| Course | PU570 Chronic Disease Epidemiology |
|---|---|
| Unit | Unit 5 |
| Paper type | Confounding analysis |
| Length | About 850 words, 6 pages double-spaced |
| Format | APA 7 student paper |
| School | Purdue University Global |
| Program | Master of Public Health (MPH) |
| Updated | October 2026 |
Free sample paper for PU570 Unit 5
From 1.52 to 1.14 in Four Adjustment Steps: A Unit 5 Confounding Analysis Apportioning a Gum Disease and Heart Disease Ratio Between the Gums and the Smoking, Diabetes and Low Income That Travel With Them
[Student Name]
Purdue University Global
PU570: Chronic Disease Epidemiology
Unit 5 Confounding Analysis
[Instructor Name]
[Date]
The cohort of 8,000 and every estimate here are composite; the confounding structure, the adjustment logic and the research cited are real.
The Starting Point
Crude ten-year risk of coronary heart disease: 9.4 percent among adults with periodontitis, 6.2 percent among adults without. Risk ratio 1.52, with a confidence interval from 1.31 to 1.76. This is the number a press release would carry.
Step One: Age and Sex
Both conditions rise with age, and periodontitis is more common in men. Adjusted risk ratio 1.38. Age alone moves the estimate more than any later variable, which is unsurprising and is the reason age adjustment is treated as automatic rather than as a finding.
Step Two: Smoking, in Three Categories
Smoking causes periodontitis and causes coronary disease, which makes it the textbook confounder here. Entered as never, former and current, with pack-years available for current and former smokers. Adjusted risk ratio 1.22. Smoking accounts for about a third of the remaining excess, and the choice of categories matters: entering smoking as a yes or no variable leaves more residual confounding than entering it as three categories with pack-years, because a 40 pack-year smoker and somebody who quit at 25 are not the same exposure.
Step Three: Diabetes
Diabetes causes periodontal disease, is caused in part by the same behaviors, and causes coronary disease. Adjusted risk ratio 1.17. The direction of this arrow is contested, since periodontal inflammation may also worsen glycemic control, which would make diabetes partly a mediator rather than purely a confounder. The analysis adjusts for it and reports the alternative, because a variable that is both cannot be handled by a single model.
Step Four: Income and Education
Low income and fewer years of schooling predict both dental care access and cardiovascular risk. Adjusted risk ratio 1.14, confidence interval 0.97 to 1.34, which no longer excludes one. Socioeconomic position is the hardest confounder here because it is measured crudely, so what remains after adjustment is a partially controlled variable rather than a controlled one. Each adjustment is a claim about a causal structure, and four adjustments in a row are four claims. The final number is only as good as the weakest of them.
What Is Left, and What It Might Be
Three possibilities for the residual 1.14, and the analysis declines to choose between them on these data.
1. A causal effect of periodontal inflammation on atherosclerosis, which is biologically plausible and is the hypothesis the field has pursued for thirty years.
2. Residual confounding by the variables already entered, especially smoking intensity and socioeconomic position, both measured with error that biases toward the exposure.
3. Unmeasured confounding by something neither variable captures, such as long-term inflammatory burden or health care engagement.
The published position is close to the second and third: a consistent association exists, and the evidence does not establish independence from shared risk factors (Lockhart et al., 2012). The meta-analytic estimate before this cohort's own data pointed the same way, with a modest association that narrows as adjustment improves (Humphrey et al., 2008).
Two Variables Deliberately Not Adjusted For
1. Number of teeth lost. It is a consequence of periodontitis rather than a cause of it, so adjusting for it would block part of the pathway under study and bias the estimate toward no effect. Adjusting for a mediator is the commonest avoidable error in analyses of this kind.
2. C-reactive protein. It sits on the proposed causal pathway between gum inflammation and vascular disease, so the same objection applies. It is reported descriptively and kept out of the model.
Confounder selection should follow the assumed causal structure rather than the variable's availability or its effect on the estimate, which is the principle the methodological literature states most plainly (VanderWeele, 2019).
What Would Settle It
Not another adjusted cohort. Three designs would help. A trial of periodontal treatment with cardiovascular endpoints, which has been attempted at small scale with inconclusive results. An analysis using a genetic instrument for periodontal susceptibility, which breaks confounding by behavior if the instrument's assumptions hold. Or a sibling design, which holds much of the shared social and genetic background constant. The analysis ends on that point because the alternative, a fifth adjustment step on the same data, will move the estimate and settle nothing.
How This Should Be Reported
1. Report the crude and all four adjusted estimates, in order, rather than the final figure alone.
2. Name the variables held out and why, since a reader cannot otherwise tell a mediator from an omission.
3. State that the final interval includes one, and resist the sentence a modest association remained.
4. Give the smoking categories, because the most common criticism of this literature is that smoking was handled too crudely to believe the adjusted figure.
