What is a case control study?
🕵️ What is a Case-Control Study?
A case-control study is a type of research study where scientists:
- First identify people who already have the outcome or disease (these are the cases).
- Then find people who do not have the disease (these are the controls).
- Then look back in time (retrospectively) to see whether they were exposed to a possible risk factor.
In simple terms (rewording your image clearly):
> A case-control study starts with the outcome (the disease or condition), and then looks backward to see if there was a certain exposure.
So it is defined by the outcome of interest, not by the exposure.
🧠 Why is it useful?
Case-control studies are especially helpful when:
- The disease is rare
- The disease takes a long time to develop
- It would be too expensive or slow to follow people forward for years
🏥 Clinical Example (1–3 sentences)
Researchers want to know whether smoking is linked to lung cancer. They identify people who already have lung cancer (cases) and people who do not (controls), then look back to see how many in each group smoked in the past. If smoking is more common in the cancer group, it suggests an association.
👶 Explanation for a 10-Year-Old (1–3 sentences)
Imagine you find kids who already have a tummy ache and kids who don’t. Then you ask both groups what they ate yesterday to see if one food might be the cause.
🧠 Memory Hook (funny, logical, interesting)
“Case first, then chase the past.” 🕵️♂️⏳
🚨 Key Takeaways
- Starts with the outcome (disease).
- Compares cases (disease) to controls (no disease).
- Looks backward in time for exposure.
- Good for studying rare diseases.
- Measures association (usually using odds ratio).
What are the advantages of a case control study?
🕵️ Advantages of a Case-Control Study
A case-control study has several important advantages:
1️⃣ Good for Rare Diseases
Because researchers start by finding people who already have the disease, it works very well for studying rare conditions.
You don’t have to wait years for enough cases to develop — you begin with them.
2️⃣ Less Expensive
Since researchers are not following large groups of people over many years, it usually costs much less than long-term cohort studies.
3️⃣ Not Time-Consuming
Case-control studies look backward in time (retrospective).
That means the outcomes have already happened, so results can be obtained much faster.
🧠 Why these advantages matter
If a disease is rare or takes decades to appear (like certain cancers), following thousands of healthy people forward would take too long and cost too much. Case-control studies solve this by starting with people who already have the disease.
🏥 Clinical Example (1–3 sentences)
Researchers want to study a rare brain tumour. Instead of following thousands of people for 30 years, they identify patients who already have the tumour (cases) and compare them with people who don’t (controls), then look back to see possible risk factors.
👶 Explanation for a 10-Year-Old (1–3 sentences)
If only a few kids in the whole school got a strange rash, you’d start by finding those kids first. Then you’d ask what they did last week instead of watching the whole school for years.
🧠 Memory Hook (funny, logical, interesting)
“Case-control = Rare, Rapid, and Reduced cost.” 🕵️♂️💨💰
🚨 Key Takeaways
- Excellent for rare diseases.
- Cheaper than long-term studies.
- Faster because it looks backward in time.
- Efficient when outcomes have already occurred.
What are the disadvantages of a case control study?
🕵️ Disadvantages of a Case-Control Study
While case-control studies are useful, they also have important weaknesses:
1️⃣ Not Good for Rare Exposures
Case-control studies work well for rare diseases, but not for rare exposures.
If very few people were exposed to something (for example, a rare chemical), it can be hard to find enough exposed individuals to compare.
2️⃣ Recall Bias (Because It Looks Backward)
Since case-control studies are retrospective (they look back in time), they rely on people remembering past exposures.
People with the disease (cases) may remember or report exposures differently than those without the disease (controls).
3️⃣ Cannot Prove Causation
Because exposure and outcome are measured after the fact, it can be hard to prove which came first (the exposure or the disease).
So case-control studies show association, not definite cause-and-effect.
4️⃣ Controls Can Be Hard to Find
Choosing the right control group can be difficult.
Controls must be similar to cases in important ways — except for having the disease.
🧠 Why these disadvantages matter
Because the study starts with the disease and looks backward, there’s more room for memory errors, selection problems, and uncertainty about timing. That limits how strong the conclusions can be.
