‘wiarre020 "Methods of sampling ‘rom a population | Heath Knowledge HealthKnowledge ge = sDvCATON PD Ato REVALDATION FROM HAST PHAST® Friday Webinars $Proesing An introductory set of 1 hour taster workshops will be launched in November. tas pideeey Methods of sampling from a population {information about a population based on results fram a subset ofthe population, without having to investigate every individual. Reducing the number of individuals {na smdy reduces the cost and workoad, and may make it casi to obtain high quality information, but his has to Be Balanesd against having a large enough sample ‘ze with enough power to detect ste attcition, (Caleultion of sample size is addressed in section 1B (sais) ofthe Pac A sylabus) {fa sample it to be used, by whatever method i is chosen i important specifically geting hard to reach groups not be registred and 2 the individual selected are representative ofthe whole population, This may involve or example, ithe electra oll fora town was used 1oidentily participa, some people, such asthe homeless, would herforeexchded from the study by defal, Ther are several diferent sampling techniques avilable, and they canbe subdivided int two groups: probability semping and non-probebiity sampling In probability random) sampling you stat wit «complete sampling fam of all eligible individuals from which you select your sample In tis way, all eligible individuals havea chance of being chosen for the sample, and you wil be mose abe to generalise the fesuls fiom you study Probability sampling methods tend to be more time-consuming and expensive than non-probaility’ sampling. In non-rababiby (no-random) sampling, you do no stat with a complete sampling frame, some individuals have no chance of boing selected. Consequently, you cannot estimate the effect of stmpling eror and there ise significant risk of ending up with 1 non-zepresenatve sample which poduces non-teneralisable sus, However, nor-pobebility sampling methods tend toe cheaper and mote convenient, and they re wef for exploratory research and hypothesis generation, Probability Sampling Methods 1. Simple random sampling In this cas each individaal is chosen eniely by chance and each member ofthe population has an equal chance, or probability, of being selected, One way of btsning a random sample so give each individual in «population a number, nd then wea Inble of random numbers to deci which individuals to incase "For example ifyou havea sampling frame of 1000 individuals, labeled 10999, use groups of thos digit from the random number table o pick your sample, So, ifthe first three numbers fom the randam number table were 09, selee the individual labelled “OM”, and oo, hitps Aww ealtnknowledge.or9.ukipublic-hoalthtextbookiteseareh-methodsta-epidamologylmethods-o-samping-population 18‘warre020 “Methods of sampling rom a population | Heath Knowledge ‘As withall probability sampling methods, simple random sampling allows the sampling err tobe calulted and duces selection bias. A specific advantage is that itis the most staightforward method of probability sampling. A disadvantage of simple random sampling is tat you may not select enough individuals with your characteristic of interest, especialy i that characterise i uncommon, It may alo be difical to define a complete sampling fame and inconvenient o contact them, especially ferent forms of conte are required (oil, phon, post) and you sample units ae scattered aver a wide geographical rea, 2 Systematic sampling Individuals are solsted at regular intervals fom the sampling fram. The interval re chosen to ensure an adequate sample size. Ifyou needa sample size fom a population of size x, you should select every én individual forthe sample, For example, ifyou waned a sample sizeof 100 from a population of 1000, select every 1000/100-~ 10" member ofthe sampling frame ‘Systematic sampling is often more convenient than simple random sampling, and itis easy to administer. However, it may also lad to bias, for example if there are underlying ptm in the order of the individuals in the sampling fame, such dh the sampling tshnigue coincides with the periodicity ofthe underlying parm. As aypothesical example, if group of students were being sampled o gain ther opinions on college facilites, bu the Student Record Departments central isto al students was arranged such thatthe sex of students alternated between male and female, choosing an even interval (every 20" student) would esl ina sample ‘ofall males o al fersales. Whilt inthis example the bas i obvious and shouldbe ely comected this may not always be the cue, {3 Stratfed sampling In this method, the population s fist divided into subgroups (or stat) who al share a similar characteristic Is used when we might reasonably expect the measurement of interest to vary betwoen the diffrent subgroups, and we want to cnsute representation fom al the subgroups. For example, in a study of stroke culcones, we tity stay the populain by sen, lo ensure equal representation of men and worse, The study amples then obtained by taking equl sempe sizes fom cach stratum. In staid sampling, it may also be appropriate to choose non-equal sample sizes from each