醫學研究進階訓練_081412

CGMH EM Kuan-Fu Chen 8/14/12

1027abstract 1.(1)(2) 2.:10:00~12:00; 10:00~12:00

1 101/8/14(W) 2 101/9/6(W) 3 101/10/4(W) 4 101/11/1(W) 5 101/12/6(W)

Proposal

6 102/1/3(W) 7 8 9 10 11

102/2/7(W) () 102/3/7(W) () 102/4/4(W) () 102/5/2(W) () 102/6/6(W) () paper

12 102/7/4(W) 13 102/8/1(W)

() () paper reference

(1) (2) (3)

2

Objec8ves Forming study ques8ons Skill of Literature review Introduc8on of study design

3


4

Ten Commandments

PICKING A RESEARCH PROBLEM

11

12

13

Picking a Research Problem (Ten Commandments) I. An8cipate the Results before Doing the First Study II. Pick an Area on Basis of Interest of Outcome III. Look for an Underoccupied Niche Has Poten8al IV. Go to Talks & Read Papers outside Area of Interest V. Build on a Theme

14

Picking a Research Problem (Ten Commandments) VI. Find a Balance btw Low-Risk & High-Risk VII. Be Prepared to Pursue to Any Depth VIII. Dieren8ate Yourself from Your Mentor IX. Do not Assume Clinical Research Is Easier X. Focus, Focus, Focus

15

Good research Design Technology the best at this 8me

Execu8on quality of procedures, care with which the protocols (e.g. few losses to follow up) how to determine: QC

Analysis Repor8ng accurately check internal consistency (when you are a reviewer) 16

Outstanding research 1. Asks important ques8ons (Who is the audience making the assessment of importance?) Advances scien8c understanding Promise of u8lity in the future Clinical outcomes or HCD (health care delivery) Heuris8c ability (leads the eld into a new direc8on): e.g. stem cell

2. Seminal result being the seed that promote whole blossoming of an area Paradigm shii

opening up a new biologic pathway, new target for disease Nobel Prize!

17

Outstanding proposal NIH Cri8que FINER Feasibility Interes8ng Novel Ethical Relevant

18

Outstanding proposal Feasibility Technical exper8se to do what proposing to do. Pa8ent recruitment demonstrate that you have enough pa8ents. Manageable in scope. Reasonable 8me and money.

Interes8ng To whom? Should be fascina8ng to others. 19

Outstanding proposal Novel Not always mean absolutely new & undone before. Not repea8ng something already known. Can use known methods. Even if two previous trials showed conic8ng results to resolve the results

Goes farther than a previous study extent what was shown before

New approach, new technique, new paradigm (e.g. genomics, proteomics) oien awarded on novelty than feasibility/likelihood 20

Outstanding proposal Ethical Relevant Scien8c clinical outcome If this lack in knowledge important to the eld? Clinical outcomes or clinical delivery outcomes? Furthering important new discoveries in the eld?

21

HOW ABOUT YOUR IDEA?

22


23

How do you search? Why not look in a textbook? Why not just read a few papers? Does doing a good search maper? What databases are available? How does one search? Not easy! Dont underes8mate Obstacle: Not found or published. 24

Where do you search? Primary, secondary vs. Public database PubMED/MEDLINE database Cochrane Central Register of Controlled Trials Other medical and health care databases EMBASE, Science Cita8on Index,, CINAHL, Psychlit

Non-English language literature

25

Where do you search? Grey literature (theses, industry les, internal reports)

References of primary sources (and references of references)

Other unpublished sources known to experts (seek by personal communica8on)

Raw data from published studies (seek by personal communica8on)

26

Dierence btw PubMed vs. Ovid MEDLINE Use PubMed Want to be always available, even if leaving CGMH O-campus and having connec8on issues. Want quick results with strategies automa8cally created for you. When looking for extremely recent cita8ons. Also want to search for gene8cs and molecular biology informa8on.

Use Ovid MEDLINE To be guided through selec8ons for a precise search based on Medical Subject Headings, subheadings, and limits.

