NURS5607 Evidence Based Practice
Overview of Quantitative Research I
Dr. Kris LOK
Assistant Professor
School of Nursing, The University of Hong Kong
Courtesy of Dr Naomi Takemura
Learning outcomes
In this lecture, students will be to understand:
• Basic concepts of research paradigm
• Basic research terms
• Quantitative research design
Research paradigm
Positivist Constructivism
Quantitative methodology Qualitative methodology
Deductive reasoning (演繹) Inductive reasoning (歸納)
General principle →prediction Specific observation → develop conclusion
1. Theory or conceptual framework 1. Observe and organize data
2. Problem of interest 2. Identify pattern, commonalities &
3. Developing hypothesis relationship
4. Collecting evidence 3. Develop conclusion / build theory
Causal relationships Results are themes, clusters of ideas, or theory
constructs
Deductive vs. Inductive Reasoning
Basic research terms
▪ Research hypotheses
▪ Concepts/constructs
▪ Variables
▪ Association/Relation
▪ Theory/Theoretical framework
Research hypotheses
• Research hypothesis (question)
✓ To study the association between X and Y
✓ To study the association between pain control and quality of life
Research hypotheses
Hypothesis Summary Type
H1: A = B There is no relationship between A and B. Null
H2: A ≠ B There is a relationship between A and B. Alternative
Here, there is a relationship, but we don’t know if it is
positive or negative.
H3: A < B There is a negative relationship between Alternative
A and B. Here, the < suggests that the
less A is involved, the better B.
H4: A > B There is a positive relationship between Alternative
A and B. Here, the > suggests that the more B is
involved, the better A.
To disapprove/reject the null hypothesis
Assuming innocent until proven otherwise, “To Err is Human”.
Types of statistical errors
Image source: unbiasedresearch.blogspot.com
Types of statistical errors
• Type I error (α)
✢ Wrongly rejecting a true null hypothesis
✢ α = 0.05 (5%) (p value)
✢ 5% the results due to random chance
✢ Set high p (<0.01) if the finding has very important implication
• Type II error (β)
✢ Wrongly accepting a false null hypothesis
✢ Power = 1- β
✢ 80% = 1 - 0.2
Concepts/ constructs
• Concepts
• Generalized idea about objects, attributes, occurrences or processes
• E.g. pain, anxiety, quality of life (particular aspect of human behavior or
characteristics)
• Constructs
• Definition that represents an abstract phenomenon or concept
• E.g. self-care, health locus of control (deliberately invented)
• Constructs are of a more complex abstraction than concepts.
Variables
• Well-defined and measurable concepts or constructs
• Two processes
1. Conceptualization (conceptual definition)
• Generating, defining the concept
2. Operationalization (operational definition)
• Tools or procedures for measuring the concept
Example of variables
Concept Variable Measurement
Health General health Self rated health
How you rate your health?
1. Very good 2. Good 3. Fair 4. Poor
Mental health Patient Health Questionnaire – 9 items (PHQ)
(Depression)
Mental health State-trait Anxiety Inventory – 20 items
(Anxiety)
Quality of Life SF-36, SF-12
Chronic disease Yes / No, ICD-9
Pain Visual analogue scale (VAS) 0-10
Blood pressure Reading from the machine
Types of variables
• Continuous variable
o Height (170, 175.5 180.2 cm…)
• Ordinal variable
o Very good, good, fair, bad
• Categorical variable
o Yes, No (binary)
o Place of birth, occupation….
Independent & dependent variables
• Independent variable (IV)
– Predict (“cause”) or influence the dependent variable
• Dependent variable (DV)
– Predicted or influenced by the IV
– Y, outcome
Examples:
• Smoking (IV) → Lung cancer (DV)
• Pain (IV) → quality of life (DV)
• Wound dressing (IV) → wound healing / complications (DV)
Extraneous variables
• Confounders, confounding effects
• Observed association was partly or fully due to confounder /
extraneous variable
• Independent variable (X)
• Dependent variable (Y)
• Research hypothesis
• What are the potential
extraneous variables /
confounders?
• Association or causation ?
• Independent variable
(X)
• Dependent variable
(Y)
• Research hypothesis
• What are the potential
extraneous variables
/ confounders?
• Association or
causation ?
• Independent variable (X)
• Dependent variable (Y)
• Research hypothesis
• If this is not a RCT: what are the potential extraneous variables / confounders
Association/ relation
• A bond connection between variables
• Cause-and-effect (causal) association
✢ Smoking causes lung cancer
✢ Need substantial evidence, powerful study design (e.g. RCT)
✢ Hill’s criteria of causation
• Functional (associative) association
✢ Correlation, association, prediction (directive)
✢ Pain was associated with lower quality of life
✢ Drinking was correlated with smoking in adolescents
✢ Poor drug compliance predicts TB complications
✢ Observational studies
Theory/ framework
• An abstract generalization that presents a systematic explanation about how
phenomena are interrelated
• Framework
✢ Theoretical framework: A framework underpins the relation between
study variables with a theory
✢ Conceptual framework: A framework underpins the relation between study
variable with information from literature
✢ Functions
• Rationales for developing research question
• Provide justification for the study
• Explain the association observed
Interventions supported by theory
• Stage of Change (SoC) /
transtheoretical model
Stage-matched smoking cessation interventions
• Pre-contemplation (not aware)
– Knowledge, perception on smoking and quitting
• Contemplation (aware but not intention)
– Self-efficacy, benefit of quitting
• Preparation
– Barriers of taking action, practical skills
• Action
– Treatments, counselling, encouragement, reminders
• Maintenance
– Replace preventions, medication, practical skills
• Relapse
– Encourage on quit attempt
Theory guided data analysis
Example for physical activity in cancer patients
REF: Min, J., Yu, YW., Lee, J. et al. Application of the theory of planned behavior to understand physical activity intentions and behavior among Korean breast cancer
survivors. Support Care Cancer 30, 8885–8893 (2022).
