Article Analysis 2
Statistics Used in Current Nursing Research Kathleen Zellner, MSN, RN, BC; Connie J. Boerst, MSN, RN, BC; and Wil Tabb, PhD
ABSTRACT Undergraduate nursing research courses should empha-
size the statistics most commonly used in the nursing litera- ture to strengthen students’ and beginning researchers’ un- derstanding of them. To determine the most commonly used statistics, we reviewed all quantitative research articles published in 13 nursing journals in 2000. The findings sup- ported Beitz’s categorization of kinds of statistics. Ten pri- mary statistics used in 80% of nursing research published in 2000 were identified. We recommend that the appropriate use of those top 10 statistics be emphasized in undergradu- ate nursing education and that the nursing profession con- tinue to advocate for the use of methods (e.g., power analy- sis, odds ratio) that may contribute to the advancement of nursing research.
A dilemma faced by teachers of nursing research in un- dergraduate baccalaureate programs is identifying the key concepts that would be most meaningful to
students as they enter professional practice. Statistics have been an integral part of nursing practice and research since the beginning of modern nursing. Florence Nightingale’s innovative use of descriptive statistics and her use of the now-famous radial graphs, sometimes called coxcombs, rep- resents an early and effective use of statistical analysis of data in nursing research. This marked the spread of the use
of statistical analysis in nursing, which migrated from the larger field of 19th-century biomedical research (Nightin- gale, 1859).
Nursing curricula include a component of statistical methods as part of nursing research courses to enable stu- dents to understand current research and contribute to its ongoing discussion. Learning statistics can be a difficult task for any student, and nursing students often struggle with learning the concept and the transfer of their knowl- edge to nursing literature (Schuster & Ritchey, 1998). Such difficulty can often deter students from furthering their nursing degree (Stranahan, 1995).
In developing statistics courses for undergraduate nurs- ing students, it is important to consider how to teach sta- tistics, what statistics to teach, and what statistics are common in the literature to provide students with a basic comprehension of data that will enable them to interpret statistical results. Therefore, it is important to identify the statistical tests and procedures needed to comprehend mod- ern nursing research.
Nursing journals frequently publish articles in which a basic understanding of statistics is assumed for an educated reading of the research. To guide nursing instructors, we aimed in this research to identify the statistical tests used most frequently in nursing research articles. We anticipate that this information could then help faculty in emphasizing the statistics nurse readers would most likely encounter.
TheoReTiCAl FRAMeWoRK
The teaching of statistics can be linked to the Conversa- tion Theory of Gordon Pask (1975). His theory focused on the premise that “learning occurs through conversations about a subject matter which serve to make knowledge explicit” (Theory Into Practice Database, n.d., Overview Section, ¶1). Pask’s (1975) theory serves as a foundational component to student learning of statistics through the application of cog- nitive learning. His methods of teaching relationships and the use of problem-solving strategies facilitate an under-
Received: October 1, 2004 Accepted: June 30, 2005 Ms. Zellner is Associate Professor and Ms. Boerst is Associate
Professor and Undergraduate Program Director, Bellin College of Nursing, and Dr. Tabb is ad hoc professor, University of Wisconsin Green Bay, Green Bay, Wisconsin.
Address correspondence to Connie J. Boerst, MSN, RN, BC, As- sociate Professor and Undergraduate Program Director, Bellin Col- lege of Nursing, 725 South Webster Avenue, PO Box 23400, Green Bay, WI 54305-3400; e-mail: connie.boerst@bcon.edu.
