IHP 525 Module 3 Milestone One Example

Milestone One: Framing the Statistical Evaluation of Bayview's Community Blood Pressure Program

[Student Name]

Southern New Hampshire University

IHP 525: Biostatistics

Module Three Milestone One

[Instructor Name]

[Date]

Milestone One: Framing the Statistical Evaluation of Bayview's Community Blood Pressure Program

A good statistical analysis begins with a clear question and a precise description of the data that will answer it. This milestone frames the evaluation of Bayview County's community blood pressure program. It states the research question and hypotheses, describes the study design and data sources, defines each variable by type and role, explains key decisions about how outcomes will be measured and outlines the analysis and its limits.

Background

The program enrolled 240 adults with uncontrolled hypertension, defined as systolic pressure of 140 mm Hg or higher at two readings, through six barbershops and four churches. Trained community health workers measured blood pressure monthly, encouraged medication adherence and referred participants to clinicians who could adjust treatment. The comparison group consists of 260 adults with uncontrolled hypertension receiving usual care at county clinics during the same period, identified from electronic records and matched on age group and sex.

Research Question and Hypotheses

The primary research question is whether adults in the community program experienced a greater reduction in systolic blood pressure over six months than similar adults receiving usual clinic care. The null hypothesis is that the mean six-month change in systolic pressure does not differ between program and clinic participants. The alternative hypothesis is that the mean change differs between groups. A two-sided test is used because a program could, in principle, perform worse than usual care. Secondary questions ask whether the share of participants reaching controlled pressure, below 140/90 mm Hg, differs between groups and whether the program's effect remains after adjusting for age, sex, baseline pressure and medication use.

Study Design

The evaluation is an observational comparison of two groups with measurements at baseline and six months. Participants were not randomly assigned, so the groups may not be alike on factors that shape blood pressure, such as motivation or access to care. Adjusting for measured characteristics can reduce, but not eliminate, this concern.

Variables

Table 1 defines the variables. The outcome for the primary question is the change in systolic pressure, a continuous variable measured in mm Hg, calculated as the six-month value minus the baseline value, so negative numbers indicate improvement. The main exposure is group membership, a categorical variable with two levels. Covariates include age, a continuous variable in years; sex, categorical; baseline systolic pressure, continuous; and number of blood pressure medications at baseline, a discrete count.

Table 1. Variables for the Evaluation

Note. Variable definitions are fixed before analysis.

Keeping the Outcome Continuous

Program partners asked that the main result be the percentage of participants whose pressure became controlled. Altman and Royston (2006) explained that turning a continuous measure into two categories loses information and statistical power and treats values near the cutoff as if they were very different. A participant whose pressure falls from 175 to 145 mm Hg improves substantially yet remains uncontrolled, while one who moves from 141 to 139 becomes controlled with almost no change. Using the continuous change as the primary outcome captures the program's full effect, and the control rate is reported as a secondary outcome because clinicians and partners find it meaningful.

Estimation Rather Than Testing Alone

Gardner and Altman (1986) urged researchers to report confidence intervals for the size of effects, not just p-values, because intervals convey how large an effect might be and how precisely it has been estimated. The primary result will therefore be the difference in mean change between groups with its 95% confidence interval, accompanied by a p-value. A clinically meaningful difference is defined in advance as 5 mm Hg, which is associated with meaningful reductions in cardiovascular risk, so that results can be judged against a practical standard as well as a statistical one.

Checking Model Capacity

For the secondary outcome, logistic regression will estimate the odds of reaching controlled pressure in the program compared with usual care, adjusting for covariates. Peduzzi et al. (1996) found in simulations that logistic regression models with fewer than about ten outcome events per predictor variable produced biased and unreliable estimates. With an expected 160 or so participants reaching control across both groups and five predictors, the model has roughly thirty events per variable, comfortably above that threshold.

Comparing the Groups at Baseline

Because participants were not randomized, the analysis will first compare the groups at baseline on every covariate, using means and standard deviations for continuous variables and percentages for categorical ones. Standardized differences, the difference in means divided by the pooled standard deviation, will be reported rather than p-values, since with samples of this size small but meaningless differences can reach statistical significance and large ones in small subgroups may not. Standardized differences above about 0.1 will be flagged as meaningful imbalances to be addressed through adjustment. Preliminary records suggest program participants were somewhat younger and had slightly higher baseline pressure, differences that make adjustment essential and that could, if ignored, exaggerate the program's apparent effect through regression to the mean.

Data Quality and Missing Values

Blood pressure readings in community settings were taken with automated devices using a standard protocol, while clinic readings came from routine visits, which may differ in technique. Six-month values are missing for about 9% of program participants and an unknown share of comparison patients. These issues will be examined in the analysis plan in the next milestone.

Analysis Approach

The primary comparison will use a two-sample t-test of mean change, supported by linear regression adjusting for covariates. The secondary comparison will use a chi-square test and logistic regression. All results will be reported with 95% confidence intervals.

Conclusion

The evaluation asks a clear question, uses a continuous primary outcome, reports effects with confidence intervals against a predefined meaningful difference and plans models within the capacity of the data. Framing these choices now protects the analysis from decisions made after seeing the results.

References

Altman, D. G., & Royston, P. (2006). The cost of dichotomising continuous variables. BMJ, 332(7549), 1080. https://doi.org/10.1136/bmj.332.7549.1080

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

Peduzzi, P., Concato, J., Kemper, E., Holford, T. R., & Feinstein, A. R. (1996). A simulation study of the number of events per variable in logistic regression analysis. Journal of Clinical Epidemiology, 49(12), 1373-1379. https://doi.org/10.1016/S0895-4356(96)00236-3