Everything you need to know about... MEDICAL STATISTICS!

Summary

"Everything You Need To Know About Statistics" is a unique on-demand teaching session designed for medical professionals who want to refresh or enhance their understanding of key statistical concepts. This session, developed by Millie and Prisha, is reviewed by doctors to ensure accuracy and provides weekly tutorial content focused on core presentations and diagnostic techniques. You'll dive deep into definitions such as incidence, prevalence, and different types of studies like meta analysis, randomized control trials, cohort studies, etc. The session also covers essential statistical calculations, from sensitivity/specifinity to odds ratio and beyond. Regular updates about upcoming events are provided via email and group chats. This is a fantastic opportunity for any medical professional to learn or revisit statistics in a high-yield, practical manner.

Generated by MedBot

Description

Welcome to Teaching Things!

We're excited to bring you this high-yield teaching series, designed to help you ace both your written and practical exams.

This tutorial will focus on Statistics covering key differentials such as NNT and specificity/sensitivity to ensure you're well-prepared.

The session will be led by Millie and Prisha, both medical students in their clinical years at UCL, who are passionate about delivering practical, exam-focused content.

Don’t forget to fill out the feedback form after the tutorial—we value your input! And remember, you can access recordings of all past tutorials on our page.

Learning objectives

Understand and distinguish between different types of statistics such as incidence, prevalence, odds, sensitivity, specificity, etc.
Understand the principles and applications of different types of research studies including meta-analysis, systematic review, case control, randomised control trial, cross-sectional, and cohort studies.
Evaluate statistical data accurately and make informed medical decisions based on that data.
Understand the advantages and limitations of different types of research studies and be aware of potential biases.
Develop skills in explaining statistical significance and clinical relevance to both peers and patients.

Generated by MedBot

Similar communities

View all

Similar events and on demand videos

Computer generated transcript

Warning!
The following transcript was generated automatically from the content and has not been checked or corrected manually.

