Abstract Writing Tips

Transcription

1 Abstract Writing Tips Deb Houry, MD, MPH Associate Professor Vice Chair for Research Emory University Clinical Research Bootcamp February 2014

2 Session Objectives At the conclusion of the session, the attendee will: 1. Understand the components of an excellent abstract 2. Identify common pearls and pitfalls of abstract submissions

3 No one likes my idea Understand the abstract review process To central organization To a bunch of expert reviewers Final score Final decision made Was it poor study, poor reviewers or poor abstract review process? Was it controversial?

4 Abstract Evaluation Most societies Accept Kind of like it Kind of don t like it Reject Some will also want Extended abstract Revisions after their review

5 Abstract Evaluation Objectives Study Design Methods- Validity Methods- Reliability Statistics Results / Conclusion Impact

6 Abstract Scoring HYPOTHESIS/OBJECTIVES: (0-2) 0: No stated hypothesis OR objective 1: Stated hypothesis difficult to test OR stated objective poorly developed 2: Clearly stated testable hypothesis OR well thought out study objective Examples 0: We examined characteristics of elderly patients using our EMS system 1. The goal of this study was to determine how frequently elderly patients use our EMS system 2. Our hypothesis is that patients over 65 are two times more likely to use 911 for medical transport to the Emergency Department than patients under 65.

7 Abstract Scoring STUDY DESIGN: (0-2) 0: Chosen study design will not test the stated hypothesis OR study objective 1: Chosen study design is sub-optimal, but does test the stated hypothesis OR study objective 2: Chosen study design is the best feasible method to test the stated hypothesis OR study objective Examples 0: Using a survey to determine how well ED goal directed ultrasound improved patient care 1: Doing a retrospective chart review 2: Randomized controlled trial of patients with and without ultrasound to determine impact on survival and length of stay in patients with RUQ pain

8 Abstract Scoring METHODS- VALIDITY: (0-2) 0: Chosen study design poorly executed with critical flaws that definitely endanger the validity of the results. 1: Chosen study design executed with one or more flaws that potentially endanger the validity of the results. 2: Chosen study design executed in an acceptable manner in which results are expected to be valid. Example for retrospective chart review study 0: No descriptor of abstractor training or blinding 1: Defined inclusion/exclusion criteria and had single assistant review charts 2: Well described training, blinded abstractors, and reviewed charts with standardized data collection instrument

9 Abstract Scoring METHODS- SAMPLE SIZE AND RELIABILITY: (0-2) 0: Chosen study design poorly executed with critical flaws that definitely endanger the reliability of the results. 1: Chosen study design executed with one or more flaws that potentially endanger the reliability of the results. 2: Chosen study design executed in an acceptable manner in which results are expected to be reliable. Example 0: No inter-rater reliability (retro); small number of subjects 1: Inappropriate survey instrument or method of survey delivery 2: Well described endpoint and well-validated baseline measures (prospective): power calc

10 Abstract Scoring STATISTICS (0-2) 0: Inappropriate AND poorly described statistical methods 1: Inappropriate OR poorly described statistical methods 2: Appropriate AND well described statistical methods Examples Using parametric test to analyze non-parametric data Not reporting p-values or confidence intervals

11 Abstract Scoring RESULTS/CONCLUSION: (0-2) 0: Data critical to interpretation of the study is absent AND conclusions not supported by results 1: Data critical to interpretation of the study is not clearly presented or incomplete OR conclusions only partially supported 2: Data critical to interpretation of the study is completely presented AND conclusions fully supported Examples 0: Gives general drift of outcomes without adequate data (95% CI, p-values, etc); conclusion on non-significant data 1: Does not include information about study population (baseline data); overstates conclusions- clinical vs. stat significant relevance

12 Abstract Scoring IMPACT: (0-2) Change EM Practice for clinical studies Contribution to the current fund of knowledge for nonclinical studies

13 Sample Abstract Does a Brief Ultrasound Seminar Teach the Skills Required to Independently Perform Emergency Ultrasound

14 Sample Abstract Introduction Controversy persists regarding the best methods for teaching emergency ultrasound. The goal of this study was to evaluate the ability of a three-hour ultrasound seminar to teach the skills required to independently perform emergency sonographic studies. Hypothesis and Objective Is it clear and testable?

15 Sample Abstract Introduction Controversy persists regarding the best methods for teaching emergency ultrasound. The goal of this study was to evaluate the ability of a three-hour ultrasound seminar to teach the skills required to independently perform emergency sonographic studies WITH MORE THAN 90% ACCURACY AS ASSESSED BY AN INDEPENDENT OBSERVER.

16 ?? Sample Abstract Methods Participants at the US seminar at the ACEP 2000 Scientific Assembly were given a one-hour didactic lecture on the basics of abdominal or pelvic sonography. subsequently divided into groups of 3 and guided in the performance of limited, goaloriented examinations. The participants practiced for two-hours under the direct supervision of ultrasoundcredentialed EP s who later evaluated each participant s ability to independently perform sonographic studies. The instructors had predetermined criteria for grading the completeness of the monitored sonograms.??????

