1. Public Health Information Network (PHIN) Series II Outbreak Investigation Methods: From Mystery to Mastery

3. Series II Session III “Study Design”

4. Series II Sessions TitleDate “Recognizing an Outbreak”June 2 “Risk Communication”July 7 “Study Design”August 4 “Designing Questionnaires”September 1 “Interviewing Techniques”October 6 “Data Analysis”November 3 “Writing and Reviewing Epidemiological Literature” December 1

5. Access Series Files Online http://www.vdh.virginia.gov/EPR/Training.asp http://www.vdh.virginia.gov/EPR/Training.asp Session slides Session activities (when applicable) Session evaluation forms Speaker biographies Alternate Web site: http://www.sph.unc.edu/nccphp/phtin/index.htm http://www.sph.unc.edu/nccphp/phtin/index.htm

6. Site Sign-in Sheet Please submit your site sign-in sheet and session evaluation forms to: Suzi Silverstein Director, Education and Training Emergency Preparedness & Response Programs FAX: (804) 225 - 3888

7. Session Overview Session I summary and review: How we arrived at the research hypothesis Analytic Epidemiology Developing and testing hypotheses Study Designs: –selection, implementation, and data analysis

8. Today’s Presenters Amy Nelson, PhD, MPH Epidemiologist NC Center for Public Health Preparedness Sarah Pfau, MPH Moderator

9. “Study Design” Learning Objectives Understand what an analytic study contributes to an epidemiological outbreak investigation Understand the differences in methodology between cohort, case-control, clinical trial, and cross-sectional study designs Be able to describe the advantages and disadvantages of alternative study designs

10. “Study Design” Learning Objectives (cont’d.) Know how to assess which study design to apply during an outbreak investigation Understand how to select cases and controls in a case-control study design Know how to interpret odds ratios from case-control study data and risk ratios from cohort study data

11. Session I Review

12. Detected an outbreak of gastroenteritis at a university in Texas, based on a March 11 complaint from a student Conducted: –active case finding –chart reviews –hypothesis generating interviews 75 cases detected by March 12

13. Session I Review (cont’d.) Collected stool specimens –All negative for bacterial pathogens –Assume viral pathogen Earliest date of exposure was March 5 Conducted interviews with 7 of the earliest cases (all students) –Only commonality was 6 of 7 reported eating at deli bar of main campus cafeteria

14. Session I Review (cont’d.) Developed a leading hypothesis with the main campus cafeteria as the suspect –No source food identified Interviewed 30 of 31 cafeteria staff members –Staff member who declined interview worked at deli bar* * Today’s session includes a follow up investigation with this staff member

15. Session I Review (cont’d.) Investigators closed deli bar on March 12 th based on: –Link between 6 out of 7 interviewed cases and a common food source –Several unsanitary food handling practices identified via an environmental health investigation

16. Basic Steps of An Outbreak Investigation 1.Verify the diagnosis and confirm the outbreak 2.Define a case and conduct case finding 3.Tabulate and orient data: time, place, person 4.Take immediate control measures 5.Formulate and test hypothesis 6.Plan and execute additional studies 7.Implement and evaluate control measures 8.Communicate findings

17. Analytic Epidemiology Hypotheses

18. Descriptive vs. Analytic Epidemiology Descriptive Epidemiology deals with the questions: Who, What, When, and Where Analytic Epidemiology deals with the remaining questions: Why and How

19. Analytic Epidemiology Used to help identify the cause of disease Typically involves designing a study to test hypotheses developed using descriptive epidemiology

20. Exposure and Outcome A study considers two main factors: exposure and outcome Exposure refers to factors that might influence one’s risk of disease Outcome refers to case definitions

21. Case Definition A set of standard diagnostic criteria that must be fulfilled in order to identify a person as a case of a particular disease Ensures that all persons who are counted as cases actually have the same disease Typically includes clinical criteria (lab results, symptoms, signs) and sometimes restrictions on time, place, and person

