HLSC 4120

Quantitative research asks questions about facts, measurements and relationships eg; does X cause Y.
- It produces facts, measurements and relationships,
- It is done through a deductive reasoning approach (Theory --> Hypotheses --> Data)
- Postivisit perspective (genuine knowledge is based solely on observable, measurable, and objective facts)

Structuring Quantitative Questions --> Use PICOT criteria
P – Population (who?)
I – Intervention (what?)
C – Comparison (compared to what)
O – Outcome (measuring what)
T – Time (over what period)

Example: In elderly patients with chronic lower back pain (P), does a 12-week physiotherapy programme combined with mindfulness training (I), compared to physiotherapy alone (C), lead to greater pain reduction and improved daily function (O) at 12 weeks (T)?

Qualitative research...
- produces knowledge that is constructed through peoples’ meanings, perspectives, interpretations and experiences --> these cannot be measured numerically and thus require descriptive interpretive approaches
- It uses inductive reasoning (identify patterns and themes, to build upon theory)
- It operates in a constructivist paradigm (assumes reality is subjective or context dependant)

Structuring Qualitative Questions --> Use PEO criteria

P – Population (who?)
E – Exposure/Experience (what phenomenon or intervention?)
O – Outcomes/Context (What aspects are we exploring)

Example: "What are the experiences of university students with mild to moderate anxiety (P) when using mobile app-based CBT programmes (E) in managing their daily academic and social functioning (O)?"

Inductive Reasoning (Qualitative)
• Core Definition: "Bottom-up" logic. Researchers gather specific, empirical observations and look for recurring patterns or trends to build a new, broader theory.
• Direction of Logic: Specific Observation --> Pattern --> Tentative Hypothesis --> General Theory.
• Objective: Theory-building. The goal is to deeply understand a specific phenomenon and generate new frameworks or explanations.

Deductive Reasoning (Quantitative)
• Core Definition: "Top-down" logic. Researchers take an established theory or general principle and use it to formulate specific, testable predictions.
• Direction of Logic: General Theory --> Specific Hypothesis --> Observation --> Confirmation (or rejection).
• Objective: Theory-testing. The goal is to prove, disprove, or refine an existing theory in a new context.

- an observational research method that analyses data from a population at a single, specific point in time. Rather than tracking individuals over an extended period, researchers take a "snapshot" of the group to observe correlations between variables, assess disease prevalence, or understand current behaviours.

- These are best for ‘diagnosis questions’ or questions that focus on the accuracy of tools and/or tests.

- Eg; What is the reliability of new stroke assessment tools? Or How accurate are rapid antigen tests for COVID 19?

- observational, longitudinal research designs that follow a group of people who share a common characteristic or exposure. Researchers track these individuals over a period of time to compare health outcomes and determine how specific risk factors influence the development of a disease or condition

- These are best for ‘prognosis questions’ or questions that ask for future outcomes. These studies will track people over time.

- Eg; What factors predict recovery after ACL reconstruction? Or What is the 5-year survival rate of early detected melanoma?

- an observational research design used to identify potential risk factors or causes of a disease. Researchers compare individuals who already have a specific condition or outcome (cases) with similar individuals who do not (controls) to investigate their past exposures

- These are best for ‘aetiological questions’ or questions that explore causes or risk factors for a certain pathology

- Eg; Is shift work associated with increased cardiovascular disease? Or Do occupational factors contribute to chronic back pain? (maybe night shift v morning or manual labour v desk job)

Meta-anlyses & systematic reviews

Probability Sampling
- Every person in the source population has a known chance of selection
- This is a random selection and thus the gold standard for generalisability (stronger external validity)
- Allows for the calculation of sampling error and CI

Non-probability Sampling
- Non-random selection criteria, that is, the purposeful selection of people to meet criteria. More practical and less expensive than probability
- Prone to sampling bias and limited generalisability (poorer external validity)

1) Simple Random Sampling: Every member has equal chance of selection

2) Systematic Sampling: Selection using fixed intervals (every nth person). Systematic sampling can be affected if there are hidden patterns that align with the sampling interval (e.g., certain patient types attending at specific times)

3) Stratified Sampling: Ensuring subgroup representation. Done by the population being divided into subgroups or strata based on key characteristics, then random or systematic sampling within each stratum. Eg; stratified by age or duration or seizure.

