What drives meeting length in breast cancer multidisciplinary team discussions

Multidisciplinary team meetings, or MDTMs, have become the backbone of modern breast cancer care, bringing surgeons, medical oncologists, radiation oncologists, radiologists, and pathologists around a single table to plan treatment for every new patient. Decades of evidence show that these meetings improve decision-making quality and even survival, but they are also among the most time-consuming and expensive activities in oncology. Now, a team of Dutch researchers has found a clever way to measure exactly how long each patient discussion takes—without ever setting foot in the meeting room—by mining routinely collected hospital data. Their work, published in Breast Cancer Research and Treatment, offers a new technical blueprint for studying and streamlining one of medicine’s most labor-intensive rituals.

The study, led by Lejla Kočo of the Department of Imaging at Radboud University Medical Center in Nijmegen, together with Wendelien B.G. Sanderink, Mathias Prokop, and Ritse M. Mann, tackled a stubborn methodological problem. Traditionally, researchers who want to know how MDTMs actually function must send an observer into the room, a requirement that creates logistical headaches, alters the natural flow of discussion, and—in the Dutch legal context—typically demands informed consent from every patient discussed, something rarely feasible in routine practice. The team’s solution was to reconstruct discussion durations retrospectively by linking two existing data sources: the Netherlands Cancer Registry (NCR), managed by the Netherlands Comprehensive Cancer Organization, and the timestamps embedded in the hospital’s electronic medical record (EMR) system.

The technical trick at the heart of the study is elegantly simple. During each weekly breast cancer MDTM, a notetaker saves a digital report for every patient discussed, and the EMR (the Epic system) records the exact date and time of each “report saving instant.” The researchers reasoned that the interval between two consecutive saving instants—for two different patients—would closely approximate the duration of the first patient’s discussion. Because the first patient on the list has no preceding timestamp, all first discussions were excluded. Timestamps outside the scheduled meeting window were also discarded, since reports are sometimes edited after the meeting ends, and durations falling outside the central 95 percent of the distribution—below the 2.5th or above the 97.5th percentile—were treated as probable recording errors and removed. A calculated discussion of three seconds, for example, almost certainly reflects a record opened by accident rather than a genuine discussion.

Critically, the method complies with privacy law. Only timestamps belonging to patients already present in the pseudonymized NCR dataset were coupled to clinical data; all other timestamps were labeled NULL and could not be traced to any individual. Patients who had signed an opt-out declaration—declining use of their data for scientific research—were automatically excluded. The NCR supplied 126 clinical variables covering patient, tumor, diagnostic, and treatment characteristics, while the EMR query platform CliniQuest extracted the timing data without exposing privacy-sensitive information. Ethical approval was granted by the local medical ethical committee (CMO 2019–5371), and the key file linking patient IDs to study pseudonyms was stored separately with restricted access.

After rigorous cleaning, the final dataset comprised 1,048 tumors in 996 patients, yielding 1,487 usable discussion times spanning 2014 to 2018 at a single Dutch academic hospital. The mean patient age was 57 years, with a range of 26 to 96. Discussion durations ranged from 38 seconds to 11 minutes, with a mean of 3 minutes 35 seconds and a median of 3 minutes 1 second. The population was divided into three subgroups based on the purpose of the discussion: pre-operative planning (727 discussions), post-operative review (648 discussions), and cases in which no surgery was performed (112 discussions). Pre-operative discussions averaged 3:58, post-operative discussions averaged 3:01, and no-surgery discussions were the longest at 4:16 on average—a pattern the authors attribute to the greater complexity of decisions made before and instead of surgery.

Because the duration data were strongly non-normal—confirmed by Kolmogorov-Smirnov tests for discussion time, age, and tumor size—the team used non-parametric statistics throughout: Spearman’s rank-order correlation for continuous variables, the Mann-Whitney U-test for two-category comparisons, and the Kruskal-Wallis test for multi-category comparisons. The headline result was that pre-operative and no-surgery discussions were significantly longer than post-operative ones (p = 0.000). Within the pre-operative group, several patient and tumor characteristics mattered. Cases of invasive breast cancer took longer to discuss than carcinoma in situ (4:05 versus 3:08, p = 0.001). Poorly differentiated tumors required the longest conversations (4:14) compared with well and moderately differentiated tumors (p = 0.002). Pre-menopausal patients generated longer discussions than post-menopausal ones (4:40 versus 3:47, p < 0.001), plausibly reflecting added considerations such as fertility preservation and hormone-sensitive therapy choices. Stage 4 cases consumed the most time (5:34 versus 3:08 for stage 1, p < 0.001), and, intriguingly, discussions ran slightly longer when an MRI scan had not yet been performed (p = 0.049)—suggesting incomplete diagnostic workups force clinicians to deliberate more. Older age correlated with shorter discussions (r = -0.137, p < 0.001).

In the post-operative group, a different set of predictors emerged. Molecular subtype, BI-RADS score, and cancer stage all significantly influenced duration, and invasive cancers again drew out discussions relative to in situ disease (3:09 versus 2:15, p < 0.001). Tumor size showed a small but significant positive correlation with discussion time (r = 0.097, p = 0.017), likely because larger tumors complicate decisions about adjuvant chemotherapy, radiation, endocrine and targeted therapies, genomic testing, and germline mutation work-up. Notably, age, tumor differentiation grade, the presence of a DCIS component, tumor type, morphology, and menopausal status had no significant effect after surgery—consistent with the idea that post-operative decisions are more standardized or were largely settled in the earlier pre-operative meeting. Several factors also failed to reach significance in the pre-operative group, including molecular subtype, DCIS component, tumor type, BI-RADS score, morphology, and tumor size, underscoring that no single characteristic fully determines how long a case will take.

The practical implications are considerable. Knowing the expected duration of different case types could allow MDTM coordinators to build smarter agendas, front-loading or spreading out complex discussions to prevent meetings from overrunning and to reduce decision fatigue. The findings also caution against overly aggressive streamlining proposals, such as exempting apparently simple cases from discussion altogether. Even cases that might seem straightforward on paper—small cancers in older patients, or pure DCIS—showed non-negligible average discussion times, and the substantial overlap between subgroups suggests case complexity is genuinely reflected in meeting time. The data do lend support to one structural change, however: separating pre-operative from post-operative discussions, since the two follow distinct patterns and may benefit from different pacing. The authors estimate that within their setting, no more than 20 to 25 patients can realistically be discussed per hour, a sobering benchmark for any hospital planning meeting capacity.

The study is not without limitations, which the authors acknowledge candidly. It is a single-center analysis, and meeting formats, team composition, hospital size, and patient populations vary widely across institutions, limiting generalizability. The timestamp-based method also cannot capture the actual content of discussions, verify whether a case was tabled and deferred to a later meeting, or assess the quality of decision-making the way direct observation can. Patients may receive multiple conditional recommendations whose final resolution depends on information not yet available. These assumptions were unavoidable given the retrospective, privacy-constrained design—but they are the price of a method that requires no observer, no consent process, and no disruption to clinical workflow.

What makes the study resonate beyond breast oncology is its demonstration that existing administrative data can serve as a research instrument. Cancer registries and EMR timestamp trails exist in virtually every modern health system, and the approach could in principle be automated to generate discussion-duration reports for any hospital, providing local benchmarks for organizing and optimizing tumor boards. As breast cancer incidence rises, treatment options multiply, and staff shortages persist, the sustainability of meetings that discuss every new case is under real pressure. A method that quantifies where the minutes actually go—without adding a single task to clinicians’ workloads—may prove to be exactly the kind of quiet, data-driven innovation that helps multidisciplinary care survive its own success.