Wk 3 - Design Strategies Used in Epidemiology PDF

NURS 1113: Epidemiology Feb. 2024 M. Emanuel-Frith At the end of the session students should be able to:  Define term, research design  Identify phases of epidemiological investigation  Differentiate between observational and experimental studies; descriptive and analytic epidemiology  Explain t...

NURS 1113: Epidemiology Feb. 2024 M. Emanuel-Frith At the end of the session students should be able to:  Define term, research design  Identify phases of epidemiological investigation  Differentiate between observational and experimental studies; descriptive and analytic epidemiology  Explain the various study design strategies used in epidemiology 2  “Conceptual blueprint within which research is  “Framework that guides data collection and analysis  “Systematic plan to study a scientific problem” conducted”  Divided into two main branches – Observational Study Design and Experimental (Trials) Study Design 3  Enable researcher to logically address research question with minimal ambiguity  Play major role in determining scientific value of research study  Aid clinicians in practicing evidence-based medicine; guide health promotion activities and help health administrators make informed decisions re allocation of resources 4  Study design should be well thought of before starting investigation  Choosing inappropriate design can undermine validity of study ] 5 Descriptive Analytic Experimental Describes population atrisk Tests hypothesis derived from descriptive study Uses controlled trials to determine effects of exposure Collects and analyze data Who? Where? When? Addresses cause and effect (Why? How?) Sources of data include: census, vital statistics, clinical records Findings lead to formulation of hypothesis Design Strategies_Epi [VHWB-2023] Identifies and quantifies factors associated with disease or health event Researcher manipulates independent variable(s) and observes response [Ask students to read up on “blinding”] 6 Design Strategies_Epi [VHWB-2023] 7 Epidemiologic Study Design Algorithm 8  Observational (Non-experimental) study is sub-divided into two main categories: descriptive and analytic  Exposure is not manipulated by the investigator  Observation may be prospective, retrospective or current depending on the sub-type of the study  Experimental (Interventional) study is sub-divided into two main categories: randomized and non-randomized  Exposure is manipulated by the investigator 9 ] 10  Provides description of aspects of disease or healthrelated events in a population by person, place and time 11  Concerned with organizing and describing data according to: Person (Who)– identification of frequency of disease and those at greatest risk Place (Where) – geographical extent of disease, reservoir of agent and transmission of disease Time (When) – reveals extent of problem in terms of when and whether disease is predictable ] 12  Case study / case report  Case series  Cross-sectional  Ecological 13  Detailed report of specific features of a particular case (usually a single individual)  Most simple and basic type  In conjunction with case series, early detection of emergence of new disease or epidemic may be identified 14  Collection of cases studied  Include systemic review of the interesting and common features of small group of patients with similar diagnosis  In conjunction with case study, early detection of emergence of new disease or epidemic may be identified (provides cues pertaining to cause / course) 15  Conducted over short period – no follow-up; gives snapshot of the characteristics of participants in a single point of time  Persons studied based on being part of a group e.g. diabetics  Used to assess prevalence of a disease in a population 16  Advantages: useful for examining associations among health-related events; can be used to study several associations at once;  Disadvantages: not useful for rare conditions; cannot generalize to future time points; influenced by response bias; cannot be used to establish cause and effect relationship Design Strategies_Epi [VHWB-2023] 17  Used when data at an individual level is unavailable or, when large-scale comparisons are needed to study population level effect of exposures on a disease condition  Type data of measures are aggregates of individual level  Generally used in public health research; findings are applicable only at the population level 18  Subject to a type of confounding called Ecological fallacy - occur when researcher mistakenly assumes