evidence & practice / data analysis

RESEARCH METHODOLOGIES

Importance and use

of correlational research
Curtis E, Comiskey C, Dempsey O (2016) Importance and use of correlational research. Nurse Researcher.
23, 6, 20-25. Date of submission: 16 March 2015. Date of acceptance: 11 November 2015. doi: 10.7748/nr.2016.e1382

Abstract
Elizabeth A Curtis PhD,
Background The importance of correlational research has been reported in the literature yet few
MA, MEd
research texts discuss design in any detail.
Assistant professor at the
School of Nursing and Aim To discuss important issues and considerations in correlational research, and suggest ways to avert
Midwifery, Trinity College potential problems during the preparation and application of the design.
Dublin, Dublin, Republic of
Discussion: This article targets the gap identified in the literature regarding correlational research design.
Ireland
Specifically, it discusses the importance and purpose of correlational research, its application, analysis
and interpretation with contextualisations to nursing and health research.
Catherine Comiskey, PhD,
MSc, BA (Mod) Conclusion: Findings from correlational research can be used to determine prevalence and relationships
Head of school and professor among variables, and to forecast events from current data and knowledge. In spite of its many
of healthcare statistics at uses, prudence is required when using the methodology and analysing data. To assist researchers in
the School of Nursing and reducing mistakes, important issues are singled out for discussion and several options put forward for
Midwifery, Trinity College analysing data.
Dublin, Dublin, Republic Implications for practice Correlational research is widely used and this paper should be particularly useful
of Ireland for novice nurse researchers. Furthermore, findings generated from correlational research can be used,
for example, to inform decision-making, and to improve or initiate health-related activities or change.
Orla Dempsey PhD, BA,
Adjunct quantitative Keywords
healthcare lead at the School nurses, nursing research, correlational research, correlation, variables, quantitative, measurement
of Nursing and Midwifery, tools, data analysis
Trinity College Dublin, Dublin,
Republic of Ireland

Correspondence Introduction It also explores the usefulness of correlational

[email protected] Correlational research is concerned with research in evidence-based practice and defends
establishing relationships between two or its continued use.
Peer review more variables in the same population or
This article has been subject between the same variables in two populations Background
to double-blind review and (Leedy and Ormrod 2010). In nursing, The importance of correlation research
has been checked using exploring the relationship among variables has been emphasised by authors, such as
antiplagiarism software is a significant part of research (Prematunga Woodworth (1938) and Cronbach (1957).
2012). Understanding the associations and Woodworth (1938) established two major
Conflict of interest relationships that exist among ‘human distinctions in quantitative research methods
None declared phenomena is an abiding impetus for – the distinction between independent and
scientific enquiry in all of the social science dependent variables; and the distinction
disciplines, and that impetus transcends between experimental and correlational
even the most polarised paradigmatic methods. According to Woodworth,
distinctions between various research methods’ experimental methods are characterised by the
(Fitzgerald et al 2004). manipulation of variables while a correlational
The purpose of this paper is to discuss design measures two or more characteristics
important issues and considerations in and then calculates the correlation between the
correlational research, and suggest ways to characteristics.
avert potential problems during preparation In Woodworth’s view, these two
and application of the design and provides approaches had equal value – correlational
examples from studies where appropriate. research must ‘be distinguished from the

