Getting Past Speed Bumps & Road Blocks

Writer’s Roadmap:

Getting Past Speed Bumps & Road Blocks

Sally Rigler, MD, MPH

Office of Scholarly, Academic and Research Mentoring (OSARM)

Department of Internal Medicine

August 24, 2012

STAGE 1: Developing Your Writing Plan

Do you have data from a research project or QI project?

A B

Situation A: No data yet for a data-driven article:

Take a few minutes and jot down at least three ideas that you are interested in enough to spend some time really getting to know the literature in this area. Do you have a relevant case that comes to mind?

1)

2)

3)

Who is a faculty member with content expertise and interest who could provide some guidance and be a co-author with you?

List three journals that might be low-hanging fruit for a case report or narrative review in your area:

1)

2)

3)

Situation B: Yes, you have data for a quantitative data-driven article

1. START WRITING!

1. GET A FIRST DRAFT GOING ASAP: IT DOESN’T TO BE PERFECT

1. Draft your section headers and drop basic thoughts quickly into these sections; you know more than you think you do so just START!

Background

Methods

Results

Discussion

Tables and Figures

References

------

1. Now, take those basic sections and see if you have some obvious sub-sections of thinking inside them that can be paragraphs:

Background: “CAN YOU EXPLAIN IT TO YOUR MOTHER?”

THOUGHT #1: Condition A bad for you (painful, kills people, etc.) or it costs society a lot of money or is a marker of poor quality care etc..

THOUGHT #2: Treatment B has been proposed to help Condition A based on xyz data but results are mixed. Whatever explanation you need for why more study is needed. Think ‘equipoise’—Think “It is unknown whether….”

THOUGHT #3: We now have the opportunity to studyTreatment B in some relevant population or context to address this lack of knowledge.

THOUGHT #4: Thus, our goals were toe.g…. characterize the safety of treatment B in hospitalized inpatients… determine whether Treatment B is associated with reduced symptom burden in Condition A… determine whether a new EMR alert is associated with better consistency in delivery of B to all persons with Condition A….determine levels of patient satisfaction with the delivery mechanism and tolerability of Treatment B.

Methods:

Consider the bigger picture first, and then you can parse it out.

(e.g. We studied pre-post measurements before and after a new type of breathing treatment to delivery caffeine instantly to sleep-deprived junior faculty members? We used a new intervention using motivational interviewing to improve exercise adherence in fat fertile forty-yearold female faculty? We conducted a cross-sectional survey of job satisfaction among hospitalists who do and do not work with residents. We studied the trajectory of quarterly dyspnea measures in persons with PAH over a 2 year follow up period. We conducted a retrospective review of appendectomy outcomes from 1882 to 1889 in an urban safety net teaching hospital.)

From what population did you sample?

(You must bear in mind who was NOT in your sample)

How did you find them, where did they come from?

Who did you include/exclude?

What did you do and when did you do it?

Often, your study is more descriptive and you didn’t actually DO anything to anyone. Think “we characterized….we described…we determine what factors were associated with ….”

What did you measure?

What measures did you gather?

Are they based on validated methods for capturing that ‘thing,’ whatever it may be? How did you gather these measures? Who gathered them? Are the gathering methods reproducible? (esp for chart abstracted judgment-calls)

How did you analyze the data?

Biostatistician will help you understand this.

Descriptive statistics?

Bivariate statistics?

Multivariable models?

What test statistics were used?

Before you leave Methods, add your comments about IRB approval or and stats packages.

Results:First: who ended up in your sample.

Who was excluded and why? How many folks did you have to screen to find the ones you finally could recruit from? Of those you recruited, how many agreed and were enrolled? How many of those dropped out and why? (LOOK FOR FLOW DIAGRAM SAMPLES IN THE CONSORT STATEMENT FOR RCTS)

Descriptive statistics on who you studied:

Often, Table 1 is “Demographic and Clinical Characteristics of Study Participants”

Second, what was associated with what in bivariate analyses?

Third, what remained associated with what, when everything was accounted for all at once in multivariable models? Usually another Table.

Your study may have entirely different goals, though, so this is just an example for usual kinds of things. You might have time-to-event figures, or other things to display.

