Writing a high-impact manuscript can feel overwhelming. However, understanding common structures and methodologies can simplify the process. This allows you to focus on what truly matters: your research question and the methods you use to answer it.

This blog post will decode how clinical studies are structured so that you can adopt these approaches for your next paper.

The Struggle of Penning Down Science

During my first year of internal medicine residency, I faced the daunting task of structuring my first journal article. The challenge wasn’t the lack of data or findings; it was packaging those findings in the right way. Our approach was not structured, and we had too many results. This led to articulating those findings in a scientifically rigorous and compelling manner much harder. Hours turned into days, and days into weeks, as I grappled with structuring my thoughts, findings, and the overarching narrative. The blank document on my computer screen seemed to mock me, its blinking cursor a constant reminder of the words that eluded me.

It was during this period of struggle and introspection that I stumbled upon the works of Kurt Vonnegut. While Vonnegut might seem an unlikely muse for a clinical researcher, his insights into literary writing offered a fresh perspective that I desperately needed.

Vonnegut’s Revelation: The Hidden Structures of Narratives

Kurt Vonnegut, a literary giant, suggested that most narratives follow a few common patterns, deeply rooted in the human psyche. By plotting the protagonist’s fortune on a graph, Vonnegut revealed six primary story arcs:

1. Rags to Riches (a rising emotional arc)
2. Riches to Rags (a falling emotional arc)
3. Man in a Hole (a fall followed by a rise)
4. Icarus (a rise followed by a fall)
5. Cinderella (rise, fall, rise)
6. Oedipus (fall, rise, fall)

Drawing Parallels: The Blueprint in Clinical Research

Inspired by Vonnegut’s approach to literature, I began to wonder: could there be similar hidden structures in clinical research manuscripts? Could there be a blueprint that, if followed, could make the process of writing not just easier but also more impactful?

My ensuing journey into the world of clinical research manuscripts was revelatory. Just as Vonnegut had unearthed the hidden structures in literary writing, I began to discern patterns in clinical research manuscripts. These weren’t mere coincidences but were tried and tested structures that resonated with readers, reviewers, and editors alike.

I discovered that most clinical studies follow a similar approach: they begin with a descriptive analysis and then move into an analytic phase. In today’s post, we will explore these components and see how you can reverse-engineer successful studies to design your own.

What Makes a High-Impact Study

Delivering a well-structured clinical study can be a challenging task. That’s why I’ve spent the last several years perfecting a formula that helps streamline this process. With increasing access to comprehensive databases, future research will likely focus more on real-world data and less on controlled clinical trials. Therefore, I will focus on real-world observational studies here, taking one of my recent publications as an example.

Most successful observational studies have at least two parts: descriptive and analytic.

Let’s focus on each of these components step-by-step.

Step 1: Descriptive Analysis

The first part of any clinical study involves a descriptive analysis. This section sets the stage by describing the study participants and their characteristics. This is true regardless of the type of observational study: cross-sectional, case-control, or cohort-based study. Studies examining trends of a specific disease or condition over time also follow this structure where the first part is descriptive (e.g., studies from the National Inpatient Sample which describe trends in hospitalization for common diseases).

Example Study

Let’s examine a recent cohort study we published. In our study on diagnostic delay in psoriatic arthritis (PsA), we characterized a retrospective cohort of patients, detailing their demographics and clinical features.

• Population: The cohort included 162 patients with PsA from Olmsted County, Minnesota, with a mean age of 41.5 years and 46% female.
• Clinical Characteristics: Data collected included age at symptom onset, BMI, presence of enthesitis, and time to diagnosis.

Step 2: Analytic Phase

Once you have described your participants, the next step is to analyze differences and associations within your data. This often involves comparing groups within your cohort or against a control group (if controls are available). The analytic part of your study will typically use t-tests (or alternate methods depending on the type of data) or regression methods to explore these differences. A slightly complicated analysis is time-to-event analysis or survival analysis, which is also a type of regression method taking time into consideration.

The analytic phase is what often makes the study more interesting. Most early-stage researchers miss this part. For example, they only examine disease trends over time but do not compare it to anything. Yes, the trends in the past and present can themselves act as comparators but this is often not enough. You need a control group without the disease to show that the trend is not universal or due to some peculiarity of the data or something else that happened that year (e.g., most healthcare visits for other causes dropped during COVID). Similarly, in a cohort-study, it is optimal if we can find general population controls (without the disease of interest). If not, think if you can find two groups within the cohort that you can compare. e.g., male vs. females, with a certain comorbidity vs. no comorbidity, etc.

Example from the Study

• Analysis of Diagnostic Delay: In our cohort study on diagnostic delay in PsA, the two groups were those with diagnostic delay (≥2 years from symptom onset) vs. no diagnostic delay (<2 years from symptom onset). The study utilized logistic regression models to identify factors associated with delay in PsA diagnosis. We are still looking at the difference between the two groups with the logistic regression but we used the regression analysis to control for age and sex (making sure that the difference is not just due to the difference in age and sex between the groups).

Step 3: Additional things to consider

Simplifying Statistics

Many researchers tend to think that complicated methods are necessary to answer their research questions. While advanced techniques like clustering and causal inference have their place, most questions can be answered with simple statistical methods.

Remember, simplicity often trumps complexity in research.

While it might be tempting to use the latest sophisticated methods, straightforward statistical tools can be highly effective when the research question is precise and the data robust. This minimalist approach not only makes the study more accessible but also emphasizes the importance of the research question over the complexity of the methods.

Methods are the most common reason for studies getting desk rejected before even entering the review phase. Therefore, keeping it simple but appropriate to the study question is our best bet.

For descriptive analysis, simple statistics like means, medians, and percentages often suffice.

For the analytic phase, basic regression models are typically all you need. The key is to formulate a good research question and use the appropriate database and methods to address it.

A Compelling Narrative Around the Results

Imagine being a patient with PsA, experiencing years of pain and uncertainty before receiving a diagnosis. This emotional journey underscores the importance of timely diagnosis and the impact that research can have on patient lives. Rather than just reporting the data, the manuscript should capture the essence of this patient’s journey. See my post on how to tell compelling stories with your data here.

Practical Tips for Your Research

1. Focus on Your Research Question: Ensure your research question is clear and well-defined. This will guide your entire study design.
2. Select the Right Database: Choose a database that provides comprehensive and reliable data relevant to your research question.
3. Keep Statistics Simple: Use straightforward statistical methods to analyze your data. The complexity of the methods should align with your research needs.
4. Tell a Story: Pick 2-3 key findings, then identify a core message. This will allow you to build a strong narrative for your study.

By following a structured approach, you can decode the process of writing high-impact manuscripts. Start with a solid descriptive analysis. Proceed to an analytic phase using simple yet effective statistical tools. Always keep your research question at the forefront. This method not only simplifies the writing process but also enhances the clarity and impact of your research.

It might be tempting to use the next “revolutionary” machine learning or AI approach but don’t do it unless absolutely necessary.

This Week’s Action Step

Take a look at your current research project. Identify how you can apply a simplified statistical approach to both your descriptive and analytic phases. Revisit your research question. Ensure it is guiding your entire study design. By refining these elements, you’ll be well on your way to crafting a high-impact manuscript.

P.S.: If you are struggling to find or narrow down your research question, you can use our custom GPT to find research gaps in minutes.