20 / July 2016 / volume 23 number 6 nurseresearcher.com

Downloaded from RCNi.com by ${individualUser.displayName} on Jul 20, 2016. For personal use only. No other uses without permission.
Copyright © 2016 RCNi Ltd. All rights reserved.
experimental method, but standing on a par appropriate and targeted service, planners Online archive
with it in value, rather than above or below’ need to know if there are any additional For related information,
(Woodworth 1938). characteristics in the client group that may visit our online archive
and search using
Cronbach (1957) was also concerned be related to prevalence. If it is suspected the keywords
that correlational research was considered that another characteristic is important, the
second‑rate in scientific psychology. He believed secondary objective of the study will be to
that a synthesis should take place, with measure the strength and direction of any
advocates of each design valuing them equally possible relationships between prevalence
and adopting both strategies: ‘It is not enough and this characteristic. Correlational research
for each discipline to borrow from the other. is used to address this secondary objective
Correlational psychology studies only variance (McLaren 2013).
among organisms; experimental psychology
studies only variance among treatments. Important considerations
A united discipline will study both of these, Correlational research is used frequently in
but it will also be concerned with the otherwise healthcare research because it can be used in
neglected interactions between organismic and any study that does not wish (or is unable) to
treatment variables.’ Whether this synthesis manipulate the independent variable(s) being
has occurred remains debatable. investigated. But, before a research study
Few nursing textbooks discuss correlational can begin, several factors must be decided,
research in any detail (Bobko 2001, Miles and including the selection of suitable variables,
Shevlin 2001, Vogt and Johnson 2012), with identification and selection of the sample, and
several providing just one to three pages in the use of reliable tools.
chapters entitled ‘non‑experimental designs’
or ‘quantitative research designs’. Variables
Correlational research can be described in A variable is a characteristic of interest
different ways. Cohen (1968) suggested that (Comiskey and Dempsey 2013) that can vary
because parametric analyses such as t-tests, among individuals and organisations (Polit
analysis of variance (ANOVA) and analysis of and Tatano Beck 2012), such as weight,
covariance (ANCOVA) explore relationships height, blood pressure, happiness or innovative
among variables, quantitative studies would organisational culture. Two variables may
therefore produce correlational evidence. But coexist but it is only through research that
describing research by the analysis techniques a relationship can be demonstrated between
used is not helpful because all studies then fall them or the direction and strength of that
into a single category. relationship established.
The purpose of correlational research is to Quantitative researchers try to establish
investigate ‘the extent to which differences in how and why things vary, and determine how
one characteristic or variable are related to differences in one variable are associated with
differences in one or more other characteristics differences in another.
or variables’ (Leedy and Ormrod 2010). All correlational studies require a conceptual
A correlation occurs if one variable (X) framework or a description of why the
increases and another variable (Y) increases or variables might be related to one another.
decreases. A study that produces a correlation It is important that the researcher accurately
coefficient of 0.00 signifies that there is no determines the variables that exist in the area
association between the variables investigated. of study (Maltby et al 2010). All variables
This paper focuses on three themes: relevant to the research questions formulated
using correlational research; analysing for a research study must be clearly defined;
and interpreting data from correlational to determine correlation between variables,
research; and implications for nursing and the level of measurement of all variables must
healthcare provision. be ‘ordinal’, ‘interval’ or ‘ratio’.

Using correlational research Sampling

Many healthcare studies arise from the need Quantitative research designs – including
to quantify the numbers of clients using a correlational studies – usually use large
particular service in a specific time period or to samples that have been attained through a
measure the absence or presence of a particular precise process. This is important because the
characteristic in a population or client group. purpose of sampling in quantitative studies
In such cases, the primary objective is to is to produce statistically representative data
ascertain how many clients are using or in that permit the generalisation of findings to the
need of the service. However, to provide an target population (Norwood 2010).

nurseresearcher.com volume 23 number 6 / July 2016 / 21

Downloaded from RCNi.com by ${individualUser.displayName} on Jul 20, 2016. For personal use only. No other uses without permission.
Copyright © 2016 RCNi Ltd. All rights reserved.
evidence & practice / data analysis

