(14.6)--临床流行病学VariabilityandBias.ppt

《(14.6)--临床流行病学VariabilityandBias.ppt》由会员分享，可在线阅读，更多相关《(14.6)--临床流行病学VariabilityandBias.ppt（107页珍藏版）》请在得力文库 - 分享文档赚钱的网站上搜索。

1、1Variability and BiasProfessional English The study of the distribution and determinants of health-related states or events in specified populations,and the application of this study to control of health problems.”Epidemiology is the study of how often diseases occur in different groups of people an

2、d why.Epidemiologists examine patterns of illness in the population and then try to determine why certain groups or individuals develop a particular disease whereas others do not.Epidemiology is the study of human health in populations,rather than in individuals.Epidemiological information is used t

3、o plan and evaluate strategies to prevent illness and as a guide to the management of patients.Epidemiologic research involves observations of phenomena that occur naturally within human populations.Such an approach is unique among the medical sciences.Two features:(1)the focus on human populations

4、(2)reliance on nonexperimental observations.Advantages of Obs.StudiesNon-experimental studies offer some advantages over trials.For example:1.They cost less,and so can be quite large.2.They can address endpoints of interest,even if rare.3.Fewer ethical barriers.4.They can study treatments as actuall

5、y given in clinical practice.61854,Snow,in investigating the epidemic,began plotting the location of deaths related to Cholera(see illustration).he found that in one particular location near the intersection of Cambridge and Broad Street,up to 500 deaths from Cholera occurred within 10 days.After th

6、e panic-stricken officials followed Snows advice to remove the handle of the Broad Street Pump that the epidemic was contained.Through mapping the locations of deaths related to Cholera,Snow was able to pinpoint one of the major sources of causation of the disease and support his argument relating t

7、o the spread of Cholera.Snows classic study offers one of the most convincing arguments of the value of understanding and resolving a social problem through the use of spatial analysis.Epidemiology-Aims describe the distribution of health-related states and determinants in the population(descriptive

8、 study);discover which are the key determinants and to define the natural history of diseases(analytic study);assess the efficacy,effectiveness and efficiency of methods to prevent,cure and alleviate disease(intervention or experimental study);evaluate the process and outcome of services provided fo

9、r these purposes(health services research)9Patient Profile nA 45-year-old man began working as a production supervisor,and his employer required that he undergo a complete medical examination.His physician learned that the patients father had died of myocardial infarction at age 65.On physical exami

10、nation,the patient was moderately obese,and his blood pressure was 130/86.The remainder of the examination revealed no notable abnormalities.The patients total serum cholesterol level(non-fasting)was 242 mg/dL.Variability and Bias10nAccording to the guidelines of the National Cholesterol Education P

11、rogram(NCEP),a total serum cholesterol concentration greater than 240 mg/dL is an indication for possible pharmacologic lowering of serum cholesterol.A value of 200239 mg/dL is considered borderline and should trigger dietary intervention,and a value less than 200 mg/dL is considered normal.11nBased

12、 on the initial cholesterol results,the physician asked the patient to return in 2 weeks for further testing.On repeat measurement,the total serum cholesterol concentration was 198 mg/dL on a fasting lipid profile.nSeveral different factors could explain the observed variability in measured total se

13、rum cholesterol level.The source of this variability in the measured total cholesterol level had important implications for how the physician treated this patient.Variability in Medical Research nAppreciation of these issues is important for the interpretation and appropriate application of research

14、 findings in the clinical setting.nVariability in measurements can be either random or systematic.Increased random error(B versus A)and systematic error(D versus C).14nIt is important to consider the accuracy and precision of any measurements made in the medical setting.nIn clinical medicine and med

15、ical research,variability can occur at a number of different levels.16Variability within the IndividualnVariation can occur because of biological changes in an individual over time.nThese changes may(1)occur on a minute-to-minute basis(eg,heart rate),(2)follow a regular diurnal pattern(eg,body tempe

16、rature),(3)progress with normal development(eg,height or weight).nWhen the variation within a subject is large,a single measurement may not adequately represent the true status of that individual.nBy repeating a test,the physician may obtain a better understanding of the true value and its variabili

