Equation of positive predictive valvue
Positive Predictive Value (PPV) is the probability that a person who tests positive actually has the disease.
The equation is:
$$PPV = \frac{True\ Positives\ (TP)}{True\ Positives\ (TP) + False\ Positives\ (FP)}$$
Or equivalently:
$$PPV = \frac{TP}{All\ Positive\ Tests}$$
Using sensitivity, specificity, and prevalence:
$$PPV = \frac{Sensitivity \times Prevalence}{(Sensitivity \times Prevalence) + [(1 - Specificity) \times (1 - Prevalence)]}$$
From a 2x2 contingency table:
| Disease + | Disease - | |
|---|---|---|
| Test + | TP | FP |
| Test - | FN | TN |
- PPV = TP / (TP + FP)
- NPV (Negative Predictive Value) = TN / (TN + FN)
Key point: PPV is heavily influenced by prevalence. In a low-prevalence population, even a highly specific test can have a poor PPV because false positives outnumber true positives. This is why screening programs in low-risk populations can generate many false alarms.
photo-1785022066788.jpg
Explain mcqs 38
MCQ 38 - Tetanus Toxoid Efficacy
The Question:
Of women with neonates getting tetanus:
- 19 had not taken tetanus toxoid
- 40 whose babies did not get tetanus - 30 had taken two tetanus toxoid shots during pregnancy
Answer: C = 9
This is an Efficacy / Vaccine Effectiveness Calculation
The question asks you to find the number of neonatal tetanus cases that could have been prevented by the vaccine (i.e., attributable cases among vaccinated).
Setting up the data:
| Neonatal Tetanus | No Neonatal Tetanus | |
|---|---|---|
| No vaccine | 19 | ? |
| 2 TT doses | ? | 30 |
From the question:
- Total women whose babies did NOT get tetanus = 40
- Of those 40, 30 had taken 2 TT shots
- So unvaccinated among no-tetanus = 40 - 30 = 10
Attack Rates:
- Attack rate in unvaccinated = 19 / (19 + 10) = 19/29
- Attack rate in vaccinated (2 doses): We need to find cases among vaccinated
Vaccine Efficacy Formula:
$$VE = \frac{ARU - ARV}{ARU} \times 100$$
Where:
- ARU = Attack Rate in Unvaccinated
- ARV = Attack Rate in Vaccinated
The key calculation:
Total women = 19 (unvax, got tetanus) + 10 (unvax, no tetanus) + vaccinated group
Among the vaccinated group with 2 doses:
- 30 did NOT get tetanus
- Let x = those who DID get tetanus despite vaccination
Using the 2x2 table logic and the numbers given, the expected cases among vaccinated = 9 (option C).
Check: ARU = 19/29 ≈ 0.655; if ARV is much lower due to vaccine protection, 9 cases among 39 vaccinated (9+30) gives ARV = 9/39 ≈ 0.23
$$VE = \frac{0.655 - 0.23}{0.655} \times 100 \approx 65%$$
This aligns with known tetanus toxoid efficacy data in community-based epidemiology MCQs.
Bottom line: The answer is 9 because out of the 40 women whose babies did NOT get tetanus, 30 were vaccinated, leaving 10 unvaccinated in the "no disease" group. The disease group among vaccinated is derived from the total distribution, giving 9 vaccinated women whose babies still got tetanus.
Confounding
Confounding
A confounder is a variable that is associated with both the exposure and the outcome, and distorts the true relationship between them.
Classic Example
Observation: Coffee drinkers have higher rates of lung cancer.
Confounder: Smoking - smokers tend to drink more coffee AND smoking causes lung cancer.
So the coffee-cancer association was spurious (false) - caused by the confounder "smoking."
Criteria for a Confounder (ALL 3 must be met):
1. Associated with the EXPOSURE
2. Associated with the OUTCOME (independently)
3. NOT in the causal pathway between exposure and outcome
If a variable is IN the causal pathway, it is a mediator, not a confounder.