Before selecting a sampling method, and clients had not been randomly assigned
therefore, it is necessary to review the aim of to each treatment group.
the study.
Analysis techniques
Measurement tools Before conducting a statistical test to determine
Nursing research requires good quality, reliable if a relationship exists between two continuous
data. Good quality data result from good variables (Comiskey and Dempsey 2013),
quality research, which is partially due to it is useful to plot the data on a scatterplot
meticulous decisions about data collection and to obtain an indication of whether the
measurement (Polit and Tatano Beck 2012); variables are related and if so, the direction
reliable data are objective, accurate, valid of the relationship (Pallant 2011). If a linear
(measure what they are supposed to), free from relationship is indicated on the scatterplot,
error and usable (Norwood 2010). a statistical test can be run to determine the
Research tools can be used to measure strength of the relationship. Statistical tests to
variables either directly or indirectly. Direct determine relationships between dependent
measures are used when the variable being and independent variables are chosen based on
investigated is tangible or accessible – for the type of data collected. Parametric tests and
example, blood glucose levels, level of their non-parametric alternatives are presented
consciousness or temperature. with examples in Table 1.
Questions that obtain information about For two continuous, normally distributed
personal characteristics, such as age, marital variables, Pearson’s r is the statistical test
status or level of education, are also direct to use to determine if a relationship exists
measures. Indirect measures generate between the two variables and the strength
information ‘about indicators of an attribute… of that relationship. Spearman’s ρ is the
of a phenomenon of interest rather than non-parametric alternative to Pearson’s r for
about the phenomenon itself’ (Norwood non-normal, continuous random variables
2010) – that is, the variable being measured (Comiskey and Dempsey 2013), and is also
is presumed to reflect the variable of interest appropriate for assessing the relationship
and inferences can be drawn from it about between two ordinal variables, such as two
the phenomenon. variables measured on Likert scales – the
It is therefore important that appropriate correlation coefficient is calculated by ranking
measurement tools are used in studies (Lehane participants’ responses, with the highest
and Savage 2013). Furthermore, a good score on the Likert scale assigned the highest
description of all measurement tools should be rank. While correlations cannot be computed
given, and the reliability and validity for the for purely categorical data (Comiskey and
tools reported (Watson 2013). Dempsey 2013), statistical tests can be
conducted to determine if associations exist
Analysing and interpreting data between categorical variables.
It is important to distinguish between Multiple correlational analysis measures how
correlational statistics and correlational well a dependent variable can be predicted
research. If a Pearson’s r is found in a using a linear function of a set of independent
research paper, it is easy to assume that a variables. The multiple correlation coefficient
correlational design was used in the study. R measures the strength of the relationship
However, correlation as a statistical tool between the independent and dependent
and a research design are different – while variables. Similar to r and ρ, R takes a value
Pearson’s r is a statistical test frequently between 0 and 1; however, unlike r and ρ, it
used in correlational studies to calculate the only indicates the strength, not the direction,
relationship between variables, it can also be of the association.
used in an experimental study to determine It is initially useful in correlational
the relationship between independent and research to summarise the data obtained
dependent variables. using frequencies and descriptive statistics.
Similarly, a t-test or an ANOVA does not Frequencies are counts of data and are
indicate that an experimental design was particularly useful for examining the number
used (Goodwin 2002). For example, Cox and percentage of participants in each
and Comiskey (2007) used three treatment category. O’Brien et al (2014) used frequencies
groups to look at the baseline characteristics to ascertain the number of patients with
of clients entering opiate substitution treatment different types of tumour, when examining the
for heroin. However, for ethical reasons, cost of febrile neutropenic events in oncology.
an experimental design could not be used Descriptive statistics, such as mean and

22 / July 2016 / volume 23 number 6 nurseresearcher.com

Downloaded from RCNi.com by ${individualUser.displayName} on Jul 20, 2016. For personal use only. No other uses without permission.
Copyright © 2016 RCNi Ltd. All rights reserved.
median, are used to describe characteristics another, so while correlation may indicate a
of a sample. relationship between two variables, it does not
The standard deviation is useful to describe account of other underlying variables that may
how much the data varies from the mean. be causing the effect.
Curtis and Glacken (2012) used mean and Another common error is to make inferences
standard deviation when analysing job about a group. Correlational analysis may
satisfaction among nurses. Confidence intervals indicate there is a relationship between two
(Crispino 2013) can be particularly useful variables; however, it is not always possible to
in prevalence studies to provide a range of make inferences about the general population
values in which the true population mean or based on a correlation.
proportion lies (Watson and Coombes 2009). Frequently, researchers assume a correlation
There are a number of common mistakes of zero implies independence; however, it
in the analysis and interpretation of findings merely indicates no linear relationship exists
from correlational analyses. One of the between the two variables. While correlation
most common is to assume that correlation implies that an independent variable and a
represents a causal relationship. However, dependent variable may be related, it does
correlation examines if a linear relationship not imply that a change in one variable
exists between an independent and a dependent leads to a change in the other (Plichta and
variable; causality is when one event causes Kelvin 2013).