17 95% CI=71-99% Sample Abstract Baseline information Results 57 participants were enrolled, 22 in the abdominal and 35 in the pelvic session. 2 participants had significant previous experience (>100 scans), but the majority (45/57) had performed less than 10 ultrasounds prior to the workshops; 46% of the participants had no experience. During the abdominal US 91% were able to fully scan the aorta, 67% the right kidney, 57% the left kidney, 71% the gallbladder, and 86% were able to perform a complete FAST evaluation. In the pelvic section, 94% and 77% of participants were able to independently identify the uterus and adnexa. Previous scanning experience did not relate to better performance at the conclusion of the workshop.

18 Conclusion Sample Abstract A three-hour session, consisting of didactic teaching and a workshop, provides sufficient experience to develop 57-91% accuracy in emergency ultrasound performance as judged by predetermined criteria.

19 Sample Abstract Objectives Study Design Methods- Validity Methods- Reliability Statistics Results/ Conclusion Impact

20 Abstract Success PROSPECTIVE Better Studies Basic RCT Cohort Higher acceptance Nothing is perfect Other Survey Case Series Case Control Before-After Observational Cross-sectional Nonrand comparative SAEM Submissions 1999

21 Your turn Skim abstract And prepare to vote

22 Abstract #1 STUDY OBJECTIVE: Our purpose was to describe the history, physical, and laboratory findings in women with ovarian torsion (OT).

23 HYPOTHESIS/OBJECTIVES: (0-2)

24 METHODS: A retrospective chart review was conducted at 2 urban teaching hospitals. All women admitted from 1984 to 1999 with surgically proven OT were included in the study.

25 STUDY DESIGN: (0-2)

26 VALIDITY: (0-2)

27 SAMPLE SIZE AND RELIABILITY: (0-2)

28 RESULTS: The 87 women ranged in age from 14 to 82 years (mean 32 years). Twelve were pregnant, 15 were postmenopausal, and 7 were posthysterectomy. Thirty-five (40%) had prior pelvic surgery; 18 of these (21% of the total) had undergone tubal ligation. Twenty-two (25%) women had a history of an ovarian cyst. Sixtyfive (75%) patients were seen in the emergency department.

29 Pain characteristics were variable: the onset was sudden in 51 (59%); "sharp" or stabbing in 61 (70%); and radiated to the flank, back, or groin in 44 (51%) patients. Only 3 had peritoneal signs at presentation. The majority of patients (70%) had nausea or vomiting. Fever was rare (2 patients). OT was considered in the admitting differential diagnosis in 41 (47%) patients. An enlarged ovary (>5 cm) was found in 77 (89%) patients at surgery. Only 26 patients had surgery within 24 hours. In 8 (9%) patients, detorsion was possible; of these, 3 had surgery within 24 hours.

30 STATISTICS (0-2)

31 CONCLUSION: The diagnosis of OT is often missed and ovarian salvage is rare. Pain characteristics are variable and objective findings are uncommon in OT.

32 RESULTS/CONCLUSION: (0-2)

33 IMPACT: (0-2)

34 In or out? Total score

35 Score it again Pick a score excellent- accept- consider for oral, award, etc 3- very good- accept 2- fair (not embarrassing)- accept if room 1- poor- do not accept even if all the room in the world

36 Common reasons for rejection 1. Trivial study question (the so what test) or lack of a useful message 2. Lack of originality 3. Serious methodologic flaws 4. Results that are statistically significant but not clinically significant 5. Study is underpowered

37 A Really Bad Day

38 Abstract #2 (if enough time) Background: Motorcycle helmet laws vary greatly in the US. Georgia does mandate helmet use, but does not specify whether these helmets must be standard or regulation helmets. Previous research has suggested that these phony helmets will not protect motorcyclists from head injury. These motorcyclists may be at risk for head injury despite helmet use if they are not wearing a standard helmet. Objective: The objective of this study was to determine the frequency of phony helmet use among motorcyclists in Georgia, a state that mandates motorcycle helmet use.

39 HYPOTHESIS/OBJECTIVES: (0-2)

40 Methods: A prospective observational study was conducted over a six-month period. Varied locations, times of day, and days of the week were selected over several months to ensure a representative sample. Two trained resident researchers used a standardized checklist and observed motorcyclists and patterns of helmet use. Helmets were considered phony based on prior published studies: less than an inch thick, whether or not the entire ear was covered, unusual surfaces (canvas, baseball caps), and whether there were attachments to the helmet.

41 STUDY DESIGN: (0-2)

42 VALIDITY: (0-2)

43 SAMPLE SIZE AND RELIABILITY: (0-2)

44 Results: A total of 263 motorcyclists were observed during the study period. All motorcyclists were wearing helmets (100%). 23 patients were wearing phony helmets (8.7%) and another 16 (6.1%) were wearing helmets that the research assistants were unsure if they were phony or standard, so up to 39 (14.8%) of motorcyclists could be wearing substandard helmets.

45 STATISTICS (0-2)

46 Conclusion: The majority of motorcyclists did wear standard motorcycle helmets, but there was still a significant minority who wear phony motorcycle helmets. These motorcyclists are at risk for head injury and future research should be targeted at interventions for this population.

47 RESULTS/CONCLUSION: (0-2)

48 IMPACT: (0-2)

49 In or out? Total score

50 Score it again Pick a score excellent- accept- consider for oral, award, etc 3- very good- accept 2- fair (not embarrassing)- accept if room 1- poor- do not accept even if all the room in the world