L5. Study Designs PDF

05 Population Medicine Prof. Suhail Doi Study Designs 15th January 2020 Maryam T Aboughalia Aseel S Hassona & Mohamed M Tawengi 1 This document resorted to: 1. “Study Designs” Lecture Slides 2. “Understanding Evidence in Health Care” Chapter 1 3. “FIRST AID for USMLE, 2020” Section II, P 255-264 Gre...

05 Population Medicine Prof. Suhail Doi Study Designs 15th January 2020 Maryam T Aboughalia Aseel S Hassona & Mohamed M Tawengi 1 This document resorted to: 1. “Study Designs” Lecture Slides 2. “Understanding Evidence in Health Care” Chapter 1 3. “FIRST AID for USMLE, 2020” Section II, P 255-264 Greetings all, the main focus of this sheet is study designs. We came across this topic multiple times, which marks its importance in clinical research. Going through the sheet, we will recap few concepts needed, and mainly focus on study designs, this sheet should be sufficient for our need. For a condensed revision of clinical epidemiology, you can revisit sheet 20 of unit 1 “introduction to EBM and Clinical Epidemiology”. You may also read the sheet “Study Designs” written by batch 2025 for detailed explanation of study designs. Statistical hypothesis: Null (H0): a statement suggesting that there is no difference, or association between two or more variables. H0 is tested for possible rejection under the assumption that the hypothesis is true. Example: “there is no association between sodium intake and hypertension.” Alternative (H1): a statement suggesting that there is some difference, or association between two or more variables, and contrary to the null hypothesis, the observations are the result of a real effect. Example: “increased sodium intake leads to increased blood pressure.” Outcomes of statistical hypothesis testing: Correct result ▸ Stating that there is an effect or difference when one exists (null hypothesis rejected in favour of alternative hypothesis) → "Power" ▸ Stating that there is no effect or difference when none exists (null hypothesis accepted/not rejected). Incorrect result *Blue shading = correct result ▸ Stating that there is an effect or difference when none exists (null hypothesis incorrectly rejected in favour of alternative hypothesis) → "Type I error" ▸ Stating that there is no effect or difference when one exists (null hypothesis is not rejected when it is in fact false) → "Type II error" α → You accused an innocent man. β → You blindly let the guilty man go free. 2 Prevalence, Incidence, and Duration: ▸ Prevalence is how many people in a sample group have a condition at a certain point in time. # 𝑜𝑓 𝑒𝑥𝑖𝑠𝑡𝑖𝑛𝑔 𝑐𝑎𝑠𝑒𝑠 𝑃𝑟𝑒𝑣𝑎𝑙𝑒𝑛𝑐𝑒 = (𝑎𝑡 𝑝𝑜𝑖𝑛𝑡 𝑖𝑛 𝑡𝑖𝑚𝑒) 𝑡𝑜𝑡𝑎𝑙 𝑝𝑜𝑝𝑢𝑙𝑎𝑡𝑖𝑜𝑛 ▸ Incidence is how many people will newly acquire a condition in a given period of time. 𝐼𝑛𝑐𝑖𝑑𝑒𝑛𝑐𝑒 = # 𝑜𝑓 𝑛𝑒𝑤 𝑐𝑎𝑠𝑒𝑠 (𝑝𝑒𝑟 𝑢𝑛𝑖𝑡 𝑜𝑓 𝑡𝑖𝑚𝑒) ℎ𝑒𝑎𝑙𝑡ℎ𝑦 𝑝𝑜𝑝𝑢𝑙𝑎𝑡𝑖𝑜𝑛 𝑎𝑡 𝑟𝑖𝑠𝑘 ▸ Duration is how long a given condition lasts, on average. These three terms are related by the formula: 𝑃𝑟𝑒𝑣𝑎𝑙𝑒𝑛𝑐𝑒 = 𝐼𝑛𝑐𝑖𝑑𝑒𝑛𝑐𝑒 × 𝐷𝑢𝑟𝑎𝑡𝑖𝑜𝑛 Key Fact! For diseases of short duration (e.g. common cold): Prevalence ≈ Incidence For chronic diseases (e.g. diabetes): Prevalence > Incidence; due to large # of existing cases Prevalence & Incidence are used to calculate: Risk; the likelihood of contracting a disease, it is calculated form incidence rate. Morbidity; the rate of disease in a population, reported in terms of prevalence & incidence rates. 3 Study Designs: All study designs have similar components; ① a defined population from which groups are selected and studied, ② exposures or interventions that are applied to different groups of subjects, and ③ the measured outcomes. The study design is broadly described as analytic if it compares groups of subjects and assesses association between exposures and outcomes, or descriptive if it simply measures the frequencies of factors in a population. ▸ Taxonomy of Study Designs ▸ Descriptive vs Analytic Descriptive studies Analytic studies ▸ Hypothesis generating. ▸ Hypothesis testing. ▸ No comparison groups. ▸ Comparison groups. ▸ Describe characteristics of disease (outcome) or ▸ Determine the association between exposure exposure (risk factor) with regards to populations, and outcome for decision making, is the geographic distribution and frequency variations over intervention associated with the outcome? time. Experimental studies (decide ▸ Case