1 A Writer’s Algorithm or Papers Without (too much) PainUZ-UCSF Collaborative Research Programme Writing Workshop July 2017 Amy J. Markowitz, JD
2 Today's agenda A framework for: The Writer’s Algorithm- Presenting research Teaching writing skills The Writer’s Algorithm The basics of good writing habits The sections of a paper The basics of effective self-editing Managing Reviews
3 Today’s style Please interrupt frequentlyShare with others - research and writing are not solitary pursuits - at least not mostly
4 Most research involves a simple findingA > B Weight of experimental mice > Weight of control mice Bone density with TT genotype > Bone density with tt genotype Survival after surgery > Survival with medical therapy Health care in UK > Health care in US
5 THE ONLY RULE KNOW YOUR DATA
6 Is Your Data? NEW? TRUE? IMPORTANT?
7 Who cares? Who is your audience? Which journals are “reach” schools?Where have you published before? - Have you served as a reviewer?
8 Choosing a journal Strategy 1 Strategy 2Ask senior colleagues for likely, possible, and “reach schools” Send to one of the “possibles” Send to a “likely” if in a hurry Strategy 2 Find the most equivalent published article Aim a notch lower? Use JANE (Journal/Author Name Estimator
9 “Effects of Hypoxia on Kv1.5 Channels.Manuscript Title Based on the research question Interesting (catchy), dynamic and conclusive, rather than descriptive* “Hypoxia Inhibits Kv1.5 Channels in Rat Pulmonary Artery Smooth Muscle Cells” is preferable to “Effects of Hypoxia on Kv1.5 Channels. Remember the magic words Randomized, blinded, prospective, etc. * Except JAMA and NEJM (!)
10 Getting Started on the TitleState the research question Is A associated with B in population C? Is tai chi associated with falls in older women? Draft a title Effect of tai chi on risk of falls in older women Use clear, precise descriptions
11 Improving a Boring TitleState the main finding Older women randomized to tai chi had a lower risk of falls and better quality of life than those randomized to a wait list Add interesting design features Practicing tai chi reduces risk of falls in sedentary older women. A randomized trial
12 Polishing a Good Title Make it interesting Don’t go overboardTai chi prevents falls and improves quality of life in sedentary elderly women. A randomized trial Don’t go overboard Kick-butt older women don’t fall: A randomized trial
13 Title Checklist Are the title and the research question closely related? Is the title objective in tone? Are special features of the study mentioned? (cross-over, prospective, randomized, special populations) Illustrates the elements of the Title check list in Warren’s book: Are the title and the RQ closely related? Is the title objective in tone? Are special features of the study mentioned? (cross-over, prospective, randomized, special populations)
14 Good Title Tai chi prevents falls and improves quality of life in sedentary elderly women. A randomized trial Illustrates the elements of the Title check list in Warren’s book: Are the title and the RQ closely related? Is the title objective in tone? Are special features of the study mentioned? (cross-over, prospective, randomized, special populations)
15 The 4 basic parts of an abstract, paper, or presentationIntroduction: Why would it matter if you could show that A > B? Methods: How you will show that A > B. (Effect size: Comparing A with B) Results: Show that A > B. Discussion: What is the implication, now that we know that A >B?
16 Begin Before the BeginningScribble or type a list of topics, themes, ideas, conclusions, in any order Work for about 15 minutes and then reward yourself with a latte, and a quick peek at Coke on the Beat! Or Ruvheneko** **Must know what the hip kids are watching!
17 Create a Scaffold Using the Instructions for Authors contained on the Web site of every journal, set up the major headings/sections of the paper You are now not looking at a blank screen and can treat yourself to a snack and another latte – or a unicorn frapp (blech)!
