NURS 700 Week 6 Written Assignment 3: Evaluation Methods and Statistical Analysis Example

Reading and Planning the Numbers: Statistical Methods for Evaluating Lung Cancer Screening Outreach in Rural Primary Care

Introduction

The first two assignments for this course located evidence on lung cancer screening outreach and planned the evaluation of a nurse-led program in four rural clinics of the Blue Ridge Family Health network. This paper addresses the statistics behind both tasks. Its first half interprets the statistical results reported in the studies that informed the program, because a DNP-prepared nurse must judge what those numbers can and cannot support. Its second half sets out the analysis plan for the program's own data, matching each measure to a method suited to it.

Reading the Numbers in the Literature

Odds Ratios and Confidence Intervals

DiCarlo et al. (2026) randomized primary care patients flagged as possibly eligible to telephone outreach with decision counseling or to usual care and reported screening referral or order rates of 16% and 9%, an odds ratio of 1.82 with a 95% confidence interval of 1.37 to 2.41. An odds ratio compares the odds of an outcome in two groups; a value of 1.82 means the odds of a referral or order were 82% higher with outreach. The confidence interval marks out the values the trial's data would plausibly support, and because the entire interval lies above 1, the difference is statistically significant. The same trial reported a completion odds ratio of 1.45 with an interval from 0.999 to 2.09. That lower bound sits just below 1, so the completion result is at the edge of significance and should be read as suggestive rather than settled. The comparison shows why a reader should look at the interval, not only the point estimate.

Survey Weighting and Population Estimates

Jia et al. (2026) analyzed 5,987 respondents to the 2024 National Health Interview Survey to estimate screening uptake for roughly 37.5 million American adults who have ever smoked. That leap from a few thousand people to tens of millions is possible because the survey uses a complex sampling design, and survey weights adjust each respondent's contribution so the sample represents the population. The authors used survey-weighted logistic regression and found that adults eligible under federal criteria had 4.54 times the odds of past-year screening compared with adults eligible under neither criterion, with an interval of 3.44 to 5.98. Without weighting, the estimates would describe the sample, not the nation.

Correlation Is Not the Same as Cause, or as Absence of Cause

Cho et al. (2026) examined whether states with more screening facilities had higher screening prevalence and found correlations of 0.0144 with the number of facilities and 0.0424 with facility density, neither statistically significant. A correlation coefficient ranges from negative 1 to 1, and values this close to zero indicate essentially no linear relationship between the variables at the state level. The finding is useful for the project because it suggests that adding scanners is not the answer. It does not prove that access never matters, since state averages can hide differences between counties and between individual patients, an example of why conclusions drawn from aggregated data must be applied to individuals with care.

Precision and Sample Size

Two studies illustrate how sample size governs what a result can show. In a pilot randomized trial of a culturally tailored navigation intervention for Chinese and Korean Americans, screening completion was 45% in the intervention group and 31.3% in the control group, but the trial included only 36 participants (Katcher et al., 2026). A difference of nearly 14 percentage points sounds important, yet with so few participants the uncertainty around it is wide, and the trial is best read as a promising signal that justifies a larger study. By contrast, Wernli et al. (2026) randomized 1,837 patients and reported that stepped reminders raised annual adherence with a relative risk of 1.59, and the interval around that estimate ran only from 1.47 to 1.72. The pilot's difference may be real but cannot yet be trusted; the large trial's difference is both large and precise, and only the second is strong enough to change practice.

The Analysis Plan for the Outreach Program

Descriptive Statistics

The evaluation will begin by describing the eligible population with counts and percentages for categorical variables, such as sex, insurance type, smoking status and clinic, and with means or medians for continuous variables, such as age, pack-years and driving distance to the scanner. These descriptions will be repeated for patients who completed screening and those who did not, so the team can see at a glance whether completers differ from non-completers before any formal test is run. Medians will be used for distance, which is likely to be skewed by a few patients who live very far away.

Run Charts for Monthly Monitoring

Monthly screening completion and annual adherence will be plotted on run charts throughout the program. Run charts allow the team to see change as it happens, and simple rules separate real signals from random variation: for example, a run of six months in a row sitting above the pre-program median points to a genuine shift rather than chance. These charts will guide the quarterly adjustments planned in the program evaluation.