To build a search strategy in steps and by trying mul8ple combina8ons. 27

Dierence btw PubMed vs. Ovid MEDLINE

28

MeSH heading Medical Subject Headings Maintained by Na8onal Library of Medicine (NLM)

Subheadings: tree or mapping PubMed: [sh] MEDLINE: /su[ ] If no picked, all selected

29

MeSH heading Strength of MeSH Secondary topic pre-linked to a MeSH term by an indexer Standardiza8on Aor8c rupture vs. Ruptured aorta vs. Rupture of the thoracic aorta

Focus/major topic Subheadings Explode Find MeSH below in the tree Dierence between PubMed vs. Ovid MEDLINE 30

MeSH heading Limita8on Not for all (e.g. Aor8c dissec8on) Not for recent cita8ons Not every word in abstract New MeSH not retroac8ve

31

Text word searching Helpful if no MeSH term, and to supplement MeSH Strength Instant index Up-to-date terminology Every word in abstract searchable

Challenge Easy to miss informa8on Dierent words for same topic Cant use Explode, Focus, or Subheadings 32

Text word searching Searching for varia8ons Trunca8on character (*): Sep* Synonyms: Sepsis, bacteremia Bri8sh spellings: Baecteremia Break up phrases Sep8c shock sep* AND shock ($ = * ? OR #)

Search MeSH as major topic (focus) then scan results to see varia8on in 8tles

Display MeSH scope note PubMed: Entry Terms Ovid MEDLINE: [i] in mapping or tree display; look for used for list 33

34

Search process (1/4) Analyze the ques8on Iden8fy cri8cal topics, keeping only essen8al ones (fewest possible)

Search each topic separately Most important topic rst

Create separate result sets for MeSH and for text words Why? Dierent thought processes: MeSH Use mapping; check Tree display to decide whether to explode Make major topic (focus) decision Check Subheadings for exact match to search topic. 35

Search process (2/4) Create separate result sets for MeSH and for text words (cont) Text words Type synonyms, Use trunca8on character, Dierent spellings, spacings, etc. Combine word varia8ons using OR (RU486 or RU 486 or RU-486).

Be sure topics in text word search match topics in MeSH If MeSH is major topic, search text words in 8tle only (major point).

36

Search process (3/4) Combine set numbers Use OR to combine words or set numbers expressing SAME TOPIC. Computer merges results and eliminates duplicates.

Use AND to combine words or set numbers expressing DIFFERENT topics Computer nds cita8ons with both topics (or all, if >2)

Use parentheses to avoid logic errors PubMed: (#6 OR #9) AND (#12 OR #15) Ovid: 1 and (5 OR 8)

37

Search process (4/4) Too many irrelevant cita8ons? Make one or more topics a major point (focus).

LIMIT to target large search results beper: Review ar8cles Core Clinical Journals NEJM, JAMA, and other Valid clinical evidence limits: Publica8on Type, Subject Subset Search all MeSH and text word varia8ons match topics.

Look at MeSH Headings assigned to the best results. If see another good MeSH, add it to your search. 38

39

PubMed tutorial Help and tutorials Searching format(Boolean) MeSH Saving strategies (NCBI) Saving collec8ons Clinical queries Matcher Limits Links Related ar8cles hpp://www.nlm.nih.gov/bsd/disted/pubmed.html 40

(Journals) Conference proceedings Bibliographies of related studies Bibliographies of systema8c reviews Grey literature (theses, industry les, internal reports)

Hand searching

42

43

HAVE YOU DONE A THOROUGH REVIEW? 44


45

Terminology (Jargon) Prospec8ve vs. Retrospec8ve Case vs. control Prospec8ve vs. Retrospec8ve cohort Cohort? Causal inference Validity: accuracy, bias (lack of internal validity) Reliability: precision, reproducibility, variability, random error Counterfactual 46

Study Designs in Clinical Research Case report Case series Cross-sec8onal Case control Nested case control Case cohort Cohort 47

Causal Diagram

48

Simpsons Paradox

49

50

51

Causal Diagram

52

Causal Diagram S&S Vaccination

Viral Load+ BCP Antiviral + Demographics

Severity

Time since onset

53

Causal Diagram (exam from Dr. Chao) Immunity DM physician factor comorbidity mortality

social economic status age recent blood sugar control severity (MEDS) Others??54

Study pyramid

55

Lancet 2002; 359: 57-61

What is the study design?