EVERYTHING YOU NEED TO KNOW ABOUT STATISTICS Millie and Prisha Slidand Arighnireated by Shazia Here’s what we do: ■ Weekly tutorials open to all! ■ Focussed on core presentations and teaching diagnostic technique If you’re new here… ■ Bstudentsl students, for medical ■ Reviewed by doctors to ensure W elcome to accuracy T eaching ■ We’ll keep you updated about our Things! upcoming events via email and groupchats!Statistics & Ethics - is this high yield?What do we need to know for Statistics? Deﬁnitions: Calculations: 1. Incidence 1. Sensitivity/ Speciﬁcity 2. Prevalence 2. Odds Types of studies: 3. Odds ratio 1. Meta analysis 4. Absolute risk 2. Systematic review 5. Relative risk/ Risk ratio 3. Case control 6. Absolute risk reduction 4. Randomised control trial 7. Relative risk reduction 5. Cross sectional 8. Number needed to treat/ Number needed to harm 6. Cohort 9. Positive predictive value/ Negative predictive valueMatch the following A. Number of children who have polio in the 1. Incidence UK in 2024 B. Number of children who developed polio 2. Prevalence in the UK in 2024 Match the following A. Number of children who have polio in the 1. Incidence UK in 2024 B. Number of children who developed 2. Prevalence polio in the UK in 2024Deﬁnitions 1. Incidence: Number of new cases of characteristic that develop in a population in a speciﬁed time period 2. Prevalence: Proportion of population who have speciﬁc characteristic in a given time period regardless of when they ﬁrst developed the characteristicWhat’s the Randomised Control T rial? 1. 5000 people are enrolled in a study. 1000 are given a cholesterol lowering drug and 4000 are given a placebo. The participants are followed up over 10 years for incidence of an MI. 2. 200 women with breast cancer and 200 without breast cancer are enrolled in a study. They are asked about use of hormonal contraceptive in the past 10 years to determine an association. 3. A sample of 600 medical students asked about current use of Anki in their studies. 4. 500 people following a vegetarian diet and 500 people who eat meat are followed for 10 years and analysed for the development of cardiovascular disease.What’s the Randomised Control T rial? 1. 5000 people are enrolled in a study. 1000 are given a cholesterol lowering drug and 4000 are given a placebo. The participants are followed up over 10 years for incidence of an MI. 2. 200 women with breast cancer and 200 without breast cancer are enrolled in a study. They are asked about use of hormonal contraceptive in the past 10 years to determine an association. 3. A sample of 600 medical students asked about current use of Anki in their studies. 4. 500 people following a vegetarian diet and 500 people who eat meat are followed for 10 years and analysed for the development of cardiovascular disease.T ypes of studies 1. Randomised control trial: - Experimental, prospective, longitudinal study - Allows analysis of efficacy of intervention - Sample population divided into placebo and intervention groups - Can be blind or double blind - Limitations: time consuming, ethical issues, must ensure demographics of sample population apply to the person who is considering the intervention- Intention to treat analysis: means including all patients who were enrolled in the study in the analysis even if they drop out of study to mimic real world outcomes e.g. non-compliance - Composite primary outcome: combining multiple endpoints into single outcome e.g. death and MI used as single endpoint - P-value: describes how likely it is that the results we get are due to random chance - we normally say that if it is less than 5% we are happy to say we have signiﬁcant results and the results are not due to random chance - Conﬁdence interval: the ranges of values within which 95% of the data within a certain dataset liesWhat’s the Case Control Study? 1. 300 medical students are given caffeine pill, and 400 students are given placebo tablets, and are analysed for exam performance in 6 months. 2. A sample of 1000 nurses are asked about their current exercise habits. 3. 300 people living in London (high pollution levels) and rural Wales (low pollution levels) are followed for 20 years and investigated for development of asthma or COPD. 4. 800 people with lung cancer and 600 people without cancer are investigated for past smoking history to determine if there is an association.What’s the Case Control Study? 1. 