🏥 Clinical Example (1–3 sentences)
Researchers study whether a rare pesticide causes Parkinson’s disease. Patients with Parkinson’s may over-report past pesticide exposure because they are searching for a reason for their illness. Meanwhile, finding a control group with similar farming backgrounds but no disease may be challenging.
👶 Explanation for a 10-Year-Old (1–3 sentences)
If you ask kids who already have a tummy ache what they ate last week, they might guess or blame something. And it might be hard to find other kids who are exactly the same except for the tummy ache.
🧠 Memory Hook (funny, logical, interesting)
“Case-control can recall wrong, can’t prove strong, and controls take long.” 🕵️♂️⏳
🚨 Key Takeaways
- Not useful for rare exposures.
- Risk of recall bias.
- Cannot prove causation (only association).
- Controls can be difficult to identify.
- More vulnerable to bias than prospective studies.
Explain what selection bias is in a case control study and how it can occur.
🕵️ What is Selection Bias in a Case-Control Study?
Selection bias happens when the cases and controls are chosen in a way that makes them not truly comparable.
In a case-control study, this can occur if the control group is not representative of the same population that produced the cases. If cases and controls differ in important ways (other than the disease), the results may be misleading.
🧠 How can it happen?
Selection bias can occur if:
- Controls are chosen from a different setting than cases (e.g., hospital patients vs general population).
- Controls are healthier, wealthier, or from a different age group than cases.
- The way participants are recruited depends on exposure status.
When this happens, the study may falsely suggest an association — or hide a real one.
🏥 Clinical Example (1–3 sentences)
Researchers study whether alcohol use is linked to liver disease. They select cases from a hospital liver clinic but choose controls from a gym. Since gym members may drink less alcohol than the general population, the study might exaggerate the association between alcohol and liver disease.
👶 Explanation for a 10-Year-Old (1–3 sentences)
Imagine comparing kids with tummy aches from the school nurse’s office to kids picked from the soccer team. The soccer team might be healthier overall, so the comparison wouldn’t be fair.
Explain how selection bias in a case control study can be minimised in a known group.
🕵️ How to Minimise Selection Bias in a Case-Control Study (Using a Known Group)
📌 What is a “Known Group”?
In this context, a known group is a clearly defined and identifiable population that exists independently of the study — for example, people listed in a population registry, members of a health database, or residents of a specific geographic area.
It is a population where:
- The source of participants is clearly defined.
- The population can be described and sampled from.
- Cases and controls both come from the same underlying group.
✅ Step-by-step: How selection bias is minimised
1️⃣ Clearly Define the Cases
Cases should be selected using clear inclusion and exclusion criteria (for example, schizophrenia diagnosed using DSM research criteria).
This ensures everyone labelled as a “case” truly has the same condition.
2️⃣ Choose Comparable Controls
Controls should be similar to cases in every way except for the disease.
Importantly, controls should be chosen independent of exposure status — meaning we do not select them based on whether they were exposed or not.
They should represent the background exposure rate that would be expected in the population from which the cases came.
3️⃣ Use a Known Population
A known group is a clearly defined population observed over time (for example, residents of a specific town or members of a health registry).
Good ways to sample controls from a known group include:
- Random sampling from a population register
- Random digit dialing of phone numbers when no formal list exists
This helps ensure controls truly come from the same population as the cases.
4️⃣ Nested Case-Control Studies
When a case-control study is conducted within a larger cohort study, it is called a nested case-control study.
This further reduces bias because cases and controls come from the same well-defined group.
🧠 Why this works
If both cases and controls come from the same clearly defined population and are selected randomly (not based on exposure), the comparison becomes fairer and more accurate.
Poor control selection can cause major errors, falsely increasing or decreasing the association between exposure and disease.
🏥 Clinical Example (1–3 sentences)
Researchers studying pesticide exposure and Parkinson’s disease use a national farming registry. Cases are farmers diagnosed with Parkinson’s, and controls are randomly selected farmers from the same registry who do not have the disease. Because both groups come from the same defined population, selection bias is reduced.
