捕獲-再捕獲分析.ppt

1、Capture recapture analysis捕獲-再捕獲分析,Keith Sabin, PhD, MPH DHHS/CDC/GAP,What is it for?為什么?,Capture-recapture analysis is used for counting the total number of people in a population using two or more incomplete lists of those people 捕獲-再捕獲分析用于兩組或多組非完整名單來(lái)對(duì)某個(gè)人群進(jìn)行計(jì)數(shù) Why should I be interested?為什么我對(duì)此感興趣?

2、 Evaluating surveillance systems 評(píng)價(jià)監(jiān)測(cè)系統(tǒng) Magnitude of issues 問題的大小,Overview概要,Origin of method方法的來(lái)源 Application to epidemiology - why is it useful for us? 方法的應(yīng)用-為什么對(duì)我們有用? Principles原則 Conditions for using capture-recapture methods應(yīng)用捕獲-再捕獲的條件 Methods方法 Two sources兩個(gè)來(lái)源 Multiple sources多個(gè)來(lái)源 Limitations不

3、足之處,Origins of capture-recapture analysis捕獲-再捕獲分析方法的來(lái)源,Origins in demography來(lái)源于人口學(xué) 1662 - used to estimate the population of London 1662年用來(lái)估計(jì)倫敦市的人口 1783 - Laplace used to estimate population of France 1783年laplace用于估計(jì)法國(guó)的人口 1949 - Sekar and Deming used to estimate birth rate and mortality in India 19

4、49年Sekar和Deming用于印度的出生率和死亡率 Subsequently most often for estimating wildlife populations 隨后通常用于野生動(dòng)物數(shù)量 More recently applied to epidemiology (Wittes 1968) 近期被用于流行病學(xué),Application of capture-recapture analysis to human epidemiology捕獲-再捕獲分析在人類流行病學(xué)上的應(yīng)用,Evaluating completeness of a surveillance source 評(píng)價(jià)監(jiān)測(cè)資

5、料的完整性 Passive surveillance被動(dòng)監(jiān)測(cè) Registers登記 Refining incidence and prevalence estimates from surveillance systems or population surveys改進(jìn)監(jiān)測(cè)系統(tǒng)或人群調(diào)查得到的發(fā)病率和患病率估計(jì) Used for cancers, stroke, homelessness, mental illness, drug use, congenital disorders, infections運(yùn)用于癌癥、中風(fēng)、無(wú)家可歸、精神疾患、吸毒、生天性疾患和感染,Principles原則,

6、Two or more sources (lists) of cases a given disease 某一疾病的兩組或多組來(lái)源病例 Sources considered random capture samples in population 認(rèn)定來(lái)源病例為一總體的的隨機(jī)捕獲樣本 Cases can be matched by unique identifiers 病例之間可以用唯一識(shí)別匹配 Estimate total number of cases that are not captured by any source from the matched and unmatched估計(jì)未

7、被任何病例來(lái)源（匹配和未匹配）捕獲的病例總數(shù),Critical assumptions/conditions重要假設(shè)條件,1. Population is closed 人群是封閉的 methods exist for open populations有用于開放人群的方法 2. Individuals captured on both occasions can be matched兩次被捕獲的個(gè)體間可以匹配 3. Capture in the second sample is independent of capture in the first兩次捕獲之間是相互獨(dú)立的 4. Probabi

8、lity of capture is homogeneous across individuals每個(gè)個(gè)體被捕獲的可能性是一致的 Homogeneity of individuals個(gè)水平上 Homogeneity of lists名單水平上,Application to humans在人群上的應(yīng)用,“Capture” = appearing on a “l(fā)ist”“捕獲”出現(xiàn)在“名單”上 “re-capture” = linking by identifying individuals appearing on both lists by criteria name, date of bir

9、th etc “再捕獲”同時(shí)出現(xiàn)在兩個(gè)名單位并符合標(biāo)準(zhǔn)（姓名、生日等）的連接 “Trap fascination” ”陷阱誘惑“ if you feed the animal they are more likely to be caught again 如果你喂食，動(dòng)物被抓獲的可能就增大 laboratory confirmed cases are more likely to be reported in other systems實(shí)驗(yàn)室確證的病例在其他系統(tǒng)被報(bào)告可能就更大 “Trap avoidance” ”陷阱逃逸” if you scare the animal they will

