Background/Objectives: This study uses Sankey diagrams to analyze treatment pathways in patients with peripheral artery disease (PAD), which is a vascular condition characterized by atherosclerotic occlusion of the arteries, particularly in the lower limbs, affecting up to 14% of the general population. This study focuses on the treatment pathways that lead to amputation versus those that do not, utilizing the STARR dataset and the All of Us dataset. Methods: The study utilized Sankey diagrams to visualize treatment pathways, highlighting the progression from initial treatments to outcomes. Odds ratio analysis was performed to quantify the association between treatment pathways and outcomes. Recognizing potential confounders, analyses were conducted by filtering patients with PAD into subgroups based on these coexisting conditions. Sankey diagrams were then generated for each sub-cohort to visualize treatment pathways. Results: Pathways including antiplatelet and lipid-lowering treatments accounted for 56% of non-amputation cases in the STARR data and 50% in the All of Us data. Amputation pathways frequently included revascularization procedures, representing 15% of amputations in the STARR data and 20% in the All of Us data. Confounder analysis revealed that most amputated PAD patients were over 50 years old and had one or more conditions, such as diabetes, hypertension, or hyperlipidemia. Conclusions: These visualizations provide insights into treatment pathways and their associations with outcomes in PAD patients, highlighting the potential impact of specific treatments on amputation and non-amputation cases. Future work should build on these findings by incorporating predictive models using machine learning techniques to further explore and quantify these relationships.
Cardiovascular surgeries, though lifesaving, can lead to complications requiring further interventions, including amputations. This study uses Sankey diagrams to analyze treatment pathways in patients with peripheral artery disease (PAD), focusing on those that lead to amputation versus those that do not. The analysis demonstrates pathways that begin with initial treatments, such as anti-platelet therapy and lipid lowering treatment often routes to non-amputation. However, as the flows progress through the diagram, they diverge, leading to distinct outcomes and suggesting areas for early intervention. Amputation pathways are more likely to include revascularization procedures. Confounder analysis, including the risk factors of PAD has been performed to support the insights we obtain from the PAD visualizations. These visualizations highlight patterns in patient care and emphasize the importance of tailored treatment strategies to reduce amputations, offering insights to predict likely outcomes based on treatment sequences.
UNSTRUCTURED Background/Objectives: This study uses Sankey diagrams to analyze treatment pathways in patients with diabetes, focusing on amputation versus non-amputation outcomes. The study utilizes the All of Us dataset to investigate treatment patterns and their associations with amputation. Methods: Sankey diagrams visualized treatment sequences, displaying progression from initial treatments to subsequent therapies and outcomes. Logistic regression identified factors associated with amputation outcomes, calculating odds ratios to quantify relationships. Recognizing potential confounders, analyses were conducted by filtering patients into subgroups based on significant factors. Sankey diagrams were generated for statistically significant sub-cohorts to visualize treatment patterns. Results: In the All of Us dataset, metformin-based regimens were most common among patients without amputation, with metformin alone or combined with sulfonylureas and repaglinides predominating. Among patients who underwent amputation, GLP1 receptor agonist sequences were more frequent and showed significantly higher associations with amputation versus metformin (OR=4.53, 95% CI: 3.91-5.25, p<0.001; GLP1-insulin combination OR=8.38). Analysis identified significant amputation-associated factors: heart failure (OR=1.45), smoking (OR=1.63), male gender (OR=1.88), and Black race (OR=4.16, with 20.4% amputation rate vs 8.7% overall). Age-stratified analysis revealed patients aged 40-50 with amputation had 39.58% GLP1 agonist use versus 10.28% without amputation. Conclusions: Our observational findings demonstrate associations between treatment sequences and amputation outcomes, though causality cannot be established. Future research might incorporate machine learning models to predict amputation risk based on treatment sequences and patient factors, while accounting for treatment timing and duration, potentially enabling more personalized treatment approaches for at-risk diabetic patients.
Cardiovascular surgeries, though lifesaving, can lead to complications requiring further interventions, including amputations. This study uses Sankey diagrams, Sunburst plots, and Icicle plots to analyze treatment pathways in patients with peripheral artery disease (PAD), focusing on those that lead to amputation versus those that do not. We found that non-amputation outcomes commonly involve anti-platelet and lipid-lowering medications, while amputation pathways, although less frequent, often share similar initial treatments but diverge later, suggesting areas for early intervention. These visualizations highlight patterns in patient care and emphasize the importance of tailored treatment strategies to reduce amputations, offering insights to predict likely outcomes based on treatment sequences.
3 October 2026 in Artificial Intelligence in the Life Sciences
ObjectiveTo analyze treatment pathways in diabetic patients and identify factors associated with amputation. We used Sankey diagrams to visualize treatment sequences leading to amputation versus non-amputation outcomes.MethodsThe All of Us dataset has been used for the analysis, which was accessed through the researcher workbench on November 21st, 2024, the dataset includes participant data collected through July 1, 2022, in the United States. Logistic regression identified patient factors associated with amputation. We then generated Sankey diagrams for statistically significant subgroups to visualize distinct treatment patterns.ResultsMetformin-based regimens were commonest in patients without amputation. Conversely, sequences involving GLP-1 receptor agonists were more frequent in the amputation group and associated with higher amputation odds (OR=2.43; GLP-1 with insulin, OR=6.78). Significant risk factors included heart failure (OR=1.99), smoking (OR=1.63), male sex (OR=2.88), and Black race (OR=4.57). Patients younger than 40 had lower amputation odds (OR=0.48).ConclusionsOur findings highlight strong associations between treatment pathways, patient characteristics, and amputation, though causality cannot be established. These visualized pathways can inform future machine learning models to predict risk and personalize treatment for at-risk diabetic patients.