Portinari: A Data Exploration Tool to Personalize Cervical Cancer Screening

Sagar Sen, Antônio H. Ribeiro, Raquel C. de Melo-Minardi, Wagner Meira, Mari NigardarXiv 2017 · DOI · arXiv

Introduction

Public health screening programs for cervical cancer represent complex socio-technical systems involving medical practitioners, laboratories, and patients. While organized screening has significantly reduced cancer incidence, current systems typically rely on a "one-size-fits-all" triage algorithm. These algorithms determine follow-up intervals based on previous test results, but often fail to account for individual symptoms or personal risk perceptions.

The standard cervical cancer screening triage algorithm used in Norway.

Patients frequently deviate from these rigid guidelines—some returning earlier due to symptoms like unexplained bleeding, while others delay follow-ups. Understanding these deviations is crucial for personalizing care. We present Portinari, a web-based data exploration tool designed to query and visualize the diverse trajectories of patients within the Norwegian screening system.

The Portinari System

Portinari addresses the challenge of exploring massive longitudinal datasets by combining graph-based data modeling with intuitive visual interfaces. The tool is built on three primary components:

  1. Graph Database Backend: Utilizing Neo4j, the system transforms decades of screening records into a directed property graph. Nodes represent medical events (exams, diagnoses), and edges represent chronological sequences. "Multi-hop" edges are implemented to optimize complex temporal queries, such as identifying patients who developed cancer within a specific window following a particular test result.
  2. Visual Query Interface: Users can define patient cohorts by "drawing" a sequence of events on a drag-and-drop canvas. This interface translates visual patterns into Cypher queries to fetch matching trajectories from the database.
  3. Trajectory Visualization: Results are displayed using Sankey diagrams, where the width of the flow represents the volume of patients moving between different clinical states over time.
The Portinari interface, featuring the visual query canvas (a) and Sankey diagram visualizations (e, f) of patient cohorts.

Dataset and Methodology

The tool was validated using a high-quality dataset from the Cancer Registry of Norway, encompassing approximately 5 million exam records from over 900,000 women between 1992 and 2014. To protect patient privacy, the data was anonymized using the ARX tool, which perturbs dates while preserving the underlying behavioral and clinical trends essential for epidemiological research.

Key Insights from Patient Trajectories

Exploring the data with Portinari revealed several critical insights into the limitations of standard triage:

  • The Guideline Gap: Even among women who strictly adhered to the three-year screening interval after normal results, a small fraction (0.01%) still developed squamous cell carcinoma. This suggests that the standard window may be insufficient for certain high-risk individuals.
  • Symptom-Driven Behavior: The tool identified a cohort of over 4,000 women who returned for testing within just 30 days of a normal result. Many were subsequently diagnosed with polyps or HPV+, indicating that clinical symptoms often override automated screening invitations.
  • Risk Perception and Adherence: Patients receiving an HPV-positive result showed significantly higher adherence to one-year follow-ups compared to those with other abnormal results. This suggests that the perceived gravity of an HPV diagnosis strongly influences patient behavior.

Results and Future Directions

Portinari successfully validated that the majority of women (88.2%) following the Norwegian triage algorithm maintained normal results. However, its true value lies in identifying outliers—such as the 39 specific cases where cancer occurred despite guideline adherence—providing a vital starting point for clinical review.

While subgraph matching in large graphs remains computationally intensive, future development will focus on integrating genetic and lifestyle data to further enrich patient trajectories. Ultimately, the goal is to move from retrospective exploration to predictive modeling, identifying high-risk individuals before they deviate from optimal screening paths.

Summary updated

Abstract

Socio-technical systems play an important role in public health screening programs to prevent cancer. Cervical cancer incidence has significantly decreased in countries that developed systems for organized screening engaging medical practitioners, laboratories and patients. The system automatically identifies individuals at risk of developing the disease and invites them for a screening exam or a follow-up exam conducted by medical professionals. A triage algorithm in the system aims to reduce unnecessary screening exams for individuals at low-risk while detecting and treating individuals at high-risk. Despite the general success of screening, the triage algorithm is a one-size-fits all approach that is not personalized to a patient. This can easily be observed in historical data from screening exams. Often patients rely on personal factors to determine that they are either at high risk or not at risk at all and take action at their own discretion. Can exploring patient trajectories help hypothesize personal factors leading to their decisions? We present Portinari, a data exploration tool to query and visualize future trajectories of patients who have undergone a specific sequence of screening exams. The web-based tool contains (a) a visual query interface (b) a backend graph database of events in patients' lives (c) trajectory visualization using sankey diagrams. We use Portinari to explore diverse trajectories of patients following the Norwegian triage algorithm. The trajectories demonstrated variable degrees of adherence to the triage algorithm and allowed epidemiologists to hypothesize about the possible causes.
BibTeX
@misc{sen2017portinari,
  author = {Sagar Sen and Antônio H. Ribeiro and Raquel C. de Melo-Minardi and Wagner Meira and Mari Nigard},
  title = {{Portinari: A Data Exploration Tool to Personalize Cervical Cancer Screening}},
  year = {2017},
  eprint = {1704.00172},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1704.00172},
}
Figure