stratum. For example, in a study ofthe heath ‘uteomes of musing stain a county ifthere ae thee hospitals each with diferent numb of nursing staf hospital A has OO nurses, hospital B has 1000 sed hospital € bas 2000), thet would be appropriate to choose the sample mabers Lom etch hosptl proportionally (eg, LO fom hospital, 20 froma hospital Band 40 from hospital C), This ensures a more realistic and accurate estimation of the health outcomes of muses aroes the county, whecses simple random sampling would ‘ovenaepresent nurse fom hospitals Aan B, The fet thatthe sample was sated should be taken into acount atthe analysis stage ‘Staified sampling improves the accuracy and represntativenss ofthe resls by reducing sapling bas, However, it requires knowledge of the appropriate characteristics of the sampling fame (the deals of which are no always available) and it canbe dificult decide which characters) to stray by 4. Clustered sampling Ta custered sample, subgroups ofthe population ae used as the sampling unit, rater than individuals. Th population i divided int subgroups, ows a chistes, ‘which are randomly selected tobe included in the sty. Clusters are usally already defined, for example individual GP practice or towns could be ented as luster. single-stage cluster sampling, ll members of th chose cusers are thea included inthe study. In two-stage cluster sampling, selection of individuals fim each clases then randomly selected snnually in Englands» good example of (one-stage) custer sample, All members ofthe selected houssbolde (chster) are inched in the survey." inclusion, Castering shouldbe takes ito account nthe analysis, The General Household survey, which is undertaken ‘Cluster sampling can be more efficent that simple random sampling, especially where a study takes place vera wide geographical region. For instance, tis easier nviduals in afew GP practices than few individuals in many diferent GP practices. Disadvantage include an increased isk o clusters are not representative ofthe population, resting in an ineresed sampling er. concat oso as, ithe chosen Non-Probabilty Sampling Methods 1. Convenience sampling Convenience sampling is perhaps the easiest method of sampling, because participants are selected based on availability and willingness to take prt. Useful sults can be oblained, bu the resus are prone to significant hits, because those who vlunicr to ake pst may he different from those who choose not to (volutes its), and the sample may not be representative of ther characterises, such as age or sex. Noe: volutes bis is ask of all non-probablity sampling methods 2. Quota sampling ‘This method of sampling i fen wed by marke researchers Interviewers are given «quoi of subjects of specified type wo atlemp a reruit For example, an inteviewer might be told ogo ou and selet 20 adult men, 20 adult women, 10 teenage girls and 10 teenage boys so that they cold interview them about het. television viewing Mell the quotas chosen would proportionally represen thecharcterstics ofthe underlying population ‘While his has the advantage of being relatively staighforwatd and potentially representative, the chosen sample may not he representative of other characterises ‘hat weren't considered (a consequence ofthe non-random nate of sampling)” 4. Judgement (or Purposive) Sampling hitps Aww ealtnknowledge.or9.ukipublic-healthtextbookiteseareh-methods/ta-epidamologylmethods-o-sampling-population 218‘warre020 “Methods of sampling rom a population | Heath Knowledge ‘Also known as selective, or subjective, sampling, this tecague relies om the judgement ofthe researcher when choosing whe to ask o participate, Researchers may implicitly thus choose a “representative” sample to suit their neds, or specifically approach individuals with certain characterises. This approach soften used by the media wen canvasing the publi for options andin qualitative esarch Sudgement sampling has the advantage of bing time-and cost-effective to perform whilst erltng in a range of respouses (particularly usefl in qualitative researc, However, in adition to volunteer bia, i alo prone to eran f judgement by the researcher and the Gndings, whilst being potentially broad, wil not nevessrly be representative ‘This method i commonly use in socialsciences when investigating hard-to-reach groups Existing subjects are asked to nominate further subjects known to them, so the sample inreassin size like a rolling snowball. For example, when carying out a survey of risk behaviours amongst intravenous drug users, parejpans may be askod to nominate other ust tobe interviewed, ‘Sowall sampling canbe effective when a sampling fame is diicult to identify. However, by selecting ends and acquainanees of subjects already investigated, ‘here signeant risk of selection bias (choosing large numberof people with sila characteristics o views othe naive iene, ‘Basin sampling ‘Ther re five important poteatial sources of bias hat shouldbe considered when selecting «sample, respective ofthe method used, Sampling bias may be introduced when 1, Any pre-tgreed sampling ules are deviated from 2. People in and--reach groups are omited 3. Selected individuals ae replaced with ters, for