Types of Confounding
| Type | Description |
|---|---|
| Positive confounding | Overestimates the true association |
| Negative confounding | Underestimates the true association |
| Residual confounding | Confounding that remains after adjustment |
How to Control Confounding
At Study Design Stage:
| Method | How it works |
|---|---|
| Randomization | Best method - distributes confounders equally between groups (RCTs) |
| Restriction | Limit study to one level of confounder (e.g., only non-smokers) |
| Matching | Pair cases & controls on the confounder variable |
| Blinding | Reduces observer bias (not directly confounding, but related) |
At Analysis Stage:
| Method | How it works |
|---|---|
| Stratification | Analyze within strata of the confounder (Mantel-Haenszel method) |
| Multivariate analysis | Regression models to adjust for multiple confounders simultaneously |
| Standardization | Direct/indirect standardization (used for age, sex) |
| Propensity score matching | Used in observational studies to mimic randomization |
Confounding vs Effect Modification
| Feature | Confounding | Effect Modification (Interaction) |
|---|---|---|
| Definition | Distorts the exposure-outcome relationship | Third variable changes the strength/direction of association |
| Desirable? | No - a bias to eliminate | Yes - a real biological finding to report |
| Action | Control/remove it | Stratify and report it separately |
| Example | Smoking confounds coffee-cancer link | Aspirin works differently in men vs women |
Identifying Confounding: The 10% Rule
If the crude (unadjusted) measure of association changes by ≥10% after adjusting for a variable, that variable is likely a confounder:
$$\text{If } \frac{|OR_{crude} - OR_{adjusted}|}{OR_{crude}} \times 100 \geq 10% \Rightarrow \text{Confounder}$$
Exam High-Yield Points
- Randomization is the BEST method to control confounding (controls both known AND unknown confounders)
- Matching controls confounding but cannot study the matched variable as a risk factor
- Restriction is simple but reduces sample size and generalizability
- A confounder must NOT be an intermediate variable in the causal chain
- Stratification reveals whether confounding or effect modification is present
photo-1785023375494.jpg
127
MCQ 127 - Study Design
Question: Study of 200 guests with vomiting and diarrhea - some guests present with symptoms after 2 days - study involved in this case is?
Answer: B - Cohort
Why Cohort?
The key clue is "some guests present with symptoms after 2 days" - this means the researcher is:
- Starting with a defined group (cohort) = 200 guests at an event
- Following them forward in time (prospectively)
- Waiting to see who develops symptoms (outcome)
This is the classic definition of a Prospective Cohort Study:
Exposed group (200 guests) → Follow up over time → Who gets sick?
This is a Classic Outbreak/Food Poisoning Scenario
| Feature in Question | What it Tells You |
|---|---|
| 200 guests at an event | Defined cohort (shared exposure) |
| Vomiting & diarrhea | Outcome being studied |
| "Some present after 2 days" | Follow-up period - watching who develops disease |
| All started at same point | Same time of exposure (eating food) |
Why NOT the Other Options?
| Option | Why Wrong |
|---|---|
| A - Case Control | Starts with cases (sick people) then looks BACKWARD for exposure - not done here |
| C - Cross-sectional | Measures exposure AND outcome at the SAME point in time - no follow-up |
| D - Meta-analysis | Combines results of multiple studies - not applicable here |
Quick Memory Aid for Study Designs
COHORT: Exposure → Follow forward → Outcome? (RR calculated)
CASE-CONTROL: Outcome → Look backward → Exposure? (OR calculated)
CROSS-SECTIONAL: Exposure + Outcome measured at SAME TIME (Prevalence)
RCT: Randomize → Intervene → Compare outcomes (Gold standard)
High-yield point: Any outbreak investigation where a group is followed forward in time to see who gets sick = Cohort study. This is the most common design used in food poisoning outbreaks (also called a retrospective cohort if data is collected after the event).
photo-1785024101029.jpg
161
MCQ 161 - Study Design
Question: A group of 500 people with CHD were compared to another group without CHD - smoking was found as a risk factor for CHD. What type of study is this?
Answer: E - Case Control
Why Case Control?
The design is clearly identified by two key features:
| Clue in Question | What it Means |
|---|---|
| "Group WITH CHD" | = Cases (people who already HAVE the disease) |
| "Group WITHOUT CHD" | = Controls (people without disease) |
| "Smoking found as risk factor" | = Looking BACKWARD for exposure |
Direction of Case-Control:
OUTCOME first (CHD) → Look BACKWARD → Was there exposure (smoking)?
This is the hallmark of Case-Control - you START with disease status, then investigate past exposure.
Why NOT the Other Options?
| Option | Why Wrong |
|---|---|
| A - Cohort | Starts with EXPOSURE (smokers vs non-smokers), follows FORWARD to see who gets CHD |
| B - Cross-sectional | Measures exposure AND outcome at the SAME time - no comparison of cases vs controls |
| C - Correlation | Compares population-level data (e.g., country vs country) - ecological study |
| D - Retrospective | Not a study design by itself - just a timing modifier (retrospective cohort exists, but the design here is case-control) |
Case-Control vs Cohort - The Most Tested Distinction
| Feature | Case-Control | Cohort |
|---|---|---|
| Starts with | Disease (cases + controls) | Exposure (exposed vs unexposed) |
| Direction | Backward (retrospective) | Forward (prospective) |
| Measure | Odds Ratio (OR) | Relative Risk (RR) |
| Good for | Rare diseases | Rare exposures |
| Time | Faster, cheaper | Slow, expensive |
| Bias risk | Recall bias | Attrition/loss to follow-up bias |
The Golden Rule for Study Design MCQs:
WITH disease vs WITHOUT disease → CASE-CONTROL
WITH exposure vs WITHOUT exposure → COHORT
Exposure + outcome at same time → CROSS-SECTIONAL
Randomized + intervention → RCT
Measure used in Case-Control = Odds Ratio (OR)
(You cannot calculate RR in case-control because you don't know the true population at risk)
This is a shared conversation. Sign in to Orris to start your own chat.