How Tableau helps spot inflammation in diabetes patients — before it gets serious
When most people think about diabetes, they reflect on blood sugar. But there's a side that doesn't get talked about nearly as much, how it quietly makes the body more vulnerable to inflammation.
I've spent time working with patient health data, and one thing that kept coming up was how much useful information was just sitting there, distributed across different columns in a spreadsheet, impossible to act on quickly. This post is about how bringing that data together and visualizing it in Tableau can actually help clinicians catch problems earlier.
A Review: Diabetes and Infection
In diabetes, how well the immune system functions. The white blood cells that normally fight off bacteria don't work as effectively. So, diabetic patients are more likely to get infections, things like urinary tract infections, skin infections, foot wounds that won't heal, and respiratory illnesses, and they tend to take longer to recover from them.(5)
Research shows the risk of infection-related hospitalization is roughly 2 to 4 times higher in people with diabetes compared to those without it. (1,3) That's a significant difference, and it shows up clearly once you start looking at the data the right way.
We have data everywhere in healthcare settings to identify at-risk patients. The challenge is being able to see it clearly and quickly enough to do something about it.
The dataset behind the study
The analysis took a reference of well-known public dataset, the UCI Diabetes over 100 US Hospitals dataset, which covers over 100,000 patient encounters across US hospitals from 1999 to 2008.⁴ It includes things like blood sugar control levels, lab results, age, how many times a patient was admitted, and what medications they were on—enormous data, but noisy in its raw form.
Data Cleaning: the understated part that matters most
Before any chart gets made, the raw data has to be tidied up. This is where most of the actual work happens, and it's easy to underestimate how much difference it makes. Here's the basic flow:
Check the data — Find missing values, duplicates, and fields that don't make sense
Fix formats — Make sure dates, numbers, and categories are consistent throughout
Fill gaps carefully — Missing BMI or lab values filled in using reasonable estimates
Build new columns — Create an infection flag and a simple risk score from the cleaned fields.
And then what Data reveals
Patient Risk Distribution by HbA1c Category
Once everything is cleaned and organized, two patterns stand out immediately.
Source: UCI Diabetes 130-US Hospitals dataset.(4)
This chart illustrates the distribution of patients across HbA1c (blood sugar control) categories and their associated risk levels. Patients classified in the Very High HbA1c category showed the highest risk score (45.18%), indicating a greater likelihood of adverse clinical outcomes. The Borderline category also demonstrated elevated risk (46.52%), followed by Poorly Managed (44.15%). Patients with Well-Controlled HbA1c levels had the lowest observed risk (41.70%).
Here, Data Analytics turned and visualized numbers clearly
Tableau comes in. Once the cleaned data goes into Tableau, you can build a dashboard that lets a clinician answer the question "who needs attention right now?" in about a minute — without digging through any spreadsheets.
A few things make a real difference in how usable a dashboard like this actually is. The risk thresholds stay adjustable — so if a clinician wants to lower the cutoff slightly based on their patient population, they can do that on the fly without needing anyone to rerun the data. Clicking on an age group or risk category in one chart automatically filters everything else on the screen. It sounds small, but in a busy clinical setting, removing even one unnecessary step adds up.
One thing clinicians expect is that they don't want a tool that needs a tech person to run it. So everything stays in their hands. If they want to tweak who counts as high risk, they can do it themselves. Click on an age group, and the whole dashboard updates around it. No extra steps. In a clinic where the day never slows down, that kind of simplicity is the whole point.
Why This Matters in the Real World
A doctor or nurse gets a few minutes per patient or sometimes less. A dashboard like this isn't there to second-guess clinical judgment. It's there so the right picture is already in front of you before you walk into the room.
If someone's blood sugar has been running badly for months, their white cell count is up, and they've had two admissions in the last 90 days (6). That combination shouldn't require digging. It should just be visible. The data exists. It's sitting in the system. What the visualization does is stop it from staying buried.
REFERENCES & DATA SOURCES
1. Zhou K, Lansang MC. Diabetes Mellitus and Infection. Endotext [Internet]. Updated June 2024. https://www.ncbi.nlm.nih.gov/books/NBK569326/
2. Casqueiro J et al. Infections in patients with diabetes mellitus. Indian J Endocrinol Metab. 2012;16(S1):S27–36.
3. Kim MK et al. Diabetes Mellitus and Infectious Diseases. Endocrinol Metab. 2024. https://pmc.ncbi.nlm.nih.gov/articles/PMC12436038/
4. Strack B et al. Diabetes 130-US Hospitals 1999–2008 [Dataset]. UCI ML Repository. https://archive.ics.uci.edu/dataset/296/diabetes+130-us+hospitals+for+years+1999-2008
Also on Kaggle: kaggle.com/datasets/brandao/diabetes
5. Centers for Disease Control and Prevention. Diabetes and Your Immune System. https://www.cdc.gov/diabetes
6. World Health Organization. Diabetes Fact Sheet. https://www.who.int/news-room/fact-sheets/detail/diabetes