Choosing a Research Design?
Do you have an intervention?
This drug works!
No! Yes!
Observational Experimental
studies studies
▪ Prospective study
▪ Retrospective study/case-control study ▪ True experimental study
▪ Cross-sectional study ▪ Quasi-experimental study
What is the best research design?
Research question: To study the effects of blood pressure lowing pill on
reducing cardiovascular risks:
• Within group comparison
- Check the medical history / ask the patients in the SOPD / pre and
post questionnaires
• Between group comparison
- Check CVD events between patients who taken pills and not taken
pills
• Manipulate the intervention by giving some patients pills or placebo
True experimental study
• Randomized control trial (RCT)
• Randomly allocate patients to intervention or control group
• Provide intervention and placebo
• Measurement
– Pretest and post test
– Post test only
• Example
– Intervention group: BP pill
– Control group: placebo pill
– Follow-up: 12 month CVD events required medical consultation or hospitalization
Recap: Terms of RCT
• Randomization, random allocation, random assignment
• Allocation concealment
• Blinding
• Intervention (manipulation)
• Control
– No any treatment
– Placebo
– Usual care
– A different active treatment
• Intention-to-treat analysis (ITT)
Randomization
Randomization
• To balance
✢Groupsize
✢Known prognostic factors
✢Unknown prognostic factors
• To remove investigator bias in the allocation of participants
• Guarantees valid statistical inference
Different randomization methods
• Individual randomization
o Individual subject as unit of randomization
• Cluster randomization
o Social units, geographic units, or clusters as unit of randomization
• Block/stratified randomization
o Balance the characteristics between groups
• Permuted block randomization
o Multi-sites/ multi-centers
o Balance between groups
Allocation concealment
• Procedures for protecting the randomization process so that the
treatment to be allocated is not known before the patient is entered
into the study
o Commonly done using sequentially numbered, opaque, sealed envelopes
(SNOSE) that are prepared by someone outside the study team and not
involved in recruitment or data collection.
• If the investigator is aware of the sequence, it can lead to bias in
participant selection.
Blinding
• Masking of treatment identity
• Use of blinding techniques is largely for randomized control trials
• Aims to minimize bias
• Involves several components (e.g. single or double blinding)
What if there’s no blinding?
No blinding
• Potential to have high risk of bias
• Wrong estimation of intervention effects
• Intervention not possible to be blinded
Control group
• No any treatment
• Placebo: Deliver similar intervention but without the active ingredients
• Usual care
• A different active treatment
Purpose of control group
To allow discrimination of patient outcomes caused by the test drug from
outcomes caused by other factors, such as the natural progression of the
disease, observer or patient expectations, or other treatment.
REF: ICH E10 (Step 5, 2001): Choice of Control Group and Related Issues in Clinical Trials
Uncontrolled vs. Controlled
Percentage of positive findings
Therapeutic Area Uncontrolled Controlled
Psychiatric (Foulds, 1958) 83% 25%
Antidepressant 57% 29%
(Wechsler et al., 1965)
Antidepressant (Smith et al., 1969) 58% 33%
Respiratory distress syndrome 89% 50%
(Sinclair, 1966)
Rheumatoid arthritis (O’Brien, 1968) 62% 25%
REF: Spiker (1991)
Intention-to-treat analysis (ITT)
• Includes all randomized patients in the groups to which they were
randomly assigned
• Regardless of their adherence with the intervention
• Regardless of the treatment they actually received
• Regardless of the subsequent withdrawal from treatment or deviation from
the protocol
• ITT analysis includes every subject who is being randomized.
Intention-to-treat analysis (ITT)
• Per-protocol analysis is just a subset of the participants who
completed the study without any major protocol violations
• Per-protocol analysis can be performed as a secondary analysis to
evaluate the impact of the possible placebo effect, or non-compliance
etc.
Intention-to-treat analysis (ITT)?
REF: Cheung DST, Takemura N, Lam TC, et al. Feasibility of Aerobic Exercise and Tai-Chi Interventions in Advanced Lung Cancer Patients: A Randomized Controlled
Trial. Integrative Cancer Therapies. 2021;20.
Quasi-experimental study
• Have manipulated the intervention
• Randomization is not possible
▪ No randomization (have control group)
▪ No control group
• Measurement
▪ Pre-post
▪ Post only
• Example
▪ Intervention: BP pills to eligible patients who want the pills
▪ Control: patients who did not take the pills
Essential Elements of a Good Trial Design
Essential elements of
What if not, …
a good trial design
A. 17% over-estimate
1. Control treatment effect
2. Randomization B. 30-41% over-estimate
treatment effect
3. Blinding
C. Treatment effect not
measurable
REF: Schulz et al. (1995). JAMA
Checkpoint: True or False?
Randomization guarantees similar characteristics between groups.
We must have controls for estimating treatment effects.
Randomized controlled design should be the choice for all purposes.
Any questions?
⎼ If you are too shy to ask questions/ give me comments.
• Please go to the following Google Form
• https://forms.gle/H1BMGh4PJvUAjeaRA
• It is entirely anonymous