65

An assessment of the incremental value of the ABCD2 score in the emergency department evaluation of transient ischemic attack.

STUDY OBJECTIVE:We study the incremental value of the ABCD2 score in predicting short-term risk of ischemic stroke after thorough emergency department (ED) evaluation of transient ischemic attack. METHODS:This was a prospective observational study of consecutive patients presenting to the ED with a transient ischemic attack. Patients underwent a full ED evaluation, including central nervous system and carotid artery imaging, after which ABCD2 scores and risk category were assigned. We evaluated correlations between risk categories and occurrence of subsequent ischemic stroke at 7 and 90 days. RESULTS:The study consisted of 637 patients (47% women; mean age 73 years; SD 13 years). There were 15 strokes within 90 days after the index transient ischemic attack. At 7 days, the rate of stroke according to ABCD2 category in our study was 1.1% in the low-risk group, 0.3% in the intermediate-risk group, and 2.7% in the high-risk group. At 90 days, the rate of stroke in our ED was 2.1% in the low-risk group, 2.1% in the intermediate-risk group, and 3.6% in the high-risk group. There was no relationship between ABCD2 score at presentation and subsequent stroke after transient ischemic attack at 7 or 90 days. CONCLUSION:The ABCD2 score did not add incremental value beyond an ED evaluation that includes central nervous system and carotid artery imaging in the ability to risk-stratify patients with transient ischemic attack in our population. Practice approaches that include brain and carotid artery imaging do not benet by the incremental addition of the ABCD2 score. In this population of transient ischemic attack patients, selected by emergency physicians for a rapid ED-based outpatient protocol that included early carotid imaging and treatment when appropriate, the rate of stroke was independent of ABCD2 stratication.

66


67

Placing physician orders at triage: the effect on length of stay.

STUDY OBJECTIVE:Emergency department (ED) crowding is a signicant problem nationwide, and numerous strategies have been explored to decrease length of stay. Placing a physician in the triage area to rapidly disposition low-acuity patients and begin evaluations on more complex patients is one strategy that can be used to lessen the effect of ED crowding. The goal of this study is to assess the effect of order placement by a triage physician on length of stay for patients ultimately treated in a bed within the ED.METHODS:We conducted a preexperimental study with retrospective data to evaluate patients with and without triage physician orders at a single academic institution. A matched comparison was performed by pairing patients with the same orders and similar propensity scores. Propensity scores were calculated with demographic and triage data, chief complaint, and ED capacity on the patient's arrival.RESULTS:During the 23month study period, a total of 66,909 patients were sent to the waiting room after triage but still eventually spent time in an ED bed. A quarter of these patients (23%) had physician orders placed at triage. After a matched comparison, patients with triage orders had a 37-minute (95% condence interval 34 to 40 minutes) median decrease in time spent in an ED bed, with an 11-minute (95% condence interval 7 to 15 minutes) overall median increase in time until disposition

68


69

US emergency department performance

on wait time and length of visit.

