Immortal time and healthy-adherer bias in a composite cardiac rehab cohort, and the line between what that design supports and what only a trial could, set out for NU505 Unit 3. Searches like "nu 505 unit 3 assignment example", "nu505 unit 3 sample" and "nu505 unit 3 example" land here.
Which Claims Survive? A Study Design Critique of a Registry Cohort Reporting 38 Percent Lower Mortality With Cardiac Rehabilitation
[Student Name]
Purdue University Global
NU505: Clinical Epidemiology and Population Health Promotion
Unit 3 Assignment
[Instructor Name]
[Date]
The registry cohort critiqued here is a composite study summary prepared for the exercise, not a published study; the methods sources and the trial evidence cited are real.
The Study
The cohort included 6,200 adults discharged from hospital after a first myocardial infarction, drawn from a regional registry. Of these, 2,480 attended at least one cardiac rehabilitation session within six months of discharge. Three-year mortality was 6.9 percent among attenders and 11.2 percent among non-attenders, a risk ratio of 0.62. After adjustment for age, sex, diabetes, ejection fraction and revascularization, the authors reported an adjusted hazard ratio of 0.64. Their abstract makes four claims: that rehabilitation reduces mortality by more than a third, that each additional session further lowers risk, that the benefit holds across subgroups, and that referral rates should rise.
Claims Table
1. "Rehabilitation reduces three-year mortality by 38 percent." Can a cohort support it? No, not as a causal effect size. A cohort can show that attenders died less often, adjusted for measured factors; it cannot give the size of a causal effect when attendance itself depends on health.
2. "Each additional session further lowers risk." Partly. A dose pattern is consistent with causation, but patients who complete more sessions are also those who stay well enough to keep coming, so the pattern is exposed to the same biases as claim 1.
3. "The benefit holds across subgroups." Partly. Consistent associations across subgroups add some weight, but every subgroup shares the same design problems.
4. "Referral rates should rise." Yes, but not because of this study. Referral is already supported by randomized evidence and guidelines; the cohort adds useful information on who is not being referred.
Immortal Time
The first threat is structural. To count as an attender, a patient had to survive long enough to reach a first session, a median of 38 days after discharge in this cohort. Deaths during those weeks could happen only to patients classified as non-attenders, because no one who died before a first session could ever be an attender. That period is immortal time for the attender group: time during which, by definition, the outcome could not occur (Suissa, 2008). Counting it as exposed time, or classifying patients by what they later did, makes the exposure look protective even if it has no effect at all (Lévesque et al., 2010).
The fix is known, and the authors' own appendix applies one. In a landmark analysis starting at six months after discharge, which compares only patients alive at that point, classified by whether they had attended by then, the risk ratio shrinks from 0.62 to 0.81. Roughly half of the apparent benefit disappears once the immortal period is handled, which shows how much of the headline figure the design produced.
Healthy-Adherer Bias
The second threat is harder to fix. Attenders in this registry were younger, more often insured and had fewer comorbidities than non-attenders. Adjustment handles the measured differences. It cannot handle traits that were never measured but predict both attendance and survival: motivation, health literacy, social support, the ability to arrange transport three times a week. People who adhere to one health behavior tend to adhere to others, and adherent patients often do better even when the thing they adhere to is inert. A landmark analysis cannot remove this bias, because it arises from who chooses to attend, not from how time is counted.
What a Stronger Observational Analysis Would Do
Three changes would make the registry's estimate more credible without a trial. First, define exposure and follow-up from the same starting point for everyone, either with a landmark, as the appendix does, or by treating attendance as a time-varying exposure so that each patient counts as unexposed until the day of the first session. Second, address selection with more than regression adjustment: comparing patients referred to programs near their homes with patients referred to programs far away uses distance as a source of variation in attendance that is less tied to motivation. Third, report a negative control outcome, such as deaths from accidents or cancer, which rehabilitation should not affect. If attenders also die less often from causes rehabilitation cannot plausibly touch, the association is likely driven by who attends rather than by what attending does.
None of these would turn the cohort into a trial. Each would narrow the distance between the reported association and the causal claim the abstract makes, and each could be done with data the registry already holds.