18 Put on the Sorting Hat Insert fragments from the scribbled list into the scaffolding sections, eg, background? result? discussion? Pen a meaningful topic sentence for the fragments. Note: meaningful = an original idea that sets up the issue to be discussed in that section or paragraph Continue to fill in the space under the topic sentences by moving entries around, and by adding entries from the scribbled list Open Endnote or other reference library and troll around
19 Put on the Sorting Hat (continued)Note ideas for tables, boxes, figures Re-check rules for authors as to formatting requirements Note areas that require further thought or discussion Go for a run or a bike ride
20 Write an Introduction Do not reinvent the wheel - go back to the grant, proposal, RFP for the RESEARCH QUESTION Content: The introduction is your promise to the reader Use a writing resource, style manual, dictionary, grammar guide American Medical Association Style Manual Merriam Webster online dictionary Stedman’s Medical Spellchecker and dictionaries Nuts and Bolts of Scientific Writing- Constance Baldwin, PhD
21 Content of IntroductionThe introduction is your promise to the reader (in 3 paragraphs or less - PRESENT TENSE FOR ESTABLISHED KNOWLEDGE) Describe: Background, the raison d’etre of the study (why the problem was compelling) State of the field/ Gap in the knowledge (relevant literature to date) The reason your findings will be relevant, and (if you’re feeling brave) the contribution you have made Close with a “road map” of what the reviewer/reader will find in the paper: Hypothesis, Design, Sample, Methods
22 JAMA. 2016;315(19): doi: /jama BACKGROUND In 2010, it was estimated that nearly half of US adults near the end of life were unable to make decisions for themselves about whether to accept life-prolonging technologies. Family members or other individuals are asked to serve as surrogate decision makers for these often difficult decisions. To effectively function in this role, surrogates require a clear understanding of the likely outcomes of treatment. Numerous studies over the last decades indicate that surrogates of patients with advanced illness often have optimistic expectations about prognosis. This is problematic because optimistic expectations are associated with more use of invasive treatments in dying patients and delayed integration of palliative care. Clinicians cite unrealistic expectations by surrogates as one of the most important barriers to high-quality end-of-life care in seriously ill patients.
23 Gap in the Knowledge Although misperceptions about prognosis by surrogates are well documented, an important barrier to progress is the lack of a clear understanding of the causes of these misperceptions. … Several small qualitative and simulation-based studies among surrogates in intensive care units (ICUs) suggest that such considerations may contribute to physician/surrogate discordance about prognosis, but these hypotheses have not been empirically validated in surrogates making actual decisions for incapacitated patients. Without clarity about factors that contribute to discordance, it will be difficult to design interventions targeting the key barriers to effective conversations about prognosis.
24 Road Map We therefore sought to quantitatively determine the prevalence of physician-surrogate discordance about prognosis and to qualitatively understand reasons that surrogates sometimes hold beliefs about prognosis that differ from physicians’ assessments.
25 Introduction: The Final TestAfter reading it, could someone not familiar with the field understand… Why you did the study How it advances the current state of the evidence
26 Improve the IntroductionFocus on the big picture – not too much detail Don’t state the obvious – “death is bad” Summarize prior research – don’t describe studies one-by-one Emphasize problems with prior research that your study addresses Be objective – avoid hostility and overstatement Avoid jargon and acronyms Save the detail for the discussion Don’t state the obvious – death is bad, CVD is the most common cause of death among men in the US, etc. Summarize – too often authors say “A cohort study by xxx showed yyy”. A second case-control study by xxx showed yyy. State the problems with prior research in a way that seamlessly leads to the strengths of your study design – emphasize the problems that your design will address. Use objective language (avoid words like bad, poor, meaningless, etc.), emphasize the problem with prior studies, but don’t be too hostile (especially to specific investigators who might wind up reviewing your manuscript), and don’t overstate the greatness of your own study design. The introduction has to suck in your readers. Jargon and acronyms are off-putting to most readers and defeat your purpose.
27 Shorten the IntroductionDo you have more than 4 paragraphs? Is some of the material tangential or extraneous? Did you describe prior research in too much detail? Is the description of your study more than 2 sentences? Brevity is a major virtue.
