A quantitative research article can be fairly easy to follow until you reach the Results section. Suddenly, there are means, standard deviations, confidence intervals, t values, F statistics, odds ratios, degrees of freedom, and p values competing for attention.

It can be tempting to scan for p < .05, decide whether the findings were statistically significant, and move on. That shortcut leaves out much of what clinicians need to know. DNP students need enough statistical literacy to determine what researchers measured, whether the analysis fits the research question and data, how large the observed effect was, how much uncertainty surrounds it, and whether the finding is meaningful enough to influence practice (Grove & Cipher, 2025).

Start With the Research Question and Study Design

Before interpreting a statistical test, identify what the researchers were trying to determine. Look at who was studied, what intervention or exposure was examined, which outcomes were measured, and whether the researchers were describing a population, comparing groups, examining relationships, or trying to predict an outcome.

The structure of the data affects the analysis. Independent samples contain different participants in each group, while paired samples involve related observations, such as measurements taken from the same participants before and after an intervention. Level of measurement also influences which statistical methods are appropriate (Grove & Cipher, 2025).

A simplified guide to several common analyses looks like this:

Research question	Common analysis
Compare a continuous outcome between two independent groups	Independent-samples t test
Compare a continuous outcome in the same participants at two time points	Paired-samples t test
Compare means across three or more independent groups	One-way ANOVA
Compare categorical frequencies or proportions	Chi-square or Fisher’s exact test
Examine a linear relationship between two continuous variables	Pearson correlation
Examine an ordinal or ranked relationship	Spearman correlation
Predict a continuous outcome using one or more predictors	Linear regression
Analyze time until an event occurs	Survival analysis

This is only a starting point. Real studies may involve repeated observations, multiple predictors, confounding variables, missing data, or more complex designs. The practical question is whether the selected analysis fits the study design and the structure of the data.

Read the Descriptive Statistics First

Descriptive statistics tell you who was studied and what happened within the sample before researchers make inferences about a larger population.

Start with the sample size and participant characteristics. Look at age, clinical characteristics, baseline measurements, and whether comparison groups appear reasonably similar. Continuous variables are often summarized using means and standard deviations when the distribution is reasonably symmetrical, while medians and interquartile ranges may be more informative for skewed data. Categorical variables are generally reported using counts and percentages (Grove & Cipher, 2025).

Suppose an intervention study reports that mean systolic blood pressure decreased from 148 mm Hg to 134 mm Hg. That result already tells you something clinically useful. Inferential statistics then help evaluate the uncertainty surrounding the observed difference and what conclusions can reasonably be drawn beyond the study sample.

Make Sure the Analysis Fits the Data

Statistical tests are not interchangeable. Researchers need to consider whether the data structure and assumptions are appropriate for the selected analysis.

Relevant issues may include whether observations are independent or paired, the level of measurement, the distribution of the data, sample size, and variability across groups (Grove & Cipher, 2025). An independent-samples t test, for example, is designed for two unrelated groups, while a paired-samples t test is appropriate for related observations.

Readers do not need to reproduce every calculation to appraise a study. They do need to recognize when the statistical method does not make sense for the design and data the researchers collected.

Stop Treating p < .05 as the Answer

The p value is probably the most familiar statistic in healthcare research, and it is often given more meaning than it carries.

A p value does not tell you the probability that the null hypothesis is true. It does not tell you the probability that the findings were caused by chance, and it does not tell you whether an effect is large or clinically important (Wasserstein & Lazar, 2016).

In simplified terms, the p value describes how incompatible the observed data are with a specified statistical model, usually including the assumption that the null hypothesis is true. Many studies use an alpha level of .05, so a result below that threshold is conventionally described as statistically significant.

That threshold should not substitute for interpretation. Two otherwise similar studies reporting p = .049 and p = .051 should not suddenly be treated as fundamentally different bodies of evidence. Likewise, a nonsignificant result does not automatically establish that no effect exists. Sample size, effect magnitude, variability, and statistical power all influence whether a study can detect an effect reliably (Grove & Cipher, 2025).

Find the Effect Estimate and Confidence Interval

After identifying what was compared, find the number that describes the magnitude of the finding. Depending on the study, that may be a mean difference, risk difference, relative risk, odds ratio, hazard ratio, correlation coefficient, regression coefficient, or standardized effect size.