4) Cluster Sampling: Sampling groups rather than individuals. The population is divided by pre-existing units (schools, hospitals etc) and then random selection within the clusters occurs

1) Purposive Sampling: Participants selected to match the study aims and objectives. This may mean multiple inclusion and exclusion criteria. This means you can only generalise to the population defined by the selection criteria

2) Convenience / Availability / Snowball Sampling: A sample is drawn from a conveniently accessible source which may not be representative of a population. This may limit external validity but can give high internal validity (assuming methodology is sound). Eg; a single hospital outpatient clinic

3) Quota sampling: non-random selection from defined population subgroups (this is the key difference from stratified population sampling), eg; approximately 30% of rural population. We must consider whether this quota criteria is justified by the research aim/question and whether the selection within quotas was random.

Random Error
- This is the unpredictable error that varies in size and direction, affects precision and cannot predict when/how they occur
- Sources of random error include: natural biological variation, measurement imprecision (scale imprecision, instrument limitations), observer variability, environment factors (room temperature, time of day, recent activities), participant state (fatigue, anxiety, hunger affecting performance)

Systematic error or bias
- This is the predictable error that is consistent in direction, affects accuracy and causes systematic distortion
- Sources of systematic error: Faulty equipment (instruments not properly calibrated), consistent observer error (always rounding up), selection processes, information gathering (how data is collected, who collects it and from who), confounding variables

Bias refers to any process occurring at any stage of an investigation tending to produce results which depart systematically from true values. The three types of bias include;
- Selection Bias
- Information Bias
- Confounding Bias (occurs when an unmeasured third variable distorts the true relationship between an exposure and an outcome)

This is a distortion in data analysis caused by non-random sampling or improper participant selection. It occurs when the selected group does not accurately represent the target population, leading to flawed conclusions. Main types include:

SAMPLING BIAS – where the sampling frame doesn’t represent the target population

VOLUNTEER BIAS– self-selection creating systematic differences, (e.g., highly passionate or highly health-conscious)

BERKON'S BIAS – how hospital samples may differ from community populations.

any systematic error or deviation from the truth that arises during the collection, recording, or measurement of data. It can occur due to faulty instruments, flawed participant recall, or skewed information processing, ultimately leading to incorrect conclusions.

Recall Bias – memory accuracy differs by exposure or outcome status

Observer bias – assessors knowledge influences measurements

Measurements bias – systematic error in data collection instruments

Response bias – participants provide socially desirable answers

Accuracy is the degree of how close a measurement is to the true value. It relates to how well a measuring system can measure the true value. It is affected by systematic errors

Precision is the consistency of repeated measurements or the ability to reproduce results , where higher precision will produce smaller variability and smaller random error

Reliability is the consistency of measurement across repeated applications --> Related to precision. There are three main types:

Test-Retest Reliability: Assesses consistency across time. It measures whether a test yields the same results when administered to the same group of people on two separate occasions.

Internal Consistency: Assesses consistency across items. It checks if multiple questions designed to measure the same concept within a single test or survey produce similar answers.

Inter-Rater Reliability: Assesses consistency across observersesearchers. It evaluates the degree of agreement between different judges or raters observing and scoring the same phenomenon.

- Type I Error (false positive) or when we incorrectly reject the null
- Type II Error (false negative) or when we fail to detect a real difference

When appraising statistics in papers we look at the P-value

- P Values show the strength of evidence against the null or ‘no effect’

- Smaller p-values (P<0.05) indicates stronger evidence for a real effect

- Importantly the P-value tells us if theres evidence for a difference, not how much that difference is --> to find this we need to examine effect sizes and Confidence Intervals (CI)

Study Power – the studies ability to detect a true effect if it exists (sensitivity). A typical power is 80% which thus concedes 20% chance of a false negative occurring (type II error --> often due to insufficient sample size) Factors affecting Power:
- The significance level or (α)
- The effect size being assessed (the magnitude of difference or strength of relationship --> think r value)
- The frequency of the outcome
- The sample size

A confidence interval (CI) is a range of values, derived from sample data, that is likely to contain the value of an unknown population parameter (e.g., mean or proportion) with a specified probability, usually 95% or 99%.

