# What are confidence intervals?

In the study by Smith (2013) in this issue on readmissions in patients with congestive heart failure (CHF), some results were reported with confidence intervals, abbreviated CI. For example, across studies, a statistically significant 20% (95% CI 831) reduction occurred in mortality from all causes when telemonitoring was used. This type of statistic has been used increasingly in reports of nursing research, and readers should have a general understanding of what this statistic communicates. Confidence intervals provide more information than p values (West & Dupras, 2012). This is important when we are using data to make evidence-based practice decisions. When we use statistical significance, such as the p value, we are saying the results occurred greater than chance alone or that the difference was greater than zero. But p values do not indicate the size of the difference (Young & Lewis, 1997). A relatively small change can be statistically significant and yet have little clinical significance. That is where the confidence intervals can help.

What Do Confidence Intervals Mean?

A confidence interval is a range of values (also may be given as a number) that estimates the percentage something exists in the population based on the sample studied. It tells us how precise the data are (Clarke, 2012). Prior to conducting the study, the researcher chooses the confidence level, which in the Smith (2013) study was 95%. This reflects a significance level of 0.05. What the 95% means is that the resulting interval reported (8%-31%) will occur in multiple samples of the population approximately 95% of the time. The other confidence level that is frequently used is 99%. This reflects a 0.01 significance level (Dorey, 2010).

With these figures, researchers are indicating they are 95% or 99% confident the true value from the populations is within the confidence interval given. In the study under discussion, Smith (2013) noted the reduction in all-cause mortality related to telemonitoring was 20%, with 95% confidence the reductions in mortality from all causes attributed to telemonitoring in similar studies would be 8%-31%. This does not mean values in future studies may not fall outside the interval limits, just that they are considered improbable (Veldhuizen, PaskerDeJong, & Atsma, 2012).

The confidence interval is used to support the reliability of this estimate. It is important to note the author is not predicting the actual true parameters of the population. Certain factors may affect the confidence interval, including size of the sample, level of confidence, and population variability. A larger sample typically will lead to more confidence in the findings or, in other words, larger samples usually demonstrate a smaller or tighter confidence interval. For example, a confidence interval of 3 is more reliable than a confidence interval of 20 (Clarke, 2012).

Interpreting Confidence Intervals

When reviewing results that include confidence intervals, the reader should be concerned about two things: a CI that contains a value that indicates no change, and whether the range of values is of clinical significance. If the values in the range include (or cross) zero, this indicates no statistically significant change. For example, if there was a 95% CI of -5 to 10, this would suggest the treatment was not effective. Also, if the effect of the treatment was small, perhaps only 1% improvement, the reader would question the benefit of such a treatment (Fethney, 2010). Clinical significance will vary depending on what is being studied and requires clinical judgment.

The confidence interval provides a range of values that we can believe, with a given level of confidence, contains the true value of a variable in the larger population that is being estimated by the sample in a particular study. When comparing studies on a common subject, such as in the Smith (2013) article, comparing confidence levels is considered to be a better way of summarizing the results than comparing the results and the p values (Dorey, 2010).

REFERENCES

Clarke, J. (2012). What is a CI? Evidence-Based Nursing, 15(3), 66.

Dorey, F.J. (2010). Confidence intervals: What is the real result in the target population? Clinical Orthopaedics and Related Research, 468(11), 3137-3138.

Fethney, J. (2010). Statistical and clinical significance, and how to use confidence intervals to help interpret both. Australian Critical Care, 23(2), 93-97.

Smith, A.C. (2013). Effect of telemonitoring on re-admission in patients with congestive heart failure. MEDSURG Nursing, 22(1), 39-44.

Veldhuizen, I., Pasker-DeJong, P., & Astma, E (2012). Significance or relevance: What do you use in large samples? About p values, confidence intervals, and effect sizes. Transfusion, 52(6), 1169-1171.

West, C.P., & Dupras, D.M. (2012). 5 ways statistics can fool you-Tips for practicing clinicians. Vaccine. Advance online publication, doi: 10.1016/j.vaccine.2012.11.086

Young, K.D., & Lewis, R.J. (1997). What is confidence: Part 1:The use and interpretation of confidence intervals. Anna/s of Emergency Medicine, 30, 307-310.

Lynne M. Connelly, PhD, RN, is an Associate Professor and Director of Nursing, Benedictine College, Atchison, KS. She is Research Editor for MEDSURG Nursing.