Presenting Public Health Data2009 CDC/CSTE Applied Epidemiology Fellowship OrientationJulie Magri, MD, MPH (EIS 98, PMR 00)EIS Field Assignments Branch, CDD, OWCDIntroductionThis used to be two different lectures!How to set up and deliver an CDC-style presentationUnderstanding tables, graphs, and chartsMerged into “Presenting Public Health Data”Appropriate use of tables, graphs, charts for type of data and intended messageSuccessful communication of scientific findings to audiences in presentationsObjectivesDescribe format of a 10-minute oral scientific presentation Discuss elements that contribute to or detract from an effective slide presentationUnderstand appropriate ways to display different types of public health dataOrganizing and Delivering an Oral Scientific PresentationStructure of a 10-minute Oral Scientific PresentationTitleBackground MethodsResultsDiscussionAcknowledgmentsQuestion and answer periodTitle Slide (10-15 sec)Title should includeSubjectLocationTime periodYour name Your affiliationAppropriate logosSay “Good morning / afternoon / evening”Elevated Fall-Related Mortality Rates New Mexico, 19992004Aaron M. Wendelboe, PhDNew Mexico Department of HealthOWCD, CDD, EIS Field Assignments BranchBackground (1-2 min)Usually a few slidesEngage audienceSet stage for outbreak investigationProvide rationale for planned studyEssential information (only) about diseaseEstablish relevance to public healthInclude a slide describing study objectivesMethods (1-2 min)Usually a few slidesDescribe study design(s)Define a case and describe case findingTell how controls were selected if CC studyDefine cohort if cohort studySay what laboratory tests were usedDescribe any environmental investigation methodsResults (3-4 min)Usually several slidesEmphasize most important findingsDescribe characteristics of study participantsInclude descriptive epidemiologic results and analytic resultsUse mixture of text, tables, figures, photos as appropriate to your dataDiscussion (2-3 min)Interpretation of findingsDont repeat resultsPrioritize findings from most to least importantLink findings to study objectives Put findings into context with previous studiesLimitations slide (only the important ones)Conclusions slide(s) based on your findingsRecommendations slide(s) Control measuresDirections for future studiesAcknowledgments (10-15 sec)Recognize coauthors and contributorsMore screen time at end of presentationOrganize by agency Same logos as on title slideOMB disclaimer (dont read aloud)“The findings and conclusions in this presentation are those of the authors and do not necessarily represent the views of CDC.”Your last words = “Thank You”Creating Effective SlidesEffective SlidesAre uncluttered, clear, visibleDont distract the audience Use informative titles“Characteristics of Study Participants”“Risk Factors for Illness” Not “Results 1, “Results 2”Use bolded, sans serif font (Arial, Tahoma)Have simple, high-contrast, consistent color schemes Typical CDC Color SchemeDark blue background (R=0,G=0,B=102), yellow titles, white text, highlighted text in yellow or bright blue AdvantagesHigh contrast as long as room is darkProjects well on EnvisionDisadvantagesPoor contrast if room not darkLimits available colors for graphsPasting in graphics can be more complexOther Effective Color Schemes?nLightest gray background, dark typenAdvantagesHigh contrast even in a brightly lit roomProjects well on Envision Allows wider choice of colors for graphsPasting in graphics can be easiernAvoid pure white backgroundHard on eyes in a dark roomCauses flare on EnvisionColor-Blind “Friendly” Presentations Avoid red-green color combinationsIf must use red, use yellowish red (R=255 / G=82 / B=0) instead of pure redAvoid red characters / lines on dark backgroundMake text and lines as big or thick as practicalUse high-contrast color schemeStart with a Slide MasterSet font, size, boldness, color scheme, capitalization, bullet size/shape/hierarchy Slides created from the Master will automatically be formatted correctlyMake the Master before creating your slidesApplying the Master retroactively to existing slides doesnt work wellSave as template for future useRecommended Fonts and SizesSans serif font, all titles and text boldedFor Arial (bolded):Titles 36 ptMain bullets 28 ptSub-bullets 28 pt if room, otherwise 24 ptAvoid sub-sub bullets (re-format)Keep text / title size consistent across slidesStop PowerPoint from changing text size as you typeGo to Tools, AutoCorrectUncheck “Autofit body/title text to placeholder”Effective Text SlidesOrder of slide text matches order of scriptKey words only, not complete sentences810 lines maximumBulleted text better than numbered items in most casesParallel structure (all verbs, all nouns, etc) Things to AvoidVisual clutter from too many colors Unbolded, serif