10、avoid the trap 如果嚇唬動(dòng)物，動(dòng)物將不會(huì)被逮 a person cant appear on community injecting drug user registry if they are in prison 被關(guān)押在監(jiān)獄的人不可能出現(xiàn)在社區(qū)IDU登記冊(cè)上,僅在第一次 被捕獲,僅在第二次 被捕獲,未被捕獲,第一次和第二次 均被捕獲,未被捕獲,Two sources兩次捕獲樣本,Source A,Source B,x12,x11,x21,x22?,1 included in source 在捕獲樣本中 2 not included in source不在捕獲樣本中,Captu

11、re (Source A) and recapture (source B)捕獲（樣本A）和再捕獲（樣本B）,Estimation估計(jì),If sources independent: P(A+ if B+) = P(A+ if B-) 如果是獨(dú)立樣本：,Capture (Source A) 是針對(duì)難以接近人群的抽樣方法 Description of RDS; RDS的描述 Lessons learned from Vietnam; 越南實(shí)踐中得到的經(jīng)驗(yàn)教訓(xùn),Probability Sampling 概率抽樣 (Simple單純隨機(jī), Systematic系統(tǒng), Cluster整群),Gold

12、Standard-Best methods for sampling But, do not reach hidden populations: 金標(biāo)準(zhǔn)-最好的抽樣方法,但不能接近隱匿人群 No sampling frame沒有抽樣方案 Stigmatized被歧視 Would need huge sample sizes in order to capture a hidden population 需要很大的樣本量才能找到隱匿人群 Expensive費(fèi)錢,Sampling Methods to Reach Hidden Populations 針對(duì)隱匿人群的抽樣方法,Time-Locati

13、on (TLS), Venue-Based 在場(chǎng)所的時(shí)間-空間抽樣法 -Major Bias: Only captures those who are visible 主要偏倚: 僅能找到顯性人群 Snowball滾雪球 -Major Bias: Not representative of the population (tendency for in-group affiliation, volunteerism and masking) 主要偏倚：代表性不好（組內(nèi)從屬傾向，自愿性）,Background on RDSRDS背景,Developed by D. Heckathorn and

14、R. Broadhead with IDUs in Connecticut and in Yaroslavl, Russia; 由D. Heckathorn和R. Broadhead在 美國(guó)Connecticut州和俄羅斯Yaroslavl的IDU中研發(fā)而成 Sampling vs. Recruitment strategy;抽樣 vs 招募策略 Different from other chain referral methods because it can give us point estimations with standard errors.不同于其他鏈?zhǔn)酵婆e方法,因?yàn)榭梢越o出一

15、個(gè)點(diǎn)估計(jì)和標(biāo)準(zhǔn)誤,How RDS Works如何做RDS,Use of a dual system of recruitment through the use of incentives. 運(yùn)用雙重招募系統(tǒng),并給予報(bào)酬(小的刺激) Use of recruitment quotas. 運(yùn)用招募限額 Use of peers to recruit peers. 運(yùn)用同伴招募同伴 Use of links between recruiters and recruits. 運(yùn)用招募人與被招募人之間的聯(lián)系,The Theory Behind RDSRDS的理論要點(diǎn),Uses prinicples o

16、f First Order Markov Theory 運(yùn)用Markov鏈?zhǔn)嚼碚?Long referral chains 推舉長(zhǎng)鏈 Final sample will be independent of those selected as “seeds” 最后的樣本獨(dú)立于剛開始的”種子” Final sample will be similar to the population of the network from which you are recruiting 最后的樣本將與你所招募捐的人群相近,Wave 1 Wave 2 Wave 3 Wave 4 Wave 5,Wave 1 Wa

