Standardizing Clinical Trials Workflow Representation in UML for International Site Comparison

Elias Carvalho, Madhav Kishore Jayanti, Adelia Portero Batilana, Andreia M. O. Kozan, Maria João Rodrigues, Jatin Shah, Marco Rocha Loures, Sunita J. Patil, Philip Payne, Ricardo PietrobonPLoS 2010 · DOI

Abstract

This paper describes a method for modeling clinical trial workflow in UML and standardizing these workflow diagrams through a UML profile. In the increasingly global environment of clinical trials, the standardization of workflow modeling is a necessary precursor to conducting a comparative analysis of international clinical trials workflows.

Introduction

As clinical trials become increasingly globalized, researchers face a significant operational hurdle: lack of standardization across international sites. Even when following the same protocol, a trial site in Brazil may handle patient intake and data entry differently than a site in the United States. These variations introduce "noise" into clinical data, complicating the reproduction of results and the comparison of efficiency across research centers.

The core of this issue is the absence of a standard computational language to describe ground-level operations. While terms like "phlebotomy" and "drawing blood" are synonymous to clinicians, they are distinct to computational systems. Without a standardized model, identifying bottlenecks—such as slow data entry or inefficient physician consultations—remains a manual and error-prone process. This research proposes using Unified Modeling Language (UML) to bridge this gap.

Methodology

The researchers utilized UML, a standard in software engineering, to model the "business logic" of clinical trials. The approach was divided into three primary phases:

  1. Ethnographic Observation: The team conducted 53 hours of direct observation at two Brazilian clinical trial sites (specializing in oncology and rheumatology). Time-motion studies recorded the duration and sequence of tasks performed by various staff members.
  2. UML Activity Diagrams (AD): Standard UML 2.0 was used to create visual flowcharts of these observed processes, mapping the sequence of activities from patient arrival to visit completion.
  3. The UML Profile Innovation: Because standard UML is generic, the authors developed a specialized "UML Profile." This set of extensions, or "stereotypes," allowed them to tag tasks with domain-specific metadata, including time-related attributes (duration, delay, fail rates), statistical distributions (e.g., Gamma or Poisson distributions for task timing), and specific actors (e.g., Clinical Research Coordinators, nurses, or physicians).
UML Activity Diagram illustrating a standardized clinical trial workflow.

The modeling was executed within the Eclipse environment using the UML2 plugin, ensuring that the resulting diagrams were compatible with discrete event simulation software.

Key Insights and Analysis

UML as a Communication Bridge

The study demonstrates that UML is a powerful tool for "Healthcare Business Process Re-engineering." It provides a visual language that is accessible to clinicians while remaining rigorous enough for simulation engineers.

Simulation-Ready Modeling

By embedding statistical distribution parameters directly into the UML profile, the researchers moved beyond static flowcharts. These models are designed for direct integration into simulation engines like Arena or AnyLogic. This allows trial managers to predict how changes in staffing or protocol might affect overall trial speed before implementing them in the real world.

Identifying Operational Bottlenecks

The ethnographic data revealed that tasks requiring the "full commitment" of a coordinator—where multitasking is impossible—created significant idle time for other staff members. Furthermore, the study quantified the different "tempos" of clinical roles: nurses act as high-frequency engines of the trial, while physicians often experience long idle periods followed by bursts of complex activity.

Results

The analysis identified several key patterns and friction points in the clinical trial workflow:

  • Role-Specific Workloads: Nurses performed the highest frequency of tasks with the most consistent durations. Administrative assistants exhibited a "U-curve" workload, being busiest at the start and end of patient visits.
  • Physician Downtime: Physicians experienced the most downtime at the beginning of the workflow, suggesting that efficiency could be improved if preparatory tasks were fully completed by other staff before physician involvement.
  • Workflow Friction Points: Four main failure points were identified: tasks requiring 100% coordinator attention, manual transfer of paper notes to electronic systems (a high-error zone), deviations from Standard Operating Procedures (SOPs), and a lack of integration between disparate IT systems.

Conclusion and Future Work

While this study establishes a robust methodology for modeling clinical trials, it faces limitations such as a small sample size and the remaining "semantic gap" where standardized vocabulary is not yet enforced within the UML diagrams. Future work will focus on integrating controlled medical vocabularies into the UML profile and developing automated interfaces to import these models directly into simulation software. Ultimately, this framework provides the necessary foundation for the "holy grail" of this research: a rigorous, automated comparison of clinical trial efficiency across international borders.

Summary updated

Abstract

This paper describes a method for modeling clinical trial workflow in UML and standardizing these workflow diagrams through a UML profile. In the increasingly global environment of clinical trials, the standardization of workflow modeling is a necessary precursor to conducting a comparative analysis of international clinical trials workflows.
BibTeX
@inproceedings{carvalho2010standardizing,
  author = {Elias Carvalho and Madhav Kishore Jayanti and Adelia Portero Batilana and Andreia M. O. Kozan and Maria João Rodrigues and Jatin Shah and Marco Rocha Loures and Sunita J. Patil and Philip Payne and Ricardo Pietrobon},
  title = {{Standardizing Clinical Trials Workflow Representation in UML for International Site Comparison}},
  booktitle = {PLoS},
  year = {2010},
  doi = {10.1371/journal.pone.0013893},
}
Figure