28 Write the Easy Parts FirstGood bets for knocking off sections are the Methodology and Results sections Methods: Carefully track the research protocol, and if you repeat or reproduce a part of the protocol as stated in your original proposal, do not paraphrase or change verbiage (PAST TENSE FOR WHAT YOU DID)
29 Methods: How will you show that A > B?Who (what) did you study? What, if anything, did you do to them? How did you make your measurements? How did you compare A with B? Statistical tests to “show” that A ≠ B
30 Methods Checklist: 4 ElementsDesign: Define the type, e.g., retrospective, case-control, RCT, prospective Subjects: population, inclusion/exclusion criteria, controls Measurements: survey instruments, assays, physical measurements Analysis: Statistical plan and rationale
31 Methods: How to make sadzaPut mealie-meal in pot. Add cold water to make a paste. ... Keep stirring until rakukwata. Cover pot, reduce heat and let it simmer for 15 minutes. After 15 minutes, add more mealie-meal bit by bit and mona sadza. Be sure to do it well. PREDICTORS, OR INDEPENDENT VARIABLES COME BEFORE OUTCOME VARIABLES OR DEPENDENT VARIABLES INDEPENDENT VARIABLES ANSWER THE QUESTION "WHAT DO I CHANGE?". - DEPENDENT VARIABLES ANSWER THE QUESTION "WHAT DO I OBSERVE?". - CONTROLLED VARIABLES ANSWER THE QUESTION "WHAT DO I KEEP THE SAME?"." HOW DID YOU MEASURE THE PREDICTOR VARIABLES? CHART REVIEW? SELF-REPORT VIA QUESTIONNAIRE? IN A CLINICAL TRIAL- INCLUDE A SEPARATE PARAGRAPH DESCRIBING THE INTERVENTION AND CONTROL ARMS, THEIR ASSIGNMENT NEXT, DESCRIBE HOW YOU MEASURED OUTCOMES, AND YOUR CRITERIA FOR DEFINING THE OUTCOMES- SEVERITY, CHRON ORDER OF APPEARANCE, ETC. LEAVE OUT OBVIOUS DETAIL, OR DETAILS THAT ARE EASILY INFERRED, EG, DON’T NEED THE N OF BOTH SEXES, ONLY 1 TEST YOUR “APPROPRIATE” LEVEL OF DETAIL BY IMAGINING HOW YOU WILL WORD A LIMITATIONS SECTION IN THE RESULTING PAPER
32 Methods: How to make sadzaMix 225g of cornmeal with 250ml of water. Bring 750ml of water to the boil in a pan, then turn down the heat and add the cornmeal mixture, stirring all the time. Cook for five minutes, then gradually add the remaining cornmeal. Transfer to a bowl or serve in pan. PREDICTORS, OR INDEPENDENT VARIABLES COME BEFORE OUTCOME VARIABLES OR DEPENDENT VARIABLES INDEPENDENT VARIABLES ANSWER THE QUESTION "WHAT DO I CHANGE?". - DEPENDENT VARIABLES ANSWER THE QUESTION "WHAT DO I OBSERVE?". - CONTROLLED VARIABLES ANSWER THE QUESTION "WHAT DO I KEEP THE SAME?"." HOW DID YOU MEASURE THE PREDICTOR VARIABLES? CHART REVIEW? SELF-REPORT VIA QUESTIONNAIRE? IN A CLINICAL TRIAL- INCLUDE A SEPARATE PARAGRAPH DESCRIBING THE INTERVENTION AND CONTROL ARMS, THEIR ASSIGNMENT NEXT, DESCRIBE HOW YOU MEASURED OUTCOMES, AND YOUR CRITERIA FOR DEFINING THE OUTCOMES- SEVERITY, CHRON ORDER OF APPEARANCE, ETC. LEAVE OUT OBVIOUS DETAIL, OR DETAILS THAT ARE EASILY INFERRED, EG, DON’T NEED THE N OF BOTH SEXES, ONLY 1 TEST YOUR “APPROPRIATE” LEVEL OF DETAIL BY IMAGINING HOW YOU WILL WORD A LIMITATIONS SECTION IN THE RESULTING PAPER
33 Measurements: A Way to OrganizePredictors before outcomes Medical presentation History, physical, simple lab, complex matters Explain odd decisions or missing data “Appropriate” level of detail - know your audience PREDICTORS, OR INDEPENDENT VARIABLES COME BEFORE OUTCOME VARIABLES OR DEPENDENT VARIABLES INDEPENDENT VARIABLES ANSWER THE QUESTION "WHAT DO I CHANGE?". - DEPENDENT VARIABLES ANSWER THE QUESTION "WHAT DO I OBSERVE?". - CONTROLLED VARIABLES ANSWER THE QUESTION "WHAT DO I KEEP THE SAME?"." HOW DID YOU MEASURE THE PREDICTOR VARIABLES? CHART REVIEW? SELF-REPORT VIA QUESTIONNAIRE? IN A CLINICAL TRIAL- INCLUDE A SEPARATE PARAGRAPH DESCRIBING THE INTERVENTION AND CONTROL ARMS, THEIR ASSIGNMENT NEXT, DESCRIBE HOW YOU MEASURED OUTCOMES, AND YOUR CRITERIA FOR DEFINING THE OUTCOMES- SEVERITY, CHRON ORDER OF APPEARANCE, ETC. LEAVE OUT OBVIOUS DETAIL, OR DETAILS THAT ARE EASILY INFERRED, EG, DON’T NEED THE N OF BOTH SEXES, ONLY 1 TEST YOUR “APPROPRIATE” LEVEL OF DETAIL BY IMAGINING HOW YOU WILL WORD A LIMITATIONS SECTION IN THE RESULTING PAPER
