Evidence Informed Health Practice: Paradigms, Designs, and Statistical Foundations
Foundations of Evidence Informed Health Practice
Evidence informed health practice for CMHL 1001 involves understanding the philosophical and methodological underpinnings of research to provide the best clinical care.
A flipped classroom approach is utilized for this unit, emphasizing that being prepared for tutorials is essential for maximizing learning outcomes.
Understanding the "lens" or worldview of a researcher is a practical necessity; without identifying these assumptions, a practitioner cannot properly weigh evidence or judge if a study is trustworthy.
Research Paradigms: Ontology and Epistemology
A research paradigm is a worldview or a set of assumptions accepted before a research question is even formulated.
Paradigms are built upon two foundational pillars:
Ontology: This asks, "What is reality?" It questions if there is a single objective truth or if reality is messy, personal, and socially constructed.
Epistemology: This is the follow-up question, "How can we know that reality?"
How a researcher answers these metaphysical questions determines if they fall into one of two major camps: Positivism or Interpretivism.
The Positivist Paradigm: The Objective Lens
Positivism is the classic "lab coat" approach and serves as the default setting for much of health science.
Ontology of Positivism: Assumes there is one single, objective reality regardless of human thoughts or feelings. For example, gravity exists whether or not one believes in it, and a broken bone on an X-ray is an objective fact regardless of a patient's emotional state.
Epistemology of Positivism: Because reality is objective, it can be measured. This leads directly to Quantitative Research.
Characteristics of Positivist Research:
Focuses on hard numbers and data.
Involves testing a hypothesis.
Seeks to control every variable in a laboratory setting to prove a definitive cause and effect.
Requires rigidity to avoid "feelings" contaminating the data, especially when proving if a drug lowers blood pressure.
The Interpretivist Paradigm: The Multi-Faceted Lens
Interpretivism views reality as a series of "shades of gray" rather than black and white.
Ontology of Interpretivism: Assumes that reality is multiple and socially constructed.
Example: The biological signal of pain may be identical (positivist view), but the reality of that pain differs for a marathon runner (who sees it as progress) compared to a cancer patient (who sees it as decline).
Reality is defined by culture, background, and context.
Epistemology of Interpretivism: Focuses on understanding meaning rather than measuring rules. This leads directly to Qualitative Research.
Characteristics of Interpretivist Research:
Uses words, interviews, and small samples.
Explores the "why" and "how" of a situation.
Does not aim to control the environment; instead, it dives into the messiness of the lived experience.
The Levels of Evidence Pyramid
Modern medicine, including drugs and surgeries, is largely built on the positivist foundation to prove that interventions work.
The hierarchy of evidence ranks study designs based on their ability to prove causality ().
Systematic Reviews
Situated at the very top of the pyramid.
Classified as Secondary Research rather than a new primary study.
It involves finding every study ever conducted on a specific topic (e.g., ibuprofen for tension headaches), including good, bad, and "ugly" studies, and mathematically synthesizing them into a definitive answer.
Its position at the top is due to its ability to reduce the error found in any single isolated report.
Randomized Controlled Trials (RCTs)
The "gold standard" for Primary Research.
Based on the concept of the Counterfactual: The ideal (but impossible) scenario where you see what happens to a patient if they took a drug and compared it to what happened to the exact same patient if they did not take the drug at the same moment in time.
RCTs use a "fake version" of this scenario by taking a large group and randomly splitting them into statistically identical groups.
One group receives the intervention (the drug), and the other receives a placebo or usual care.
Randomization is the "magic sauce" that washes away messy variables like diet, age, and lifestyle, ensuring any difference in outcome is caused by the drug.
Observational and Longitudinal Designs
Moving down the pyramid includes Cohort Studies and Case Control Studies.
These involve more uncertainty because without randomization, researchers cannot be sure if the result is due to the treatment or because the patients were healthier to begin with.
Cross-sectional Studies: Analogous to a single slice of bread pulled from a loaf (time).
It captures one moment in time.
It can identify prevalence (how common something is).
It cannot prove cause; for instance, it might show that smokers have high blood pressure, but it cannot determine if smoking caused the pressure or if the stress of high pressure caused the smoking.
Research Question Frameworks
The research design must follow the specific question being asked. Three main tools assist in framing these questions:
PECO (Population, Intervention/Exposure, Comparison, Outcome):
Used for interventions, such as testing if a specific splint helps wrist pain.
Typically indicates a quantitative design like an RCT.
PECO (Exposure variant):
Used when it is unethical to randomize exposure.
Example: You cannot ethically force half a population to live near a highway to see if it causes asthma.
Results in an observational design like a Cohort Study to watch natural occurrences.
PCO (Population, Context):
Used for interpretivist/qualitative research.
Example: Understanding the experience of grief for young widows.
Utilizes designs like Phenomenology (the study of the lived experience).
Sampling Methodologies
The relationship between a population and a sample is compared to a pot of soup and a spoonful:
The Theoretical Population is the whole pot.
The Sample is the spoonful. The spoonful must taste exactly like the pot (representative).
If the sample does not represent the population, the results are considered "trash."
Probability Sampling
Ideally used in positivist, quantitative research.
Simple Random Sampling: Pulling names out of a hat.
Stratified Random Sampling: Ensuring enough people are pulled from specific subgroups, such as rural versus city residents.
Goal: Generalizability (applying findings from people to an entire country).
Non-Probability Sampling
Used when randomization is impossible or inappropriate.
Snowball Sampling: Used for hard-to-reach groups (e.g., intravenous drug users). No public database exists, so researchers build trust with one person and ask for referrals.
Purposive Sampling: Used in qualitative research to select specific experts. A random 20-year-old is useless for studying stroke recovery; researchers purposively select stroke survivors.
Convenience Sampling: The "danger zone." This involves grabbing whoever is nearby (e.g., standing outside a mall with a clipboard).
It is cheap and easy but introduces massive sampling bias.
Example: People at a mall at on a Tuesday are likely retirees or unemployed and do not represent the general working population.
Statistical Errors and Power
Power is likened to the resolution of a camera or the brightness of a flashlight. It determines how many people are needed in a study to see an effect if it exists.
High-effect interventions (e.g., using a parachute to prevent injury) require very small samples to prove they work.
Subtle effects (e.g., a improvement in muscle mass) require massive samples to distinguish the result from random noise.
An Underpowered study (too few participants) leads to statistical errors.
Type 1 Error (False Positive)
The study concludes the drug works ("Eureka!") when, in reality, it does not.
This is considered a massive ethical failure because patients may receive useless or harmful treatments.
Researchers set statistical bars, or values, very high to avoid this.
Type 2 Error (False Negative)
The study concludes the treatment failed when it actually works.
This is usually caused by inadequate statistical power (too small a sample).
Home Workout Example: A study on home exercises for elderly muscle mass may show no significant difference because the group was too small to catch a real, albeit small, benefit.
Consequence: A potentially cheap and accessible intervention is discarded incorrectly.
Mantra: "Absence of evidence is not evidence of absence."
Critical Appraisal Workflow
To critically appraise research, one should follow a specific workflow:
Identify the Lens: Is the researcher using a positivist or interpretivist approach?
Match Design to Question: Did they use an appropriate design (e.g., RCT for causality)?
Evaluate the Sample: Was the "soup" spoonful representative or just a convenience sample from a mall?
Assess Power: Was the study powerful enough to trust the answer, or is a Type 2 error likely?
A final consideration is the persistence of bias: even in randomized computer-coded studies, humans write the code, choose the questions, and interpret the outcome. Therefore, total objectivity may be an impossibility.