17、ve 2 Wave 3 Wave 4 Wave 5,Wave 1 Wave 2 Wave 3 Wave 4 Wave 5,Wave 1 Wave 2 Wave 3 Wave 4 Wave 5,Wave 1 Wave 2 Wave 3 Wave 4 Wave 5,Wave 1 Wave 2 Wave 3 Wave 4 Wave 5,A Long Referral Chain: Jazz Musicians in New York City,Selection of Seeds,Example in Hai Phong Vietnam越南海防的例子,Final Sample size: 420 I

18、DUs in Hai Phong and Saigon; 418 CSWs in Saigon and 220 in Hai Phong 最后的樣本量: 海防和西貢的420名IDU,西貢的418名CSW,和海防的220名CSW Recruitment process招募過程 20 seeds selected by peer educators 同伴教育員選擇20個(gè)種子 Three coupons to each participant 每個(gè)參加對(duì)象發(fā)三張卡片 Participants asked to recruit their peers 要求參加對(duì)象招募其同伴 Time: March J

19、une, 2004時(shí)間: 2004年3-6月 Three sites (Hai Phong); Four sites (Saigon) 海防的三個(gè)地點(diǎn);西貢的四個(gè)地點(diǎn),Eligibility Criteria入選標(biāo)準(zhǔn),CSWs: Women, 18 years or more, living or working in Hai Phong or Saigon;女性,18歲及以上,在海防或西貢生活或工作 Has sold sex for money in the last 30 days; 在過去的30天內(nèi)以性換錢 Has a green coupon (except seeds); 有一張綠卡

20、片(除種子外) Has provided consent. 知情同意 IDUs Women (Saigon only) or Men,18 years or more, living in Hai Phong or Saigon; 男性或女性,18歲及以上,住在海防或西貢,西貢僅做女性 Has injected drugs during the last 30 days; 在過去30天內(nèi)注射過毒品 Has a yellow coupon (except seeds);有一張黃卡片(除種子外) Has provided consent.知情同意,Coupon: Front Side卡片正面,LI

21、FE-GAP project: For Your Health and Safety Payment coupon Address_ Telephone:_ (You can call to make an appointment in advance) You will receive 15,000 VND for each person who you recruit and enrolls into the study (you may recruit up to 3 persons) ID number: Please call us in advance. You must pres

22、ent this coupon for payment,Coupon: Back Side卡片背面,Networks of CSWs in Hai Phong,A network in Hai Phong,Seed,Initial Lessons from Vietnam越南的初步經(jīng)驗(yàn),Seeds should have high degree-initial focus group may be important;種子應(yīng)該具有高學(xué)歷,最初的中心小組很重要 No slow down mechanism to end RDS; 沒有減速機(jī)制來(lái)停止RDS Need for security-In

23、terviewers have no choice of whom they interview; 安全保障-調(diào)查員無(wú)法選擇調(diào)查對(duì)象 Managing multiple sites can be difficult; 同時(shí)管理幾點(diǎn)地點(diǎn)有困難 Managing coupon numbers;卡片號(hào)碼管理 No way to control for those who recruit faster.沒法控制那些快速招募者,Initial Lessons from Vietnam (Cont),Difficult to discourage recruiters from selling coupo

24、ns or giving them out in a non random way; 沒法控制招募人出售卡片或以非隨機(jī)的方式發(fā)出卡片 Non response information difficult to obtain (incentives picked up by friends, recruiters do not return for secondary incentive) 無(wú)應(yīng)答信息很能獲取(報(bào)酬被朋友拿走,招募人不回來(lái)再取報(bào)酬),Philosophical objection?哲學(xué)上的異議?,Capture-recapture is fun, so it must be ep

25、idemiology! 捕獲-再捕獲很有意思,所以客觀上一定是流行病學(xué)! But, as epidemiologists we are interested in 但是作為流行病學(xué)家,我們對(duì)三間分布感興趣 Time, place and person Capture-recapture does not capture time - it is a static tool which relies on lists which correspond to prevalence of a chronic disease (e.g. diabetes) or long time periods f