Regression will tell us how well independent variables predict an outcome and the strength of relationships. When we are critically appraising, we need to see whether the authors claim prediction or causation and if confounders were appropriately controlled. In research papaers you will often see adjusted and unadjusted regression models

- unadjusted models assesses on the relationship between predictor and outcome without addressing confounding variables

- Adjusted models will account for confounding variables (that is variables that may affect the dependant variable)

Respect
- This recognises the inherent dignity, autonomy and rights of the participants involved. This looks like voluntary participation, informed consent, cultural respect and the right to withdraw

Research Merit and Integrity
- This is a balancing act where the research must be worthwhile and methodologically sound (merit) and conducted by competent and honest researchers in pursuit of knowledge and understanding (integrity).

Justice
- This ensures fairness in recruitment, burden and benefit. No group should be unfairly excluded or exploited, and the benefits of research should be distributed equally

Beneficiance
- Researchers must maximise potential benefits and minimise potential harms, this looks like ongoing risk-benefit assessments and stopping or modifying research if risk increases.

Risk and Benefit
- Researchers must identify, minimise, and justify risks (physical, psychological, social, privacy related). Benefits must outweigh risks, and risks must be proportionate to the research purpose.

Informed Consent
- Participants must receive clear, accurate information about the study’s purpose, procedures, risks, benefits, and their rights. Consent must be voluntary, ongoing, and free from coercion.

Autonomy
- Participants must be free to make their own decisions about participation. This includes the right to withdraw without penalty and the right to refuse data use.

Privacy and Confidentiality
- Researchers must protect personal information through secure storage, de identification, and limiting access. Data must only be used for the purposes consented to.

Grounded Theory

Thematic Analysis

Phenomenology

Thematic Analysis is a flexible method for identifying, analysing and interpreting patterns (themes) within qualitative data. It does not aim to build a theory or describe lived experience in philosophical depth but rather organise meaning.

Thematic analysis is more a set of data analysis techniques rather than its own methodology.

Types of techniques include:
- Code reliability (coding as identifying themes and themes are presented as topic summaries)
- Codebook Approaches (combination of reflexive and code reliability)
- Reflexive Approaches (later theme development from coding)

This a systematic methodology used to generate a theory directly from data. Researchers collect data (e.g., interviews, observations), code it line by line, identify categories, and build an explanatory model grounded in participants’ experiences. That is – generating theory from data from the ‘ground up’.

There are three schools:
1) Classical Grounded Theory (The Glaserian School)
2) Straussian Grounded Theory (The Corbinian School)
3) Constructivist Grounded Theory (The Charmazian School)

1. Classical Grounded Theory (The Glaserian School)
Originating directly from the founders, this approach relies on a mostly unstructured, inductive process where theory is meant to organically "emerge" from the data.
• Core Principle: Researchers are advised to keep an "empty mind" by minimizing preconceived ideas or literature reviews before gathering data.

2. Straussian Grounded Theory (The Corbinian School)
Developed by Anselm Strauss and Juliet Corbin, this variant is more structured and allows the researcher to bring existing professional knowledge and formal theories into the study.
• Core Principle: Employs systematic coding paradigms (such as open, axial, and selective coding) to establish connections between conditions, contexts, actions, and consequences.

3. Constructivist Grounded Theory (The Charmazian School)
Pioneered by Kathy Charmaz, this school takes an interpretive approach that rejects the idea of a single, objective reality waiting to be discovered.
• Core Principle: Acknowledges that the resulting theory is a co-construction created through the interaction between the researcher and the participants.

Grounded theory is the right approach when little is known about the phenomenon, a new angle is needed or when the aim is to understand social processes (how and why people act as they do). This is because it starts with a broad area of interest rather than a tighly specified hypothesis.

A methodology focused on understanding the lived experience of individuals. It explores the essence or meaning of a phenomenon as experienced from the first person perspective. The goal is to describe the meaning or essence of this experience both in terms of what was experienced and how it was experienced.

There are two types:
1) Descriptive phenomenology – calls for an exploration of phenomena through direct interaction between researcher and the objects of the study.