300 medical students are given caffeine pill, and 400 students are given placebo tablets, and are analysed for exam performance in 6 months. 2. A sample of 1000 nurses are asked about their current exercise habits. 3. 300 people living in London (high pollution levels) and rural Wales (low pollution levels) are followed for 20 years and investigated for development of asthma or COPD. 4. 800 people with lung cancer and 600 people without cancer are investigated for past smoking history to determine if there is an association.T ypes of studies 1. Case Control: - Group with disease compared to group without disease and retrospectively analysed for exposure to variable/ risk factor e.g. smoking for MI - Allows for analysis of one outcome for multiple risk factors, control for confounding variables, allows study of rare outcomes - Limitations: recall bias in retrospective analysis, correlation doesn’t necessary imply causationWhat’s the Cross-sectional Study? 1. 80 patients following a vegan diet and 70 patients following a non-vegetarian diet are followed up over a period of 30 years for development of osteoporosis. 2. 100 patients with Clostridium Difficile and 100 patients without Clostridium Difficile are investigated for prior use of antibiotics. 3. A sample of 60 Neurosurgical Consultants are asked whether they like having medical students in their clinic. 4. 70 patients are given gold-standard quadruple TB treatment and 50 patients are given 3 medications and 1 placebo medication. They are followed up for negative TB testing and resolution of symptoms.What’s the Cross-sectional Study? 1. 80 patients following a vegan diet and 70 patients following a non-vegetarian diet are followed up over a period of 30 years for development of osteoporosis. 2. 100 patients with Clostridium Difficile and 100 patients without Clostridium Difficile are investigated for prior use of antibiotics. 3. A sample of 60 Neurosurgical Consultants are asked whether they like having medical students in their clinic. 4. 70 patients are given gold-standard quadruple TB treatment and 50 patients are given 3 medications and 1 placebo medication. They are followed up for negative TB testing and resolution of symptoms. T ypes of studies 1. Hi 2. Hi 3. Cross sectional: - Observational survey at one point in time i.e. how many people have condition - Quick, easy - Limitations: Need large sample size, difficult to draw correlations, causationsWhat’s the Cohort Study? 1. 900 chemotherapy patients are surveyed about experiencing pruritic symptoms following treatment. 2. 100 patients with Infective endocarditis and 100 patients without infective endocarditis are assessed for IV drug use in the past 5 years. 3. 100 medical students who regularly attended sport night and 100 medical students who occasionally attend sports night are investigated for alcoholic behaviour in 10 years. 4. 800 patients with Multiple Myeloma are given hydrocortisone along with their chemotherapy, and 700 are given a placebo. They are assessed for D&V in the week following receipt of treatment.What’s the Cohort Study? 1. 900 chemotherapy patients are surveyed about experiencing pruritic symptoms following treatment. 2. 100 patients with Infective endocarditis and 100 patients without infective endocarditis are assessed for IV drug use in the past 5 years. 3. 100 medical students who regularly attended sport night and 100 medical students who occasionally attend sports night are investigated for alcoholic behaviour in 10 years. 4. 800 patients with Multiple Myeloma are given hydrocortisone along with their chemotherapy, and 700 are given a placebo. They are assessed for D&V in the week following receipt of treatment.T ypes of studies 1. Hi 3. Hi 4. Cohort study: - Understanding cause of disease - Groups exposed to different factors and followed up to look for conditions e.g. smokers followed up in 10 years for incidence of MI - Allows the study of multiple outcomes - Limitations: time consuming, patients can be lost at follow upT ypes of studies 2. Fi 3. F 4. G 5. Meta -analysis: - Lots of studies analysed and data combined to assess efficacy/impact of intervention/ condition - Limitations: time consuming, different techniques, must ensure all studies are comparable 6. Systematic review: - In depth analysis of relevant research surrounding topic, and assessing quality, conclusions and reliabilityHow should we calculate speciﬁcity and sensitivity? Patients with colonPatients without Total cancer colon cancer Positive CEA 925 7 932 Negative CEA 75 893 968 1000 900 1900 Speciﬁcity: Sensitivity: 1. 