22. Developing Hypotheses A hypothesis is an educated guess about an association that is testable in a scientific investigation Descriptive data provide information to develop hypotheses Hypotheses tend to be broad initially and are then refined to have a narrower focus

23. Example Hypothesis: People who ate at the church picnic were more likely to become ill –Exposure is eating at the church picnic –Outcome is illness – this would need to be defined, for example, ill persons are those who have diarrhea and fever Hypothesis: People who ate the egg salad at the church picnic were more likely to have laboratory- confirmed Salmonella –Exposure is eating egg salad at the church picnic –Outcome is laboratory confirmation of Salmonella

24. TX Case Study Hypothesis and Study Design

25. What do we know? We know… –75 cases reported by March 12 – all were students who lived on campus –All of cases sampled (n=7) reported eating at the main campus cafeteria 6 of 7 at the deli bar –No illness among off-campus students, faculty, or community members

26. What do we suspect? We suspect the main cafeteria is the source of infection We also suspect the deli bar, in particular, may be the outbreak source

27. What can we “prove”? Can we “prove” it? –Need to conduct an analytic study –Need a testable hypothesis

28. What hypothesis will be tested? 1.Main cafeteria vs. other dining halls? 2.Within the main cafeteria: –Deli bar vs. grill? –Salad bar vs. hot entrée? 3.Food item vs. food handler at deli bar?

29. Our Hypothesis A viral infection was spread by food(s) or beverage(s) served at the university’s main cafeteria between March 5 and 10.

30. Source Population Students (12,000) 2400 on campus 2000 meal plan Town residents (39,000)

31. Which Study Design?

32. Types of Analytic Studies Two main categories: 1.Experimental 2.Observational 1.Experimental – exposure status is assigned 2.Observational – exposure status is not assigned

33. Experimental Studies Can involve individuals or communities Assignment of exposure status can be random or non-random The non-exposed group can be untreated (placebo) or given a standard treatment Most common is a randomized clinical trial

34. Experimental Study Examples Randomized clinical trial to determine if giving magnesium sulfate to pregnant women in preterm labor decreases the risk of their babies developing cerebral palsy Randomized community trial to determine if fluoridation of the public water supply decreases dental cavities

35. Observational Studies Three main types: 1.Cross-sectional study 2.Cohort study 3.Case-control study

36. Cross-Sectional Studies Exposure and outcome status are determined at the same time Examples include: –Behavioral Risk Factor Surveillance System (BRFSS) - http://www.cdc.gov/brfss/http://www.cdc.gov/brfss/ –National Health and Nutrition Surveys (NHANES) - http://www.cdc.gov/nchs/nhanes.htm http://www.cdc.gov/nchs/nhanes.htm Also include most opinion and political polls

37. Cohort Study

38. Definition of a Cohort In Epidemiology, “Any designated group of individuals who are followed or traced over a period of time.” Last, JM. A Dictionary of Epidemiology, 3 rd ed. New York: Oxford University Press, 1995

39. Cohort Studies DiseaseNo Disease Study Population Exposed Non-exposed No DiseaseDisease Exposure is self selected Follow through time

40. Cohort Studies: Prospective vs. Retrospective ExposureOutcome ProspectiveAssessed at beginning of study Followed into the future for outcome RetrospectiveAssessed at some point in the past Outcome has already occurred

41. Cohort: Forward Directionality Cohort Timeline ExposureOutcome Exposed Ill? Not Ill? Research Timeline

42. Cohort Study Steps in a Cohort Study: 1.Identify Cohort 2.Administer Questionnaire 3.Analyze Data

43. Cohort Study Step 1 – Identify cohort –Entire student body (n=12,000) –On-campus students (n=2400) –On campus students with meal plan who mainly ate at cafeteria (n=2000) –Sick students (n=75)  Do not select cohort so that either everyone is exposed or everyone is diseased