February 2007, Vol. 46, No. 2 55
STATISTICS USeD IN NURSING ReSeARCH
standing of statistics. Pask’s work provided the early steps for later use of learning theories related to teaching statis- tics. An appropriate assumption made on the basis of Pask’s theory is to teach statistics so students have the advantage of performing statistics while learning the concepts.
liTeRATURe RevieW
Johnson (1984) and Knapp and Miller (1987) wrote two of the early articles addressing the incorporation of statistical concepts and biometrical methods into undergraduate nurs- ing curricula. Johnson (1984) argued that many nursing students have high levels of anxiety regarding their math- ematical ability and recommended that statistical concepts be taught simultaneously with a nursing research course. Knapp and Miller (1987) recommended that in addition to requiring college algebra, statistics, and an introduction to biomedical computing, nursing programs must ensure that these courses are relevant to nursing. Similarly, Stranahan (1995) concluded that being enrolled in both classes simulta- neously is more beneficial to learning. In addition, Schuster and Ritchey (1998) agreed and found that an interdisciplin- ary statistics course incorporating the mathematics of both statistics and nursing research enabled students to inter- pret “the meaning of the statistics within the context of the nursing research literature” (p. 34).
Johnson (1984), Knapp and Miller (1987), Beitz and Wolf (1997), and Robinson (2001) asserted that to make statis- tics meaningful, the examples used in teaching statistical techniques need to be relevant to learners. Beitz and Wolf (1997) recommended that to make statistical concepts seem less abstract, educators should build in concrete activities that make the concepts more meaningful. According to Be- itz (1998), students’ greatest confusion in using statistics is in transferring statistical knowledge and selecting the most appropriate statistical test for a given situation. To that end, Beitz developed an organized table that lists the statistical tests central to nursing research and provides students with an overview of the definition and application of a variety of statistical measures. Robinson (2001) acknowledged Beitz’s work and developed the Guidelines for Statistical Analysis (p. 137), as well as the Golden Rules for Statistical Analysis Adequacy (p. 138).
Authors agree that knowledge of statistics is a vital and essential skill, necessary for the progression of the nurs- ing profession, and needed for students to be able to read, interpret, and integrate nursing research (Robinson, 2001; Schuster & Ritchey, 1998; Taylor & Muncer, 2000).
The literature offers little information regarding which, if any, statistical techniques are used most frequently and which should therefore receive greater emphasis in statistics cours- es. Although Beitz (1998) developed a chart listing the major statistical tests, she provided no explanation of the criteria for a test to be considered major. If frequently used tests could be identified, taught, and emphasized in undergraduate pro- grams, educators could spend more time providing application opportunities and less time on seldom used or higher level sta- tistics that are beyond the needs of beginning learners.
Graduate nursing research courses could build on this knowledge, as Knapp and Miller (1987) suggested. Gradu- ate courses could explore the lesser known but equally as important and higher level statistical techniques. The pur- pose of this article was to identify the most frequently used statistics in nursing research today so that educators would be able to share the results with beginning nursing research- ers, enabling them to apply and critique this information.
A content analysis of current nursing research provides many useful results. It informs readers of current topics of professional interest and summarizes the principle meth- ods used in research. It identifies the topics that might be highlighted in classroom discussions to empower students to become more literate readers of nursing journals. This study joins a number of analyses of statistical methods in professional research (Beitz, 1998; Polit & Sherman, 1990; Robinson, 2001; Taylor & Muncer, 2000), which identify the key statistical tools used in research. As such, it serves as a guide to educators about appropriate classroom topics and a resource to learners of essential research skills.
MeThoD
We reviewed a selection of 462 articles from 13 nursing journals published in 2000. In choosing the journals for in- clusion in the study, we believed it desirous to incorporate those with wide readership, as well as those publishing ex- tensive scholarship. The nursing journals used in this study were identified through a three-step process. First, we used the “Brandon/Hill Selected List of Print Nursing Books and Journals” (B/H) by Hill and Stickell (2002). The B/H is a “guide for nurses and librarians who find themselves responsible for choosing a collection of current nursing lit- erature” (Hill & Stickell, 2002, p. 100); it also includes a comprehensive list of journals identified as key components of any library.
