| Course | PHLT 8520 Advanced Epidemiological Methods |
|---|---|
| Week | Week 10 |
| Paper type | Validation check paper |
| Length | About 1,020 words, 6 pages |
| Format | APA 7 student paper |
| School | Walden University |
| Program | PhD in Public Health |
| Updated | October 2026 |
Free sample paper for PHLT 8520 Week 10
Testing the Cohort's Own Assumptions: A Shift Diary Substudy of Timekeeping Accuracy and a Negative Control Using Partners' Night Shifts, With What a Null Result Would and Would Not Mean
Student Name
Walden University
PHLT 8520: Advanced Epidemiological Methods
Instructor Name
Month Day, Year
Testing the Cohort's Own Assumptions: A Shift Diary Substudy of Timekeeping Accuracy and a Negative Control Using Partners' Night Shifts, With What a Null Result Would and Would Not Mean
Two Assumptions, Two Checks
The composite night-shift cohort rests on two assumptions that earlier weeks stated but could not test. The first is that timekeeping records classify night work well enough, with errors unrelated to the outcome. The second is that, after adjustment, nothing about a woman's household circumstances produces both night work and preterm birth. This paper reports one check of each: a shift diary substudy that measures how often timekeeping misclassifies exposure, and a negative control exposure that asks whether household circumstances alone could create an association. All figures are composite and written for this sample.
Check One: The Shift Diary Substudy
Design
Two hundred twelve pregnant employees in rostered jobs agreed to keep a daily shift diary for eight weeks between gestational weeks 6 and 20, recording the start and end of every shift actually worked, including informal swaps. Diaries were collected weekly by a research nurse, before delivery, so the diary record could not be shaped by the outcome. Each woman's diary and timekeeping records were compared for the same weeks.
Results
Timekeeping missed night work for four women, all of whom covered a colleague's night without a recorded swap, and recorded night work for six women who had given their nights away. Sensitivity was 0.96 and specificity 0.95. The errors were similar among women who later delivered preterm and at term, although with only 19 preterm births in the substudy that comparison is imprecise.
Table 1
Timekeeping Compared With Shift Diaries for Any Night Shift in the Window (Composite Figures)
| Timekeeping record | Diary shows night work | Diary shows none | Total | |
|---|---|---|---|
| Night work recorded | 96 | 6 | 102 | |
| No night work recorded | 4 | 106 | 110 | |
| Total | 100 | 112 | 212 | |
Note. Composite figures. Sensitivity 96 of 100 (0.96); specificity 106 of 112 (0.95).
What the Accuracy Means for the Estimate
If the errors are nondifferential, they should, on average, dilute the risk ratio. Applying the substudy's sensitivity and specificity to the Week 9 illustrative counts, following the simple correction that Lash et al. (2014) describe, moves the risk ratio from 1.80 to about 1.90. The correction is small because timekeeping is accurate, and it moves the estimate away from the null, as expected. Misclassification of exposure is therefore not a plausible explanation for a large association, and the earlier concern about dating error in Week 7 remains the more important measurement question.
Check Two: A Negative Control Exposure
Logic
A negative control exposure is a variable that shares the suspected sources of bias with the real exposure but cannot plausibly cause the outcome (Lipsitch et al., 2010). A partner's night shifts fit that description. A household under financial strain might lead both partners to take nights, and the same strain could raise preterm risk. But a partner's work hours have no direct biological route to the timing of the pregnant woman's delivery. If partners' night shifts were associated with preterm birth after adjustment, household circumstances would be the likely reason, and the main estimate would be suspect.
Who Can Be Studied
Partners' schedules exist only where both partners work rostered jobs at the same health system. In the first-birth cohort, 540 pregnancies meet that condition. The model relates any partner night shift during the same gestational window to preterm delivery, adjusted for the same covariates plus the woman's own night work, since couples often coordinate schedules and the two exposures are correlated.
Result
The adjusted risk ratio for any partner night shift is 1.07, with a 95% confidence interval from 0.71 to 1.62.
What the Null Would and Would Not Mean
The result is reassuring, but only within limits that need to be stated plainly.
What It Supports
The point estimate is close to 1. If household strain severe enough to produce the main association were present, it would be expected to push partners toward nights as well and to show up as an association here. The absence of a clear signal is consistent with the main estimate not being driven by household circumstances.
What It Cannot Show
First, the interval is wide. An upper limit of 1.62 does not exclude a moderate association that, if real, would point to residual confounding. With 540 pregnancies, the check has little power, and a null from a small study is weak evidence. Second, a negative control detects only confounders it shares with the main exposure. Strain that leads the woman, but not her partner, to work nights would be invisible to it. Third, couples who both work at the system are not typical; they may have more stable incomes than couples in which one partner works elsewhere, so the confounding in this subgroup may differ from the cohort as a whole. Fourth, a partner's nights may not be truly inert: disrupted household sleep could affect the pregnant woman, which would bias the negative control away from the null and make a null result harder to obtain, not easier.
Where the Checks Leave the Cohort
The diary substudy makes exposure misclassification an unlikely explanation for a strong association and suggests the main estimate may be slightly understated. The negative control finds no sign of household confounding, while being too small and too narrow to rule it out. Neither check addresses the differential dating identified in Week 7, which remains the leading candidate for inflating the estimate. Hernán and Robins (2016) describe validation of this kind as testing how well data emulate the target trial; here, the emulation of exposure measurement passes, the emulation of exchangeability passes weakly, and the outcome measurement has not yet been tested.
