Why am I inflamed?
In today’s technologically advanced society, the application of data science in healthcare has paved the way for a deeper understanding of our quantified selves. In this article, we will delve into an analysis conducted on the relationship between High-Sensitivity C-reactive protein (CRP) and inflammation levels, how it could be monitored via digital and at-home healthcare, and its impact on revolutionizing the traditional paradigm of healthcare in America.
Meet Bernadette, a single mom who, like many, yearns for a deeper understanding of her well-being. Bernadette, who is always on the go working a job and taking care of her three sons, doesn’t have the time for doctor’s visits for regular health and wellness checkups. Instead, she leverages SiPhox Health’s at-home blood biomarker tests and Apple HealthKit data to proactively monitor her family’s health.
One of the most important biomarkers of health for her family is inflammation being that her middle child struggles with rheumatoid arthritis. She wants to be able to assess the effect of anti-inflammation drugs her child has been using recently to know if levels of high-sensitivity C-reactive protein are indeed being reduced. Current academic research establishes this blood biomarker as being the cause of inflammatory conditions.
Luckily, SiPhox health has a web app that not only allows her to seamlessly access her child’s blood biomarker tests results for C-reactive protein but also has a feature to view predictive AI/ML insights and trends in her son’s data along with others with rheumatoid arthritis. By training AI/ML models on large de-identified datasets they can be used to produce predictive insights on blood biomarker trend. Using this data as feedback, she’s able effectively manage her son’s condition by making lifestyle changes like adjusting her son’s diet.
One of the ML-generated insights she’s particularly interested is the duration and intensity of C-reactive protein levels. She views that on average C-reactive protein levels generally decrease by half as inflammation subsides until it reaches a steady baseline. Let’s go more in depth into how this ML insight is generated.
Data Pre-Processing:
- The data is loaded from a JSON file into RStudio.
2. Missing values are trimmed the capped with the mean value of the same variable (in order to maintain the same distribution)
3. The data is plotted on a scatterplot. A non-linear least squares algorithm is used to adjust the parameters and minimize the sum of the squared differences between the observed data points and the corresponding values predicted by the model function. Exponential decay curve is visualized from initial parameters.
4. The summary statistics of the model function will be calculated.
RMSE: 5.834
Get Jet Mejia’s stories in your inbox
Join Medium for free to get updates from this writer.
MSE: 34.036
MAE: 4.626
R-Squared: 87.2
5. The line of best fit will be visualized between actual and predicted values of inflammation.
The data and insights can now be visualized for Bernadette’s family to receive personalized feedback to improve their health.
In her case, Bernadette is able to visualize the effect of dietary changes, supplements, and other lifestyle factors on her family’s blood biomarkers. Further integrating Apple HealthKit to compute sleep, heart rate, blood pressure, and activity level only enriches the digital representation of their quantified selves.
In the future, they will be able to access insights on their data via a conversational AI interface. They will be able to speak to their health data, kind of like having a personalized AI health assistant ready to answer any inquiries you have on your health.
By then, someone could say, “Why am I inflamed?” to their phone instead of another unnecessary doctor’s visit.
