Data Study on Diabetes Treatment
The skills the author demoed here can be learned through taking Data Science with Machine Learning bootcamp with NYC Data Science Academy.
This project was conducted to analyze data from public reviews and deduce possible factors that might be responsible for the effectiveness ratings in diabetes treatment. A significant proportion of the world population is diabetic, according to the World Health Organization, 8.5% of the current world population is diabetic and in the, U.S, 9.4% amounting to 30.3 million people live with diabetes.
METHODS
Although there are many health service stores online, WebMd was chosen for this project because of its independence and comprehensive database. It is an online source for health-related information with huge user review database that includes the name, date, gender, duration of treatment, age range, comments and most importantly ratings based on effectiveness, ease of use and satisfaction.
Scrapy, an open source website scraping framework was used to extract data of 3 diabetes drugs. The drugs are Avandia Oral, Metformin HCl and Invokana. All the information needed appear on one URL of WebMD which makes Scrapy a perfect tool to scrap the website. Upon download, data cleaning was the first exercise that needed to be carried out and was done before I analyzed the data. Below is the image of the data frame before cleaning.
Some features that are of importance to this analysis that needed to be cleaned are the condition column which has attached to it some links, and duration column which was downloaded as a string(text) and need to be converted to integers. Upon cleaning, a new feature was added, created from averaging the duration feature, average duration.
Data RESULT AND DISCUSSION
Effectiveness reviews of the drugs, the core of this project, were analyzed alongside the satisfaction ratings.
Bar plot of Avandia's satisfaction and effectiveness ratings
Bar plot of Avandia's satisfaction and effectiveness ratingsBelow is how the effectiveness and satisfaction ratings of Avandia drug stack together:
Avandia's ratings
Avandia's ratingsBar plot of Metformin's satisfaction and effectiveness ratings
Bar plot of Metformin's satisfaction and effectiveness ratingsMetformin's ratings
Metformin's ratingsBar plot of Invokana's satisfaction and effectiveness ratings
Bar plot of Invokana's satisfaction and effectiveness ratingsInvokana's ratings
Invokana's ratingsApart from metformin with slightly more than average review, the drugs are highly rated in terms of effectiveness. Satisfaction ratings vary from drug to drug, users are highly unsatisfied with Avandia and Metformin while Invokana has a relatively favorable satisfaction.
Plot of age-range and duration against effectiveness ratings of Avandia drug
Plot of age-range and duration against effectiveness ratings of Avandia drugBar Plot of age-range and duration against effectiveness ratings of Metformin drug
Bar Plot of age-range and duration against effectiveness ratings of Metformin drugPlot of age-range and duration against effectiveness ratings of Invokana drug
Plot of age-range and duration against effectiveness ratings of Invokana drugFrom the plots in Figures 5,6 and 7, duration of treatment produced more interesting visuals than age-range of the user. Effectiveness of the drugs tends to increase with continued usage of all the drugs. Users of Avandia for over two years period gave a very high effectiveness ratings while those that treated their condition with either metformin or Invokana for at least three months rated the effectiveness 4 stars and above.
Bar plot of different categories of age-range analyzed
Bar plot of different categories of age-range analyzedPlot of gender distribution
Plot of gender distributionThe 2 figures above show the demographic make-up of the users, majority of diabetic patients are between the ages of 45 to 74 years, and are mostly females.
CONCLUSION
Scraping and analyzing WebMD reviews of three type II diabetes drugs has given some insight into public perception of the treatment. Prior to carrying out this project, one might assume a 5-star rated effective drug would translate to customer satisfaction, but this analysis has shown that it is not always the case and both can be independent of each other.
Further work will include sentiment analysis of the comments which could help elucidate what is responsible for customer dissatisfaction.
Based on the reviews, effectiveness improves with increased duration of treatment, while age-range of the patient does not affect potency of the drugs.
Analyzing these reviews can help medical practitioners make informed decisions based on real-life feedback as against the usual practice of relying solely on the information supplied by the Pharmaceutical companies and the FDA. More importantly is the insight this will give the producing pharmaceutical companies of these 3 drugs and other diabetic drug manufacturers, on the factors that affect the effectiveness of these drugs. In marketing, companies should consider the messaging, and in production, manufacturers should consider the formulations, active pharmaceutical ingredients and excipients since most consumers will be advised to stay on the drugs for a long period of time.
Author
Oluwole Alowolodu
About Author
