Data Science Intern
Analyzed the relationship between GP prescriptions and pharmaceutical residues in sewage water for a public-health project, and built interactive Tableau dashboards that let non-technical stakeholders explore the data across two districts.
- ▸Independently analyzed multi-year prescription data and two rounds of sewage residue measurements across two Dutch districts
- ▸Built interactive Tableau dashboards for a mixed audience of clinicians, researchers, and environmental and policy stakeholders
- ▸Investigated links between prescriptions, consumption, and measured residues in sewage
- ▸The data did not show a simple, clean correlation; the work was exploratory and worked around real data-quality limitations
Context
Researchable is a Groningen-based data company. I interned there on a public-health project studying pharmaceutical residues in water, part of a regional network bringing together healthcare, the water sector, government, and research. Pharmaceutical residues that pass through people and into the sewage system are an environmental and drinking-water concern, and the project wanted to understand how prescribing patterns relate to what actually shows up in the water.
What I Did
I worked with two data sources for two districts: multi-year records of what general practitioners prescribed, and two rounds of measurements of pharmaceutical residues in sewage samples from the same areas. I analyzed both independently and then looked for relationships between them.
I built interactive Tableau dashboards so that people without a data background (clinicians, water authorities, policymakers) could explore prescribing behavior and measured residues side by side, broken down by year, age group, and medication type. The two districts were chosen for contrast: one with a strong focus on healthy living and one without, to see whether that difference was visible downstream.
What I Found
The honest result is that the data did not show a simple, clean correlation between prescription levels and measured residues. The work was exploratory, and it ran into the usual real-world problems: sparse measurements (two sampling rounds), differences in how the two data sources were collected, and confounding factors between prescription, consumption, and what ends up in sewage. The dashboards were still useful as a way for a mixed group of stakeholders to look at the same evidence and reason about it together.
What I Learned
Making complex data legible to people from completely different domains is a real skill, and a different one from the analysis itself. It also reinforced the importance of being clear about what the data can and cannot support: presenting an exploratory finding as a firm conclusion would have been the wrong call here.