STUDY OBJECTIVE:Prolonged emergency department (ED) wait time and length of visit reduce quality of care and increase adverse events. Previous studies have not examined hospital-level performance on ED wait time and visit length in the United States. The purpose of this study is to describe hospital-level performance on ED wait time and visit length. METHODS:We conducted a retrospective study of a stratied random sampling of 35,849 patient visits to 364 nonfederal US hospital EDs in 2006, weighted to represent 119,191,528 visits to 4,654 EDs. Measures included EDs' median wait times and visit lengths, EDs' median proportion of patients treated by a physician within the time recommended at triage, and EDs' median proportion of patients dispositioned within 4 or 6 hours. RESULTS:In the median ED, 78% (interquartile range [IQR], 63% to 90%) of all patients and 67% (IQR, 52% to 82%) of patients who were triaged to be treated within 1 hour were treated by a physician within the target triage time. A total of 31% of EDs achieved the triage target for more than 90% of their patients; 14% of EDs achieved the triage target for 90% or more of patients triaged to be treated within an hour. In the median ED, 76% (IQR 54% to 94%) of patients were admitted within 6 hours. A total of 48% of EDs admitted more than 90% of their patients within 6 hours, but only 25% of EDs admitted more than 90% of their patients within 4 hours. CONCLUSION:A minority of hospitals consistently achieved recommended wait times for all ED patients, and fewer than half of hospitals consistently admitted their ED patients within 6 hours.

70


71

Laryngospasm during emergency department ketamine sedation:

OBJECTIVE:The objective of this study was to assess predictors of emergency department (ED) ketamine-associated laryngospasm. METHODS:We performed a matched analysis of a sample of 8282 ED ketamine sedations (including 22 occurrences of laryngospasm) assembled from 32 prior published series. We sequentially studied the association of each of 7 clinical variables with laryngospasm by assigning 4 subjects with no laryngospasm each subject with laryngospasm while matching for the remaining 6 variables. We then used univariate statistics and conditional logistic regression to analyze the matched sets. RESULTS:We found no statistical association of age, dose, oropharyngeal procedure, underlying physical illness, route, or coadministered anticholinergics with laryngospasm. Coadministered benzodiazepines showed a borderline association in the multivariate but not univariate analysis that was considered anomalous. CONCLUSIONS:This analysis of the largest available sample of ED ketamine-associated laryngospasm did not demonstrate evidence of association with age, dose, or other clinical factors. Such laryngospasm seems to be idiosyncratic, and accordingly, clinicians administering ketamine must be prepared for its rapid identication and management. Given no evidence that they decrease the risk of laryngospasm, coadministered anticholinergics seem unnecessary.

72


73

3

4

Cohort study3 events or cases occur among 8 people, of whom 5 are ever exposed

g

y

Persons

Exposed are solid lines, unexposed are dashed Dots are events

Time5 6

Study design: Cohort study 74

1

Epidemiologic Methods 752

Nested case-control study(Incidence Density Sampling)Compare 3 cases to 3 non-cases (at time of event)

Case c

Persons

Examples:Lung cancer: JAMA 1950;143:329-36 AIDS: Ann Intern Med 1983;99:145-51

Time7

Study design: Nested case-control Basic Idea for the SamplingAt each time at which one or more events are75

Persons

Study design: Nested case-cohort Case cohort studyCompare 3 cases to non-case person-time in random subcohortSubcohort

November 20, 2006

s to time of

Persons

Discard non-event person-time outside subcohort Examples:HCV: Hepatology 2003;38:50-6

9-36

3;99:145-51

Time7 8

76

Study design: Nested case-cohort Case-cohort example - samplingCOHORT 10,275 CASES 1155Step 1: Select the SUBCOHORT (random sample of the full cohort) Step 2: Select a random sample of the cases

November 20, 2006

SUBCOHORT 612 92

CASES 529CASES outside the subcohort44

1943

73

456

77

Laryngospasm during emergency department ketamine sedation:

OBJECTIVE:The objective of this study was to assess predictors of emergency department (ED) ketamine-associated laryngospasm. METHODS:We performed a matched analysis of a sample of 8282 ED ketamine sedations (including 22 occurrences of laryngospasm) assembled from 32 prior published series. We sequentially studied the association of each of 7 clinical variables with laryngospasm by assigning 4 subjects with no laryngospasm each subject with laryngospasm while matching for the remaining 6 variables. We then used univariate statistics and conditional logistic regression to analyze the matched sets. RESULTS:We found no statistical association of age, dose, oropharyngeal procedure, underlying physical illness, route, or coadministered anticholinergics with laryngospasm. Coadministered benzodiazepines showed a borderline association in the multivariate but not univariate analysis that was considered anomalous. CONCLUSIONS:This analysis of the largest available sample of ED ketamine-associated laryngospasm did not demonstrate evidence of association with age, dose, or other clinical factors. Such laryngospasm seems to be idiosyncratic, and accordingly, clinicians administering ketamine must be prepared for its rapid identication and management. Given no evidence that they decrease the risk of laryngospasm, coadministered anticholinergics seem unnecessary.