relationships identified at group level data to be true for individuals  Useful for comparing health in populations across countries or at different times e.g. measles rates; initial investigation of causal hypothesis 19  Provides information about new disease  Provides clues to identifying new cases  Helps to determine extent of public health problem  Identifies population at greatest risk  Obtains description of problem that can be readily communicated  Helps in planning and resource allocation  Identifies avenues for research into cause and effect 20 https://www.youtube.com/watch?v=Jd3gFT0-C4s 21  Why and how diseases or other health events occur  Deals with cause and effect  Analytical studies: Case control Cohort 22 Focus on outcome to exposure (identified by disease and observed for risk factors / exposure). Two groups of participants:  Cases (persons with particular disease or health event)  Controls (persons from general population without disease or health event) who are appropriate match for cases Matching reduces confounding (confounders equally distributed) 23  Used to determine degree of association between various risk factors and outcomes  Factors that affect risk of a disease are called exposures 24 Outcomes Study begins Exposed Unexposed Disease (cases) Compare rates of exposure Exposed Unexposed No disease (controls) 25 S T A R T I N G P O I N T  Retrospective study – looks in past for possible exposures persons might have had as risk factor  Subject to information bias (recall bias and observer bias) when collecting exposure data  Inexpensive, efficient and less time consuming  Useful to identify exposures that are harmful or beneficial e.g. commonly used in food borne outbreak investigations 26  Suitable for rare diseases and diseases that have long latency periods  Main outcome measure is Odds Ratio (OR)  OR = odds of being exposed (case) ÷ odds of being exposed (control)  OR = odds of disease exposed ÷ odds of disease unexposed 27 2 x 2 (contingency) tables are use in calculations of Odds Ratio  Contingency tables are used to analyse relationships between exposure and outcome  Outcome is listed in the columns  Exposure is listed in the rows  Cell data are counts (frequency) 28 Cases Control s Exposed a b Not exposed c d A study was carried out to determine association between coffee drinking and pancreatic cancer. There were 750 cases of which 450 reported drinking coffee regularly. 450 controls were selected of which 200 reported consuming coffee regularly. What is the: - exposure variable - outcome variable - association between coffee drinking and Design Strategies_Epi [VHWB-2023] OR = [a/b] / [c/d] OR = ad / bc Cases Control s Coffee No coffee 450 200 300 250 450 Exposure variable 750 – coffee drinkers Outcome variable – pancreatic cancer Association: [450 x 250] / [200 x 300] = 1.875 Coffee drinkers are 1.9 times more likely 29 to develop pancreatic cancer than those who do not drink coffee  Classify patients into two groups (cohorts) based on exposure status: exposed group unexposed group Patients identified by risk factor and observed for disease (Do NOT have the disease at start of study)  Also known as Longitudinal or Follow-up study 30 Study begins Time Outcome s 31  Cohorts followed over time to see who develops disease in both groups  The numbers of newly occurring (incidence) cases of disease are recorded and compared between groups  Study may be prospective (monitor groups over time) or retrospective / historical (looks back in time at groups) 32  Prospective Cohort Study (example)  1967 study initiated to assess association between asbestos exposure and lung cancer deaths in USA & Canada.  Exposed group – 17,800 asbestos workers  Unexposed group – general population of males of same age group  Analysis of lung cancer deaths among the two groups from 1967 to 1975 was done 33  Retrospective Cohort Study (example) 1965 study conducted to assess association between asbestos exposure and lung cancer deaths in USA Exposed group – asbestos workers during 1948-1951 Unexposed 1951 group – cotton textile workers during 1948- Analysis of lung cancer deaths among the two groups from 1948-1963 was done 34 Comparison between Prospective and Retrospective Cohort Studies Prospective Retrospective More expensive Less expensive Time consuming Less time consuming Not efficient for diseases with long latency period Allows for good exposure data Efficient for diseases with long latency period Exposure data may be inadequate Better data on confounding May have