This project analyzes NHS antidepressant prescription data to understand trends and predict costs. The initial analysis explores regional and monthly variations in prescription volumes and costs, identifies key antidepressant drugs influencing these trends, and investigates the case of Venlafaxine due to its significant impact on cost despite lower volume. Finally, after the challenge was submitted, a further analysis was performed personally where several machine learning models are built and evaluated to predict antidepressant prescription costs based on relevant features.
The analysis revealed interesting trends in antidepressant prescribing, including a general increase in prescription volume but a decline in overall costs. Seasonal patterns were observed, with costs peaking in the first quarter. Sertraline hydrochloride and Amitriptyline hydrochloride were found to be the most frequently prescribed drugs, while Venlafaxine had a disproportionately high cost per item, suggesting potential factors like brand or formulation influencing its price. Regional variations in both prescription volume and cost were also identified. In the machine learning phase, the LightGBM Regressor model demonstrated the best performance in predicting prescription costs, indicating that a non-linear model is more suitable for this dataset compared to Linear Regression and SVR.