78

Cohort study vs. Survey study Cohort Study Purpose Example Sampling Generalizability Power/sample size Causal inference NHANES Convenient Narrow Common Survey Study survey NAMCS/NHAMCS Systema8c Broad Depend on sampling method Non-response bias sampling error 79

Feasibility/diculty Cohort retainment

Placing physician orders at triage: the effect on length of stay.

STUDY OBJECTIVE:Emergency department (ED) crowding is a signicant problem nationwide, and numerous strategies have been explored to decrease length of stay. Placing a physician in the triage area to rapidly disposition low-acuity patients and begin evaluations on more complex patients is one strategy that can be used to lessen the effect of ED crowding. The goal of this study is to assess the effect of order placement by a triage physician on length of stay for patients ultimately treated in a bed within the ED.METHODS:We conducted a preexperimental study with retrospective data to evaluate patients with and without triage physician orders at a single academic institution. A matched comparison was performed by pairing patients with the same orders and similar propensity scores. Propensity scores were calculated with demographic and triage data, chief complaint, and ED capacity on the patient's arrival.RESULTS:During the 23month study period, a total of 66,909 patients were sent to the waiting room after triage but still eventually spent time in an ED bed. A quarter of these patients (23%) had physician orders placed at triage. After a matched comparison, patients with triage orders had a 37-minute (95% condence interval 34 to 40 minutes) median decrease in time spent in an ED bed, with an 11-minute (95% condence interval 7 to 15 minutes) overall median increase in time until disposition

80

Comparison of nested studies Nested case-control Case may serve as control for previous case Individual may be a control at several 8mes Case-cohort Case in subcohort is control for all previous cases Case outside subcohort is never a control Non-cases in subcohort are controls for all 8mes at risk

Does not provide data to directly es8mate Provides data to es8mate the reference reference hazard hazard rate (subcohort, excluding non- subcohort cases) Specied outcome Time-dependent exposures need to be measured only at 8me individuals are selected Cannot obtain a popula8on risk factor prevalence es8mate Useful for dierent outcomes Time-dependent exposures need to be measured at all 8mes in subcohort Baseline sample allows es8ma8on of prevalence of risk factors 81

A cross-ethnic comparison on incidence of suicide in Taiwan.

BACKGROUND: Suicide rates vary widely across nations and ethnic groups. This study aims to explore potential factors contributing to inter-ethnic differences in suicide rates.

Method: Study subjects came from a psychological autopsy study conducted in Taiwan, including 116 consecutive suicides from two aboriginal groups and Taiwanese Han; 113 of them each matched with two living subjects. Gender-, age- and method-specic suicide rates, population attributable fraction (PAF) of suicide for ve major risk factors, help-seeking before suicide and emergency medical aid after suicide were compared between the three ethnic groups.

RESULTS: One aboriginal group (the Atayal) had signicantly higher adjusted rate ratios (RR) of suicide than the other aboriginal group (the Ami) [RR 0.20, 95% condence intervals (CI) 0.12-0.34] and the Han (RR 0.26, 95% CI 0.16-0.40). Such differences can be explained by higher PAFs of suicide for three major risk factors (substance dependence, PAF 47.6%, 95% CI 25.5-64.2; emotionally unstable personality disorder, PAF 52.7%, 95% CI 32.8-69.0; family history of suicidal behaviour, PAF 43.5%, 95% CI 23.2-60.2) in this group than in the other two groups. This higher suicide rate was substantially reduced from 68.2/100 000 per year to 9.1/100 000 per year, comparable with the other two groups, after stepwise removal of the effects of these three risk factors. Suicide rates by self-poisoning were also signicantly higher in this group than in the other two groups.