What the Randomized Evidence Says
Trials give a useful check. A Cochrane-based meta-analysis of 63 randomized trials with 14,486 participants found that exercise-based cardiac rehabilitation reduced cardiovascular mortality, with a relative risk of 0.74, and hospital admissions, but found no significant effect on total mortality over a median follow-up of 12 months (Anderson et al., 2016). The trials measured a shorter period and different populations, so the comparison is not exact. But a cohort showing 38 percent lower total mortality, against trials showing no significant total mortality effect, is exactly the gap one would expect if the cohort estimate were inflated by immortal time and healthy-adherer bias.
What the Cohort Does Well
The study deserves credit for what a registry can do that a trial cannot. It documents who is not referred and who does not attend, in a real regional population rather than a selected trial sample. Its finding that uninsured and older patients attend less often is useful regardless of the mortality estimate, because it identifies a referral and access problem a health system can act on. A nurse navigator program aimed at older and uninsured patients could be evaluated against the same registry, which is a better use of these data than a mortality claim. Polit and Beck (2021) describe this as a strength of observational designs: they show how care is actually distributed.
Conclusion
The clinical decision does not depend on this study. Referral to cardiac rehabilitation after myocardial infarction is already supported by randomized evidence for cardiovascular outcomes, hospital admissions and quality of life. The question the critique answers is narrower: what number to quote to patients. The cohort's 38 percent should not be quoted as the benefit of attending. A more honest statement, drawing on the trials, is that rehabilitation lowers the chance of dying from heart disease and of returning to the hospital, and improves how people feel, while its effect on overall survival over a year or two is less certain.
References
Anderson, L., Oldridge, N., Thompson, D. R., Zwisler, A.-D., Rees, K., Martin, N., & Taylor, R. S. (2016). Exercise-based cardiac rehabilitation for coronary heart disease: Cochrane systematic review and meta-analysis. Journal of the American College of Cardiology, 67(1), 1-12. https://doi.org/10.1016/j.jacc.2015.10.044
Lévesque, L. E., Hanley, J. A., Kezouh, A., & Suissa, S. (2010). Problem of immortal time bias in cohort studies: Example using statins for preventing progression of diabetes. BMJ, 340, b5087. https://doi.org/10.1136/bmj.b5087
Polit, D. F., & Beck, C. T. (2021). Nursing research: Generating and assessing evidence for nursing practice (11th ed.). Wolters Kluwer.
Suissa, S. (2008). Immortal time bias in pharmacoepidemiology. American Journal of Epidemiology, 167(4), 492-499. https://doi.org/10.1093/aje/kwm324
How this NU505 Unit 3 example is structured
Claims organize the critique rather than design features, because the prompt asks which claims survive. Each claim in the composite registry cohort's abstract is tested against what a cohort can establish: temporal order, an association adjusted for measured factors and a dose pattern, but not the size of a causal effect when attendance depends on health. Immortal time comes first because it is structural and fixable, and the appendix's landmark figure shows how much of the association it produced. Healthy-adherer bias follows and is harder to fix. A section proposes stronger observational analyses, and the randomized evidence is then set beside the cohort. The critique credits what the registry does well, and the close says what number to quote to patients.
Get an NU505 Unit 3 example written to your instructions
Share the article your Unit 3 prompt assigns, or its citation, plus the questions and the grading rubric your section uses. The critique is organized around that study's own claims, sorting what its design can carry from what it cannot. A first custom sample is provided at no cost and is ready within 24-48h. 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.
NU505 Unit 3 questions, answered
What is immortal time bias in plain terms?
It arises when one group, by definition, must survive a period to be classified as exposed. Patients counted as rehab attenders had to live until their first session, so any deaths before then could only fall in the comparison group. That makes the exposure look protective even if it did nothing. Landmark analysis or time-varying exposure definitions are standard ways to correct it.
Does the critique conclude that cardiac rehab is ineffective?
No, and that distinction matters for the grade. The critique argues that this cohort overstates the size of the benefit, not that the benefit is zero. Referral after a heart attack is recommended by major cardiology guidelines on the strength of broader evidence. The example keeps the clinical recommendation intact while questioning the specific number the authors promoted.
Which claims can a cohort study legitimately support?
Cohorts are strong on temporal order, on describing who receives an exposure, and on associations adjusted for measured factors. They can suggest a dose pattern. They cannot fully separate the exposure from reasons people chose it when those reasons also affect the outcome. Sizing a causal effect under those conditions usually requires randomization, or careful emulation of a trial with explicit assumptions.