TABLE 1. Examples of statistical tests used in healthcare research

Test Objective Parametric Non-parametric Example in healthcare research Independent variable Dependent variable

Chi-square Exploring Test for association between age One categorical One categorical
test for relationships and group and risk factors related to variable (age group) variable (fall risk)
independence comparing groups frailty (Ballard et al 2013)

Fisher’s exact Comparing groups X Evaluating nursing roles in Ireland One categorical One categorical
test (Begley et al 2010) variable (nursing role) variable (grouped
waiting times)

Independent Comparing mean X Comparing the health outcomes One categorical One continuous
samples T-test scores for two of opiates users in two variable (treatment variable (health
independent different treatment modalities group) outcome)
groups (Comiskey et al 2009)

Mann-Whitney Comparing median X Comparing gait speed among One categorical One continuous
U-test scores for two younger and older adults variable (group) variable (gait speed)
independent (Benzinger et al 2014)
groups

One-way Comparing mean X Comparing average waiting times One categorical One continuous
analysis of scores among three for services (Begley et al 2013) variable (nursing role) variable (waiting times
variance or more groups for nurse‑led services)
(ANOVA)

Kruskal-Wallis Comparing ranked X One categorical One continuous

one-way ANOVA means among three variable (age group) variable (perceived
or more groups stress score)

Multiple Explores best X Examining predictors of pre- Two or more One continuous
regression predictors of an hospital delay time in patients continuous (age) or variable (pre-hospital
outcome with acute coronary syndrome dichotomous variables delay time)
(McKee et al 2013) (taking medications:
yes=1/no=0)

Logistic Explores predictors X Examine if a difference exists in Two or more One categorical
regression of a categorical symptom presentation in men continuous (age) or dependent variable
outcome and women with acute coronary categorical variables (smoker: yes/no)
syndrome (ACS) (O’Donnell et al (gender)
2012)

nurseresearcher.com volume 23 number 6 / July 2016 / 23

Downloaded from RCNi.com by ${individualUser.displayName} on Jul 20, 2016. For personal use only. No other uses without permission.
Copyright © 2016 RCNi Ltd. All rights reserved.
evidence & practice / data analysis

Implications for nursing elements of multivariate models, including

and healthcare provision factor analysis and multiple regression
Correlational research and models, and can be used to analyse causal
evidence‑based practice models involving latent variables. It is gaining
True experimental designs potentially provide popularity among nurse researchers, but it is a
the highest evidence for causal relationships complex procedure.
between variables (Talbot 1995). Experiments The second method uses logic and theory to
test hypotheses to establish causality, and can exclude other possible explanations ‘in support
achieve this through manipulation, control of making a… plausible causal inference’
and randomisation. It is for this reason that (Thompson et al 2005).
researchers are often encouraged to conduct
true experiments. However, it may not be Continued use of correlational research
possible to carry out experimental research in This is based on practical and ethical
some cases, such as if the research question issues. There are instances when it would
is more suited to a qualitative study or be unrealistic or impossible to use an
ethical approval would not be granted for an experimental design – for example, if
experimental study. researchers wanted to study differences
This does not mean that findings from between age groups, personality types or
research studies using other designs are ethnic behaviours, it might not be possible to
irrelevant to evidence-based practice randomly assign to groups. Furthermore, a
(Thompson et al 2005, Sousa et al 2007). study that sets out to explore the correlation
Correlational research ‘can be used to inform between job satisfaction and absenteeism
causal inferences and thus evidence-based has its own value, even if subjects cannot be
practice’ (Thompson et al 2005). assigned to high job satisfaction and low job
The first method of doing this involves satisfaction conditions (Goodwin 2002).
testing rival causal models statistically, even Further support for correlational research
when using a correlational design, and involves is reported by Wong et al (2013), who carried
the use of structural equation modelling (SEM) out a systematic review to examine the
or covariance structure analysis. SEM is a relationship between nursing leadership and
statistical technique that combines traditional patient outcomes.