34 Analysis Don’t list tests: explain why you used themTell how measurements became variables How did you estimate the effect size? How did you determine the precision and significance of the effect size? Univariate Multivariate (say what you adjusted for) AT THE TIME OF SUBMISSION, YOU MAY STILL BE IN ANALYSIS PHASE YOU CAN STILL DESCRIBE YOUR ANALYTIC PROCESS MAKE SURE THAT THE EFFECT SIZE CORRESPONDS TO YOUR HYPOTHESIS - IF YOU’RE COMPARING A PERCENTAGE CHANGE IN AN OBJECTIVE VARIABLE (BLOOD PRESSURE) ON THERAPY A VS THERAPY B, THE EFFECT SIZE IS THE MEAN DIFFERENCE IN THE PERCENTAGE CHANGE BETWEEN THE GROUPS AND YOUR EXPLANATION OF “PERCENTAGE CHANGE” DISCUSS ADJUSTMENTS THAT YOU MADE, AND BE CONSISTENT THROUGHOUT THE PIECE IN DESCRIBING
35 Results: Showing that A > BMake sure the main result is obvious Don’t bury it in the middle of a long paragraph, an 8 x 6 table, or a complex figure Use alternative “definitions” of A and B Different measurement techniques or times “Multivariate” adjustment In various subgroups
36 Results, continued Order of presentation: Collected sample firstFollow order of hypothesis, chronology or design elements Simple results before complex Strongest findings first Use subsection headings as a roadmap
37 Results (leading to Discussion) (Pull out the scribbled list again)Framing the Content (GENERALLY, PAST TENSE FOR WHAT YOU FOUND) What grabbed you about your results? Was there an expected or unexpected finding? If you are presenting something new, build the case in a logical order – eg, is this study the result of a long line of similar research that is “confirmatory, but”? Is it presenting a new theory to explain an old phenomenon? Is it rebutting a long-held belief in the field? Does it have implications for research policy or social policy? Will it be a useful “tear-out” with pragmatic clinical utility?
38 Pocock SJ, Stone GW. (2016) The Primary Outcome Fails — What NextPocock SJ, Stone GW. (2016) The Primary Outcome Fails — What Next? N Engl J Med. 375:971-9.
39 Peacock SJ, Stone GW. (2016) The Primary Outcome Fails — What NextPeacock SJ, Stone GW. (2016) The Primary Outcome Fails — What Next? N Engl J Med. 375:
40 Discussion Content Answer the Research Question in Sentence 1 of Paragraph 1 Highlight key findings as they relate to the study’s purpose (what can you say about your hypothesis/null?) Evaluate findings in relation to literature Discuss limitations Conclude with recommendations
41 Remind them of the question; Signal that it has been answeredTHE QUESTION POSED IN THE INTRO: Clinicians cite unrealistic expectations by surrogates as one of the most important barriers to high-quality end-of-life care in seriously ill patients. … Although misperceptions about prognosis by surrogates are well documented, an important barrier to progress is the lack of a clear understanding of the causes of these misperceptions.
42 The Answer THE ANSWER THAT OPENS THE DISCUSSION:Among a diverse cohort of critically ill patients at high risk of death, we found a high prevalence of physician-surrogate discordance about prognosis. The discordance was associated with both misunderstandings by surrogates of physicians’ assessment of prognosis and by surrogates holding different beliefs about a patient’s prognosis compared with what they perceived to be the physician’s assessment of prognosis.
