Clinical Note Bloat Reduction for Efficient LLM Use

Jordan L. Cahoon; Chloe Stanwyck; Asad Aali; Rachel Madding; Sulaiman S. Somani; Emma Sun; Yixing Jiang; Renumathy Dhanasekaran

cs.CYMar 21, 2026

Clinical Note Bloat Reduction for Efficient LLM Use

Authors: Jordan L. Cahoon, Chloe Stanwyck, Asad Aali, Rachel Madding, Sulaiman S. Somani, Emma Sun, Yixing Jiang, Renumathy Dhanasekaran, +1 more

Organizations: Department of Biomedical Data Science, Stanford University, Stanford, CA · Department of Pathology, Stanford University, Stanford, CA · Department of Anesthesiology, Perioperative and Pain Medicine, Stanford University, Stanford, CA · Department of Radiology, Stanford University, Stanford, CA · Department of Obstetrics and Gynecology, Stanford University, Stanford, CA · Department of Computer Science, Stanford University, Stanford, CA · Division of Gastroenterology and Hepatology, Stanford University, Stanford, CA · Weill Cancer Hub West

Abstract

Background: Clinical notes contain extensive duplicated text from templates, copy-paste, and auto-populated fields ("note bloat"), diluting clinical signal, limiting longitudinal context, and increasing large language model (LLM) costs. Methods: TRACE removes note bloat using note-level EHR metadata to identify templated and copied content, with frequency-based de-duplication when metadata are unavailable. We evaluated TRACE using blinded physician span review and gold-standard templated-text annotations across four cohorts spanning liver transplant, obstetrics, and inpatient populations at multiple health systems (5.3M notes). We compared zero-shot LLMs and embedding-based classifiers using original and TRACE-processed notes for 20 information extraction tasks and prediction of 5-year survival, postpartum hemorrhage, and 30-day readmission. Results: Only 0.3-6.6% of removed text was flagged as author-generated; TRACE captured 86% of annotated templated characters. Information extraction F1 differences averaged by cohort ranged from -0.009 to +0.004; task-specific prediction F1 differences ranged from -0.011 to +0.018. Among 1,000 randomly sampled Stanford Health Care patients, TRACE reduced chart text by 47.3% (742.7M characters), averaging 220,167 fewer tokens per patient. Using 2024 encounter volumes at a large tertiary academic center and one query per encounter, projected three-year net savings ranged from $1.00M to$ 13.58M across evaluated model pricing schemes, including initial and annual TRACE processing costs. Conclusion: TRACE substantially reduces clinical note redundancy while preserving information extraction and prediction performance. Underused EHR metadata can reduce LLM inference costs, expand usable longitudinal context, and support scalable clinical AI.

Figures & tables

Figure 1: TRACE Method. (A) TRACE identifies templated and copied text using Epic Clarity attribution metadata (Reference Module) and flags additional redundant text blocks using frequency-based de-duplication (Frequency Module). (B) Example TRACE-annotated note showing templated and copied spans with sources identified by the Reference and Frequency Modules.

Figure 2: TRACE downstream tasks. (A) TRACE utility is evaluated through two downstream tasks: Information Extraction (IE) and Clinical Outcome Prediction. In IE, a large language model performs inference over batched notes and updates the prediction as new notes are added. In clinical outcome prediction, the most recent 32K tokens (embedding-based) or up to 1M tokens (zero-shot) are used. Notes can either be used for zero-shot inference or embedded and used to train a classifier. (B) Datasets for downstream evaluation.

Figure 3: TRACE facilitates the analysis of templated and copied text in 1,000 randomly sampled patients (A) Distribution of text removal by TRACE on full note timelines. (B) Number of characters identified as Templated, Copied, and the intersection. (C) Number of characters identified with Reference and Frequency Modules. Counts reflect flagged characters, while the intersection refers to characters identified by both methods. Because copied spans are retained at their first occurrence, flagged characters exceed removed characters. (D) Percent of templated and copied text for the 24 most common note types in Stanford Health Care.

Figure 4: TRACE-processed notes achieve non-inferior performance in information extraction (IE) compared with original notes (A) IE performance across tasks in hepatology and obstetrics for original and TRACE Notes. The X axis indicates the F1 score for the respective cohort. The Y axis indicates the specific task. (B) Distribution of text removed in patient timelines across four real world cohorts. (C) Paired F1 differences (TRACE – Original), averaged across IE tasks, by within-cohort quartile of the percentage of characters removed by TRACE (Q1, least; Q4, most). Points show observed differences; error bars show 95% percentile bootstrap confidence intervals (1,000 replicates).