Discussion:

First paragraph: Our goal was to determine XYX. We found that MAIN PUNCHLINE GOES HERE WITHOUT NUMERIC DETAILS.

Next, expand a littleto focus on things you want to be sure the reader takes away. “Specifically, we found that……blah blah blah, as shown in Table 1 where the row precentages show the degree of aerosolized caffeine uptake when stratified by whether the faculty member is a usual tea drinker or coffee drinker.” Here you may repeat some numeric values and statistical significance levels, but only for selected important items.

You will refer to your tables and help the reader understand how they are to read the table easily, but DO NOT REPEAT EVERYTHING IN THE TABLE. Otherwise, there is no point in having a Table and it can just go solely in the text.

NEW THOUGHT: How do your results fits in with those already reported in the literature in this field?

“These results confirm those of Dr. Seuss et al who reported blah blahblah….” Or “These results stand in stark contrast to those of previous investigators who found that faculty members tolerated aerosolized caffeine well. Indeed, we found a 95% mortality rate among junior faculty members participating in this trial.

NEW THOUGHT: What are the limitations of your study?

“Our results apply only to junior faculty physicians in a university health center and may not apply to those in private practice settings. In addition, our study time-frame inadvertently overlapped with implementation of a new EMR which increased catechol levels among all study subjects. Indeed, the substantial increase in serum catecholamine observed in the placeborecipients suggests that all subjects were under increased stress. An interaction between EMR implementation and excess mortality cannot be ruled out but could not be determined from this study design.”

FINAL CONCLUSION: Brief paragraph just to summarize how the punch line fits into the big picture. What should be done different, if anything, knowing this information? Tailor this to the audience for the journal.

1. REFERENCES

Tip: Add enough of a trigger of a reference to remember what paper it is, as you go. E.g. (Jones et al, 2010, JAMA, benzo in hip fx) Better yet, add the citation in as you write. It is very hard to go back later and remember what thought went with what citation later on.

GET ENDNOTE ON YOUR COMPUTER HERE HOME!

But keep only one single endnote file for each topic or you will drive yourself absolutely batty if you have different versions going on different computers. I speak from experience. This is the reason my hair is gray. Really.

1. GO BACK NOW AND POLISH YOUR PARAGRAPHS:

Once you have ‘dropped’ all your thoughts into bullets or messy sentences into paragraphs, go back and clean them up. The punch-line goes first, and then supporting information follows. It can feel very counter-intuitive because it is not deductive in sequence.

The way you reason: “The heart pumps blood through the body, and without adequate delivery of blood, the tissues become starved for oxygen, and if that happens, organs begin to fail, and if that happens, you cannot survive. Hence, you need a reasonably functioning heart to survive.”

The way you need to write it: “The health of the heart is critical to all other functions. Blood flow to keep tissues functioning is necessary. If this blood flow becomes inadequate, organs will not receive adequate oxygen and may begin to fail. Ultimately, death may occur.”

Duh. But put the ‘Hence’ thought FIRST, and then add the supporting thoughts AFTERWARDS. It may feel quite backwards to you initially until you get used to it.

Tip: If in doubt about a paragraph being too long, it is. Break it up. Later, you may be able to cut out more thoughts and wordiness, and re-consolidate.

1. Next: An iterative process of getting feedback from your mentors & improving your draft, word-smithing and shortening it.

You will incorporate some suggestions and comments. Adjudicating between differingadvice can be tricky…

1. Write the abstract last after you have a really concise and very solid understanding of what your ‘punch line’ is going to be

(This is just my personal opinion…some folks start with it first)

1. Format for the journal you have selected near the end of the writing process, using Instructions to Authors. Watch word limits and limits on how many tables and figures you can have.

DO NOT BLOW OFF THE DETAILS IN THE INSTRUCTIONS TO AUTHORS—THEY MEAN THEM!!! ALL OF THEM!!!

If they require formatting guidelines, use them:

The Equator Network has all the formatting guidelines for all types of papers:

1. Get final ‘OK’ from all co-authors before you submit! Decide issues of order of authorship based on contribution and standard rules of authorship. You will need signatures and contact information for all authors.

1. Submit! Use a cover letter to explain briefly why your paper is important. You may be allowed to suggest potential reviewers.
   
2. What happens next?