Segmented Regression for the Interrupted Time Series

At the end of the year, the main outcome, which is each month's share of registry patients with a completed scan, will be analyzed with segmented regression across a year of pre-program months and a year of program months. The model estimates the level and slope of the outcome before the program and any immediate change in level and change in slope after it begins. Splitting the series this way keeps an improvement that had already begun from being credited to the program, a distinction a simple before-and-after comparison cannot make. With only 24 monthly points, the model will be kept simple, and autocorrelation between adjacent months will be checked.

Comparisons and Predictors

The proportion of patients completing screening in the year before the program and the year of the program will be compared using a chi-square test, which is appropriate for comparing proportions between independent groups. To learn who the program reaches and who it misses, a logistic regression will examine whether completion during the program year is associated with age, sex, insurance, smoking status, clinic and driving distance. Results will be reported as odds ratios with 95% confidence intervals, interpreted in the way described above. Because the network expects roughly 900 eligible patients, the model will include only a handful of predictors, so that each has enough outcome events to estimate reliably.

Missing Data

Missing pack-year histories are themselves part of the problem the program addresses. Patients with missing smoking data cannot be classified as eligible, so the evaluation will report the proportion with complete histories at baseline and at the end of the year, and will describe how many patients were added to the registry as their histories were completed, so that changes in the denominator are visible rather than hidden.

Statistical and Clinical Significance

A result can be statistically significant but too small to matter, or large and important but uncertain in a small sample. The evaluation will therefore report effect sizes, such as the absolute change in percentage points and the number of additional patients screened, alongside p-values and confidence intervals. For a rural network, a rise from 8% to 20% completion would mean more than one hundred additional patients screened each year, a change that matters regardless of how small the p-value is.

Qualitative Data Alongside the Numbers

Not every evaluation question is statistical. Interviews with the outreach nurses and a sample of patients at six months will explain why the numbers look as they do, for example why patients who agreed to screening during a call did not keep their appointment. These interviews will be summarized by theme and reported beside the quantitative results, so that a disappointing number comes with an explanation the team can act on.

Truthful Reporting

Statistics can be used to clarify or to mislead. Proverbs warns that a false witness will not go unpunished (English Standard Version Bible, 2001/2016, Proverbs 19:5). In evaluation, a false witness can take the form of a selectively chosen time frame, a hidden change in the denominator or a nonsignificant result presented as a success. The analysis plan commits in advance to the outcomes, methods and time frame, so that the results, whatever they show, can be reported honestly.

Conclusion

Reading the numbers in the lung cancer screening literature shows why confidence intervals, weighting, aggregation and sample size shape what a study can support. The analysis plan for the network's program applies the same principles, pairing each measure with a suitable method, attending to missing data and reporting both statistical and clinical significance, so that the program's results will be trustworthy.

References

Cho, B., Stallings-Smith, S., & Kim, H. (2026). State disparities in lung cancer screening uptake despite facility availability in the United States, 2024-2025: A cross-sectional ecological study. Cancer Control, 33. https://doi.org/10.1177/10732748261494041

DiCarlo, M., Shimada, A., Sifri, R., Johnson, M., Keith, S. W., Zhan, T., Jackson, H., Indictor, A., Burgess, N., Mahon, I., Pisano, E., & Myers, R. E. (2026). Increasing lung cancer screening referral/order and completion rates in primary care through patient outreach and shared decision making. Journal of Medical Screening. Advance online publication. https://doi.org/10.1177/09691413261479224

English Standard Version Bible. (2016). Crossway. (Original work published 2001)

Jia, Y., Zhao, J., & Rivera, M. P. (2026). Lung cancer screening uptake among adults newly eligible under the 2023 American Cancer Society guideline compared with USPSTF-eligible adults: A nationwide cross-sectional study. Chest. Advance online publication. https://doi.org/10.1016/j.chest.2026.09.005

Katcher, J. G., Wen, K. Y., Shusted, C. S., Liang, J., Koh, K., Juon, H. S., & Barta, J. A. (2026). Health beliefs and lung cancer screening uptake among eligible Asian Americans in Philadelphia, Pennsylvania, United States. Preventive Medicine Reports, 67, Article 103521. https://doi.org/10.1016/j.pmedr.2026.103521

Wernli, K. J., Anderson, M. L., Palazzo, L., Luce, C., Bezman, N., Rogers, K., Chin, M., Gao, H., Ralston, J., Su, Y. R., Carter-Bawa, L., Triplette, M., Tuzzio, L., Vasavada, A., & Green, B. B. (2026). Health communication and stepped reminders interventions for lung cancer screening: A randomized clinical trial. JAMA Internal Medicine. Advance online publication. https://doi.org/10.1001/jamainternmed.2026.3666