44. Cohort Study Step 2 – Administer questionnaire –Step 2a – Determine exposure status Main cafeteria Deli bar Food item –Step 2b – Determine disease status Importance of well-defined case definition

45. Cohort Study Example Dartmouth University: 698 (13.8%) of 5060 students had conjunctivitis in spring 2002 To identify risk factors... –web-based questionnaire set up –E-mail sent to 3682 undergraduates –No data entry-rapid analysis 1832 (50%) responded An outbreak of conjunctivitis due to atypical Streptococcus pneumoniae. N Engl J Med. 2003;348 (12):1112-21.

46. Cohort Study Step 3 – Analyze data IllNot IllTotal ExposedABA+B UnexposedCDC+D Risk Ratio[A/(A+B)] [C/(C+D)]

47. Interpreting a Risk Ratio The risk ratio is the ratio of the risk of disease in exposed individuals to the risk of disease in unexposed individuals RR=1.0 = no association between exposure and disease RR>1.0 = positive association RR<1.0 = negative association

48. Interpreting a Risk Ratio Example: Outbreak of conjunctivitis at Dartmouth College RR = 2.5 Disease = Conjunctivitis Exposure = Attended a fraternity / sorority party “Students who attended a fraternity or sorority party had 2.5 times the risk of having conjunctivitis compared to students who did not attend a fraternity or sorority party”

49. Cohort Study Example Recent norovirus outbreaks on cruise ships Attempt to interview all passengers Collect food history information MMWR: December 13, 2002 / 51(49);1112-1115

50. Cohort Study Examples Shigellosis among swimmers in a Georgia park – Used park registry to identify park visitors Whirlpools and Methicillin-Resistant Staphylococcus aureus –Occurred on a college football team

51. Question & Answer Opportunity

52. 5 Minute Break

53. Case-Control Study

54. Case-control Studies Study Population Cases Controls No Exposure Had Exposure

55. Case-Control: Backwards Directionality Cohort Timeline Research Timeline ExposureOutcome Ill Exposed? Not exposed?

56. Case-Control Study Steps in a Case-Control Study: 1.Identify the source population 2.Establish a case definition and select cases 3.Select controls 4.Analyze data

57. Case-Control Study Step 1 - Identify source population Represents the population that gives rise to the cases; is similar to a cohort study In the Texas case study, the source population is: all students who live on- campus

58. Case-Control Study Step 2 – Establish a case definition and select cases A standard set of criteria for deciding disease status –Clinical criteria, time, place, and person

59. Case Study Case Definition Study Hypothesis A viral infection spread by food(s) or beverage(s) served at the university’s main cafeteria between March 5 and 10 Case definition Vomiting or diarrhea (≥3 loose bowel movements during a 24-hour period) Onset on or after March 5, 1998 On-campus student seen at an emergency room or the Student Health Center

60. Case-Control Study Step 3 – Select Controls Represent source population –On-campus students in Texas case study Collect same exposure information as for cases

61. Case-Control Study Step 3 – Select Controls (cont’d.) Sources of controls –Random sample –Friends

62. Case-Control Study CasesControls ExposedAB UnexposedCD Odds Ratio(A/C) / (B/D) = (A*D) / (B*C) Step 4 – Analyze Data

63. Interpreting an OR The odds ratio is interpreted in the same way as a risk ratio: OR=1.0 = no association between exposure and disease OR>1.0 = positive association OR<1.0 = negative association

64. Interpreting an OR Example: Hepatitis A outbreak investigation OR = 12.0 Disease = Hepatitis A Exposure = Eating at restaurant X in April 2003 “Among those with Hepatitis A, the odds of having eaten at restaurant X in April, 2003 is 12 times greater among the cases than among the controls.”