example if they are dificult to contact 4, There are low response rates 5. An ou-ofate lists used as the sample fame (or example, iit excludes people who have recently moved to an area) urlter potential problems with stnping strategies ae covered in chaper 8 of his section (“Sources of variation, meaurement and control) References 1. Ben-Shlomo ¥, Brookes S, Hickman M. 2013. Lecture Notes: Epidemiology, Evidence-based Medicin and Public Health (6th ed), Wiey-Blackwll, Oxford 2: hupwwv stats gl. ukistepsslossary/sampling ml - Accessed 804/17 “Nested case-conto studies ” ‘Methods of allocation in intervention studies» Navigation 1 Ue oftotne vil an ea statisti ods the dssibution of disease in time and place and by pron + Numero, denominator and populations ik 2 Time atk + Methods for Summatsng Data 2 nseac and prevalence ising st nd indirect sandaistion 9 Years of Life Lost +» Measutes of disease burden (evenchased and time-based) and population atribuable risks including dentifiation of comparison groups appropriate o Public Heats + Sources of vraton, its measurement and contol + Common crorsin epidemiological measurements, thee effects on numerator and denominator data and thee avoidance hitps uv ealtnknowledge.or9.ukipublic-halth.textbookiteseareh-methods/ta-epidamologylmethods-o-samping-population 38‘warre020 Methods of sampling rom a population | Heath Knowledge 19 Concepts and measures of isk + Effect measures including ods ratios, rt ais and risk ati (eatve ris) 1 Associaton and Causation + Biases and Confounding + Interactions, methods for assessment of effet modification + States to allowiajust for confounding in design and analysis +> The design applications, strengths and weaknesses of descriptive studies and ecological stadiee + Design applications, stents and weaknesses of eros-sctonal, analytical studies (including eshoy, case-control and nested case-control studies), and intervention studies (including randomised controled rials) > Analysis orhouth and disease in smal acas + Vly, reliability and generalisailty * Intention o treat analysis 1 Clustered dats - effets on sample size and approaches to analysis +> Numbers needed to teat (NNT) - calculation, interpretation, advantages and disadvantages ne series designs +9 Nested case-control studies + Methods of sampling fom a population +9 Methods of allocation in intervention studies 9 Timestend analysis, + The design of documentation for recording survey data, construction of valid questionnaires and methods for validating observational eciques +9 Construction of valid questionnaires + Methods for validating observational techniques + Stuies of disease prognosis +9 The ethics and etiquette of epidemiological serch + Appropriate use of statistical methods in the analysis and interpretation of epidemiological studies, including life able analysis + Epidemic theory (effective & base eproduction numbers, epidemic thresholds) & techniques for analysis of infectious disease data (construction & use of ‘epidemic curves, generation mumbers, exceptional reporting & identification of significant clusters) + Systematic reviews, motbods fr combining data from several studies, and meta-analysis + lestonicbbtographical databases and ther imi 9 Grey literature + Publication bias +» Evidence bated medicine and policy 19 The hcrrchy of research evidence - from well conducted meta-analysis downto sl ease series + The Cochrane collaboration + Understanding of asic isues and terminology in he design, conduct, analysis and ine twin studies, linkage and association studies ion of population-based gentic association suds, including > Appendix Our most popular content + Public Heath Textbook + ensiyng and managing internal an exter stakcholer interes + Management modes and theories associated with motivation, leaderakip and change management, an thee aplicain o practical situations and problems + istry Reference Values (ORV), current itary gol, commendations, guidlines andthe evidence for them + Section 1 The theoreti perspectives and methods of enauiry ofthe seenzsconcemed wih human behaviour + Tacquaits in health (eg. by region, ens, soe-economie poston or gender) and in access to health care, cluding thet causes + The impact of political, economi, socio, envionment and ober external influences + todution to study designs intervention studies and randomised controled ils + 2h-Prineples and Practice of Health romesion + Paamerc and Non-paraescels for comparing two oF mor OURS Recently updated content hitps Aww ealtnknowledge.or9.ukipublic-healthtextbookiteseareh-methods/ta-epidamologylmethods-o-sampling-population 45‘warre020 Methods of sampling rom a population | Heath Knowledge + 36 Applications +33 Populations + 3h. Sihness and Health + Sl - Understanding the Theory and Process of Strategy Development ‘+ 2a Epidemiological Paradigms + la- Epidemiology + 2d -Genaies 4+ 26 Diagnosis and Screening + ld The Principles of Qualitative Methods + Ie Approehes to the assessment of helt care neo utilisstion and outcomes, nd the evaluation of health and health ere PHAST® Penge hitps Aww ealtnknowledge.org.ukipublic-halthtextbookiteseareh-methodsta-epidamologylmethods-o-sampling-population 55