ADS Capstone Chronicles Revised
23
Figure 17 Most Common Reactions
Headers in all tables were promoted, and minor processing steps were performed in thePowerQueryEditor,includingreplacing negligible text characters and creating custom columns to map ordinal values to their categorical counterparts. Relationships between the tables were established in the model view (Figure 16). Figure 16 Power BI Model View Relationships
5.3 Model Deployment in Streamlit An application was built using Streamlit (app.py)bydeployingthepretrainedrandom forest model inaninteractiveapplicationin the community cloud, increasing the accessibilityofthesurveillancesystemfrom this project. Streamlit allows seamless integration between GitHubrepositorycode and app deployment on the web (Streamlit, 2024).Theapplicationcanbeaccessedhere: https://pharma-drug-surveillance-8hihsqcwj 6jlv6tpki4s8z.streamlit.app/ The application contains two features: an interactive prediction of ADR outcome based on patient and drug information (Figure18)andastaticgraphdisplayoftop tendrugsforeachoutcomeforthereporting period(Table7reportsatableversionofthe top5drugsforbrevity).Thecombinationof thesetwofeaturesallowforprecisionhealth decisions and high-level public health surveillance. Adisclaimerisincludedatthe top of the tool: “Thissystemdoesnotstore anyinformationfromyouandshouldnotbe used to make healthcare decisions without guidancefromahealthcareprofessional.By using this tool,youconfirmyouunderstand this information.”
5.2.2 Dashboard Filters Thedashboardenablesuserstofilterrecords and visuals based on generic drug name, patient sex, and outcome severity level. Users can filter by drugbyeithersearching or scrolling through the dropdown list. To filter by sex, users select the "Female" or "Male" barfromthehorizontal"PatientSex Distribution" bar plot. For filtering by severity level, users choose the correspondingbarfor"Serious,""Death,"or "NotSerious"inthe"NumberofReportsby The most common adverse reactions reported in the dataset include "off label use"(9.4%ofallreports),"nausea"(6.9%of all reports), and "diarrhea" (6.3% of all reports; Figure 17). When filtering by sex, thetopthreereactionsforfemaleswere“off labeluse,""nausea,"and"fatigue,"whilefor males,thetopthreereactionswere"offlabel use," "pyrexia," and "diarrhea." Severity Level" bar plot. 5.2.3 Dashboard Insights
173
Made with FlippingBook - Online Brochure Maker