missing data on confounding factors 35  Unlike case-control study, incidence can be readily calculated  Main outcome measure is relative risk (RR) i.e., how much does exposure increase risk of the disease  RR = risk of disease (exposed) ÷ risk of disease (unexposed)  Same as, RR = cumulative incidence (exposed) ÷ cumulative incidence (unexposed) 36 Disease (Outcome) Status Yes No a b TOTAL RR > 1 indicates exposure associated with an increased risk of the disease RR < 1 indicates exposure is protective (lower risk) RR = 1 indicates risk is the same (no association) Exposure Yes a+b status No c d c+d TOTAL a+c b+d a+b+c+ A one year study: Cohort of oral contraceptive use and d bacteriuria among women aged 16-49 years old. Calculate RR Disease (Outcome) Status TOTAL Yes No Contraceptive Yes 27 455 482 use No 77 1831 1908 Design Strategies_Epi [VHWB-2023] RR = [a/a+b] / [c/c+d] RR = [27/482] / [77/1908] = 0.056 / 0.04 = 1.4 Risk for bacteriuria is 1.4 times greater in persons on OCP than those not on OCP 37 Advantages  Time sequence can be determined (causality)  Multiple outcomes can be studied simultaneously  Useful in studying childhood diseases – can commence from infancy Disadvantages  Prone to selection bias; very expensive and time consuming for studying rare diseases and outcomes that has long follow-up periods; may have high drop-out rate 38 Case-Control The different groups are identified according to their health outcomes (whether or not they have the disease) Cohort Participants groups are classified according to their exposure status (whether or not they have the risk factor) Measurement – Odds Ratio (OR) Measurement – Relative Risk (RR) 39 40  Used mostly for assessing a new treatment. Patients used as unit of study.  Two types of clinical trials: treatment trial (e.g. drugs, lifestyle modification) and prevention trial (e.g. exercise programme, vaccine) 41  Considered  Patients as the gold-standard of study designs identified by disease; (like case control)  Participants are randomly assigned to a control group and an experimental (interventional) group  Randomization in RCT avoids confounding and minimizes selection bias 42    Experimental group gets the exposure/treatment which can be an agent involved in causation, prevention or treatment of a disease Control group receives no treatment, a placebo or another standard of care treatment depending on the objective of the study Groups are then followed prospectively to see who develops the outcome of interest Eligible Population T Randomisation I Interventio n (exposed) Diseas e Non-intervention, placebo or other intervention e.g. current standard of care No disease Diseas e 43 No diseas e M E  Disadvantages:  Most expensive and time-consuming study design  Researchers often face issues with the integrity of randomization due to refusals, drops outs, crossovers, and non-compliance  Cannot be used if intervention borders on unethical ground (carries greater ethical responsibilities) Design Strategies_Epi [VHWB-2023] 44 45  Do not use random assignment  Choice of who receives the intervention is decided in some other way such as according to where or when they were recruited, or by the sequence in which they enter the study  Has the potential to introduce selection bias as group allocation is selected by the researcher 46 Field Trials:  Used mostly for assessing preventive agents. Healthy individuals used as units of study 47 Community Trials:  Communities used as units of study  Intervention applied in one community (not individuals) and not in the other  Effect  E.g. of intervention compared between communities effect of water fluoridation on dental caries 48  Bias – “Any systematic error in the design, conduct or analysis of a study that results in a mistaken estimate of an exposure’s effect on the risk of disease” – [Ask class to read up on types of biases]  Confounding – unmeasured third variable(s) related to the exposure of interest and outcome influence result of study [  Matching – participants are matched on some variable that may affect the dependent (outcome) variable 49 Design Strategies_Epi [VHWB-2023] Gordis, L. (2009). Epidemiology. Baltimore: Elservier MPH Notes (2010). Department of Community Health and Psychiatry, UWI, Mona Munnangi, S. & Boktor, S. (2019). Epidemiology of study design. Retrieved February 18, 2020 from: https://www.ncbi.nlm.nih.gov/books/NBK470342/ WHO (n.d.). Epidemiological studies. Retrieved from: https://www.who.int/ipcs/publications/ehc/216_disinfectants_part_4.pdf?ua=1 50