43 Styling Your DiscussionUsing the topic sentences, write stand-alone paragraphs following the “I.A.C.” rubric: Idea (the topic sentence) Analysis (the clinical, microbiologic, biochemical, social, economic, explanation of the result) Concluding sentence which sums up the analysis, and often will serve as a transition to the next paragraph
44 Segues About Context and Other Interesting ThingsThese results are consistent with… Our result suggest… We believe our findings… Why might our results differ from…? We made several other observations…
45 Segues About Context and Other Interesting ThingsFrom White et al: The pattern of our findings fits with insights from social psychology about biases in individuals’ risk perceptions. A noteworthy finding from our semistructured interviews is that…
46 Hammer the ImplicationsFrom White et al: There are at least 2 clinical implications of our findings. First, given the high rates of discordance about prognosis, clinicians communicating with surrogates of patients with advanced critical illness should… . Second, when clinicians recognize that surrogates’ expectations about prognosis diverge from their own, they should…
47 Anticipate Possible CriticismCareful, not defensive, explanation Anticipate critique of your methodology or study design and present the reasoning behind your choices Your design and study criteria were well thought out in the beginning – now is not the time to have a crisis of confidence
48 Play the Strengths against the LimitationsThis study has several strengths. First, we used a mixed methods approach…We validated our qualitative findings through…The study was conducted in a large and diverse cohort. Additionally we interviewed actual surrogates facing actual decisions rather than using simulation, which established the clinical relevance of our findings.
49 Set Up Each Limitation This study also has several limitations. Although we recruited a diverse cohort, our sample was limited to one region of the country…
50 Conclude With a Send-offA conclusion is not a repetition Take the bully pulpit, and set a course Set a research agenda; get others interested in your field Create some controversy that is well-founded on the basis of your findings
51 Some Practical Advice Return to the Instructions for Authors and re-check formatting requirements, word length, formatting of references, suggested number of references, advice about graphics, the works Print hard copy of the manuscript, and proof it for substance by reading it aloud once, making hard copy corrections (you will be amazed at what you will find to self-edit) Then, and only then, run spell check Wait a day, re-read, and with a sigh of relief, hit the send key to your co-authors, or friendly readers
52 Credit Where It I$ Due "This project was supported by Grant Number KL2 RR from the National Center for Research Resources (NCRR).”
53 Responding to Reviews, Internal and External
54 The Hurt Locker Read through the reviews twiceYou will be offended by everything the first time, and begin to appreciate some of the merits by the second time Consult with your co-authors
55 Why an initial rejection?T not true R not relevant A not appropriate I not interesting N not novel
56 Take a step-wise approachBegin the explanatory Response to Editor letter simultaneously with your revisions – YOU ARE WRITING TO THE EDITOR, NOT THE REVIEWERS Tone = respectful, but not obsequious Address each comment, in numeric order, citing page and line; as relevant, add the actual text to the letter once finalized in the manuscript Where logical, group comments about the same issue to address efficiently
57
58 What if the reviewer has completely missed the point?Consider whether: You have presented the idea abstrusely; try rewriting unless this undermines the integrity of the idea The reviewer (generally an expert in the field, wed to their point of view) a) has a vested interest in your being incorrect, b) has just been proven wrong by your results, or c) did not read the paper carefully If you believe the reviewer is incorrect, write an evidence-based “brief” to the Editor explaining why you decline to revise in accord with the reviewer’s point In any of the cases of a), b), or c) you will nonetheless need to explain to the journal editors why it is that you have chosen not to revise in accordance with the reviewers comment, and be prepared to support your point of view
59 Re-consult Check your intra-reviewer-rater reliabilityWhat if MOM and DAD disagree???? Get your gumption up, be thoughtful, and make a decision - you must resubmit - now’s the time…
60 Greg’s Reviewer’s Misread
61 Original Figure 3 and Legend
62 Abundance of Caution
63 Response and Lessons LearnedDealing with easy fixes, eg, adding the “n=“ Dealing with the complimentary bombshell
64 Shave and a Haircut Address stylistic editorial comments after the substantive revisions Stylistic issues frequently relate to length Reduce to tabular or graphic demographic descriptions of subjects, or less intrinsic data and descriptors Do not repeat in text what is best presented in a table or figure Give your co-authors one last shot, WITH A DEADLINE, then: pull the trigger You will be able to make some headway in reducing bloat, but it is likely more efficient to ask a colleague (or professional editor) to give the manuscript a trim.