font like Times New RomanALL CAPS (HARD ON THE EYES)Pseudo-3D charts and graphsAnimation (no flying objects; slide builds=OK)Clip art that serves no purposeUnnecessary grid lines in figuresNecessary lines that are too thinAll PowerPoint design templatesMatched DHS list of residents since 2001 to statewide TB registry from 2000-2003Genotyped M. tuberculosis isolatesSearched genotype databasesProvided onsite TB screening to shelter residentsCase Finding MethodsMATCHED DHS LIST OF RESIDENTS SINCE 2001 TO STATEWIDE TB REGISTRY FROM 2000-2003GENOTYPED M. TUBERCULOSIS ISOLATESSEARCHED GENOTYPE DATABASESPROVIDED ONSITE TB SCREENING TO SHELTER RESIDENTSCASE FINDING METHODSAvoid Distracting TemplatesAvoid Distracting Templatesn nMatched DHS list of residents since 2001 to statewide TB registry from 2000-2003n nGenotyped M. tuberculosis isolatesn nSearched genotype databasesn nProvided onsite TB screening to shelter residentsCase Finding MethodsnMatched DHS list of residents since 2001 to statewide TB registry from 2000-2003nGenotyped M. tuberculosis isolatesnSearched genotype databasesnProvided onsite TB screening to shelter residentsCase Finding MethodsMatched DHS list of residents since 2001 to statewide TB registry from 2000-2003Genotyped M. tuberculosis isolatesSearched genotype databasesProvided onsite TB screening to shelter residentsCase Finding MethodsPhotos and Clip Art Tips Should serve a purposeNo copyrighted materials without permissionNo photos of identifiable people unless releaseNo photos of your kids or your petsClip art cautionsSimplest is most effectiveCheck in Slide Show to make sure it is not animatedData MethodsManagementDouble enteredMissing data: completed using other sourcesAnalysisExcluded if: contraindications, unknown vaccination history, reported “no take” Kaplan-Meier survival curvesMultivariate logistic regression : Odds Ratios (OR), 95% Confidence intervals (95%CI)(Avoid clip art that serves no purpose)Animals and Farm EnvironmentFecal samples from sheep, goats, and large animal speciesHide swabs from sheep and goatsEnvironmental samples(Photos are good if they serve a purpose)Stakeholder InputFocus groupsOpen-ended interviewsBrief written surveyStakeholders needs incorporated into evaluation design(This stock photo adds nothing)Traceback InvestigationDistributorSlaughterhouseFarmProduct(OK to use clip art that serves a purpose)Traceback InvestigationDistributorSlaughterhouseFarmProductDont Use AnimationTables, Graphs, and ChartsTableSet of data arranged in rows and columns containing quantitative informationFrequency of occurrence of some event or characteristic in different subgroupsOften used for person or clinical dataUseful for demonstrating patterns, exceptions, differences, and other relationships in the dataBasis for preparing graphs and chartsTypes of Tables1-variable table (frequency distribution)Range of values of a single variableNumber of observations with each value2-variable tableCounts shown according to 2 variables at once3-variable table Counts shown according to 3 variables at onceSyphilis Cases by AgeUnited States, 1989CasesAge Group (years)Number Percent 551,2782.9Total44,081100.0Example of a One-Variable TableSyphilis Cases by Age and SexUnited States, 1989Age GroupNumber of Cases by Sex(years)Male Female Total551,1471311,278Total26,00618,07544,081Example of a Two-Variable TableExample of a Two-Variable (Contingency) TableExample of a Two-Variable TableExample of Three-Variable Table Syphilis Cases by Age, Sex, and RaceUnited States, 1989too complicated for a slide! Another example of a table thats too complicated for a slide presentationExample of Unnecessary GridlinesFoodCases (n=15)Controls(n=30)Odds Ratio(95% CI)No.(%)No.(%)Potatoes15(100)19(63)undef(1.9 undef)Cole Slaw12(80)16(53)3.5(0.7 19.7)Pudding12(80)14(47)4.0(0.8 22.5)Beef13(87)19(63)3.8(0.6 29.4)Chicken11(73)19(63)1.6(0.3 7.8)Case-Control Study ResultsCases (n=15)Controls(n=30)Odds RatioFoodNo.(%)No.(%)(95% CI)Potatoes15(100)19(63)undef(1.9 undef)Coleslaw12(80)16(53)3.5(0.7 19.7)Pudding12(80)14(47)4.0(0.8 22.5)Beef13(87)19(63)3.8(0.6 29.4)Chicken11(73)19(63)1.6(0.3 7.8)Case-Control Study ResultsGraphDisplays quantitative data using a set of coordinatesRectangular graph (x,y) most commonMethod of classification along horizontal (x) axis Frequency along vertical (y) axis Most often used for time dataRatePrevalenceNumber of cases by week or yearArithmetic-Scale Line GraphMost common typeIntervals on x axis are equalIntervals on y axis are equal (not necessarily the same as x axis)Scale breaks should be avoided if possibleUseful to show trends in numbers or rates over timeArithmetic Scale Line Graph PitfallsUnnecessary gridlinesToo many data series on one graphLines too thin, text unboldedNo title or axis labelsUnequal intervalsNewly Diagnosed