5.7- Discussion

In this next module, we’re going to talk about the discussion section of your manuscript. The discussion section, in terms of the writing, gives you the most freedom of any of the other written sections. It gives you the most chance to put your good writing on display. Of course, since there is so much flexibility, it is also the most challenging to write. So follow your rules for good writing, all the rules that we’ve been talking about in this class, and you’ll really shine in your discussion section. Remember, that we were represented the introduction section as a cone that went from broad to narrow. In the discussion section, you are inverting that cone. The introduction section ends with the question asked. The discussion section starts by answering that question. So you start with the most narrow piece of information, and then you broaden. You start by answering the question asked, then you’re going to support your answer with your own data and with other people’s data from the literature. So you give all the lines of evidence, say how your results fit into the context of the literature. Then, you’re going to defend your conclusions. This is the obligatory limitations section of your discussion. You want to anticipate the criticisms that readers may have and defend your conclusions against these criticisms. And then at the end, you’re going to go very broad. You want to give the big-picture implications of your research. In other words, the discussion section should impart what your results mean and why should anyone care. This last point is important. You need to make sure that readers outside of your immediate niche area in science can understand why your study matters. You have to make them care. Here’s one way of organizing the discussion. It’s a bit disciplined specific exactly what goes in that discussion, but here are the general elements that you should have. You want to start the discussion section with something like, we found that. And then you answer the question that you proposed at the end of the introduction section. You’re going to explain what the data mean at a very high level, and you want to clearly and explicitly state if the findings are novel. You may also have some key secondary findings. Oftentimes, when we do a study to find one thing, we also find some other interesting findings in the process. So you might state those other key secondary findings. After you give all your findings, you’re going to put it on contexts. Here’s where you can get into some detail that I didn’t want you to put in your introduction section. You can give possible mechanisms or pathways that might explain what you’re seeing in your data. If the research was on human subjects, you might talk about the potential underlying biological pathways. You also compare your results with other people’s results. How do your results fit with the existing literature? Do your results confirm others’ results, or are they contrasting? Then, as you’re getting towards the end of your discussion, you need to have at least one paragraph on strengths and limitations. I’ll talk a little bit more about the limitations paragraph in a bit here. You also want to spend some time saying what’s next. Here’s where you can talk about results that need to be confirmed in future studies or unanswered questions or future directions. Then, you want to give this so what, the implicate, speculate or recommend. This is where you’re saying, hey, why should anyone care about my findings. If you’re doing something in the basic sciences, tie it to humans, tell your readers why they should care. And then, it’s nice in the discussion section to have one paragraph at the end where you’ll have a strong conclusion. Some journals have a separate section for a conclusion. But if your journal doesn’t, just wrap up your discussion section by restating your main findings and giving some kind of final take-home message for your readers. Here are some tips on the discussion section. Again, showcase your good writing, use the active voice, tell it as much as you can like a story. Start and end with the main finding. The first sentence of your discussion section should be something like, we found that, where you give an answer to the main question, aim, or hypothesis of the paper. And then you want to wrap up your discussion section by restating that main finding. It’s okay to repeat it because you want to emphasize it to your reader. Be very careful in your discussion section that you don’t travel too far from your data. Near the end of the discussion, it’s okay to give some speculation, big picture implication. It’s okay to step away from your data and speculate a bit towards the end. But when you’re drawing your main conclusions, you need to make sure that you are telling the reader what you actually found, not what you hoped to find. This is one of the reasons that I like to look at tables and figures first, before I read anybody’s discussion section. I want to make my own judgment about what I think the data show. Authors sometimes see only what they want to in their data. So make sure you’re not reaching too far from your data. Another common problem is that sometimes authors will start discussing things that they don’t have data on. So they’ll go on for paragraphs about something that they didn’t even measure in their study. So again, don’t travel too far from your data. Another key tip on discussion sections is pay attention to the limitation section. One major mistake authors make in the discussion section is that they don’t spend enough time crafting a good strengths and limitations section. They often just write whatever generic limitations they can think of. But the mark of a good paper, for me, is when authors have anticipated my criticisms and concerns. As I’m reading a paper, I’m thinking about all the potential holes in it. If I then get to the limitations section in that paper and the authors proceed to address all my major concerns, I’m impressed. Even if they don’t have a good answer for those limitations, just the fact that they recognize them goes a long way to improving my confidence in the study. That’s the mark of a good paper for me. Because what you see more often is that authors write the limitations sections too generically. They just throw in a bunch of none specific boiler plate limitations that could apply to any study. And they fail to acknowledge the limitations that are most critical to the validity of this specific study. If you can anticipate what the reader will criticize in your paper, if you can anticipate that and beat them to the punchline, that goes a long way to increasing reviewers’ and editors’ confidence in your study. And then finally, the last thing on the discussion section is that make sure your take-home message is clear and consistent. Again, I think this gets to that oftentimes we want the data to show one thing, we were hoping for a certain result. And the data don’t quite show what we want them