P r a c t ic e M a t t e r s R e s e a r c h 101

Sample size in quantitative research Sample size will affect the significance of your research.

By Susan B. Fowler, PhD, RN, CNRN, FAHA, and Valerie Lapp, PhD, RN, NEA-BC, CPN

Editor’s note: This is part o f the American Nurse Today Research 101 series. To read other articles in the series, visit americannursetoday.com/category/Researcbl01.

You’ve probably been asked (or have asked) the question: How many subjects do I need for my re­ search study? That’s your sample size—the number of participants needed to achieve valid conclusions or statistical significance in quantitative research. (Quali­ tative research requires a somewhat different approach.

In this article, we’ll answer these questions about sample size in quantitative research: Why does sample size matter? How do I determine sample size? Which sampling method should I use? What’s sampling bias?

Why does sample size matter? When sample sizes are too small, you run the risk of not gathering enough data to support your hypotheses or expectations. The result may indicate that relation­ ships between variables aren’t statistically significant when, actually, they are. You also may be missing sub­ jects who might give a different answer or perspective to your survey or interview. Samples that are too large may provide data that describe associations or relation­ ships that are due merely to chance. Large samples al­ so may waste time and money.

How do I determine sample size? Larger sample sizes typically are more representative of the population you’re studying, but only if you collect data randomly and the population is heterogeneous. Large samples also reduce the chance of outliers. How­ ever, large samples are no guarantee of accuracy. If your population of interest is homogenous, you may need only a small sample.

If you’re studying subjects over longer periods of time, as in longitudinal designs, you can expect subject attrition. Know your population and how responsive they may be to repeated questionnaires and interven­ tions. Even if you’re not conducting a longitudinal study, be realistic about how many people would agree to par­ ticipate in research.

For a pilot study (a small-scale version of a bigger study testing the efficacy of an intervention), you’d usually need around 30 subjects, although that number varies according to different experts.

No matter the type of study you’re conducting, take into account time (yours and the subjects’), subject co­ operation, and resources (such as statistical assistance, access to subjects, managerial support for your study, and co- or sub-investigators).

I

r

I

i

Power analysis Power analysis is a robust way to determine sample size and decrease the risk of type II errors (false-negative conclusions that a finding was due to chance when ac­ tually it was the result of the intervention). A power analysis calculation includes a significance criterion, ef­ fect size, and power to arrive at a sample size. The sig­ nificance criterion is referred to as alpha and usually is set at 0.05, which means that in 5 of 100 situations the result would be due to chance and not the intervention. Effect size (usually described as small, moderate, or large) is the magnitude or strength of the relationship between the variables you’re studying. In nursing, we often propose that variables moderately affect one another

AmericanNurseToday.com May 2019 A merican Nurse Today 61

or are correlated. For example, when on­ cology nursing stud­ ies about the effec­ tiveness of symptom management interven­ tions were combined and analyzed, a mod­ erate to large effect was found. Power (1- beta) usually is set at .80, which means that there’s a 20% risk of committing a type II error. (See Feel the power?)

Which sampling method should I use? The sampling method isn’t the same as the sample. It’s the proce­ dure you’ll use to select study participants. We’ll look at two sampling methods: nonproba­ bility and probability.

Nonprobability sampling Convenience sampling and snowball sampling are common nonprobability methods. Convenience samples consist of people who are easily accessed and volunteer; however, the sample may not be representative of the population of interest in your study. Convenience sampling is considered the weakest form of sampling.

With snowball sampling, participants are referred by other participants. This method can be used when you have difficulty locating participants. For example, when interviewing undocumented immigrants, the researcher gains the trust of a few participants and relies on them to identify other undocumented immigrants who might participate.

Probability sampling With probability sampling, everyone in an identified population has an equal chance of being in the sam­ ple. You can use a variety of approaches, including simple random, stratified random, multistage cluster, and systematic random sampling. For example, system­ atic random sampling of patients on a medical-surgical floor for an intervention study may include selecting every sixth room number. (Visit bit.ly/2FZLzYX to learn

more about types of probability sampling.)

What’s sampling bias? Sampling bias can occur w hen a partic­ ular overrepresen­ tation or underrep­ resentation of the population occurs. For example, if a re­ searcher wants to study which method of education is more effective by gender in reducing hospital readmissions, the num ber of men and wom en should be evenly distributed. Bias occurs when the researcher deliberate­ ly omits or makes a conscious decision to exclude a participant w ho’s had several re­ admissions for exac­ erbation of his heart failure. Both omis­ sions reflect bias and may distort study re­

sults and underm ine the validity of the study.

What are the practice implications? As nurses becom e more involved in evidence-based practice projects and research investigations, they’ll need to understand key elements of research, such as sample size, so they can critically appraise and gener­ ate evidence. Remember that the “right” num ber of subjects in your investigation impacts statistical and clin­ ical significance support for your study findings. ★

Susan B. F o w le r is a n u rse s c ie n tis t a t O rla n d o H e a lth in O rla n d o , F lo rid a , m e n ­ t o r f a c u lt y a t T ho m as Edison S ta te U n iv e rs ity in T re n to n , N e w Jersey; a n d co n ­ t r ib u t in g fa c u lt y a t W a ld e n U n iv e rs ity in M in n e a p o lis , M in n e s o ta . V a le rie Lapp is a p ro g ra m m a n a g e r fo r n u rs in g a n d sp e cia l p ro je c ts a n d M a g n e t® c o o rd in a to r a t A rn o ld P a lm e r M e d ic a l C e n te r in O rla n d o , F lo rid a .

Selected references Faber J, Fonseca LM. How sample size influences research outcomes. Dental Press J Orthod. 2014; 19C4):27-P. Polit DF, Beck CT. Nursing Research: Generating a n d Assessing Evi­ dence fo r Nursing Practice. Philadelphia, PA: Wolters Kluwer; 2017. Schmidt SAJ, Lo S, Hollestein LM. Research techniques made simple: Sample size estimation and power calculation. / Invest Dermatol. 2018; 138(8):l678-82.

Feel the power Betty, a pediatric nurse, w ants to study th e effect o f distraction on children’s disco m fo rt d u rin g insertion o f an I.V. catheter before a procedure in th e ra­ d io lo g y d epartm ent. She reaches o u t to experts at her fa cility to help her determ ine how many subjects she needs fo r her study.

Researchers assist her using G*Power, a free o nline pow er analysis tool. (V anderbilt U niversity also has a free pow er and sample size calculation program th a t can be d o w n lo a d e d at b io sta t.m c.va n d e rb ilt.e d u /w iki/M a in / PowerSampleSize.)

W ith significance set at 0.05, a m oderate effect size o f 0.3, and pow er at .80, Betty w ill need 82 subjects (see below).

62 American Nurse Today Volume 14, Number 5 AmericanNurseToday.com

Copyright of American Nurse Today is the property of HealthCom Media and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder’s express written permission. However, users may print, download, or email articles for individual use.