Summary
Timekeeping classified night work with a sensitivity of 0.96 and a specificity of 0.95, and correcting for that error changes the illustrative estimate only slightly. Partners' night shifts showed no clear association with preterm birth, which supports but does not prove the absence of household confounding. The final methods paper will report both checks with their limits.
References
Hernán, M. A., & Robins, J. M. (2016). Using big data to emulate a target trial when a randomized trial is not available. American Journal of Epidemiology, 183(8), 758-764. https://doi.org/10.1093/aje/kwv254
Lash, T. L., Fox, M. P., MacLehose, R. F., Maldonado, G., McCandless, L. C., & Greenland, S. (2014). Good practices for quantitative bias analysis. International Journal of Epidemiology, 43(6), 1969-1985. https://doi.org/10.1093/ije/dyu149
Lipsitch, M., Tchetgen Tchetgen, E., & Cohen, T. (2010). Negative controls: A tool for detecting confounding and bias in observational studies. Epidemiology, 21(3), 383-388. https://doi.org/10.1097/EDE.0b013e3181d61eeb
The Walden PHLT 8520 Week 10 instructions
Validation weeks often ask students to design or report checks on their study's key assumptions, such as the accuracy of exposure measurement or the absence of residual confounding. Prompts may mention validation substudies, negative controls, positive controls or comparison with an external gold standard. Instructors want each check tied to the assumption it tests, results reported with uncertainty and an honest interpretation of what the check can and cannot show. A null result from a negative control is a frequent focus because students tend to overstate it. Expect four to five pages with tables. Many sections ask students to say how the checks change the overall assessment of the study and what remains untested.
How this PHLT 8520 Week 10 example is built
The paper opens by naming the two assumptions and the check for each. The diary section describes the design, then Table 1 presents a two-by-two comparison of timekeeping against diaries, with sensitivity and specificity in the note. A subsection applies those values to the illustrative counts and explains why the corrected estimate moves away from the null. The negative control section sets out the logic, defines who can be studied and reports the adjusted result. A long section separates what the null supports from four things it cannot show: low power, shared-confounder limits, an atypical subgroup and possible non-inertness. A closing section places both checks within target trial emulation and names dating error as untested.
Reading the PHLT 8520 Week 10 rubric
Validation papers are marked on design logic, correct reporting and interpretation. The sample earns design credit by matching each check to a named assumption and by collecting diaries before delivery so they cannot be shaped by outcome. Reporting credit comes from a complete two-by-two table and a correction that shows its direction. Interpretation carries the most weight, and the sample's four-part account of what a null negative control cannot show is the section graders single out. Linking the checks back to target trial elements shows integration with the term's framework. Naming what remains untested earns honesty credit. Mechanics, tables and citations complete the evaluation. Papers that call a wide-interval null "proof of no confounding" are typically marked down.
PHLT 8520 Week 10 help from the desk
Validation checks are easy to overstate. Say which assumption each check tests before reporting it. For a substudy, give the full table, not just sensitivity, and apply the values to your main estimate to show what changes. For a negative control, explain why it shares the bias but not the causal path, then report the interval and discuss power. List what a null cannot show: small samples, confounders the control does not share, unusual subgroups and any route by which the control might affect the outcome. End by stating which assumptions still lack a test. Keep the language measured. If you need a validation plan for your own data, a first one is free.
Get a PHLT 8520 Week 10 example written to your instructions
Share your Week 10 prompt and rubric, along with the study and the checks your section is asked to propose or interpret. The piece designs or reads each check, states the assumptions it rests on, and explains exactly what a null or positive result would establish. Free for your first sample; returned within 24-48 hours.
More PHLT 8520 papers and PhD in Public Health samples
- PHLT 8520 Week 1: Error Taxonomy Brief
- PHLT 8520 Week 2: Design Selection Paper
- PHLT 8520 Week 3: Confounding Brief
- PHLT 8520 Week 4: Design Proposal
- PHLT 8520 Week 5: Selection Analysis
- PHLT 8520 Week 6: Peer Challenge Reply
- PHLT 8520 Week 7: Measurement Error Brief
- PHLT 8520 Week 8: Control Strategy Comparison
- PHLT 8520 Week 9: Sensitivity Argument
- PHLT 8520 Week 11: Final Methods Paper
- PHLT 8500 Week 10: Survival Model Brief
- PHLT 8450 Week 10: Stakeholder Critique
- PHLT 8331 Week 10: Effect Size Decision Brief
- PHLT 8270 Week 10: Outbreak Response Action Brief
PHLT 8520 Week 10 questions, answered
Where can I find a free PHLT 8520 Week 10 sample paper?
The Week 10 validation check paper can be read in full here, with a diary substudy table, a negative control analysis and a careful reading of what a null result means.
What is a negative control exposure?
A variable that shares the suspected sources of bias with the real exposure but has no plausible causal effect on the outcome. An association between it and the outcome signals bias.
Does a null negative control prove there is no confounding?
No. It may be underpowered, it detects only confounders it shares with the main exposure and its subgroup may not represent the whole study.
How do sensitivity and specificity correct an exposure estimate?
They let you back-calculate the expected true counts of exposed and unexposed cases and noncases, then recompute the risk ratio. The sample's correction moves 1.80 to about 1.90.
Why use partners' night shifts as the negative control?
Partners share household circumstances, such as financial strain, that could drive both night work and preterm risk, but a partner's schedule has no direct biological route to the timing of delivery.