to, but we still are torn. And we want to tell our reader that the data showed that. We still believe in the hypothesis, and we end up waffling. So make sure you’re clear, and consistent about your take-home message. All right, now, I’m going to give you some example discussion sections. So I’ve been using this Academic Spam Study, I’m going to use it again here. It’s quite short, the discussion section, but it follows the format of a typical discussion well, so it makes a good example. I’m just going to remind you that at the end of the introduction section of that paper, they gave the objectives of their study. In the Academic Spam Study, we investigated the amount, relevance, content, and suppressibility of academic spam emails. So I’m going to be looking, For answers to each of those questions about amount, relevance, content and suppressibility at the beginning of the discussion section. So here’s their discussion section, they start with the Academic Spam Study shows that mid-career academics in New Zealand receive, on average, 2.1 spam invitations each day to publish papers and attend conferences. So we learn that the amount is 2.1 invitations per day. If you keep reading along here you’ll also learn that 16% of spam invitations are duplicates they found. And then in terms of relevancy they found that 83% were of little or no relevance to the recipient. And in terms of supressibility, the authors found that these emails are hard to suppress. So they answered basically all of the questions that they set out to answer right there in the first paragraph of the discussion section. Then they jump into the strength and weaknesses of the study. I usually would put this a little bit closer to the end, but it’s okay. So the BMJ Christmas issue which they went right into some potential strengths and weaknesses. They’re trying to be a little bit funny here, I won’t read through it but I’ll leave it there for you if you want to. So we see our obligatory limitations paragraph. Then we get the context with other studies, they talk a little bit about comparison with other studies. Again they’re being a little tongue in cheek and a little funny here. I’ll leave that for you to read on your own. And then we get the final paragraph, implications and future research, the kinds of things I told you, you should have at the end of your discussion section. Again, they’re trying to be a bit funny, I’ll just read through it here. What they’re trying to do here is to suggest that there should be future research on academic spam and then they write it in the format. They write it to mimic or mock academic spam, so they are trying to sound like a spam email. Nobel and prestigious colleagues, we are enthralled by prospect to notice the grammar, of novel research focus of academic spam, so we make a proposition to improve enlightenment of evidence. We wish greatly to start journal and convene scientific meeting that focus on academic spam, so illustrious colleagues can form interdisciplinary webs, of scientific rigour to advance knowledge. Maybe we will christen soon Journal of Advances in Interdisciplinary Academic Spam and launch with alacrity the First Annual International Symposium. Once we identify publisher and conference organizer we will email academics to join this exciting novel adventure, honorable colleagues, stay tuned. Often in the spam there’s a lot of exclamation points like that. So they’re trying to be funny and sound like or copy a, make fun of email, spam email. So that’s the entire discussion section, it’s just four sections or four paragraphs long but it has all the elements of a typical discussion section. So we’ll now jump into a non BMJ Christmas issue example. So again, before I can show you the discussion section, I need to show you the end of the introduction from that same paper. So this was a study looking at low fat diets versus low carb diets. So it was a hot topic a few years back. They performed a study designed to test the hypothesis that severely obese subjects who have other problems, either diabetes or pre-diabetes with a, have a greater weight loss. And b, would have that weight loss without detrimental effects on heart health, while on a carb restricted diet compared with a low-fat diet. So they think that the carb diet might help with weight loss, but they also want to check and make sure that when you’re eating a lot of, remember this diet has a lot of steak and cheese. That when you’re eating a lot of steak and cheese that it’s not having some negative effects on overall heart health even if they lose weight. So those were the questions asked in the introduction section. So we are hoping that the beginning of the discussion section starts with answers to those questions. Indeed, we get to the discussion section. The first sentence of the first paragraph says, we found that severely obese subjects with a high prevalence of diabetes and pre-diabetes lost more weight in a six month period on a carb restricted diet than on a fat and calory restricted diet. So in that very first sentence, they’ve started with, we found that, and then they answer the question a, the first question in their introduction section. I am only going to highlight certain parts of the discussion, just because, of course, discussions are long, so I’m not going to present the entire discussion. But that’s the start of the first paragraph, the start of the second paragraph answers the question to b. The second main question of the study was whether the low carb diet is bad for your cardiovascular health because you’re eating too much steak and cheese. And they found that it wasn’t, subjects in the low carb group had greater decreases in triglyceride levels than did subjects in the low-fat group. And nondiabetic subjects on the low carb diet had greater increases in insulin sensitivity, that’s something that’s good. And subjects with diabetes had a greater improvement in glycemic control. So now they’ve answered question b. Not only did the low carb diet have no detrimental effects on heart health. The people on the low fat diet actually seem to have done better in terms of heart health. Their lipid levels are good, their insulin sensitivity is good. Again I’m going to just highlight a certain parts of this discussion and then go on which is the part I’m not going to show you that go on to put their findings in the context of previous studies. For time sake, I’m not going to go into those but I’m going to jump to paragraph 4, because I want to jump to that limitations paragraph to dishighlight that I think they did a good job here of pointing out important limitations. One thing is that many participants in this study were taking other drugs in addition to doing the weight loss program. They were taking medications to lower cholesterol and lower blood pressure. And this could definitely affect the results of the study. Obviously, that’s going to affect whether their lipids look bad or good. So the authors bring up that limitation, and