👶 Explanation for a 10-Year-Old (1–3 sentences)
If you’re comparing kids with tummy aches to kids without, you should pick both groups from the same school — not one from school and one from a sports camp. That way, the comparison is fair.
🧠 Memory Hook
“Same source, fair course.” 🎯
🚨 Key Points
- A known group = a clearly defined population that both cases and controls come from.
- Define cases clearly using strict criteria.
- Select controls similar to cases except for the disease.
- Choose controls independent of exposure.
- Use random sampling from a known population.
- Nested case-control studies reduce bias further.
Explain how selection bias in a case control study can be minimised in an unknown group.
🕵️♀️ Minimising Selection Bias in a Case-Control Study (Using an Unknown Group)
📌 What is an “Unknown Group”?
In this context, an unknown group means there is no clearly defined population list 📋 (like a registry or database) to sample controls from.
Researchers cannot easily identify the full population that produced the cases, so they must find controls using practical methods — which increases the risk of selection bias ⚠️.
✅ Ways to Minimise Selection Bias in an Unknown Group
Because there is no defined population list, researchers try to choose controls in structured, careful ways 📏.
1️⃣ 🏠 Neighbourhood Controls
Controls are selected from the same neighbourhood as the case.
To reduce bias:
- Researchers follow a fixed pattern (for example, every third house 🚪🚪🚪).
- They use structured approaches like population registers or door-to-door canvassing 🚶♂️📋.
⚠️ Problems:
- Some people don’t respond 📵.
- Safety concerns 🚨.
- Can be expensive 💰.
2️⃣ 🏥 Hospital Controls
Controls are selected from patients in the same hospital.
Advantages:
- Easy to identify 👀.
- Similar healthcare-seeking behaviour 🩺.
- Information quality is usually good 📄.
⚠️ Problem:
Hospital patients may already have health conditions that increase hospitalisation risk — so they may not represent the general population 🌍.
3️⃣ ❄️ Snowball Sampling
Participants help identify other potential participants 🤝.
For example:
- A drug user (case) may nominate a friend who also uses drugs (another case).
- They might nominate a friend who has never used drugs (control).
Advantage:
- Useful in hidden or hard-to-reach populations 🔎.
- May reduce some types of bias when populations are difficult to identify.
⚠️ Problem:
Friends tend to be similar 👯♂️, which can introduce bias.
4️⃣ 👨👩👧 Relatives as Controls
Relatives share many traits with cases 🧬.
This can be helpful when:
- Genetic factors may confuse the results (confounding).
However:
- They may be too similar in lifestyle or environment 🏠, which can reduce fairness.
🧠 Why is selection bias a risk here?
Without a clearly defined population 📍, controls might not truly represent the population that produced the cases. If controls are systematically different, the association between exposure and disease may be falsely increased 📈 or decreased 📉.
To minimise bias:
- Use structured selection patterns 📏.
- Keep control selection independent of exposure 🚫.
- Ensure controls come from a similar background as cases ⚖️.
🏥 Clinical Example (1–3 sentences)
Researchers studying illicit drug use and psychosis cannot use a population registry 📋. They use snowball sampling, where participants help identify peers from similar social networks 🤝. While this helps access a hidden population 🔎, researchers must be careful because friends often share similar behaviours 👯♂️.
👶 Explanation for a 10-Year-Old (1–3 sentences)
If you can’t find a full list of kids at school, you might ask one kid to bring a friend 👦➡️👦. But friends are often alike, so your comparison might not be perfectly fair ⚖️.
🧠 Memory Hook
“Unknown crowd? Follow a pattern loud — or bias sneaks in proud!” 🎯🚪
🚨 Key Points
- Unknown group = no clear population list 📋.
- Use structured neighbourhood sampling 🏠.
- Hospital controls are convenient but may not represent the general population 🏥⚖️.
- Snowball sampling helps reach hidden groups ❄️🔎.
- Relatives can control for genetics but may introduce similarity bias 🧬.
- Always ensure controls are chosen independently of exposure 🚫.
What happens if you use more than one control per case?
What is Berkson’s Bais?
What is Neyman Bias?
🧪 What happens when you use more than one control per case?