HIV Cases by Race Mississippi, 20052007Newly Diagnosed HIV Cases by Race Mississippi, 20052007African American +30%All other races 34%Total +6%Newly Diagnosed HIV Cases by Race Mississippi, 20052007African American +30%All other races 34%Total +6%Histogram“Epidemic curve” in outbreak investigationsFrequency distribution of quantitative datax axis continuous, usually time (onset or diagnosis date)No spaces between adjacent columns (vs. bar)Easiest to interpret with equal class (x) intervals Column height proportional to # observations in that intervalOpening of schoolsVaccination CampaignsMeasles Cases by Week of OnsetTexas and Arkansas, 1970 - 1971ChartMethod of presenting quantitative data symbolically using only one coordinate Most appropriate for comparing data with discrete categoriesTypes of chartsbar, pie, maps, others Bar Charts4 subtypesSimpleGrouped or clusteredStacked100% bar chartCan be vertical or horizontalUsed for person, clinical, and place dataSubtype chosen depends on desired emphasisSimple Bar ChartDisplays data from a one-variable tableEach category is represented by a barCategories discrete or treated as if discrete, so adjacent bars do not touchHeight of bar proportional to frequency of events in that categoryGood for comparing size or frequency of different categories of a single variableOverweight Prevalence by Race/EthnicitySimple Bar ChartGrouped Bar ChartDisplays data from 2- or 3-variable tablesBars within a group touch, groups are separate Color the bars distinctively and use legendGood for comparing size of subgroups within and across main categories Not good for comparing totals across main categories (use simple bar chart)Confusing if lots of subgroupsOverweight Prevalenceby Sex and Race/EthnicityGrouped Bar ChartOverweight Prevalenceby Sex and Race/EthnicityDont Use 3D or PowerPoint DefaultsOverweight Prevalenceby Sex and Race/EthnicityOverweight Prevalenceby Sex and Race/EthnicityNumberOverweightGrouped Bar ChartCount DataStacked Bar ChartDisplays individual bars of each group of grouped bar chart stacked on top of one another as a single barGood for showing relative contributions of different subcategories within a given barCan be difficult to interpret because only the first component rests on flat baselinecant compare heights of the other components across categoriesOverweight Prevalenceby Sex and Race/EthnicityNumberOverweightStacked Bar ChartCount Data100% Component Bar ChartAll bars same height (100%)Components shown as proportions of the total, not actual valuesGood for comparing how components contribute to the whole within a groupNot useful for comparing relative sizes of the components across different groups because the denominator changesOverweight Prevalenceby Sex and Race/EthnicityPie ChartShows components of a wholeSize of slice/wedge = the proportional contribution of each component of the variableStart at 12:00 with biggest slice, go clockwise(“unknown” can be last wedge, even if big)Label wedges (category, %) directly if possible rather than using a legendEasier for human eye to detect differences in height (bar chart) than area (pie chart)Students Weight Status Los Angeles County, 2001n=281,630Students Weight Status Los Angeles County, 2001n=281,630Dont use 3-DMapsGeographic chartsUseful forplace data !Spot map: shows location of cases or eventsArea map: case counts, prevalence, or rates by state, county, etcEach point = 30 casesCumulative U.S. AIDS Cases, May 1985 N10,000Spot Map Obesity* Among U.S. AdultsBRFSS, 1985Adapted from: Mokdad A H, et al. J Am Med Assoc 1999;282:16, 2001;286:10. Area MapChoose Colors Carefully for Maps and ChartsDont obscure a hierarchy that does existIf hierarchy, use graded colors within a family, e.g. light blue, medium blue, dark blueLightest = “least”, darkest = “most”Dont imply a hierarchy where none existsIf no hierarchy, use equally strong colorsAvoid red-green combinations (color blindness)Overweight Prevalenceby Sex and Race/EthnicityPercentage of Tuberculosis Cases Among Foreign-Born PersonsUnited States, 2002From MMWR (Weekly) March 21, 2003Example of Colors Obscuring the Hierarchy50%25-49%25%Tips on Delivering Oral PresentationsPreparation TipsUse script, flesh out bullet pts into sentencesPractice is the key to making sure it doesnt sound scriptedPrint your script in large enough type (14-16 pt)Check script size in Notes Master or Notes viewPrint one slide and accompanying script per pageTime your presentation If you used “Rehearse slide timings” feature, go to Slide Show, Set Up Show, and uncheck “Advance slides using timings if present” 4Delivery TipsGet there earlyDont start speaking until readySpeak slowly and with sufficient volumeDont turn your back on your audienceCheck that the correct slide is projectingUse microphone correctlyBe careful with