References
Humphrey, L. L., Fu, R., Buckley, D. I., Freeman, M., & Helfand, M. (2008). Periodontal disease and coronary heart disease incidence: A systematic review and meta-analysis. Journal of General Internal Medicine, 23(12), 2079-2086. https://doi.org/10.1007/s11606-008-0787-6
Lockhart, P. B., Bolger, A. F., Papapanou, P. N., Osinbowale, O., Trevisan, M., Levison, M. E., Taubert, K. A., Newburger, J. W., Gornik, H. L., Gewitz, M. H., Wilson, W. R., Smith, S. C., & Baddour, L. M. (2012). Periodontal disease and atherosclerotic vascular disease: Does the evidence support an independent association? Circulation, 125(20), 2520-2544. https://doi.org/10.1161/CIR.0b013e31825719f3
VanderWeele, T. J. (2019). Principles of confounder selection. European Journal of Epidemiology, 34(2), 211-219. https://doi.org/10.1007/s10654-019-00494-6
What the PU570 Unit 5 assignment asks for
Unit 5 usually asks for an analysis of confounding in one association. Add confounders one at a time and report the estimate after each step, so a reader can see which variable did the work. Justify each as a confounder using the causal structure, not the size of the change it produces. Say when a variable may be both confounder and mediator, and report the alternative rather than hiding the ambiguity. Name at least two variables you deliberately did not adjust for, with reasons, since adjusting for a mediator biases toward no effect. Offer competing explanations for whatever association survives. Then say what design, rather than what further adjustment, would settle the question. Give the categories you used for any strong confounder, because a binary smoking variable leaves residual confounding that an adjusted figure hides.
Inside the PU570 Unit 5 example
The analysis here starts from the crude ratio a press release would carry, then adds age and sex, smoking in three categories with pack-years, diabetes, and socioeconomic position, reporting the ratio after each. A highlighted line observes that four adjustments are four causal claims and the result is only as good as the weakest. Three readings of the residual 1.14 follow, with the published position close to residual and unmeasured confounding. Two variables held out, tooth loss and C-reactive protein, are listed with reasons and flagged by a margin note. The analysis closes with three designs that would settle it and four rules for reporting the sequence honestly.
Where the points sit in the PU570 Unit 5 rubric
Credit in this unit follows the stepwise reporting. Showing the estimate after each adjustment scores well above reporting a single adjusted figure, because the sequence is what reveals which confounder mattered. Justifying confounders from assumed causal structure, rather than from how much they move the estimate, is rewarded. Marks go to identifying a variable that may be both confounder and mediator and reporting both readings. Holding out mediators, and saying so, carries substantial weight, as does offering competing explanations for the residual association. A closing section naming designs rather than more adjustment is credited. APA 7 citation of the substantive and methodological literature is expected.
Common PU570 Unit 5 mistakes, and how to avoid them
The commonest fault is a single adjusted number with no path to it, which hides whether smoking or income did the work. Report each step. The second is selecting confounders by whether they change the estimate, which is a rule the methodological literature rejects. Third, do not adjust for anything on the causal pathway, including biomarkers of the proposed mechanism, because that blocks part of the effect you are estimating. Avoid describing a residual interval that includes one as a modest remaining association. And resist a fifth adjustment on the same data, since the honest next step is a different design. An analysis of your own association takes two days.
Get a PU570 Unit 5 example written to your instructions
Whatever exposure and outcome the PU570 Unit 5 scenario pairs, the same sequence of diagram, strata and residual account applies. Include the dataset or article, the questions and the rubric. Within 24-48 hours a free first custom sample arrives with every confounder justified by both of its links and every adjustment decision explained. The paper above is an original model document written by our desk, not a submitted student paper and not an official Purdue University Global document.
More PU570 papers and MPH samples
- PU570 Unit 1: Hazard and Risk Discussion Post
- PU570 Unit 2: Measure Interpretation Exercise
- PU570 Unit 3: Study Design Critique
- PU570 Unit 4: Causal Reasoning Paper
- PU570 Unit 6: Seminar Reflection
- PU570 Unit 7: Surveillance Data Analysis
- PU570 Unit 8: Risk Factor Profile
- PU570 Unit 9: Prevention Strategy Brief
- PU570 Unit 10: Epidemiological Report
- PU565 Unit 5: First-Cycle Coding Memo
- PU555 Unit 5: Measurement Seminar Reflection
- PU550 Unit 5: Workforce Development Plan
- PU545 Unit 5: Shelter in Place Warning Set
PU570 Unit 5 questions, answered
Where can I find a free PU570 Unit 5 sample paper?
The whole confounding analysis, all four steps and the held-out variables, is open here.
Why adjust one variable at a time?
So the reader can see which confounder accounts for the association. A single adjusted figure conceals that entirely.
How are confounders chosen?
From the assumed causal structure, not from how much they shift the estimate. Availability is not a criterion either.
What happens if you adjust for a mediator?
You block part of the pathway you are studying and bias the estimate toward no effect, which is the most common avoidable error here.
What would settle this association?
A trial of treatment with cardiovascular endpoints, a genetic instrument, or a sibling design. Not another adjustment on the same cohort.