| Course | HLTH 5013 Epidemiology and Statistics |
|---|---|
| Module | Module 5 |
| Paper type | Epidemiologic research proposal |
| Length | 1,290 words, about 5 pages plus title and reference pages |
| Format | APA 7 student paper |
| School | American College of Education |
| Program | Master of Public Health |
| Updated | September 2026 |
Free sample paper for HLTH 5013 Module 5
Randomizing the Order, Not the People: A Research Proposal for a Stepped-Wedge Evaluation of a County Diabetes Prevention Program Across Five Clinics
Student Name
American College of Education
HLTH5013: Epidemiology and Statistics
Module 5 Assignment
Instructor Name
November 2, 2026
Background and Rationale
Earlier modules in this course analyzed type 2 diabetes in a composite county: roughly nine adults in a hundred already diagnosed and a steady stream of new cases each year. A retrospective cohort analysis of the county's lifestyle program for adults with prediabetes found that enrollment was associated with roughly 31% to 32% lower three-year risk of diabetes once the analysis accounted for the measured differences between enrollees and others. The finding is encouraging but vulnerable to unmeasured confounding, since people who choose to enroll may differ in motivation and habits.
The original Diabetes Prevention Program trial established that intensive lifestyle intervention can reduce diabetes incidence by 58% compared with placebo under trial conditions (Diabetes Prevention Program Research Group, 2002). What remains uncertain is how much benefit a community version, delivered by a county health department through primary care clinics to a diverse population, actually produces. Over the coming two years, the department intends to open the program at four additional clinic sites. This proposal uses that planned expansion to answer the question with a stronger design.
Aims and Hypotheses
The primary aim is to estimate the effect of offering the lifestyle program through primary care clinics on the three-year incidence of type 2 diabetes among adults with prediabetes. The hypothesis is that incidence will be lower during periods when a clinic offers the program than during periods when it does not.
Two secondary aims follow. The first is to estimate the effect on change in A1c and body weight at one year. The second is to measure reach, the share of eligible adults at each clinic who enroll, overall and by language, sex, age and insurance, since earlier analyses found that enrollees differed from non-enrollees and a program that reaches only the already advantaged will have limited population impact.
Design
The study will use a stepped-wedge cluster randomized design. Hemming et al. (2015) describe this design as one in which clusters, here clinics, begin in a control condition and switch to the intervention one after another at randomly assigned times, until all clusters receive it. It suits situations in which an intervention is expected to do more good than harm, so withholding it permanently would be hard to justify, and in which logistical limits prevent starting everywhere at once. Both conditions apply here.
Five clinics will take part: the four new sites and one large existing clinic that has not yet offered the program. After a six-month baseline period, one clinic will start the program every four months in a random order drawn by the county epidemiologist in the presence of the evaluation committee. By month 26, all five clinics will offer it. No clinic is denied the program; chance decides only when each one starts. Because each clinic contributes time both before and after starting, comparisons are made within clinics as well as between them, which protects against differences in clinic populations.
Population, Exposure and Outcomes
The population will be adults aged 18 to 75 attending the five clinics whose A1c falls in the prediabetes range at any point during the study and no prior diabetes diagnosis. Each adult will be assigned to the condition of their clinic at the time their prediabetes is identified. The exposure is therefore the offer of the program, not attendance, which keeps the comparison faithful to randomization in the same way an intention-to-treat analysis does in a trial.
The primary outcome is incident type 2 diabetes within three years, defined as a new diagnosis code or a laboratory A1c at or above the diagnostic threshold, drawn from the regional health information exchange. To reduce the detection bias identified in an earlier module, the analysis will also record how many A1c tests each person had, and a secondary analysis will restrict to people tested at least once a year. Secondary outcomes are change in A1c and weight at one year and program reach.
Sample Size Reasoning
Clinic records suggest each of the five clinics identifies about 180 new adults with prediabetes a year, or about 2,700 over the three-year enrollment window. With a control-period three-year risk of about 16% and a hoped-for reduction to about 12%, an individually randomized study would need roughly 1,300 people in each arm for 80% power at a two-sided alpha of .05. A stepped-wedge design must also account for the correlation of outcomes within clinics and for changes over time. Hussey and Hughes (2007) developed methods for calculating power in this design using a model with clinic effects and time effects. Using that approach with an assumed intracluster correlation of 0.01, the county epidemiologist estimates that 2,700 participants across five clinics provide adequate power for a difference of this size. Because the number of clinics is small, the proposal will report this power calculation and its assumptions in full, and results will be interpreted with attention to the width of the confidence interval rather than only statistical significance.
Analysis Plan
The primary analysis will use a mixed-effects model for the binary outcome, with a random effect for clinic and fixed effects for calendar period, so that the intervention effect is estimated after accounting for differences between clinics and for trends over time that would affect all clinics, such as a change in screening guidelines. The effect will be reported as a risk ratio and a risk difference with 95% confidence intervals. A sensitivity analysis will add individual characteristics, including age, weight status, baseline A1c, sex and coverage type, as covariates. Reach will be summarized as the proportion enrolling at each clinic and compared across language, sex, age and insurance groups. All analyses will be specified in writing before data are examined.