humorExplain charts / graphs before giving pointExplain epidemiologic associations clearlyPause before advancing to next slideReasons Not to Use a Laser PointerHave to turn away from audience to use itSome projection screens absorb the laser, so audience in room cannot see it Color-blind people cant see it (red)Can become a crutchIf your hands are shaking, pointer will show itAlternatives: Building “pointers” into slidesUsing computer cursor (arrow)Overweight Prevalenceby Sex and Race/EthnicityBuild arrow into slide instead of using pointerProviders “Very Comfortable” with HIV+ Patients ConcerningOverall(n = 60)Urban(n = 43)Rural(n = 17)Prescribing ART44 (73%)37 (86%)7 (41%)CD4 count57 (95%)42 (98%)15 (88%)Viral load57 (95%)42 (98%)15 (88%)Illicit drug use48 (80%)37 (86%)11 (65%)Sexual behavior47 (78%)37 (86%)10 (59%)Providers “Very Comfortable” with HIV+ Patients ConcerningOverall(n = 60)Urban(n = 43)Rural(n = 17)Prescribing ART44 (73%)37 (86%)7 (41%)CD4 count57 (95%)42 (98%)15 (88%)Viral load57 (95%)42 (98%)15 (88%)Illicit drug use48 (80%)37 (86%)11 (65%)Sexual behavior47 (78%)37 (86%)10 (59%)Instead of pointer, highlight data with a circleFoodCases(n=25)Controls(n=25)Odds Ratio(95% CI)Ham14121.2(0.43.5)Rolls13140.8(0.32.6)Milk12140.7(0.22.2)Ice Cream2088.5(2.330.1)Potatoes23231.0(0.17.7)Case-Control Study Resultsor highlight the data with an accent colorQuestion & Answer Period: DosDo take a moment to thinkDo give short, direct answersDo say you dont know if you dont know Do ask for clarification of question if needed Do rehearse answers to obvious questions Do bring a pen to write down multi-part questionsDo have a definite end to your answerQuestion & Answer Period: DontsDont fumble for extra slidesDont be defensive even if question hostileDont ask “Did that answer your question?”Dont thank the questioner for the questionDont rate the questionDont back away from the podium as if poisonDont hang on to podium as if life-preserverPoster PresentationsNot a journal article hung on the wallShort phrases, bulleted text not complete sentencesHighlight major points clearly so viewer can understand them without benefit of a presenterPresent information graphically where possibleFlow chart, tables, graphs, photosConsider giving take-home point for each graphicGraphics can be more complex than oral presentationGraphics not constrained by height-to-width ratio of slides for oral presentation2008 EIS Conference Poster AwardPoster PresentationsDesign layout and text size for viewing from 3 feet awayCheck size of display board before creatingMake flow of logic clearConsider numbering the sections to show the viewer the correct sequenceDecide whether to include abstractCheck conference requirementsTake-Home MessagesDecide type of data and the point you want to convey, then choose the visual accordingly (text, table, graph, chart, etc.)Well organized, practiced presentation with clear, effective slides (when used) reinforces your message and helps you communicate effectivelyGood science is more important than glitzEIS Presentation ChecklistBackgroundWas the public health issue clearly identified?Were the objectives clear?MethodsWhat was the study design?Was the case definition clear and appropriate?How were the cases identified?Were the analytic methods clearly described?ResultsWere results ordered logically (clinical, epi, lab, environmental)?Were descriptive data presented clearly (person, place, time)?Were correct measures of association and stat significance used?Did stated results logically follow the methods?EIS Presentation Checklist (contd)DiscussionWere findings put into context based on a thorough understanding of the literature?Were the most important limitations addressed (but not the unimportant ones)?Were conclusions stated clearly and ordered logically (from most to least important)?Were conclusions consistent with the data presented?Were the objectives met?Did recommendations follow logically from the data presented?Additional ResourcesMaking an epicurve:NCHS advice on creating scientific posters:Poster templates from CDC Creative Services :/intra-apps.cdc.gov/createit/service_sciposter.asp#client AcknowledgmentsRichard Dicker, MD, MScLaura Fehrs, MDRobert Fontaine, MD, MScRichard Goodman, MD, JDTom Trk, MD, MPHAndrea Winquist, MDEIS Officers past and presentEpidemiology of B. cepaciaLow virulence, morbidity and mortality rates are generally associated with B. cepacia in most patient populationsResearch suggests that investigations are necessary when multiple patients test positive for B. cepacia over a short time periodDont Display Full Sentences From ScriptEpidemiology of B. cepaciaLow virulence Low morbidity and mortality ratesInvestigate when multiple patients test positive for B. cepacia over short time Display Key Words as Signposts Instead