17、ty.nThis may also provide the clinician with information about variability or error due to the measurement technique.19nIn the Patient Profile,different results were obtained when the total serum cholesterol level was measured a second time.nIt is unlikely,however,that the fasting state alone could

18、cause such a drop in total serum cholesterol concentration.Furthermore,it is unlikely that the patient could have made the kind of dietary or other alterations in 2 weeks that would lead to the observed change in total serum cholesterol level.nVariability Related to MeasurementnTo determine which va

19、lue（198 mg/dL or 242 mg/dL）is closer to the truth,the physician in the Patient Profile would need to know whether both measurements were obtained in the same laboratory.nIn reality,the physician may not be able to discern readily which value is closer to the truth.This is one reason that programs wi

20、th guidelines that support cutoff points for clinical decision making,such as the NCEP,often recommend that elevated values be confirmed by repeated measurements over time before treatment is instituted.nVariations Within PopulationsnJust as there is variation in individuals,there is also variabilit

21、y in populations,which can be considered the cumulative variability of individuals.nBecause populations are made up of individuals with different genetic constitutions who are subject to different environmental influences,populations often exhibit more variation than individuals.22nPhysicians use kn

22、owledge about variability in populations to define what is normal and abnormal.nThe physician in the Patient Profile could refer to population survey data to learn that for 45-year-old males,a total serum cholesterol level of 200 mg/dL is close to the 50th percentile.Accordingly,the patient generall

23、y falls in the upper half of the population distribution of total serum cholesterol values.nAssuming that the measurement is correct,this could be a result of genetic factors,environmental factors,or both.nVariability in Research StudiesnIt is worthwhile to ask how the clinician would know that a to

24、tal serum cholesterol value in the upper end of the population distribution is disadvantageous.nAre these values really unhealthy?Answers may be found in studies that have linked the level of total serum cholesterol with an increased risk of cardiovascular mortality.nIn cohort studies such as the Fr

25、amingham Heart Study,a higher level of total serum cholesterol was associated with an increased risk of death from cardiovascular disease.nWhen investigators perform such studies,they cannot usually study the entire population.Instead they study subsets or samples of the population.nThis introduces

26、another source of variabilitysampling variability.nUsing a single sample of subjects to represent the population is analogous to using a single measurement to characterize an individual.26nVariability can be important when two groups are compared in a study.The goal of such studies often is to deter

27、mine whether a measurable difference exists between the groups.nWhen a research paper reports no statistically significant difference between groups,the reader must ask the following question:Was there actually no difference between the two treatments,or was the estimate of effect so imprecise that

28、the investigator could not distinguish differences between the two groups?28nGenerally,the larger the sample size,the more precise the estimate of effect.nIn studies with very large sample sizes,small differences between groups may be judged to be statistically significant but have little biological

29、 or clinical meaning.29Principles of Study Designnexpressed using confidence intervals.ndepends on sample sizePrecision(lack of random error)Validity(lack of systematic error)30Validity nThe concept of validity concerns the degree to which a measurement or study reaches a correct conclusion.A measur

30、ement or study may lead to an incorrect(invalid)conclusion because of the effects of bias.31the amount of bias can be determined by the degree to which the shots are off target in D.Unfortunately,in medical research the truth(bulls-eye)may not be known,or there may be no gold standard for comparison

31、.Consequently,the degree of bias often is difficult to determine.32The extent to which the results of an investigation accurately reflect the true situation of the study population.If the results are not valid in the study population,there is little reason to suspect that those results will apply to

32、 other populations.Internal Validity33A result obtained in a tightly controlled environment,however,may not be applicable to more general situations.External validity is the extent to which the results of a study are applicable to other populations.External validity addresses the following question:

33、Do these results apply to other patients,such as patients who are older or less economically advantaged than subjects in the study?External Validity34nExternal validity often is of particular interest to clinicians who must decide if a research finding is applicable to their clinical practice.35Bias

34、nBias is a systematic error in a study that leads to a distortion of the results.nBias,a threat to validity,can occur in any research,but is of particular concern in observational studies because the lack of randomization increases the chance that study groups will differ with respect to important c