65. Case-Control Study Examples Study to determine an association between autism and vaccination Study to determine an association between lung cancer and radon exposure Study to determine an association between salmonella infection and eating at a fast food restaurant

66. Activity 1 Please see: http://www.vdh.virginia.gov/EPR/Training.asphttp://www.vdh.virginia.gov/EPR/Training.asp for links to Session III activity questions and answers that you can print out. Answers are also provided on the next slide.

67. 5 Minute Break

68. Cohort versus Case-Control Study

69. Study Design AdvantagesDisadvantages Cohort 1.Least prone to selection bias 2.Can reasonable conclude that cause preceded disease 3.Can study several diseases at once 4.Can examine rare exposures 5.Retrospective can be low- cost 1.Prospective can be expensive, time- consuming 2.Prospective can lead to loss to follow up 3.Exposed may be followed more closely than unexposed, yielding invalid conclusions about causality Case- Control 1.Less expensive and quicker than cohort 2.Can examine the effect of multiple exposures 3.Require a smaller sample population 1.Inefficient for studying rare exposures 2.Susceptible to selection bias 3.Cannot directly estimate the risk of disease 4.Cannot study several diseases at once

70. Measures of Association

71. Activity 2 Please see: http://www.vdh.virginia.gov/EPR/Training.asphttp://www.vdh.virginia.gov/EPR/Training.asp for links to Session III activity questions and answers that you can print out. Answers are also provided on the next slide.

72. Activity 3 Choose a study design to investigate the source of the Texas outbreak Completion time: 5 minutes

73. Matching in Case-Control Studies

74. Makes one or more case and control attributes similar (e.g., age, gender, residence) An unmatched study design is usually preferred

75. Matching: Points to Consider More complex data analysis required Inability to assess role of matching factor on disease status –Do not match on exposure factor Potential for over-matching

76. Texas Dept. of Health Study Design Matched case-control study conducted among students at the university CASESCONTROLS Ill students who could be reached in the dorm room Roommates who had not become ill

77. Hypothesis Testing Results Texas case study: –Eating at the main cafeteria was not associated with illness –Eating lunch at the deli bar on March 9 or March 10 appeared to be associated with illness Individual food items could not be evaluated Conduct a larger, unmatched case-control study to identify high risk items or practices at deli bar

78. Hypothesis Testing: Validity of Results Selection / participation bias –Only 29 of 75 cases interviewed Not enough information on deli bar Matching on a risk factor –Can’t evaluate person-to-person transmission

79. Refine Hypothesis: Update TX Case Definition On-campus student with vomiting or diarrhea (≥3 loose bowel movements during a 24-hour period) Onset of disease on or after March 5, 1998 Member of the university meal plan

80. Test the Refined Hypothesis: Case-Control Study 40 cases were randomly selected from those reported to TDH by a local emergency room or the Student Health Center 160 controls who did not have nausea, vomiting, or diarrhea since March 5 were randomly selected from the university meal plan list

81. Questionnaire Administration The TX investigators decided to administer the questionnaire by telephone. Need for fast responses; subjects likely to be out of town Desire for accurate information, high response Anonymity not likely to be an issue Take measures to avoid interviewer bias

82. Test the Refined Hypothesis: Conduct Case-Control Study Texas study questionnaire administered by telephone March 15 - 23 –Called students at dormitory room or their home telephone number as recorded in university records –Follow-up phone calls were made to students, when possible or necessary

83. Test the Refined Hypothesis: Carry out Case-Control 36 cases and 144 controls were enrolled Cases included in the study were similar to all cases with respect to gender, age, year in college, and date of onset

84. Test the Refined Hypothesis: Case-Control Results ExposureOdds Ratio Salad Bar Lunch, Mar 90.8 Salad Bar Dinner, Mar 91.1 Deli Bar Lunch, Mar 911.1 Deli Bar Dinner, Mar 97.1 Deli Bar Lunch, Mar 105.7 28 (78%) of the 36 cases reported eating at the deli bar during at least one of the implicated meals.