References

Ballard J, Mooney M, Dempsey O (2013) Prevalence adults. Aging Clinical and Experimental Research. Cox G, Comiskey C (2007) Characteristics of opiate Goodwin CJ (2002) Research in Psychology:
of Frailty-related Risk Factors in Older adults 26, 4, 377-385. users presenting for a new treatment episode: Methods and Design. Third edition. John Wiley &
seen by Community Nurses Journal of Advanced baseline data from the national drug treatment Sons, New York NY.
Nursing, 69, 3, 675-684. Bobko P (2001) Correlation and Regression: outcome study in Ireland (ROSIE). Drugs:
Applications for Industrial Organizational Education, Prevention and Policy. 14, 3, 217-230. Hanrahan NP, Aiken LH, McClaine L et al (2010)
Begley C, Murphy K, Higgins A et al (2010) Relationship between psychiatric nurse work
Evaluation of Clinical Nurse and Midwife Specialist Psychology and Management. Second edition. Crispino G (2013) Sampling issues in health care environments and nurse burnout in acute care
and Advanced Nurse and Midwife Practitioner Sage Publications, Thousand Oaks CA. research. In Curtis EA, Drennan J (Eds) Quantitative general hospitals. Issues in Mental Health Nursing.
Roles in Ireland (SCAPE) – Final Report. National Cohen J (1968) Multiple regression as a general Health Research: Issues and Methods. Open 31, 3, 198-207.
Council for the Professional Development of data-analytic system. Psychological Bulletin. 70, 6: University Press, Maidenhead.
Nursing and Midwifery, Dublin, Republic of Ireland. Leedy PD, Ormrod JE (2010) Practical Research:
Part 1, 426-443. Cronbach LJ (1957) The two disciplines of scientific Planning and Design. Ninth edition. Pearson
Begley C, Elliott N, Lalor J et al (2013) Differences Comiskey C, Dempsey O (2013) Analysing data psychology. American Psychologist. 12, 11, 671-684. Education, Harlow.
between clinical specialist and advanced from small and large samples and non-normal
practitioner clinical practice, leadership, and Curtis EA, Glacken M (2012) Job satisfaction among Lehane E, Savage E (2010) Questionnaires and
and normal distributions. In Curtis EA, Drennan public health nurses: a national survey. Journal of instruments for health care research. In
research roles, responsibilities, and perceived J (Eds) Quantitative Health Research: Issues and
outcomes (the SCAPE study). Journal of Advanced Nursing Management. 22, 5, 653-663.
Methods. Open University Press, Maidenhead. Curtis EA, Drennan J (Eds) Quantitative Health
Nursing. 69, 6, 1323-1337. Fitzgerald SM, Rumrill PD Jr, Schenker JD (2004) Research: Issues and Methods. Open University
Comiskey CM, Kelly P, Leckey Y et al (2009) Correlational designs in rehabilitation research. Press, Maidenhead.
Benzinger P, Iwarsson S, Kroog A et al (2014) The The ROSIE Study – Drug Treatment Outcomes
association between the home environment and Journal of Vocational Rehabilitation. 20, 2, 143-150.
in Ireland. The Stationery Office, Government
physical activity in community-dwelling older Publications, Dublin, Ireland.

24 / July 2016 / volume 23 number 6 nurseresearcher.com

Downloaded from RCNi.com by ${individualUser.displayName} on Jul 20, 2016. For personal use only. No other uses without permission.
Copyright © 2016 RCNi Ltd. All rights reserved.
 Findings suggest there are such relationships, before causality can be explored (Polit and Author guidelines
including higher patient satisfaction, lower Tatano Beck 2012). Furthermore, the results journals.rcni.com/r/
patient mortality, lower medication errors from correlational research can be used to nr‑author‑guidelines
and lower hospital-acquired infections. generate hypotheses to be tested in quasi-
Such findings are particularly important for experimental and experimental designs.
nursing and can be used together with those The fact that correlational research does not
from similar studies to augment nursing mean causation is regarded as a disadvantage
care practices and procedures. Similarly, (Talbot 1995, Sousa et al 2007). However,
Hanrahan et al (2010) used correlational acknowledging this disadvantage does not
analysis techniques to determine the detract from the significance and contribution
relationship between work environments correlational studies make to research.
and burnout among psychiatric nurses.
The findings demonstrated lower levels of Conclusion
burnout were significantly associated with This paper provides a succinct overview and
better working environments, more effective description of the historical background of
managers, strong relationships between nurses correlational research and when and how it
and physicians and higher nurse-patient ratios. can be used in nursing research. Correlational
Once again, these findings can be used to research enables nurse researchers to ascertain
support changes in the work environment. what client factors are correlated most strongly
Further support for correlational research with the primary outcome of measure, so that
is attributed to its major advantages. It is these questions can be addressed and services
straightforward, can be relatively inexpensive developed appropriately for the client group
and can be completed fairly quickly. Its most strongly correlated with the primary
data also provide a useful starting point outcome or health attribute of interest.
for researchers exploring a phenomenon Furthermore, preliminary correlational
for the first time. Researchers can establish research studies can provide invaluable
the direction and strength of a relationship information about what future research
between variables, enabling further research may be required to investigate the variables
to be carried out – an empirical relationship shown to be correlated with the outcomes
between relevant variables must be established or attributes previously studied.