In a case-control study, researchers sometimes use more than one control for each case (for example, 2–4 controls per case).
Why?
- It increases the power of the study.
- This makes it easier to detect a real association if one exists.
- However, after about a 4:1 ratio (four controls per case), the improvement in power becomes very small.
So:
More controls = better ability to find a real effect 📈
But after 4 controls per case = very little extra benefit.
⚠️ Berkson’s Bias
Berkson’s bias happens when both cases and controls are selected from hospital patients instead of the general population.
Hospital patients often have other illnesses or risk factors, which can distort the relationship between exposure and disease.
Why this matters:
People in hospitals are not representative of the general population — they may be sicker overall.
🏥 Clinical Example (1–3 sentences)
Researchers studying alcohol and heart disease select both cases and controls from hospital wards. Since hospital patients may already have multiple health issues related to alcohol use, the study may overestimate or underestimate the true association.
⚠️ Neyman’s Bias (Incidence-Prevalence Bias / Selective Survival Bias)
Neyman’s bias occurs when the exposure affects survival, and only survivors are included in the study.
This can lead to misleading results because people who died earlier are not counted.
Example from the image (reworded clearly):
In a study of smoking and Alzheimer’s disease, smokers may die earlier from other causes before they develop Alzheimer’s. So when researchers look at people with Alzheimer’s, fewer smokers are present — making it look like smoking protects against Alzheimer’s, which is not true.
The bias happens because only survivors are studied.
🧠 Why these biases are important
- Using more controls improves statistical power.
- Berkson’s bias distorts results when hospital patients are used as controls.
- Neyman’s bias distorts results when early deaths remove exposed individuals from the study.
👶 Explanation for a 10-Year-Old (1–3 sentences)
If you only study kids who are still at school and ignore the ones who moved away, your results might be wrong. And if you only pick kids from the nurse’s office, they might all be sicker than average.
🧠 Memory Hook
“More controls = more power; Hospital fools = Berkson’s rule; Survivors only = Neyman’s story.” 🎯🏥⏳
🚨 Key Points
- More controls per case → increases power (up to about 4:1).
- Berkson’s bias → hospital-based selection distorts association. (Baker’s son in hospital )
- Neyman’s bias → survival effects distort association. (No man bias as he is already dead)
- Both can falsely increase or decrease observed associations.
Explain how measurement bias can occur in a case control study, and how to minimise it.
📏 What is Measurement Bias in a Case-Control Study?
Measurement bias (also called information bias, observation bias, or classification bias) happens when the exposure is measured or recorded incorrectly.
In simple terms (rewording your image clearly):
> It occurs when there is an incorrect determination of exposure — meaning researchers do not accurately measure who was exposed and who was not.
Because case-control studies look backward in time, they are especially vulnerable to errors in how past exposures are reported or recorded.
🧠 How can it occur?
Measurement bias in case-control studies can happen when:
- Cases remember past exposures differently than controls (recall bias).
- Interviewers ask cases and controls different types of questions.
- Medical records are incomplete or inaccurate.
- Exposure categories are misclassified.
If exposure is measured differently in cases and controls, the association can appear stronger or weaker than it really is.
🛡️ How to Minimise Measurement Bias
To reduce this bias:
- Use standardised questionnaires for everyone.
- Blind interviewers to whether someone is a case or control.
- Use objective records (e.g., lab results, registries) instead of memory when possible.
- Train interviewers to ask questions consistently.
- Clearly define exposure categories.
The goal is to measure exposure in the same way for both groups.
🏥 Clinical Example (1–3 sentences)
Researchers studying pesticide exposure and Parkinson’s disease interview participants. Patients with Parkinson’s may try harder to remember past pesticide use than controls, leading to recall bias. To reduce this, researchers use structured questionnaires and check employment records instead of relying only on memory.
👶 Explanation for a 10-Year-Old (1–3 sentences)
If you ask kids with tummy aches what they ate, they might blame something and try hard to remember. Kids without tummy aches might not think about it much. That makes the answers uneven.
🧠 Memory Hook
“Measure the same, or your results are lame.” 📏
🚨 Key Points
- Measurement bias = incorrect measurement of exposure.