85. Test the Refined Hypothesis: Case-Control Results* ExposureOdds Ratio American Cheese3.40 Swiss Cheese0.61 Ham1.50 Turkey0.95 Mayonnaise3.10 * Only participants reporting eating at deli bar during Mar 9-10 included

86. Interpretation of Analysis Results What do we know? What do we suspect? What can we “prove”?

87. What do we know? We know that 125 cases of vomiting or diarrhea have been reported – all students We know the cases were more likely to have eaten at the deli bar Mar 9 – Mar 10 We know the pathogen is not bacterial

88. What do we suspect? We suspect that cases were more likely to have eaten ham, mayonnaise, and American cheese, but results were not statistically significant We suspect the pathogen is viral We suspect the food was contaminated by cafeteria staff

89. What can we “prove”? We have “proven” the source came from the deli bar We will probably never “prove” which food was the source We may yet “prove” the pathogen is viral We may yet “prove” the food was contaminated by cafeteria staff

90. Interpretation of Analysis Results What questions do we still have? –Lab results for viral pathogen –Missing food handler who worked at deli bar

91. Texas Case Study: Additional Investigations

92. Additional Investigations: Environmental Sampling Results Water and ice samples obtained from the cafeteria on March 12 were negative for fecal coliforms Stool cultures and rectal swabs from the 23 food handlers were negative for bacteria

93. Additional Investigations: The Missing Food Handler The staff member who initially refused to be interviewed worked primarily at the deli bar She finally agreed to be interviewed on March 23, and reported slicing ham on March 9 for use at the deli bar during lunch and dinner that day and lunch on March 10 –Prepared and served sandwiches for March 9 and 10 meals –Wore gloves while slicing ham and while serving at the deli bar

94. Additional Investigations: The Missing Food Handler Denied any gastrointestinal illness during the outbreak period Reported her infant had been sick with watery diarrhea since March 7 Had not thought that hand washing was an important activity because she wore gloves during food preparation and serving

95. Additional Investigations: Viral Testing Of 18 fresh stool specimens sent to CDC, 9 (50%) had evidence of Norwalk-like virus (NLV) by reverse transcriptase polymerase chain reaction (RT-PCR) Of the four deli foods available from the implicated meals, only the March 9 ham sample was positive for the NLV RNA

96. Additional Investigations: Viral Testing NLV was also detected by RT-PCR in a stool sample from the ill infant of the food handler who prepared the deli sandwiches on March 9 The sequence of the amplified product was identical to those PCR products from the ill students and the deli ham

97. Texas Case Study Conclusions The evidence implicates the food handler as the source of the outbreak Diarrheal illness in the food handler’s child preceded the outbreak Food handler prepared ingredients and sandwiches served at the deli bar during the time that her child was ill

98. Texas Case Study Conclusions NLV was isolated from the child’s stool and was identical to that obtained from ill students and the deli ham The food handler was not necessarily ill or infected –May have transferred contamination directly from her ill infant

99. What was that again? Focus on Field Epidemiology Newsletter, “Selecting a Study Design” at http://www.sph.unc.edu/nccphp/focus/vol2/issue4/study_ design.pdf Visit http://www.sph.unc.edu/nccphp/focus/index.htm and click “Read FOCUS” for all newslettershttp://www.sph.unc.edu/nccphp/focus/index.htm

100. Session III Summary

101. An analytic study is used to test scientific hypotheses that may help support actions for specific control measures and to help prevent recurrence of a problem. A case definition with specific criteria helps you select your study population, as long as it does not include the hypothesis. Case-control studies, when conducted properly, are generally adequate and usually more efficient than cohort studies. Cohort studies may be preferable when you work with confined (e.g., easily identifiable and accessible) study populations such as on a cruise ship or at a wedding reception.

102. Session III Summary Case-control study controls need to be representative of the source population, and not matched on the exposure factor if matching is used. Risk ratios and odds ratios are the measures used to assess an association between an exposure and illness. Cohort study data analyses yield risk ratios; case-control study data analyses yield odds ratios.