Using the B/H list, we then cross-referenced those jour- nals with the Key and Electronic Nursing Journals: Charac- teristics and Database Coverage, 2001 ed. (KeNJ), a compre- hensive guide of more than 200 national and international nursing journals compiled by Margaret Allen (2001). Jour- nals are included on the KeNJ list on the basis of criteria of peer review; research percentages (i.e., percentages of research articles for the journal for a particular year); inclu- sion in the B/H, the Canadian Nursing Association list, and/ or the Nursing Research Journals Index; or being indexed in the British Nursing Index, CINAHL, PubMed/Medline, and other databases.
The final step was to identify those journals for which the annual percentage of published research articles was 40% or more. Thirteen journals, representing 11 specialty areas, were chosen using this process. See Table 1 for a complete list of the journals selected, including the percentage of re- search published annually and the number of articles re- viewed for this study. each issue of the selected journals was reviewed, and all quantitative research articles (N = 462) were scrutinized for the kinds of statistical methods used. By mutual agreement, we designed a data-reporting grid,
56 Journal of Nursing Education
zeLLNeR, BOeRST, & TABB
focusing on the specific sta- tistical methods used. Be- cause all authors were re- sponsible for data collection, interrater reliability was established via concurrent review of two previously published research articles, with group discussion and review to determine consis- tency of data identification.
ReSUlTS
After all 462 articles were reviewed, it became apparent that regardless of journal orientation or focus, authors used and consistently reported us- ing similar statistical methods. These similari- ties served as the basis for creating a set of categories of statistical methods that would summarize our re- view. The typology used by this study, although devel- oped independently, came to mirror the descriptions of statistical techniques noted in other studies (Beitz, 1998); the results are re- ported in Table 2.
Descriptive Statistics Descriptive statistics were the tools used most frequent-
ly in the nursing research articles. Measures of central ten- dency and dispersion accounted for approximately 63% of the statistics used. Authors assumed readers were famil- iar with the basic concepts of mean, median, variance, and standard deviation, as well as the use of percentiles and percentages in research. Many articles also displayed these statistics in plots or graphs. Although these tools were often carried into more formal discussions, they in themselves of- ten provided useful and relevant information gained from the nursing research. These techniques are an assumed base of knowledge for nurses and are another indication of the ongoing need for nursing students to master math- ematical skills early and comprehensively, as advocated by Knapp and Miller (1987).
inferential Statistics Most authors subjected their instrument to a measure-
ment of reliability; the most commonly applied measure- ment was Cronbach’s alpha. However, the next step of anal- ysis, the examination of the scores of the instrument, varied widely. Many authors simply tabulated the scores or plotted their frequency.
Interestingly, although most researchers used well- known statistical methods, such as t tests, analysis of vari- ance (ANOVA), and regression analysis in its many forms, individual research needs often guided authors to other, lesser known tests. Although it is possible to compile a list of common statistical tests (e.g., the top 10 list presented in Table 3), a comprehensive handbook of statistical methods is always an important reference tool to consult when read- ing nursing research (Polit, 1996).
Additional Findings Articles addressing the nature of biomedical research
(Cummings & Rivara, 2003; International Committee of Medical Journal editors, 1997) have suggested that it would be more appropriate to use odds ratios as a way of express- ing research outcomes. Where the outcomes of logistic re- gression have required it, this method of statistical report- ing was described. In reviewing the articles for this study, we found that this approach is just beginning to be used in nursing research and, in fact, would have been number 11 if the top 10 list were expanded.