they even tell us that they addressed that limitation by doing analysis where they excluded subjects on these medications and it appears that the results were robust. Down in the sixth paragraph they also tell us about another key limitation. Many participants dropped out, the attrition was high. Of course, if not many people could follow the diet, it’s really hard to draw conclusions. The diets just may be non-starters for a large number of people. So I think that was also an important limitation that they did address well. And this discussion section was seven paragraphs long, just to give you a sense of how long they might be. Here’s the last paragraph, paragraph seven. Notice that they restate their key findings in the last paragraph. They restate their answers to the two questions they were asking. Taken together, our findings demonstrate that this population lost more weight during six months on a carb-restricted diet than on a low fat diet. So that was answer to question a. The carb-restricted diet led to greater improvements in insulin sensitivity that were independent of weight loss, wasn’t just because they lost more weight. And a greater reduction in triglyceride levels in subjects who lost more than 5% of their weight. So they’ve just answered question b. They’ve just restated what they said in the opening two paragraphs. But you want to sum up, because that’s the really, that’s the big take home messages in the study that answers the questions asked. Then they go on to give some other big picture things, they give a caution. They want to say, this doesn’t mean that everybody should wash out and go on a low carb diet, because in the grand scheme of things, the weight loss was still pretty small even in the low carb group. Also, there was a high drop out rate which means the diet is hard to stick to, So based on all that, they end with a take home message to physicians. This study proves a principal and does not provide clinical guidance. Given the known benefits of fat restriction, future studies evaluating long-term cardiovascular outcomes are needed before a carb restricted diet can be endorsed. They don’t want to overstate their findings. They’re saying, there’s still some questions to be answered and we don’t want to translate our findings yet into an actual clinical recommendation. So they ended here with the restatement of there main findings, but then kind of the caution, a big picture caution. One more example, these was the paper comparing men and women with regard to self-citation. I’ll remind you that what was at the end of their introduction section, their goals were to compare the number of self-citations in men versus women and they also wanted to look a serve of secondary aim to trends in this gender gap over time. So, they ended the introduction with their statement of purpose what they are trying to do. So now, we’re hoping that they start the discussion section telling us what they found with regards to those two things. The discussion actually starts with a one sentence recap of the size of the study. I think they are trying to really emphasize the study’s major strength here. It was a much bigger study then anybody else has done. But then in the second sentence, they answer the questions asked. So one, in the last two decades for every seven self-citations by men, women cited themselves four times. A ratio of 1.7. So, how big was the gender gap? They found a gender gap of a ratio of 1.7 and what about the trends over time? This ratio rose sharply in the 1960s and the 1970s, and then sort of plateaued in the 1980s. So, they start their discussion with their main finding. I’ll just point out that they have some secondary findings. They were weeding through a big database here. They found some other interesting things. There was wide variation about self-citation in different fields. It turns out that almost 10% of all citations are self-citations, which I found interesting and surprising that’s I thought a big number. And then they give us the gender gap in terms of in absolute terms and in the statistics, given in a slightly different way. So, these are some secondary findings. So remember that sometimes after the primary finding, you’re going to give some important and interesting secondary findings, then the authors launch into possible mechanisms. Why might men self-promote more than women? That one possibility is that maybe self-promotions seems were acceptable for men than women in our society. So I’m not going to read through all of this, but here are some possible reasons that might explain the gender gap that the authors are seeing. Later, we get to the limitations. The main limitation cited is that the authors can’t say definitively why they’re seeing this difference. They’re pretty confident that the difference is real. They’ve looked at a big data set. They’re pretty confident that the finding is done well, but they can’t really say why. They can’t claim to know the causal mechanisms. They’re putting that as their big limitation and then the authors give several paragraphs of implications. I’m not going to put them all in the slide, but I’ll show you the last paragraph. So, the final paragraph ends with their main take home message. Historically, women’s academic contributions have been undervalued. And the fact that self-citations are more prominent in men than women, maybe one more reason to think that maybe metrics that involve citations shouldn’t be the only way that we’re measuring scientific impact. So this is really giving sort of the big picture implication of their work if self-citations are, and self-promotion if men are better at self-promotion and do more self-citation. Maybe we should be looking at metrics other than citations when it comes to academic hiring and promotion. So, maybe giving less weight to citations could be one strategy for reducing gender equity in the academic community. So, they’re ending here with the big picture. One last thing I want to just point out, I was reading a discussion once and this was the first two sentences or three sentences of the discussion section. So, don’t start your discussion like this. It started with this meta-analysis is subject to a number of limitations. So, don’t start your discussion section with the limitations. Start with the we found or this study shows and then bury the limitation section several paragraphs down. Finally, similar to methods and results and introduction, you’re going to use the rule for verb tenses. You’re going to use the past tense when referring to things that are already completed, study details, results, analyses and background research. So we found that, subjects may have experienced, Miller et al found. But you’re going to use the present tense when talking about what the data suggest, because the data are still suggesting that. So the greater weight loss suggests. The explanation for this difference is not clear. Potential explanations include. All of those would be in the present tense.