2) Hermeneutic Phenomenology – a philosophical approach and qualitative research methodology that seeks to understand everyday lived experiences by interpreting their underlying meanings. The goal of hermeneutic phenomenology is to identify the participants’ meanings from the blend of the researchers understanding of the phenomenon (participant generated information, other relevant sources, own understanding).

1. Bracketing (Epoché)
The researcher consciously sets aside their own biases, preconceived notions, and theoretical knowledge regarding the phenomenon. This suspension of judgment allows the researcher to examine the participant's lived experience with an open mind and pure focus.

2. Intuiting
The researcher immerses themselves entirely in the raw data (e.g., interview transcripts or diaries) to understand the participant's meaning. This requires the researcher to focus on the phenomenon until a shared understanding and connection with the participant's view is established.

3. Analyzing
The researcher breaks down the data into smaller segments or "meaning units" to code and organize. They transition the participant's everyday language into structural or psychological themes to isolate the fundamental elements of the experience.

4. Describing
The final step synthesizes the themes to craft a vivid, detailed, and accurate description of the phenomenon. The goal is to articulate the "essence" of what the experience is actually like for the individuals being studied, without adding subjective interpretations

Thematic Analysis = summarising and interpreting patterns in perceptions or experiences.
“What are the key themes or perspectives?” questions. Examples:
• What do older adults perceive as barriers to exercise?
• What themes emerge in students’ experiences of online learning?


Grounded Theory = explaining a social process or mechanism.
“Process” or “how/why does this happen?” questions. Examples:
• How do patients adapt to a new diagnosis?
• What processes shape nurses’ decision making in emergencies?

Phenomenology = capturing the essence of lived experience
“What is it like to experience…?” questions. Examples:
• What is it like to live with chronic pain?
• How do first time mothers experience early breastfeeding?

Purposive sampling
Most common and often most appropriate. Researchers deliberately recruit participants who can best answer the research question.

Theoretical sampling
Used primarily in grounded theory whereby participants are recruited based on concepts emerging during analysis.

Snowball/chain referral sampling
Useful for hard-to-reach populations where participants refer other participants.
Examples:
• Drug users
• Hidden communities

Convenience sampling
Examples:
• Workplace colleagues
• Posters
• Existing clients
Easy but often introduces bias

In qualitative research, participants are usually selected because they have direct experience of the phenomenon being studied. This is called purposive sampling.
Appropriate reasons for inclusion include:
• Experience of the illness, event, or situation under study
• Ability to answer the research question
• Possession of specific characteristics relevant to the phenomenon
• Ability to provide rich and meaningful data

Examples
If studying:
• Stroke rehabilitation → stroke survivors
• Bullying → children who have experienced bullying
• Splint use → people who wear splints

Participants may be excluded if they:
• Do not have experience relevant to the phenomenon
• Cannot provide meaningful information regarding the research question
• Cannot provide informed consent
• May be exposed to unacceptable risk

Structured Interviews
Characteristics:
• Standardised questions
• Same order
• Same wording
• Often survey-like
• Mix of open and closed questions
Strengths
• Consistency
• Easier comparison
Weaknesses
• Limited depth
• Less opportunity to explore unexpected ideas
Data richness
⭐ Lowest richness
Participants can only respond within predefined questions.
________________________________________
Semi-Structured Interviews
Characteristics:
• Interview guide used
• Flexible wording
• Flexible order
• Researcher can probe
Strengths
• Balances consistency and flexibility
• Allows exploration of emerging ideas
Data richness
⭐⭐⭐ High richness
Most commonly used qualitative interview method.
________________________________________
Unstructured Interviews
Characteristics:
• Conversational style
• No fixed question order
• Driven by participant responses
Strengths
• Deep exploration
• Participant-led
• Unexpected insights emerge
Weaknesses
• Requires skilled interviewer
• Harder to compare participants
Data richness
⭐⭐⭐⭐ Highest richness
Produces the deepest and most detailed data.

Focus groups are groups of peoples interviewed for the purpose of eliciting ideas, thoughts, and perceptions about a topic or issue. This looks like interaction between participants within a group where they respond to each other and the researcher.