Our readings also confirmed that nursing research has not included reference to the appropriateness of sample size, which often limited the generalization of study conclusions (Kachoyeanos, 1998; Polit & Sherman, 1990). Although sam- ple sizes were usually clearly defined and the methods used for sample selection discussed, a closer analysis of sample
TablE 1
Journals Selected for Inclusion in the Study, Percentage of Research Published annually, and Number of articles Reviewed
Journal Title Specialty area % of Research
Published annually No. of articles
Reviewed
American Journal of Critical Care Critical care 73 28
Applied Nursing Research Research 71 22
Cancer Nursing Cancer 69 34
Heart & Lung: The Journal of Acute and Critical Care
Critical care 58 25
Journal of Advanced Nursing Research 65 160
Journal of Community Health Community health 60 12
Journal of Nursing Administration Administration 35a 28
Journal of Nursing Care Quality Administration 41 15
Journal of Nursing Education Education 46 22
Journal of Nursing Scholarship Research 48 27
Journal of Obstetric, Gynecologic, and Neonatal Nursing
Maternal-child 40 24
Nursing Research Research 83 42
Western Journal of Nursing Research Research 73 23
a Although below the cut-off point of 40% research published annually, this journal was selected for its accessibility and wide readership.
February 2007, Vol. 46, No. 2 57
STATISTICS USeD IN NURSING ReSeARCH
size was often lacking. Readers were then required to consult standard works on research design and sampling methods to judge the adequacy of the sample for the research questions explored. Nurse researchers are still not using power analy- sis, which is used to estimate the size of a sample needed to obtain a significant result, in their discussions. Kachoyeanos (1998) noted that “In recent review of nursing research, the overall lack of adequate statistical power became quite clear” (p. 105). In addition, Polit and Sherman (1990) stated that “nursing research needs to pay greater attention to issues of power in designing their studies” (p. 368).
CoNClUSioN
The widespread use of statistical methods in a variety of nursing journals underscores the importance of including statistical skills in nursing education. A review of more than 400 nursing research articles revealed that regardless of journal orientation and focus, the same 10 statistics were re- peatedly used in approximately 80% of the research. These 10 statistics were included in the chart developed by Beitz (1998) and termed “major statistics and statistical tests” (p. 49). We contend that initial statistics and nursing research courses should emphasize these 10 statistics to strengthen students’ and beginning researchers’ understanding.
Pask’s Conversation Theory (1975), while applicable to any teachable subject, provides an extensive overview and
TablE 2
Statistics Used in the Journal articles Reviewed and Frequency of Their Occurrence
Statistic n (%)
Descriptive statistics: Measures of central tendency
Mean 261 (12.4)
Median 41 (2)
Mode 12 (0.6)
Frequency distribution 189 (9)
Graphs and plots
Bar graphs 61 (2.9)
Dot plot 16 (0.8)
Line graph 25 (1.2)
Skew 3 (0.1)
Descriptive statistics: Measures of dispersion
Variance 4 (0.2)
Standard deviation 209 (10)
Range 129 (6.1)
Percentages, percentiles, and quartiles 361 (17.2)
Inferential statistics: Parametric
Z score 4 (0.2)
t test, independent and dependent 124 (5.9)
Analysis of variance (ANOVA), all kinds 100 (4.8)
Multiple comparison/post-hoc tests (e.g., Scheffe, Tukey)
34 (1.6)
Analysis of covariance (ANCOVA) 12 (0.6)
Correlation 109 (5.2)
Cronbach’s alpha 82 (3.9)
Regression, all kinds 77 (3.7)
Odds ratio 27 (1.3)
Discriminant analysis 2 (0.1)
Factor analysis 23 (1.1)
Inferential statistics: Nonparametric
Chi square 114 (5.4)
Mann-Whitney test 18 (0.9)
Kruskal-Wallis test 9 (0.4)
Wilcoxon’s test 13 (0.6)
Fisher’s exact test 12 (0.6)
McNemar test 5 (0.2)
Power analysis 24 (1.1)
Total 2100 (100.1)
Note. Percentages do not equal exactly 100% due to rounding.