- Also called information, observation, or classification bias.
- Common in case-control studies because they look backward.
- Often caused by recall differences.
- Minimise by standardising questions and using objective data.
Explain how information bias can occur in a case control study, and how to minimise it.
📏 What is Information Bias?
Information bias (also called measurement, observation, or classification bias) happens when information about exposure or outcome is measured incorrectly.
In simple terms:
> It occurs when exposure is not determined accurately — meaning researchers misclassify who was exposed and who was not.
This leads to incorrect results.
🕵️ How can it occur in a Case-Control Study?
Case-control studies are especially vulnerable because they look backward in time (retrospective).
Information bias can occur when:
- The exposure is not clearly defined or measurable.
- Exposure is not identified the same way in both cases and controls.
- Investigators measuring exposure know who is a case (lack of blinding).
- Patients rely on memory to report past exposures (recall bias).
The most common patient-related information bias in case-control studies is recall bias, because people with the disease may search their memory more carefully than people without the disease.
For example, cases may think harder about past exposures (“Why did this happen to me?”), while controls may not remember as carefully.
🛡️ How to Minimise Information Bias
To reduce information bias:
- Make sure the exposure is clear, specific, and measurable.
- Measure exposure the same way in both groups.
- Blind investigators measuring exposure to whether someone is a case or control.
- Use objective records (medical records, lab results) instead of memory when possible.
- Use structured questionnaires to ensure consistency.
The key idea: measure exposure fairly and consistently.
🏥 Clinical Example (1–3 sentences)
Researchers studying head injury and dementia ask participants about past concussions. Patients with dementia may search their memory more deeply for head injuries than controls, creating recall bias. To minimise this, researchers use hospital records to confirm past injuries instead of relying only on memory.
👶 Explanation for a 10-Year-Old (1–3 sentences)
If you ask kids who have tummy aches what they ate last week, they might try really hard to remember and guess. Kids without tummy aches might not think much about it. That makes the answers unfair.
🧠 Memory Hook
“Bad memory = bad measurement.” 🧠📏
🚨 Key Points
- Information bias = incorrect measurement of exposure.
- Very common in case-control studies.
- Recall bias is the most common type.
- Prevent by standardising, blinding, and using objective records.
- Always measure exposure clearly and equally in both groups.
Explain how confounding bias can occur in a case control study, and how to minimise it.
🧩 What is Confounding Bias?
Confounding bias happens when a third factor (a confounder) is related to both the exposure and the outcome, and it distorts the true relationship between them.
In simple terms:
> A confounder is a “mix-up factor” that makes it look like the exposure causes the disease — when really another factor is involved.
From your image (reworded clearly):
Confounding bias can be managed during both:
- The design phase of the study
- The analysis phase of the study
So it must be controlled from the start and also checked when analysing the data.
🕵️ How can it occur in a Case-Control Study?
Case-control studies are especially vulnerable because:
- They are observational (no randomisation).
- Cases and controls may differ in important characteristics besides the disease.
- If these characteristics are related to both exposure and disease, they can distort the results.
For example:
If we study coffee and heart disease, smoking could be a confounder because:
- Smokers drink more coffee.
- Smokers have more heart disease.
So coffee may falsely appear to cause heart disease.
🛡️ How to Minimise Confounding Bias
During the Design Phase:
- Restriction (only include one category, e.g., only non-smokers)
- Matching (match cases and controls on age, sex, etc.)
- Careful selection of comparable controls
During the Analysis Phase:
- Stratification (analyse groups separately by confounder level)
- Statistical adjustment (e.g., regression analysis)
The key idea:
Control confounding both before and after collecting data.
🏥 Clinical Example (1–3 sentences)
Researchers study alcohol and liver cancer. Smoking is common in heavy drinkers and also increases cancer risk. If smoking is not controlled for, alcohol may appear more harmful than it truly is.
👶 Explanation for a 10-Year-Old (1–3 sentences)
If kids who eat more ice cream also play outside more, and they get more sunburn, you might think ice cream causes sunburn. But really, playing outside is the real reason.