26、or acute diseases (legionella)捕獲-再捕獲不”捕獲”時(shí)間, 這是一個(gè)靜態(tài)工具,領(lǐng)帶于與慢病性(如糖尿病)或長(zhǎng)時(shí)間急性病(軍團(tuán)病)的名單 Can be used for measuring broad trends by repeat analysis (Nardone et al Epidemiol Infect 2003)可以用重復(fù)分析來(lái)測(cè)量總體趨勢(shì),Practical limitations操作上的不足之處,Unique identifier has to match in all data sources 必須用唯一識(shí)別信息對(duì)所有數(shù)據(jù)來(lái)源進(jìn)行匹配 This

27、may contravene confidentiality laws這可能與保密法相抵觸 Clever statistics cant correct bad data 聰明的統(tǒng)計(jì)不能糾正不好的數(shù)據(jù) Rubbish in, rubbish out. 垃圾進(jìn),垃圾出 For chronic and expensive diseases (eg diabetes) it may be better to carry out an expensive detailed survey than to use quick and dirty methods對(duì)慢性和費(fèi)錢的疾病(如糖尿病),開展一項(xiàng)費(fèi)錢的

28、詳細(xì)調(diào)查比使用快速但很差的方法更好 it may be even more expensive to get it wrong. 如果出錯(cuò),費(fèi)用更高,Extrapolation is based on assumptions,we are assuming that the model which describes the observed data also describes the count of the unobserved individuals. We have no way of checking this assumption. This is analogous to,

29、and has the same dangers as fitting an arbitrary curve to a series of points (x,y), where x0, with the intention of estimating y at x=0. .this is analogous to the position of those who automatically assume that the k samples in our problem are independent. 我們假設(shè)描述觀察數(shù)據(jù)的模型也描述了未觀察個(gè)體的計(jì)數(shù),我們無(wú)法檢驗(yàn)這個(gè)假設(shè).這與隨意用一

30、曲線對(duì)一組數(shù)據(jù)(X,Y)進(jìn)行擬合一樣并有同樣的危險(xiǎn),當(dāng)X0時(shí),當(dāng)X=)估計(jì)Y.這與自動(dòng)假設(shè)我們的問題有K個(gè)獨(dú)立樣本的某些人的立場(chǎng)一樣. Fienberg, Biometrika 1972;59:591-603,Conclusion小結(jié),If conditions are met如果條件符合 Potential to use multiple incomplete registers and to estimate population size by capture-recapture有可能利用多組不完整的登記數(shù)據(jù),用捕獲-再捕獲的方法估計(jì)人群基數(shù) Cheaper than exhaustiv

31、e registers比徹底完全的登記少費(fèi)錢 兩個(gè)來(lái)源樣本 不可能對(duì)相關(guān)度進(jìn)行定量 Two sources兩個(gè)來(lái)源樣本 Impossible to quantify extent of dependence Requires third source Multiple sources Log-linear modelling method of choice Can adjust for dependence and variable catchability,Caveats警告,Use technique but be careful!使用技術(shù)但要小心 Dont treat this as

32、a black box method 不能當(dāng)成是一個(gè)暗箱方法 All prior knowledge should be used to formulate the model就用所有前人的知識(shí)來(lái)制定模型 Know your data!了解你的數(shù)據(jù) Not the solution to all problems Conditions often not met when applied to epidemiology There may still be heterogeneity you dont understand Complementary technique,References,

33、Wittes JT, Colton T and Sidel VW. Capture-recapture models for assessing the completeness of case ascertainment using multiple information sources. J Chronic diseases 1974;27:25-36. Hook EB, Regal RR. Capture-recapture methods in epidemiology. Methods and limitations. Epidemiologic Rev 1995; 17(2):

34、243-264 International Working Group for Disease Monitoring and Forecasting. Am J Epidemiol. Capture-recapture and multiple-record systems estimation I: History and theoretical development. 1995;142:1047-58 International Working Group for Disease Monitoring and Forecasting. Am J Epidemiol. Capture-re

35、capture and multiple-record systems estimation II: Applications in human diseases. 1995;142:1059-68 LaPorte RE, Dearwater SR, Yue-Fang C et al. Efficiency and accuracy of disease monitoring systems: Application of capture-recapture methods to injury monitoring. Am J Epidemiol 1995;142:1069-77,Recent