925/ 1000 1. 925/ 1000 2. 925/ 932 2. 925/ 932 3. 893/900 3. 893/900 4. 893/968 4. 893/968How should we calculate speciﬁcity and sensitivity? Patients with colonPatients without Total cancer colon cancer Positive CEA 925 7 932 Negative CEA 75 893 968 1000 900 1900 Speciﬁcity: Sensitivity: 1. 925/ 1000 1. 925/ 1000 2. 925/ 932 2. 925/ 932 3. 893/900 3. 893/900 4. 893/968 4. 893/968 Annoying but Calculations seNsitivity = Positive sPecificity = Negative 1. Sensitivity and Speciﬁcity Speciﬁcity: Sensitivity: - Tells you how many people who have - Tells you how many people who do not the disease actually test positive have the disease actually test negative - Calculated by true negative numbers/ - Calculated by true positive numbers/ all who actually don’t have the disease all who really have the disease - A high sensitivity means most people - A high speciﬁcity means most people who have the disease will test positive who do not have the disease will test with this test negative with this test - Makes it a good rule IN test - if you - Makes it a good rule OUT test - if you don’t test negative, you probably have don’t test positive, you probably don’t the disease have the diseaseIn a population of 10,000 people, the prevalence of Guillain-Barre Syndrome (GBS) is said to be 7%. Researchers used a new diagnostic test in this population to help identify those with the condition. Results showed the test had a sensitivity of 60% and speciﬁcity of 75%. By using this diagnostic test, how many people without GBS will test positive? A. 2325 B. 525 C. 2500 D. 6975 E. 7500 Sensitivity = true positive/ Specificity = true negative/ all who have disease all who don’t have disease In a population of 10,000 people, the prevalence of Guillain-Barre Syndrome (GBS) is said to be 7%. Researchers used a new diagnostic test in this population to help identify those with the condition. Results showed the test had a sensitivity of 60% and speciﬁcity of 75%. By using this diagnostic test, how many people without GBS will test positive? A. 2325 Have GBS Do not have Total B. 525 GBS C. 2500 D. 6975 Positive test 420 2325 2745 E. 7500 Negative test 280 6975 7255 700 9300 10000 Sensitivity = true positive/ Specificity = true negative/ all who have disease all who don’t have disease In a population of 10,000 people, the prevalence of Guillain-Barre Syndrome (GBS) is said to be 7%. Researchers used a new diagnostic test in this population to help identify those with the condition. Results showed the test had a sensitivity of 60% and speciﬁcity of 75%. By using this diagnostic test, how many people without GBS will test positive? A. 2325 Have GBS Do not have Total B. 525 GBS C. 2500 D. 6975 Positive test 420 2325 2745 E. 7500 Negative test 280 6975 7255 700 9300 10000How should we calculate Positive and Negative Predictive Value? Patients with colonPatients without Total cancer colon cancer Positive CEA 925 7 932 Negative CEA 75 893 968 1000 900 1900 PPV: NPV: 1. 925/ 1000 1. 925/ 1000 2. 925/ 932 2. 925/ 932 3. 893/900 3. 893/900 4. 893/968 4. 893/968 - Tells you how - Tells you how many people who many people whoHow should we calculate Positive and Negative Predictive Value? Patients with colonPatients without Total cancer colon cancer Positive CEA 925 7 932 Negative CEA 75 893 968 1000 900 1900 PPV: NPV: 1. 925/ 1000 1. 925/ 1000 2. 925/ 932 2. 925/ 932 3. 893/900 3. 893/900 4. 893/968 4. 893/968Calculations 1. Sensitivity and Speciﬁcity 2. Positive and Negative Predictive Value PPV NPV - Proportion of people who test - Proportion of people who test positive who are actually negative who are actually negative positive - Calculated by true negative - Calculated by true positive numbers/ all who are actually numbers/ all who test positive negative - Makes it a good way of - Makes it a good way of measuring measuring the the accuracy/usefulness of a test accuracy/usefulness of a test Y our turn! A new diagnostic test for myasthenia gravis is being assessed. A total of 4700 patients are recruited for the study, including 2500 healthy volunteers. Of the 2200 patients, 2000 patients receive a true-positive result, while 2200 patients receive a true-negative result. What is the positive predictive value of the new test? A. 90.9 B. 87 C. 42.6 D. 91.7 E. 88 PPV = true positive/ all who NPV= true negative/ all who Y our turn! test positive test negative