🧠 Memory Hook
“Confounder = the sneaky third wheel.” 🚲
🚨 Key Points
- Confounding bias = distortion from a third variable.
- Common in observational studies like case-control studies.
- Managed during both design and analysis phases.
- Control by restriction, matching, stratification, or regression.
- Always ask: “Is there a third factor mixing things up?”
What measure of effect is used for case control studies?
Odds ratio.
Odds ratio is a measure of effect used for case control studies. What is it? How do you calculate it?
🎲 What is an Odds Ratio (OR)?
An odds ratio (OR) is a measure of effect used mainly in case-control studies to show how strongly an exposure is associated with an outcome.
Because case-control studies start with people who already have the disease (cases), we cannot calculate risk directly (since we don’t know the total population at risk). So instead, we calculate odds.
📊 What are “odds”?
Odds are different from probability.
- Probability = chance something happens.
- Odds = probability it happens ÷ probability it does NOT happen.
So:
> Odds = Probability it happens / Probability it doesn’t happen
Example:
The probability of rolling a 4 on a die is 1/6.
The probability of not rolling a 4 is 5/6.
So the odds of rolling a 4 are 1/5.
🧮 How is the Odds Ratio calculated?
In a case-control study, we compare:
- The odds of exposure in cases
to - The odds of exposure in controls
Formula (using the classic 2×2 setup):
If:
- a = exposed cases
- b = exposed controls
- c = unexposed cases
- d = unexposed controls
Then:
> OR = (a × d) / (b × c)
Excavate x Unicorn
———————————-
Exconvict x Uncas (incas)
This tells us how much more (or less) likely exposure is among cases compared to controls.
📈 How do we interpret OR?
- OR = 1 → No association
- OR > 1 → Exposure associated with higher odds of outcome
- OR < 1 → Exposure may be protective
Also:
- Always check confidence intervals (CI) and p-values for statistical significance.
- When the outcome is rare, OR ≈ Relative Risk (RR).
🏥 Clinical Example (1–3 sentences)
Researchers study smoking and lung cancer. They compare smokers and non-smokers among people with lung cancer (cases) and people without lung cancer (controls). If the OR is 4, it means smokers have 4 times the odds of lung cancer compared to non-smokers.
👶 Explanation for a 10-Year-Old (1–3 sentences)
If more kids with tummy aches ate candy yesterday than kids without tummy aches, we compare the “odds” of candy-eating in both groups. If the odds are much higher in the tummy-ache group, candy might be linked.
🧠 Memory Hook
“Case-Control counts the Odds to measure the Effect.” 🎲📊
🚨 Key Takeaways
- OR is the main measure of effect in case-control studies.
- Odds = probability it happens ÷ probability it doesn’t.
- OR compares odds of exposure in cases vs controls.
- OR > 1 = increased odds.
- OR < 1 = protective effect.
- In rare diseases, OR approximates RR.
What is relative risk and how is it calculated?
📊 What is Relative Risk (RR)?
Relative Risk (RR) is a measure of effect that tells us how much more (or less) likely an event is to happen in one group compared to another group.
It compares:
> The risk in the exposed (or treatment) group
to
The risk in the unexposed (or control) group
Unlike odds ratio (used in case-control studies), relative risk is used in cohort studies and randomized controlled trials, where we can directly measure incidence (how often something happens).
🧮 How is Relative Risk calculated?
First, calculate the risk (probability of the event) in each group:
- Risk in exposed group = EER (Experimental Event Rate)
- Risk in control group = CER (Control Event Rate)
Then:
> RR = EER ÷ CER
EER = a/(a+b) —> Exposed that got it / (everyone Exposed)
CER = c/(c+d) —> controls that got it / (everyone not exposed)
📈 How to interpret RR
- RR = 1 → No difference between groups
- RR > 1 → Exposure increases risk
- RR < 1 → Exposure decreases risk (protective)
For example:
- RR = 2 → Event is twice as likely in exposed group
- RR = 0.5 → Event is half as likely in exposed group
🏥 Clinical Example (1–3 sentences)
In a study of a new cholesterol drug, 10% of patients taking the drug have a heart attack, while 20% in the control group do.