35、haracteristics.36nBias often is subdivided into different categories,based on how bias enters the study.nThe most common classification divides bias into three categories:1.Selection bias 2.Information bias 3.Confounding.37nNegative bias(towards the null)nPositive bias(away from the null)nSwitch-ove

36、r bias(change of direction of association)Not easily be measuredCannot be reduced by increasing sample size.Sample sizeSystmatic errorRandom errorError Overestimation of a risk ratioUnderestimation of a risk ratio41A variety of procedures can be used to select subjects for a study.Usually,it is not

37、possible to include all individuals with a particular disease or exposure in a study,so a sample of subjects must be chosen.Is a distortion in the results of a study that arises because of the manner in which the subjects are sampled.The selection process may increase or decrease the chance that a r

38、elationship between the exposure and disease of interest will be detected.Selection bias42nThe sample selected does not represent the target population.nStatistical method does not help with this problem.nWe need to understand what let to the bias and what is needed to avoid it.44na casecontrol stud

39、y of total serum cholesterol level as a risk factor for developing myocardial infarction.Suppose that of patients who have a myocardial infarction those with very high total serum cholesterol levels are more likely to die suddenly than those with lower serum cholesterol levels.nUnder these circumsta

40、nces,a comparison of patients surviving myocardial infarction with controls will underestimate the true association between elevation in total serum cholesterol level and risk of developing myocardial infarction.47Information biasnA distortion in the estimate of effect due to measurement error or mi

41、sclassification of subjects on one or more variablesn(occurs during data collection)48nIf the errors in Misclassification of exposure or disease status are independent of the level of the other variable,is termed nondifferential.neg:Regardless of disease status,respondents may underreport intake of

42、foods with high fat content because they think low-fat diets are more acceptable to the investigator.nA bias toward the null hypothesis or toward no association.Effect with noEffect with nomisclassificationmisclassificationExposedExposedUnexposedUnexposedEffect with No Misclassification of Exposure5

43、0Effect with noEffect with nomisclassificationmisclassificationExposedExposedUnexposedUnexposedEffect withEffect withmisclassificationmisclassificationEffect with Non-differentialMisclassification of Exposure51nDifferential misclassification occurs when the misclassification of one variable depends

44、on the status of the other.nThis type of misclassification can result in either an underestimate or overestimate of the relationship.53nAs a way to reduce misclassification and to improve accuracy of study measurements,investigators increasingly are using biological markers.nUse of biological marker

45、s is important in observational studies.These markers are important methodologically because they can serve to reduce misclassification by allowing more accurate assessment of exposure or disease status.54nThus,use of biological markers offers many advantages,particularly an improved assessment of e

46、xposure and a more homogeneous definition of disease.nNevertheless,because use of these markers does not eliminate the possibility of information bias,caution in interpretation is still warranted.55ConfoundingRefers to the mixing of the effect of an extraneous variable with the effects of the exposu

47、re and disease of interest.Confusion of effectsNot due to any“fault”in design of study.OR=?OR=?OR=?OR=9/4OR=17/27OR=17/2759Characteristics of a Confoundernassociated with disease(in nonexposed)nassociated with exposure(in source population)nnot an intermediate causeHypothetical data on diabetes and

48、dementiaDiabetesDementiaYesNoTotalYes4006001000No1009001000Total50015002000OR=6.0Hypothetical data on age and dementiaAgeDementiaYesNoTotal80+380620100045-791108901000Total49015102000OR=4.96Confounding-concept in epidemiology Confounder AgeOutcomeDementiaHypothetical data on age and diabetesAgeDiabe

49、tesYesNoTotal80+900100100045-791009001000Total100010002000OR=81.0Confounding-concept in epidemiology Confounder AgeExposure underInvestigationDiabetesOutcomeDementiaHypothetical data on diabetes and dementiaDiabetesDementiaYesNoTotalYes2080100No90810900Total1108901000in those aged 45-79OR=2.25Hypoth

50、etical data on age and dementiaDiabetesDementiaYesNoTotalYes360540900No2080100Total3806201000in those aged 80+OR=2.67Confounding-concept in epidemiology Confounder AgeExposure underInvestigationDiabetesOutcomeDementiaConfounding what effect?OC use MI SmokingOC users smoke more heavily thannon-usersR