report: when you report an individual case that is Observational decide who gets the intervention) the initial point of discovering new diseases and making (don’t who gets the ▸ RCT new treatment. intervention) ▸ Case series: report of a group of similar patients. ▸ Meta-analyses of RCTs ▸ Cross-sectional ecologic study: aggregates data across ▸ Cohort ▸ Case-control countries and see difference between populations. ▸ Cross-sectional 4 ▸ Descriptive Epidemiologic Studies Studies which identify patterns/trends in disease occurrence, but cannot confirm any causality, it describes the patterns in relation to 3 main variables: Person: characteristics (age, sex, occupation) of the individuals affected by the outcome. Place: geography (residence, work, hospital) of the affected individuals. Time: when events occurred (diagnosis, reporting, testing). Types of descriptive studies are: case report or case series and cross-sectional ecologic studies. These studies are usually quick and inexpensive; as they use already available databases with efficient allocation of resources. However, descriptive studies are always retrospective and, thus, cannot determine causality and may produce misleading conclusions; as a result of bias or confounding. ▸ Analytic Studies An analytic study can be experimental if it is based on active allocation of study intervention by the researcher, or observational in case of passive involvement of the researcher. In most analytic studies, the research hypothesis involves a possible association between exposure to causal factor and some health outcome of interest. The various types of analytic research designs are distinguished by three aspects, which we call: “axes of analytic epidemiological study designs” ▪ Directionality It refers to temporality of measurement (which variable was measured first) rather than direction of analysis. 5 ▪ Type of outcome A cohort study would usually focus on incident outcomes, while cross-sectional and case control studies usually focus on prevalent outcomes. ▪ Sample selection Researchers can rarely or never study the entire population of interest; thus, they choose a sample of subjects for their study that is representative of the whole population, those samples are either selected by exposure -samples of exposed & unexposed subjects/exposure to two different treatments- (cohort studies), or outcome , usually when the outcome is rare -comparing prior exposure for sample of subject with outcome ‘cases’ & other lacking the outcome ‘controls’- (case-control studies). While cross-sectional studies begin by selecting a sample population and then obtaining data to classify individuals as having or not having the outcome. Directional ity Sample selection Outcome Axis of Analytical Study Design Observational studies Cohort Case-control Cross-sectional Exposure first (Its history Outcome first (presence of is not under researcher’s Both simultaneously disease) control) Exposure Outcome Random/all Incident (new cases) Prevalent Prevalent *Summary of axes of analytic epidemiological study designs A) Observational Analytic Studies Primarily hypothesis testing studies that use comparator groups to determine if exposure affects outcome. 1) Cross-sectional It is a correlational study of individuals in which outcome and exposure status are measured simultaneously in a given population, it answers general questions about association. Cross-sectional study can be thought of as providing a ‘snapshot’ of the frequency and characteristics of a disease in a population at a point in time. Cross sectional studies answer diagnostic questions ‘is PET a sensitive and specific test in diagnosing coronary artery disease compared to coronary angiography?’ 6 2) Case control It compares a group of individuals with a specific disease ‘cases’ with a group of individuals free of the disease ‘controls’, then, the history of each group to a particular exposure or characteristic of interest is collected and compared. Findings are collected in a 2x2 table and odds ratio (OR) is calculated to reflect the association between the hypothesised exposure and the disease being studied. 𝑂𝑅 = 𝑎/𝑐 𝑎𝑑 = 𝑏/𝑑 𝑏𝑐 OR >1 → exposed individuals have higher risk of developing the disease. OR 1 → exposed individuals have higher probability/risk of developing the outcome. RR