RR = 0.10 ÷ 0.20 = 0.5.
This means the drug halves the risk of heart attack compared to control.
👶 Explanation for a 10-Year-Old (1–3 sentences)
If 2 out of 10 kids wearing helmets fall off their bikes, but 4 out of 10 without helmets fall, we compare the risks. The helmet group has half the risk of falling.
🧠 Memory Hook
“Relative Risk = Risk Compared to Risk.” 📊⚖️
🚨 Key Points
- RR compares risk in two groups.
- RR = EER ÷ CER.
- Used in cohort studies and RCTs.
- RR > 1 = increased risk.
- RR < 1 = protective effect.
- Different from odds ratio (used in case-control studies).
Calculate odds ratio and relative risk for this set of data.
Provide a 1-sentence interpretation of your findings.
See image.
What type of analysis is used in case control studies and why is it used?
📊 What is Data Analysis in a Case-Control Study?
In a case-control study, data analysis is done to:
1️⃣ Control for confounders (other variables that may distort the relationship).
2️⃣ Measure the strength of association between exposure and outcome — usually using the Odds Ratio (OR).
Because case-control studies start with the outcome (cases vs controls), we mainly analyse how exposure differs between the two groups.
🧮 How is it done?
First, researchers:
- Compare exposure between cases and controls.
- Calculate the Odds Ratio (OR) to measure association.
- Adjust for confounding factors using statistical methods.
The goal is to determine whether exposure is associated with higher or lower odds of the disease.
📈 What Types of Analyses Are Used?
1️⃣ Simple Linear Regression
- Used when both variables are continuous (e.g., blood pressure and age).
- The outcome variable must be normally distributed.
- Not commonly used for classic case-control studies since the outcome is usually yes/no.
🧠 Memory hook: “Linear loves lines and continuous signs.” 📏
2️⃣ Multiple Regression
- Used when there is more than one independent (explanatory) variable.
- Helps control for confounders.
- For example: adjusting for age, sex, and smoking simultaneously.
🧠 Memory hook: “Multiple = many movers behind one outcome.” 🎛️
3️⃣ Logistic Regression
- Most commonly used in case-control studies.
- Used when the outcome is dichotomous (yes/no).
- Example: disease vs no disease.
- Produces an adjusted Odds Ratio.
🧠 Memory hook: “Logistic = yes/no logic.” 🔘
4️⃣ Conditional Logistic Regression
- Used in matched case-control studies.
- Applied when cases and controls are paired (e.g., matched by age or sex).
🧠 Memory hook: “Conditional = matched and attached.” 🔗
🏥 Clinical Example (1–3 sentences)
Researchers studying smoking and lung cancer calculate the odds ratio to measure the association. They then use logistic regression to adjust for age and occupational exposure to chemicals. If the adjusted OR remains high, smoking is strongly associated with lung cancer.
👶 Explanation for a 10-Year-Old (1–3 sentences)
If you want to see whether eating candy is linked to tummy aches, you compare how many kids with tummy aches ate candy versus kids without tummy aches. Then you adjust for other things like whether they also drank soda.
🚨 Key Points
- Analysis measures association using Odds Ratio (OR).
- Confounders are controlled using regression.
- Logistic regression is most common (because outcome is yes/no).
- Conditional logistic regression is used for matched studies.
- Purpose = find the true association after adjusting for other variables.
-
CAP 1: Intro to Evidence-Based Practice19
-
CAP 2: Intro to Data and Statistics18
-
CAP 2.2: Statistical Test Types (REVISE)23
-
CAP 2.3: Reliability & Validity (REVISE)21
-
CAP 2.4: P-values, Confidence Intervals3
-
CAP 2.5: Type 1 and 2 Errors, Power, Causation7
-
CAP 3: Bias19
-
CAP 4: Randomised Controlled Trials15
-
CAP 5: Case Control Studies15
-
CAP 6: Cohort Studies26
-
CAP 7: Diagnostic Studies25
-
CAP 8: Systematic Reviews & Meta-Analyses36
-
CAP 9: Economic Analyses17
-
CAP 10: Qualitative Research25
-
CAP 11: Network Meta-Analysis24