103. Next Session September 1st 1:00 p.m. – 3:00 p.m. Topic: “Designing Questionnaires”

104. References and Resources Begier EM, Barrett FK, Mshar PA et al. Body Shaving, Whirlpools, and Football: An Out break of Methicillin-Resistant Staphylococcus aureus Cutaneous Infections in a College Football Team-Connecticut, 2003. Presented at the 53 rd Annual Epidemic Intelligence Service Conference. Atlanta, GA. April, 2004. Centers for Disease Control and Prevention (1992). Principles of Epidemiology: 2 nd Edition. Public Health Practice Program Office: Atlanta, GA. Centers for Disease Control and Prevention "Gastroenteritis at a University in Texas" http://www.phppo.cdc.gov/phtn/casestudies/classroom/gastro.htmhttp://www.phppo.cdc.gov/phtn/casestudies/classroom/gastro.htm Gordis, L. (2000). Epidemiology: 2 nd Edition. W.B. Saunders Company: Philadelphia, PA. Gregg, M.B. (2002). Field Epidemiology: 2 nd Edition. Oxford University Press: New York.

105. References and Resources Hennekens, C.H. and Buring, J.E. (1987). Epidemiology in Medicine. Little, Brown and Company: Boston/Toronto. Iwamoto M, Hlady G, Jeter M et al. Shigellosis among Swimmers in a Freshwater Lake-Georgia, 2003. Presented at the 53 rd Annual Epidemic Intelligence Service Conference. Atlanta, GA. April, 2004. Kleinbaum, D., Sullivan, K., and Barker, N. (2003). ActivEpi Companion Textbook. Springer-Verlag: New York. Last, J.M. (2001). A Dictionary of Epidemiology: 4 th Edition. Oxford University Press: New York. McNeill, A. (January 2002). Measuring the Occurrence of Disease: Prevalence and Incidence. Epid 160 lecture series, UNC Chapel Hill School of Public Health, Department of Epidemiology.

106. References and Resources Morton, R.F, Hebel, J.R., McCarter, R.J. (2001). A Study Guide to Epidemiology and Biostatistics: 5 th Edition. Aspen Publishers, Inc.: Gaithersburg, MD. North Carolina Center for Public Health Preparedness. March 2005 Public Health Information Network session: “Descriptive and Analytic Epidemiology.” http://www.sph.unc.edu/nccphph/phtin/index.htm University of North Carolina at Chapel Hill School of Public Health, Department of Epidemiology, and the Epidemiologic Research & Information Center (June 1999). ERIC Notebook. Issue 2. http://www.sph.unc.edu/courses/eric/eric_notebooks.htm http://www.sph.unc.edu/courses/eric/eric_notebooks.htm.

107. References and Resources University of North Carolina at Chapel Hill School of Public Health, Department of Epidemiology, and the Epidemiologic Research & Information Center (July 1999). ERIC Notebook. Issue 3. http://www.sph.unc.edu/courses/eric/eric_notebooks.htm http://www.sph.unc.edu/courses/eric/eric_notebooks.htm University of North Carolina at Chapel Hill School of Public Health, Department of Epidemiology, and the Epidemiologic Research & Information Center (September 1999). ERIC Notebook. Issue 5. http://www.sph.unc.edu/courses/eric/eric_notebooks.htm http://www.sph.unc.edu/courses/eric/eric_notebooks.htm University of North Carolina at Chapel Hill School of Public Health, Department of Epidemiology (August 2000). Laboratory Instructor’s Guide: Analytic Study Designs. Epid 168 lecture series. http://www.epidemiolog.net/epid168/labs/AnalyticStudExerInstGuid2 000.pdf http://www.epidemiolog.net/epid168/labs/AnalyticStudExerInstGuid2 000.pdf