every clinic throughout, the rollout becomes a stepped design: each clinic contributes time before and after the program begins, and the random order protects against the possibility that the clinics most eager to start also serve healthier patients. No one is denied the program permanently, which removes the main ethical objection to a trial. The design still takes years and requires consistent measurement, but it turns an expansion the department was going to make anyway into evidence.

Measuring the Outcome the Same Way for Everyone

Whatever design is chosen, the outcome has to be measured the same way in both groups, or the comparison is biased before it begins. Diagnosed diabetes depends on being tested. If program participants see a clinician more often, they may be tested more often, so their diabetes is found sooner and their recorded incidence looks higher, the opposite of the usual concern. If non-participants rarely see a clinician, their diabetes may go undiagnosed and their recorded incidence may look lower. Either pattern distorts the result. The county can reduce this bias by defining the outcome as an A1c of 6.5% or higher on any test, or a new diabetes diagnosis, and by checking how many A1c tests each group had during follow-up. If testing frequency differs sharply, the analysis should compare only people tested at least once a year, and report both results so the board can see how much the choice matters.

Recommendation

No single design answers the board's question perfectly within its budget. The department should do two things. Now, it should conduct a retrospective cohort analysis using the health information exchange, adjusting for measured confounders and matching on the propensity to enroll, and report the result as an association with its limits clearly stated. Going forward, it should randomize the order in which new clinics start the program so that the expansion itself produces stronger evidence. It should not undertake a standard randomized trial or a case-control study, and it should use cross-sectional data only to track who is reached. Together, these choices balance rigor, ethics and cost for a local agency.

References

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

Grimes, D. A., & Schulz, K. F. (2002a). An overview of clinical research: The lay of the land. The Lancet, 359(9300), 57-61. https://doi.org/10.1016/S0140-6736(02)07283-5

Grimes, D. A., & Schulz, K. F. (2002b). Cohort studies: Marching towards outcomes. The Lancet, 359(9303), 341-345. https://doi.org/10.1016/S0140-6736(02)07500-1

Schulz, K. F., & Grimes, D. A. (2002). Case-control studies: Research in reverse. The Lancet, 359(9304), 431-434. https://doi.org/10.1016/S0140-6736(02)07605-5