IHP 525 Module 9 Final Project Example

Final Statistical Report: How Well the Bayview County Blood Pressure Program Worked

[Student Name]

Southern New Hampshire University

IHP 525: Biostatistics

Module Nine Final Project

[Instructor Name]

[Date]

Final Statistical Report: How Well the Bayview County Blood Pressure Program Worked

This report sets out, for the county health department and its board, the full statistical evaluation of the community blood pressure program. It describes how the evaluation was designed, how data were analyzed, what was found and how confident we can be. It is written so that the methods could be repeated by another analyst and the conclusions checked against the numbers shown.

Executive Summary

Over six months, adults in the program lowered systolic blood pressure by an estimated 6.1 mm Hg more than similar patients in usual care; the 95% interval runs from 3.9 up to 8.3. Among those followed up, 46% of participants and 29% of usual-care patients reached a reading below 140. Results held when missing data were handled in different ways. Women appeared to benefit less than men, but that estimate is imprecise. Expansion is supported, preferably with a staggered rollout that allows a stronger test.

Background and Question

Roughly one Bayview County adult in three lives with hypertension, and most of that group are not yet at goal. The program offers twelve weekly group sessions on diet, activity and medicines, a pharmacist medication review and a home monitor. The evaluation asked whether, among adults with uncontrolled hypertension at county clinics, program participation was associated with a larger six-month drop in systolic pressure than usual care, after accounting for differences between groups.

Design and Reporting Standard

The evaluation compared people who enrolled with patients from the same clinics who met the same eligibility criteria but did not enroll, which makes it an observational cohort study. The STROBE statement (von Elm et al., 2007) is a checklist of items that reports of observational research should include so readers can judge validity, among them the setting, eligibility, variables, sources of bias, handling of missing data and study size. This report addresses each of those items in turn.

Setting, Participants and Variables

Data came from five county clinics between January and December of last year. Eligible adults were 30 to 75 years old with a recent clinic systolic reading of 140 or higher. The outcome was change in systolic pressure from enrollment to six months, taken from routine clinic readings. The exposure was program enrollment. Four covariates were recorded at enrollment: age, sex, starting systolic reading and how many pressure medicines each person took. The secondary outcomes were reaching a reading below 140, change in diastolic pressure and intensification of medicines.

Study Size and Missing Data

The analysis plan showed that about 106 people per group would give 80% power to detect a 5 mm Hg difference; the study included 240 participants and 260 comparison patients. Six-month readings were missing for 22 participants and 29 comparison patients. Sterne et al. (2009) explained that when missing values relate to recorded characteristics, analyzing only complete records can bias estimates, and that multiple imputation, done carefully, can reduce that bias. Twenty imputed data sets were created using all model variables plus clinic and program contacts.

Statistical Methods

The primary analysis was linear regression of six-month change on program participation, adjusted for the four covariates. Control was modeled with logistic regression. Secondary results were judged against a Bonferroni threshold of 0.008. Sensitivity analyses repeated the primary model using complete cases, a pessimistic assumption of no change for missing participants and exclusion of two clinics with different measurement routines. All analyses followed the plan written before data were examined.

Results

Participants were slightly older and started about 2.6 mm Hg higher than comparison patients. Unadjusted, their pressure fell 11.2 mm Hg against 4.6 mm Hg. After adjustment, the difference was 6.1 mm Hg in favor of the program. The adjusted odds ratio for control was 2.14, diastolic pressure fell 2.8 mm Hg more and medicines were intensified less often, though that last result did not pass the Bonferroni threshold.

Table 1. Main Results of the Evaluation

Note. Unadjusted group values are among completers; adjusted effects use imputed data.

How Robust Are the Findings?

The complete-case estimate was 6.4 mm Hg, the pessimistic estimate 5.3 and the estimate without the two clinics 6.0. Every one of these intervals excluded zero. Among women the benefit was 3.9 mm Hg (-0.2 to 8.0) against 8.1 among men, an exploratory finding. Separately, home monitors showed 80% sensitivity and 90% specificity against ambulatory readings, adequate for program use but likely to produce many false positives in the general population.

Interpreting the Estimate

Gardner and Altman (1986) argued that the purpose of a study is to estimate how big an effect is, with a statement of precision, rather than to deliver a yes-or-no verdict. Read that way, the evaluation's answer is that the program probably lowers systolic pressure by somewhere between 4 and 8 mm Hg relative to usual care. A drop of that size across a population is linked in large trials to meaningful reductions in stroke, so even the lower end is worth having.

What the Pilot Should Have Told Us

Last year's pilot of 60 people reported a 5 mm Hg difference with p = 0.07, and some leaders concluded the program had failed. Altman and Bland (1995) cautioned that a nonsignificant result from a small study shows only that the evidence is insufficient. The present evaluation, eight times larger, found an effect close to the pilot's estimate. The lesson for the department is to judge future pilots by their intervals and power, not by whether p fell below 0.05.

Limitations

Because people chose to enroll, motivation or health awareness may differ between groups in ways adjustment cannot capture, so the results show association rather than proof of cause. Clinic readings vary in technique. Imputation assumes missingness is explained by recorded variables. Follow-up was six months, and durability is unknown.

Recommendations

First, expand the program to the four additional clinics, as the benefit is consistent and practically meaningful. Second, stagger the start dates across clinics in random order so that the next evaluation can compare clinics before and after launch, a stronger design than the current comparison. Third, gather feedback from women participants on session timing and content, and report their results separately at twelve months. Fourth, extend follow-up of the current cohort to one year.

Conclusion

The program was associated with a clinically meaningful drop in blood pressure that held across analyses. Uncertainty remains about cause, durability and effects among women, and the recommendations are designed to reduce each of those uncertainties while bringing the benefit to more residents.

References

Altman, D. G., & Bland, J. M. (1995). Absence of evidence is not evidence of absence. BMJ, 311(7003), 485. https://doi.org/10.1136/bmj.311.7003.485

Gardner, M. J., & Altman, D. G. (1986). Confidence intervals rather than P values: Estimation rather than hypothesis testing. BMJ, 292(6522), 746-750. https://doi.org/10.1136/bmj.292.6522.746

Sterne, J. A. C., White, I. R., Carlin, J. B., Spratt, M., Royston, P., Kenward, M. G., Wood, A. M., & Carpenter, J. R. (2009). Multiple imputation for missing data in epidemiological and clinical research: Potential and pitfalls. BMJ, 338, Article b2393. https://doi.org/10.1136/bmj.b2393

von Elm, E., Altman, D. G., Egger, M., Pocock, S. J., Gøtzsche, P. C., & Vandenbroucke, J. P. (2007). Strengthening the reporting of observational studies in epidemiology (STROBE) statement: Guidelines for reporting observational studies. BMJ, 335(7624), 806-808. https://doi.org/10.1136/bmj.39335.541782.AD