Centrality in Collaboration: A Novel Algorithm for Social Partitioning Gradients in Community Detection for Multiple Oncology Clinical Trial Enrollments

Benjamin Smith,Tyler Pittman,Wei Xu

2024-11-06

Abstract:Patients at a comprehensive cancer center who do not achieve cure or remission following standard treatments often become candidates for clinical trials. Patients who participate in a clinical trial may be suitable for other studies. A key factor influencing patient enrollment in subsequent clinical trials is the structured collaboration between oncologists and most responsible physicians. Possible identification of these collaboration networks can be achieved through the analysis of patient movements between clinical trial intervention types with social network analysis and community detection algorithms. In the detection of oncologist working groups, the present study evaluates three community detection algorithms: Girvan-Newman, Louvain and an algorithm developed by the author. Girvan-Newman identifies each intervention as their own community, while Louvain groups interventions in a manner that is difficult to interpret. In contrast, the author's algorithm groups interventions in a way that is both intuitive and informative, with a gradient evident in social partitioning that is particularly useful for epidemiological research. This lays the groundwork for future subgroup analysis of clustered interventions.

Social and Information Networks,Methodology,Other Statistics

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: **How to identify the collaboration network among oncologists by analyzing the referral patterns of cancer patients between different clinical trials, and provide a basis for subsequent clinical trial participation?**

Specifically, this study focuses on cancer patients who have not been cured or relieved after receiving standard treatment and are usually candidates for clinical trials. These patients may be suitable for other clinical trials after participating in one clinical trial. And a key factor affecting whether patients continue to participate in subsequent clinical trials is the structured collaboration between oncologists and the primary responsible doctors. Therefore, this study aims to reveal the structure of these collaboration networks through social network analysis (SNA) and community detection algorithms.

### Main problems:
1. **Identify the collaboration network among oncologists**: Analyze the referral patterns of patients between different clinical trials to identify the collaborative relationships among oncologists.
2. **Evaluate the effectiveness of existing community detection algorithms**: Compare the performance of Girvan - Newman, Louvain and the Smith - Pittman algorithm developed by the author in identifying the collaboration network.
3. **Provide guidance for future research**: Lay the foundation for future subgroup analysis and intervention cluster research, especially of great significance for epidemiological research.

### Method overview:
- **Data sources**: Use simulated cancer clinical trial data, including 515 clinical trials of 2,970 patients, involving 41 principal investigators.
- **Analysis methods**: Apply social network analysis (SNA) and three community detection algorithms (Girvan - Newman, Louvain and Smith - Pittman) to identify the collaboration network based on the referral patterns of patients between different intervention types.
- **Evaluation indicators**: Evaluate the effectiveness of community detection algorithms through modularity (\(Q\)), which is defined as follows:

\[
Q=\frac{1}{m} \sum_{i, j}\left(A_{ij}-\frac{k_{i}^{\text {out }} k_{j}^{\text {in }}}{m}\right) \delta\left(c_{i}, c_{j}\right)
\]

where:
- \(m\) is the number of edges (i.e., the number of patient referrals),
- \(A_{ij}\) is the number of connections between nodes \(i\) and \(j\),
- \(k_{i}^{\text {out }}\) and \(k_{j}^{\text {in }}\) are the out - degree and in - degree of node \(i\) and node \(j\) respectively,
- \(\delta\left(c_{i}, c_{j}\right)\) is an indicator variable indicating whether nodes \(i\) and \(j\) belong to the same community.

### Results and discussion:
- **Girvan - Newman algorithm**: Each intervention is regarded as an independent community, with a low modularity (\(Q = 0.044\)), and the results are not interpretable.
- **Louvain algorithm**: Successfully identifies four different working groups, with a high modularity (\(Q = 0.177\)), but it is difficult to explain the practical significance of these groupings.
- **Smith - Pittman algorithm**: Identifies eight communities, six of which are composed of a single intervention, and two contain multiple interventions, with a modularity of \(Q = 0.08\), and the results are more interpretable, showing the connectivity between interventions and the gradient distribution of the collaboration network.

In conclusion, through comparing the effectiveness of different community detection algorithms, this study proposes a new algorithm (Smith - Pittman), which can more intuitively reveal the collaboration network among oncologists and provides a powerful tool for further research.