Advantages
• Interaction stimulates discussion
• Multiple viewpoints
• Efficient data collection
• Useful for sensitive topics with peer support

Limitations
• Groupthink
• Dominant participants
• Social desirability bias
• Some participants may not express true opinions

Data saturation is a point in qualitative research when gathering additional information or interviewing more participants no longer yields new themes, codes, or insights

It is controversial because it is inconsistently defined, difficult to demonstrate and not appropriate for all qualitative methodologies, particularly phenomenology. Researchers should explain what they mean by saturation, how they determined it during data collection and why recruitment ceased.

Qualitative data analysis involves what known as coding. Coding is the process of organising and labelling sections of qualitative data so patterns can be identified.

A simple generic coding procedure is:
Step 1: Familiarisation
• Read and re-read interview transcripts.
• Become immersed in the data.
• Note initial impressions.

Step 2: Open Coding
• Identify meaningful words, phrases, sentences, or passages.
• Assign labels (codes) that capture their meaning.

Example
Participant says: "I was scared to exercise after my knee dislocation." = The possible code: Fear of re-injury

Step 3: Group Similar Codes
Related codes are combined into categories.
Codes Category
Fear of re-injury = Psychological barriers
Lack of confidence = Psychological barriers
Anxiety about movement= Psychological barriers

Step 4: Develop Themes
Categories are examined for broader patterns.

Qualitative research cannot be evaluated by the same criteria as quantitative as it is interpretive and context-dependant. Unlike Quantitative research criteria (internal/external validity, reliability, objectivity etc), qualitative uses four criteria each asking a different question about quality (These are explored in trustworthiness).

Rigour refers to the overall quality, thoroughness and methodological soundness of a qualitative study. It reflects how carefully the study was designed, conducted and reported to ensure the findings are credible and meaningful

Unlike quantitative research, rigour is not judged by validity, reliability and statistical significance. Instead, it is demonstrated through transparent and systematic research methods.

Examples of strategies that improve rigour include:
• Appropriate sampling
• Reflexivity
• Clear data collection methods
• Transparent analysis
• Use of participant quotations
• Multiple researchers (investigator triangulation)
• Audit trails

Trustworthiness is the framework commonly used to judge the quality of qualitative research. It was proposed by Lincoln and Guba (1985) and consists of four criteria:

1. Credibility
Question: Do the findings accurately represent participants' experiences?

2. Transferability
Question: Can readers determine whether the findings apply to another setting?

3. Dependability
Question: Is the research process logical, well documented and consistent?

4. Confirmability
Question: Are the findings grounded in the data rather than researcher bias?

Think of rigour as the overall quality of the research, while trustworthiness is one framework used to evaluate whether that rigour has been achieved.

Mixed methods research combines quantitative and qualitative methods within a single study (or program of research) to provide a more comprehensive understanding of a research problem.

Researchers should justify mixed methods by explaining why one method alone is insufficient.

1. Complementarity
One method complements the other.
Example:
• Quantitative: Patients' anxiety scores decrease.
• Qualitative: Interviews explain why anxiety decreased.

2. Triangulation
Both methods investigate the same phenomenon from different perspectives to cross-check and increase confidence in the findings
Example:
• Survey results
• Interview findings
If both reach similar conclusions, confidence increases

3. Development
One method informs the other.
Examples:
• Qualitative interviews develop a questionnaire.
• Survey findings identify participants for interviews.

4. Expansion
Different methods answer different parts of the research question.
Example:
• Does rehabilitation improve function? (Quantitative)
AND
• How do patients experience rehabilitation? (Qualitative)

5. Explanation
One method explains unexpected findings from the other.
Example:
• Quantitative results show poor adherence.
• Qualitative interviews explain the barriers to adherence.

Purpose (Why are both methods included?)

Priority (Which method is more important?)

Sequence (When were the methods conducted --> Can be sequential or concurrent / parallel)

Integration is the defining feature of mixed methods research. If quantitative and qualitative components are simply conducted separately, it is not truly mixed methods.
Integration means the two datasets influence each other.

Ask:
• Did one method influence the other?
• Were datasets compared?
• Were findings combined in the discussion?
• Was a joint interpretation presented?
• Did the authors use joint displays (tables combining quantitative and qualitative findings)?

If the answer is yes, integration has occurred.