Consider two hypothetical interventions intended to reduce systolic blood pressure. One large study reports an average reduction of 1 mm Hg with p < .001. A smaller study reports an average reduction of 8 mm Hg with p = .06. Looking only at statistical significance would favor the first intervention even though the second study estimated a much larger effect.

The confidence interval adds information about the precision of that estimate. Consider:

Mean systolic blood pressure difference: -8 mm Hg, 95% CI [-12, -4].

The estimate suggests an average reduction of 8 mm Hg, and the relatively narrow interval remains entirely on the side of a reduction. Now compare it with:

Mean difference: -8 mm Hg, 95% CI [-18, 2].

The point estimate is identical, but the second interval is substantially wider and includes values consistent with a large reduction, little difference, or a small increase. The possible effect remains clinically interesting, but the estimate is much less precise.

The null value is generally 0 for difference measures such as mean difference or risk difference and 1 for ratio measures such as relative risk, odds ratio, and hazard ratio. Confidence intervals should be interpreted alongside the effect estimate rather than reduced to whether they cross the null value.

Statistical reporting guidance recommends emphasizing effect estimates and measures of precision, including confidence intervals, instead of relying exclusively on significance testing (Lang & Altman, 2015).

Statistical Significance Is Not Clinical Significance

Statistical significance addresses the evidence against a specified null hypothesis under the statistical model. Clinical significance asks whether the magnitude of the finding is important enough to affect patients or practice.

Imagine an intervention that reduces average pain from 6.4 to 6.2 on a 10-point scale. With a sufficiently large sample, that 0.2-point difference could be statistically significant even if patients barely perceive the change.

When an established minimal clinically important difference or similar threshold exists, it can help determine whether an observed effect is likely to be meaningful. Clinical importance also depends on the intervention itself. A modest benefit may be worthwhile when an intervention is safe, inexpensive, accessible, and easy to implement, while the same magnitude of benefit may be less compelling when treatment carries substantial risk, cost, or burden.

Statistical results provide information for clinical judgment, but interpretation also requires attention to patient outcomes, harms, feasibility, and context.

Keep the Statistics Connected to the Study Design

Sophisticated statistical analysis cannot compensate for fundamental weaknesses in research design. Randomized trials, quasi-experimental studies, cohort studies, case-control studies, cross-sectional studies, and descriptive studies answer different questions and carry different risks of bias.

Regression can account statistically for measured variables included in a model, but it cannot turn an observational study into a randomized trial or adjust for an important confounder that was never measured. A large sample cannot eliminate selection bias, and a very small p value cannot rescue unreliable measurement.

Missing data can also affect interpretation, particularly when participants who drop out differ systematically from those who remain. When reviewing a study, consider how much data were missing, whether losses differed between groups, and how the researchers handled them (Grove & Cipher, 2025).

The larger question is whether the design, measurement, analysis, and data quality support the conclusion the authors are making.

Use a Consistent Reading Sequence

Quantitative research becomes easier to interpret when the Results section is approached in roughly the same order each time. Start with the research question and study design, then look at the sample and descriptive statistics so you understand who was studied and what happened before inferential analysis was applied.

Next, ask whether the statistical method fits the data and design. Find the effect estimate and confidence interval to evaluate direction, magnitude, and precision, then consider the p value as one part of the interpretation.

Finally, return to the clinical question. Is the effect large enough to matter? Is the study vulnerable to important bias or confounding? Does the sample resemble the population where the findings might be applied? Are the results consistent with other evidence, and would the intervention be feasible in practice?

I may never calculate an ANOVA by hand after graduate school, but I expect to spend the rest of my career deciding whether quantitative evidence is strong enough to influence patient care. Learning how to make that judgment is the statistical skill worth keeping.

References

Grove, S. K., & Cipher, D. J. (2025). Statistics for nursing research: A workbook for evidence-based practice (4th ed.). Elsevier.

Lang, T. A., & Altman, D. G. (2015). Basic statistical reporting for articles published in biomedical journals: The “Statistical Analyses and Methods in the Published Literature” or the SAMPL guidelines. International Journal of Nursing Studies, 52(1), 5–9. https://doi.org/10.1016/j.ijnurstu.2014.09.006

Wasserstein, R. L., & Lazar, N. A. (2016). The ASA’s statement on p-values: Context, process, and purpose. The American Statistician, 70(2), 129–133. https://doi.org/10.1080/00031305.2016.1154108