36、 examples of application to field epidemiology,Legionnaires disease. Infuso et al Eurosurveillance 1998;3:48-50; Nardone et al 2003;131:647-54 Malaria. Van Hest et al. Epidemiol Infect 2002; 129:371-7 Measles. Van den Hof et al Pediatr Inf Dis J 2002; 21:1146-50 Acute flaccid paralysis. Whitfield Bu

37、ll WHO 2002;80:846-851 Pertussis deaths. Crowcroft et al Arch Dis Child 2002;86:336-8 Intussception after rotavirus vaccination. Verstraeten et al Am J Epidemiol 2001;154:1006-1012 Tuberculosis. Tocque et al Commun Dis Public Health 2001;4:141-3 Salmonella outbreaks. Gallay et al Am J Epidemiol 2000

38、; 152:171-7 AIDS. Bernillon et al Int J Epidemiol 2000;29:168-174 Meningitis. Faustini et al. Eur J Epidemiol 2000;16:843-8,Special thanks to Nancy Crowcroft Health Protection Agency London Many of the capture-recapture analysis slides come directly from her class at Epi-Et.,THANK YOU!,RDS: Advantag

39、es,Ease of field operations Little for formative research/mapping Target members recruit for you Reach less visible segment of population Good external validity (found in other studies-still waiting to see in Vietnam) Minimal number of additional questions needed Computer software available Lower Co

40、st (Still waiting to see),RDS: Limitations,Population must be a network; Must be able to verify group membership; Must track links between recruiters and recruits-coupon management; Incentives; Very difficult to deal with selective non response bias.,Option 1: Use RDS with Institutional Data,Capture

41、-recapture requires two samples of the population, only one of which need be representative. If an institutional database is available, only a single number is required to “recapture” the population. Example: # of Registered NEP members,Example of Capture-Recapture,Capture: During the study period,

42、police recorded contacts with 86 injectors. The detective who provided this information said he was “confident that this is almost all the shooters in town.” Recapture: During the study period, 388 were interviewed using RDS. Overlap: 32 respondents were in both the police and the RDS samples. Estim

43、ated population size:,Estimating the Number of Jazz Musicians in NYC using the Logic of Capture/Recapture,Capture: Proportion of NYC musician union members who identified themselves as jazz musicians (in response to a union member survey) = 70% (415/592). Number of musician union members in the New

44、York metropolitan area, according to union records is 10,499. Therefore, the estimated number of union jazz musicians is 7,360 = (10,499 x .70). Recapture: Proportion of all NYC jazz musicians who are union members according to a RDS study is 22%. Using estimate of number of NYC union jazz musicians

45、 and estimated portion of all NYC jazz musicians who are union members, the size of the NYC jazz musician universe is: 7,360/.223 = 33,003,Multiple sources,Wittes Method,Evaluate dependence among sources Compare two-source estimates of N If estimates different Test of independence Calculate odds rat

46、ios between cell counts of two sources within a third source If OR 1 dependence Merge dependent sources Repeat calculation of estimates with merged source,a,b,c,d,e,f,g,A,B,C,Test of independence,Test of independence,a,b,c,d,e,f,g,A,B,C,OR = cg/de,OR = 1 independence OR 1 positive dependence underes

47、timation of N OR 1 negative dependence overestimation of N,a,b,c,d,e,f,g,A,B,C,Test of independence,To solve, have to assume highest order interaction=0 i.e. the chance of being in all the lists (in c) is a simple function of the chance of being on any single or list of lesser combination Or, there

48、is nothing special about “c”,Analyze relationship between categorical variables in a contingency table Logarithm of expected frequency of a cell expressed as linear function of effects for each cell and interaction term For 3 variables A with i levels, B with j levels, C with k levels, logarithm of

49、expected frequency of cell Fijk for cell ijk is,Log-linear modeling - General, main effect A first order effect AB second order effect (interaction),Log-linear modeling - CRM,Estimates value of a missing cell in a 2k contingency table k = number of sources Missing cell = number of cases not listed by any source (m222),Log-linear modeling,No interaction: sources are independent (1 model) Interaction between 2 sources only (3 models) Interactions between pairs of sources (3 models) Interactions between all sources 2 by 2 (1 model),How to chose the best model,Aim Best f