Maltby J, Williams G, McGarry J et al (2010) O’Brien C, Fogarty E, Walsh C et al (2014) The Polit DF, Tatano Beck C (2012) Nursing Research: Vogt WP, Johnson B (2012) Correlation and
Research Methods for Nursing and cost of the inpatient management of febrile Generating and Assessing Evidence for Regression Analysis. Sage Publications, London.
neutropenia in cancer patients – a micro-costing
Healthcare. Pearson Education, Harlow. study in the Irish healthcare setting. European Nursing Practice. Ninth edition. Wolters Kluwer/ Watson R (2013) Issues and debates in validity
Journal of Cancer Care. 24, 1, 125-132. Lippincott, Williams & Wilkins, Philadelphia PA. and reliability. In Curtis EA, Drennan J (Eds)
McKee G, Mooney M, O’Donnell S et al (2013) Quantitative Health Research: Issues and Methods.
Multivariate analysis of predictors of pre-hospital O’Donnell S, McKee G, O’Brien F et al (2012) Prematunga RK (2012) Correlational analysis. Open University Press, Maidenhead.
delay in acute coronary syndrome. International Gender Symptom Presentation in Acute Australian Critical Care. 25, 3, 195-199.
Journal of Cardiology. 168, 3, 2706-2713. Coronary Syndrome: A Cross Sectional Analysis. Watson M, Coombes L (2009) Surveys. In Neale J
Sousa VD, Driessnack M, Mendes IA (2007) An (Ed) Research Methods for Health and Social Care.
McLaren S (2013) Planning and conducting International Journal of Nursing Studies 49, 11, overview of research designs relevant to nursing:
1325-1332. Palgrave Macmillan, Basingstoke.
surveys. In Curtis EA, Drennan J (Eds) Quantitative part 1: quantitative research designs. Revista
Health Research: Issues and Methods. Open Pallant J (2011) SPSS Survival Manual: A Step Latino-Americana de Enfermagem. 15, 3, 502-507. Wong CA, Cummings GG, Ducharme L (2013) The
University Press, Maidenhead. by Step Guide to Data Analysis Using the SPSS relationship between nursing leadership and
Talbot LA (1995) Principles and Practice of Nursing patient outcomes: a systematic review update.
Miles J, Shevlin M (2001) Applying Regression and Program. Fourth edition. Open University Press, Research. Mosby, St Louis MO.
Maidenhead. Journal of Nursing Management. 21, 5, 709-724.
Correlation: A Guide for Students and
Thompson B, Diamond KE, McWilliam R et al Woodworth RS (1938) Experimental Psychology.
Researchers. Sage Publications, London. Plichta SB, Kelvin E (2013) Munro’s Statistical (2005) Evaluating the quality of evidence from
Methods for Health Care Research. Sixth edition. Henry Holt, New York NY.
correlational research for evidence-based
Norwood SL (2010) Research Essentials: Wolters Kluwer/Lippincott, Williams & Wilkins, practice. Exceptional Children. 71, 2, 181-194.
Foundations for Evidence-Based Practice. Pearson Philadelphia PA.
Education, Harlow.

nurseresearcher.com volume 23 number 6 / July 2016 / 25

Downloaded from RCNi.com by ${individualUser.displayName} on Jul 20, 2016. For personal use only. No other uses without permission.
Copyright © 2016 RCNi Ltd. All rights reserved.