CONCLUSIONS: Higher rates of specic risk factors and use of highly lethal pesticides for suicide contributed to the higher suicide rate in one ethnic group in Taiwan. These ndings have implications for developing ethnicity-relevant suicide prevention strategies.

82


83

Validation of the Acute Heart Failure Index.

STUDY OBJECTIVE:Validate a clinical prediction rule prognostic of shortterm fatal and inpatient nonfatal outcomes for heart failure patients admitted through the emergency department.

METHODS:We retrospectively studied a random selection of 8,384 adult patients admitted to Pennsylvania hospitals in 2003 and 2004 with a diagnosis of heart failure as dened by primary discharge diagnosis codes. We reported the proportions of inpatient death, serious medical complications before discharge, and 30-day death in the patients identied as low risk by the prediction rule. RESULTS:The prediction rule classied 1,609 (19.2%) of the patients as low risk. Within this subgroup, there were 12 (0.7%; 95% condence interval [CI] 0.3% to 1.2%) inpatient deaths, 28 (1.7%; 95% CI 1.1% to 2.4%) patients survived to hospital discharge after a serious complication, and 47 (2.9%; 95% CI 2.1% to 3.7%) patients died within 30 days of the index hospitalization.

CONCLUSION:This prediction rule identies a group of admitted heart failure patients at low risk of inpatient mortal and nonmortal complications. Our validation ndings suggest the rule could assist physicians in making site-of-care decisions for this patient population and aid in analyzing presenting illness burden in study populations.

84


85

How common is MRSA in adult septic arthritis?

STUDY OBJECTIVE:We determine the proportion of methicillin-resistant Staphylococcus aureus (MRSA) in adult septic arthritis patients presenting to the emergency department (ED).

METHODS:This was a retrospective review in 2 urban academic EDs in northern California, one tertiary care and one public. Subjects included patients who underwent arthrocentesis in the ED from April 2006 through July 2007. We queried the microbiology laboratory databases for synovial uid cultures sent from the ED. We reviewed synovial uid culture results and corresponding synovial uid analyses and then classied positive culture results as true septic arthritis or likely contaminant. For septic arthritis cases, we reviewed medical records and abstracted presenting features. We report our ndings with descriptive statistics.

RESULTS:One hundred nine synovial uid cultures were sent from the EDs. Twentythree results (21%; 95% condence interval [CI] 14% to 30%) were positive, of which 9 were likely contaminants; 1 was from a soft tissue abscess and 1 was from bursitis. Of 12 septic arthritis cases, 6 cultures (50%; 95% CI 21% to 78%) grew MRSA, 4 (33%; 95% CI 7% to 60%) methicillin-susceptible S aureus, and 1 each (8%; 95% CI 0% to 24%) Streptococcus pneumoniae, Enterococcus faecalis, and Pseudomonas aeruginosa. Of the 6 MRSA cases, 4 were in male patients; median age of patients was 47.5 years, 3 patients had previously diseased joints, 2 patients injected drugs, 2 patients were febrile, 3 patients had previously diseased joints, median synovial uid leukocyte count was 15,184 cells/microL (range 3,400 to 34,075 cells/microL), and 5 patients received appropriate ED antibiotics.

86


87

HOW SHOULD I DESIGN MINE?

88

1027abstract 1.(1)(2) 2.:10:00~12:00; 10:00~12:00

1 101/8/14(W) 2 101/9/6(W) 3 101/10/4(W) 4 101/11/1(W) 5 101/12/6(W)

Proposal

6 102/1/3(W) 7 8 9 10 11

102/2/7(W) () 102/3/7(W) () 102/4/4(W) () 102/5/2(W) () 102/6/6(W) () paper

12 102/7/4(W) 13 102/8/1(W)

() () paper reference

(1) (2) (3)

89