TablE 3
Top 10 Statisticsa Used in the Journal articles Reviewed and Frequency of Their Occurrence
Statistic n (%)
Descriptive statistics: Measures of central tendency
Mean 261 (12.4)
Frequency distribution 189 (9)
Descriptive statistics: Measures of dispersion
Standard deviation 209 (10)
Range 129 (6.1)
Percentages, percentiles, and quartiles 361 (17.2)
Inferential statistics: Parametric
t test, independent and dependent 124 (5.9)
Analysis of variance (ANOVA), all kinds 100 (4.8)
Correlation 109 (5.2)
Cronbach’s alpha 82 (3.9)
Inferential statistics: Nonparametric
Chi square 114 (5.4)
Total 1678 (79.9)
a The top 10 statistics represent approximately 80% of all statistical measures used in the 462 articles reviewed.
58 Journal of Nursing Education
discussion of the use of learning statistics. The main empha- sis of the theory is to teach back what a person had learned; in other words, manipulation of the subject matter facili- tates learning. Although not all students learn in the same manner, using Pask’s theory would allow students to learn the subject as well as the relationships among the concepts, which aids in the understanding and application of statis- tics and how this applies to nursing.
At the same time, as nursing research becomes more sophisticated, it is clear that greater understanding of the techniques and issues of quantitative study needs to be em- phasized. The nursing profession should continue to move forward in the use of more advanced statistical analyses, including logistic regression and power analysis.
ReFeReNCeS Allen, M. (2001). Key and electronic nursing journals: Character-
istics and database coverage, 2001 ed. Retrieved November 18, 2003, from http://nahrs.library.kent.edu/resource/reports/ keyjrnls_intro2001ed.pdf
Beitz, J.M. (1998). Helping students learn and apply statistical analysis: A metacognitive approach. Nurse Educator, 23(1), 49- 51.
Beitz, J.M., & Wolf, z.R. (1997). Creative strategies for teaching statistical concepts in nursing education. Nurse Educator, 22(1), 30-34.
Cummings, P., & Rivara, F.P. (2003). Reporting statistical informa- tion in medical journal articles. Archives of Pediatrics & Adoles- cent Medicine, 157, 321-324.
Hill, D.R., & Stickell, H.N. (2002). Brandon/Hill selected list of print nursing books and journals. Nursing Outlook, 50, 100-113.
International Committee of Medical Journal editors. (1997). Uni- form requirements for manuscripts submitted to biomedical journals. Journal of the American Medical Association, 277, 927-934.
Johnson, J.M. (1984). Strategies for teaching nursing research: Strategies for including statistical concepts in a course in re- search methodology for baccalaureate nursing students. Western Journal of Nursing Research, 6, 259-264.
Kachoyeanos, M.K. (1998). The significance of power in research design (part I). MCN, 23, 105.
Knapp, R.G., & Miller, M.C., III. (1987). Some thoughts on biometri- cal training in colleges of nursing. Nurse Educator, 12(6), 5.
Nightingale, F. (1859). A contribution to the sanitary history of the British Army during the late war with Russia. London, UK: John W. Parker and Son.
Pask, G. (1975). Conversation, cognition and learning: A cybernetic theory and methodology. Amsterdam: elsevier.
Polit, D.F. (1996). Data analysis and statistics for nursing research. Stanford, CT: Appleton & Lange.
Polit, D.F., & Sherman, R.e. (1990). Statistical power in nursing research. Nursing Research, 39, 365-369.
Robinson, J.H. (2001). Mastering research critique and statistical interpretation: Guidelines and golden rules. Nurse Educator, 26, 136-141.
Schuster, P., & Ritchey, N. (1998). Teaching introductory statistics to baccalaureate nursing students. Nurse Educator, 23(5), 34.
Stranahan, S.D. (1995). Sequence of research and statistics courses and student outcomes. Western Journal of Nursing Research, 17, 695-699.
Taylor, S., & Muncer, S. (2000). Redressing the power and effect of significance. A new approach to an old problem: Teaching statis- tics to nursing students. Nurse Education Today, 20, 358-364.
Theory Into Practice Database. (n.d.). Conversation theory (G. Pask). Retrieved July 16, 2004, from http://tip.psychology.org/ pask.html
February 2007, Vol. 46, No. 2