A new diagnostic test for myasthenia gravis is being assessed. A total of 4700 patients are recruited for the study, including 2500 healthy volunteers. Of the 2200 patients, 2000 patients receive a true-positive result, while 2200 patients receive a true-negative result. What is the positive predictive value of the new test? A. 90.9 B. 87 Have MG Don’t have MG Total C. 42.6 D. 91.7 Positive test 2000 300 2300 E. 88 Negative test 200 2200 2400 2200 2500 4700 PPV = true positive/ all who NPV= true negative/ all who Y our turn! test positive test negative A new diagnostic test for myasthenia gravis is being assessed. A total of 4700 patients are recruited for the study, including 2500 healthy volunteers. Of the 2200 patients, 2000 patients receive a true-positive result, while 2200 patients receive a true-negative result. What is the positive predictive value of the new test? A. 90.9 B. 87 Have MG Don’t have MG Total C. 42.6 D. 91.7 Positive test 2000 300 2300 E. 88 Negative test 200 2200 2400 2200 2500 4700 PPV = true positive/ all who NPV= true negative/ all who Y our turn! test positive test negative A new diagnostic test for myasthenia gravis is being assessed. A total of 4700 patients are recruited for the study, including 2500 healthy volunteers. Of the 2200 patients, 2000 patients receive a true-positive result, while 2200 patients receive a true-negative result. What is the positive predictive value of the new test? A. 90.9 B. 87 Have MG Don’t have MG Total C. 42.6 D. 91.7 Positive test 2000 300 2300 E. 88 Negative test 200 2200 2400 2200 2500 4700 2000/2300 = 0.869 PPV = true positive/ all who NPV= true negative/ all who Y our turn! test positive test negative A new diagnostic test for myasthenia gravis is being assessed. A total of 4700 patients are recruited for the study, including 2500 healthy volunteers. Of the 2200 patients, 2000 patients receive a true-positive result, while 2200 patients receive a true-negative result. What is the positive predictive value of the new test? A. 90.9 B. 87 Have MG Don’t have MG Total C. 42.6 D. 91.7 Positive test 2000 300 2300 E. 88 Negative test 200 2200 2400 2200 2500 4700 2000/2300 = 0.869Calculations 2.Positive and Negative Predictive Val 3. Odds and Odds ratio Odds Odds ratio - Proportion of people exposed to a - Odds (affected)/ odds (control) risk factor in affected and non-affected groups - OR = 1: exposure doesn’t affect outcome - Used in case control studies - OR < 1: exposure reduces probability of being affected - Odds of exposure in affected group - OR > 1: exposure increases probability of = affected with exposure/ affected without exposure being affected - Odds of exposure in non-affected group = non-affected with exposure/ non-affected without exposure - NP: not over total sum!Y our turn! 900 people are enrolled in a study. 500 people are found to have Clostridium Difficile and 400 people do not have C. Diff. 392 of the people who have C.Diff had used antibiotics in the last month, and 102 of those who don’t have C.Diff had used antibiotics in the last month. Does this study suggest that antibiotic use increases the probability of C.Diff infection?Y our turn! 900 people are enrolled in a study. 500 people are found to have Clostridium Difficile and 400 people do not have C. Diff. 392 of the people who have C.Diff had used antibiotics in the last month, and 102 of those who don’t have C.Diff had used antibiotics in the last month. Does this study suggest that antibiotic use increases the probability of C.Diff infection? Odds of antibiotic use in C.Diff group = 392/108 = 3.63 Odds of antibiotic use in non-C.Diff group = 102/298 = 0.342 Odds ratio = Odds in C.Diff group/ Odds in non-C.Diff group = 10.61 As OR > 1 = Antibiotic use increases chance of C.Diff infection Anything ABSOLUTE = Anything RELATIVE = Calculations minusing involved dividing involved 1.Sensitivity and Speciﬁcity 3.Positive and Negative Predictive Value 4. Absolute Risk and Risk Ratio/ Relative Risk AR RR - Probability of an event - Probability of the event happening happening in the intervention in the intervention group as and control group respectively compared to the control group - Used in cohort and RCT studies - Calculated by AR in intervention - Calculated by number of times group/ AR in control group an event occurs/ total number - RR = 1: intervention has no effect on of people in the group risk - RR > 1: intervention increases risk - RR < 1: intervention reduces