Threats to Validity and How They Are Addressed
A stepped-wedge design is stronger than the retrospective cohort but not immune to bias. Three threats matter most. First, clinics may begin preparing before their assigned start, for example by talking up the program, which would blur the difference between control and intervention periods; the protocol will ask clinics not to promote the program before their start date and will record any early activity. Second, patients may move between clinics, carrying exposure from one condition to another; the analysis will assign each person to the clinic where prediabetes was first identified, and a sensitivity analysis will exclude people who changed clinics. Third, testing practices may change once a clinic starts the program, since staff attention to diabetes rises, which could increase recorded diagnoses in intervention periods and hide a real benefit. Recording the number of A1c tests per person, and repeating the analysis among those tested at least yearly, addresses this. Each threat and its handling will be reported in the final paper, so readers can judge the result.
Ethics, Timeline and Dissemination
The proposal will be submitted to the state health department's institutional review board. Because every clinic will offer the program and data come from existing records, the team will request a waiver of individual consent for the records analysis, with a public notice at each clinic and an option to opt out. Program participants will consent to the program itself as usual. Data will be de-identified before analysis and stored under the exchange's security rules.
The baseline period begins in month one, the final clinic starts in month 26, and three-year follow-up for the last enrollees ends in year five, with interim reports on reach and one-year outcomes each year. Results will go to the board of health, participating clinics and community partners in plain language, and to a peer-reviewed journal, whether the findings are positive, null or negative.
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
Hemming, K., Haines, T. P., Chilton, P. J., Girling, A. J., & Lilford, R. J. (2015). The stepped wedge cluster randomised trial: Rationale, design, analysis, and reporting. BMJ, 350, Article h391. https://doi.org/10.1136/bmj.h391
Hussey, M. A., & Hughes, J. P. (2007). Design and analysis of stepped wedge cluster randomized trials. Contemporary Clinical Trials, 28(2), 182-191. https://doi.org/10.1016/j.cct.2006.05.007
HLTH 5013 Module 5 instructions, in plain terms
HLTH 5013 Module 5 is usually where the course's methods come together, in a research proposal or a surveillance report. For a proposal, prompts usually ask for background and rationale, aims and hypotheses, a design suited to the question, the population, exposure and outcome definitions, a sample size or power discussion, an analysis plan, ethical considerations and a timeline. For a surveillance report, they ask for the data source, case definition, trends, comparisons and interpretation. Graders expect the design to follow from the question and the analysis to match the design. Build on data or questions from earlier modules if the course follows one topic, and check Canvas for the required proposal sections.
Inside the HLTH 5013 Module 5 example
The sample opens with a rationale built on the course's earlier findings and their key limitation, and on the original trial. Aims and hypotheses are stated precisely, including an equity aim on reach. The design is defined with a methods source, justified for this situation and scheduled clinic by clinic. Eligibility, exposure as the offer and an outcome with a check on detection bias follow. Sample size is reasoned from expected risks with a method suited to the design, and its assumptions are stated. The analysis plan names the model and each component's purpose. Ethics, consent, data protection, timeline and a commitment to report all results close the proposal.
Where the points sit in the HLTH 5013 Module 5 rubric
Research proposal rubrics usually reward a clear rationale, precise aims, a design matched to the question, well-defined measures, an appropriate analysis plan and attention to ethics. Graders check that the population, exposure and outcome are defined operationally, not just named. The design criterion rewards explaining why the chosen design fits and what biases it controls. A sample size or power discussion earns points even when approximate, if its assumptions are stated. The analysis plan must match the design, for example accounting for clustering in a cluster design. Ethics, feasibility and dissemination complete the core criteria, with APA 7 formatting throughout.
HLTH 5013 Module 5 help from the desk
Proposals often lose marks when the design and analysis do not match, such as a cluster design analyzed as if individuals were randomized. Another common weakness is an outcome named but never defined, or an exposure defined in a way that reintroduces bias. Students also skip sample size entirely. State your hypotheses before your methods. Define every variable so another analyst could reproduce it. Give your sample size assumptions even if the calculation is rough. Commit to reporting results whatever they show, and say who will receive them. For a proposal on injury, vaccination or maternal health, or a surveillance report instead, describe your topic and data along with the grading guide, and we can build a Module 5 proposal from them.
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.
More HLTH 5013 and Master of Public Health sample papers
- HLTH 5013 Module 1: Prevalence and Incidence Analysis
- HLTH 5013 Module 2: Study Design Comparison
- HLTH 5013 Module 3: Measures of Association Analysis
- HLTH 5013 Module 4: Data Analysis
- HLTH 5053 Module 2: Health Materials Critique
- HLTH 5063 Module 1: Disease Biology and Transmission
- HCI 5073 Module 2: Data Quality Analysis
- HLTH 5003 Module 5: Program Evaluation Plan
HLTH 5013 Module 5 questions, answered
What does HLTH5013 Module 5 usually ask for?
HLTH5013 frequently ends with a research proposal or a surveillance report: a question, design, population, measures, sample size reasoning and analysis plan, or a structured summary of surveillance data with interpretation. Which one you write is set by your own section.
What is a stepped-wedge design?
A cluster randomized design in which groups such as clinics start the intervention one after another in a random order until all have it, so each contributes data before and after.
Why define the exposure as the offer of a program rather than attendance?
Because it preserves the comparison created by randomization; analyzing only attenders would reintroduce self-selection bias.
Where can I find a free HLTH 5013 Module 5 sample paper?
This page has the complete Module 5 research proposal for a stepped-wedge evaluation of a county diabetes prevention program across five clinics, from aims and design to sample size, analysis, ethics and timeline.
Why does a stepped-wedge analysis include time effects?
Because clinics switch to the intervention at different times, and outcomes may change over time for other reasons. Adjusting for calendar period separates those trends from the intervention's effect.