riskY our turn! Review the results of the following randomised control trial. Is there evidence to suggest that the intervention reduces the results of a poor outcome? Poor No poor Total AR in intervention group outcome outcome = 98/1000 AR in placebo group = 392/1000 Intervention 98 902 1000 Group RR = 0.098/ 0.392 Placebo 392 608 1000 = 0.25 group 0.25 < 1 = intervention 490 1510 2000 reduces risk Anything ABSOLUTE = Anything RELATIVE = Calculations minusing involved dividing involved 1.Sensitivity and Speciﬁcity 3.ddsitive and Negative Predictive Value 4.Absolute Risk and Risk Ratio/ Relative Risk 5. Absolute Risk Reduction and Relative Risk Reduction ARR RRR - Difference between event - Relative/ comparative reduction in rates in control group vs the rates of event in intervention and treatment group control group - Calculated by AR (control) - AR - Calculated by 1-RR or ARR/AR (intervention) (control) - How much intervention - Can lead to overestimation of the reduces risk by treatment effectY our turn! You read a paper discussing a new drug which is said to reduce the chance of patients with ulcerative colitis developing colon cancer. The study recruited 2000 patients. 800 received the new drug of which 6 patients developed colon cancer. The remaining patients received a placebo of which 15 developed colon cancer. What is the relative risk reduction of developing colon cancer by taking the new drug? A. 0.5 B. 0.09 C. 0.0075 D. 0.0125 E. 0.4Y our turn! You read a paper discussing a new drug which is said to reduce the chance of patients with ulcerative colitis developing colon cancer. The study recruited 2000 patients. 800 received the new drug of which 6 patients developed colon cancer. The remaining patients received a placebo of which 15 developed colon cancer. What is the relative risk reduction of developing colon cancer by taking the new drug? A. 0.5 B. 0.09 AR (intervention) = 6/800 = 0.0075 AR (control) = 15/1200 = 0.0125 Cancer No Total C. 0.0075 cancer D. 0.0125 RR = 0.0075/0.0125 = 0.6 Intervention Group 6 794 800 ARR = 0.0125 - 0.0075 = 0.005 E. 0.4 Placebo group 15 1185 1200 RRR = 1 - RR = 1 - 0.6 = 0.4 = ARR/ AR (control) = 0.005/0.0125 = 21 1979 2000 0.4Calculations 1. Sensitivity and Speciﬁcity 3. 4ositive and Negative Predictive Value 5. 4 6. Number needed to treat and number needed to harm NNT NNH - Number of patients needed to - Number of people needed to treat treat to prevent 1 additional to cause 1 additional poor outcome poor outcome - Want the highest value possible - Want the lowest value possible - Calculated by 1/ARI (absolute risk - Calculated by 1/ARR increase if there is an increase) Tutor 1: Amelia Snook Tutor 2: Prisha PahariyaY our turn! A 26-year-old patient asks you about a study she read investigating the efficacy of stopping smoking in reducing ﬂares of her Crohn's disease. You review the study so that you can answer her questions effectively. The study found there were fewer cases of ﬂares of Crohn's disease in non-smokers compared to smokers. The absolute risk of ﬂare-ups in non-smokers was 0.15 versus 0.85 in smokers. What is the number needed to treat in this study to prevent a ﬂare-up of Crohn's disease? A. 142.9 B. 7 C. 0.7 D. 70 E. 1.4 NNT = 1/ARR Y our turn! A 26-year-old patient asks you about a study she read investigating the efficacy of stopping smoking in reducing ﬂares of her Crohn's disease. You review the study so that you can answer her questions effectively. The study found there were fewer cases of ﬂare-ups in non-smokers was 0.15 versus 0.85 in smokers. What is the number needed to treat in this study to prevent a ﬂare-up of Crohn's disease? A. 142.9 B. 7 ARR = AR (control) - AR C. 0.7 (intervention) D. 70 ARR = 0.85 - 0.15 = 0.7 E. 1.4 NNT = 1/0.7 = 1.42 Cheat sheet! Sensitivity = True Positive/ All Who Have Relative Risk (RR) = AR (intervention)/ AR Condition (control) Speciﬁcity = True Negative/ All Who Don’t Have Condition Absolute Risk Reduction (ARR) = AR (control) - PPV = True Positive/ All Who Test Positive AR (intervention) NPV = True Negative/ All Who Test Negative Relative Risk Reduction (RRR) = 1 - RR or Odds Ratio (OR) = Number With Exposure/ ARR/ AR (control) Number Without Exposure NNT = 1/ARR Absolute Risk (AR) = Number of Times Event Occurs/ Total Number Within Group NNH = 1/ARI Tutor 1: Amelia Snook Tutor 2: Prisha Pahariya THANKS FOR W TCHING! on Medall and see you next week! Tutor 1: Amelia Snook Tutor 2: Prisha Pahariya