Figure 5: TRACE enables efficient clinical outcome prediction. A) ROC curves for Postpartum Hemorrhage and 30-day Readmission B) F1 scores with 95% bootstrapped confidence intervals for zero-shot Overall Survival prediction (*) and embedding-based PPH and readmission prediction C) Paired F1 differences (TRACE minus Original) by within-cohort quartile of the percentage of characters removed by TRACE (Q1, least; Q4, most). Points show observed differences; error bars show 95% percentile bootstrap confidence intervals (1,000 replicates).

Supplementary Figure 1: Example of TRACE span identification. (A) Source template. (B) Corresponding de-identified note with templated, copy-forward, and overlapping spans highlighted. Edits have been made to ensure de-identification, changing some wording and minor information while retaining identified spans. Copy-forward spans are retained at their first occurrence. Highlights show identified spans, while removal depends on module-specific length thresholds.

Panel of Figure 7 — Supplementary Figure 2: Prompt used with GPT-4.1 to draft candidate templated and structured-data spans for the metadata-assisted reference annotations.

Panel of Figure 8 — Supplementary Figure 3: Prompt used for information extraction for the first batch

Panel of Figure 9 — Supplementary Figure 4: Prompt used to update extracted values after the first batch

Panel of Figure 10 — Supplementary Figure 5: Prompt used for zero-shot prediction of outcomes.

				Written by
Method		Structured	Template	note author	Indeterminate
Annotator 1
TRACE	Reference	$0.06$	91.27	$1.96$	$6.71$
		(0.00–0.18)	(85.58–95.69)	(0.57–4.06)	(2.87–11.82)
	Frequency	$3.08$	90.11	$0.29$	$6.52$
		(1.05–6.13)	(82.93–94.70)	(0.00–0.93)	(2.86–12.07)

Supplementary Table 1: Annotation category composition for TRACE (Reference and Frequency selection) versus Random selection. Values are percentages of labeled characters (95% confidence intervals), reported separately by annotator. The highest percentage in each row is shown in bold.

Minimum span (characters)	Method	Precision (95% CI)	Recall (95% CI)
$\geq$ 25	Reference	0.9568 (0.9303, 0.9747)	0.8580 (0.7725, 0.9152)
$\geq$ 25	TRACE	0.8789 (0.8139, 0.9254)	0.8763 (0.7967, 0.9299)
$\geq$ 50*	Reference	0.9614 (0.9343, 0.9792)	0.8423 (0.7546, 0.9024)
$\geq$ 50*	TRACE	0.8770 (0.8055, 0.9265)	0.8643 (0.7816, 0.9207)
$\geq$ 75	Reference	0.9623 (0.9330, 0.9812)	0.7675 (0.6645, 0.8429)
$\geq$ 75	TRACE	0.8930 (0.8228, 0.9395)	0.7946 (0.6927, 0.8663)

Supplementary Table 2: Template-detection precision and recall across minimum span thresholds in the 100-note annotated subset. Reference, Reference Module only; TRACE, Reference and Frequency Modules. Recall was evaluated against the same annotated templated spans (

\geq

50 characters) at every threshold. TRACE precision is conservative because templated text flagged only by the Frequency Module is not attribution-linked and therefore counts as a false positive. Values are estimates (95% confidence intervals from 10,000 note-level bootstrap replicates).

Minimum span (characters)	Corpus removal (95% CI)	Timeline removal (95% CI)
$\geq$ 25	48.97% (46.53%, 51.45%)	40.58% (39.73%, 41.42%)
$\geq$ 50*	47.25% (44.77%, 49.76%)	38.85% (38.01%, 39.67%)
$\geq$ 75	45.06% (42.60%, 47.57%)	36.44% (35.59%, 37.27%)

Supplementary Table 3: Text removal across minimum span thresholds in the 1,000-patient task-agnostic sample. Corpus removal is the percentage of all characters removed; timeline removal is the mean per-patient percentage removed. Values are estimates (95% confidence intervals from 10,000 patient-level bootstrap replicates).

type	mean	std	min	25	50	75	max
Progress Notes (n=120467)	0.51	0.3	0.0031	0.23	0.5	0.8	1.0
Telephone Encounter (n=34780)	0.36	0.31	0.0013	0.091	0.25	0.6	1.0
Patient Instructions (n=15399)	0.72	0.28	0.0078	0.53	0.82	0.95	1.0
Consults (n=12202)	0.29	0.24	0.0039	0.1	0.21	0.4	1.0
H&P (n=4633)	0.31	0.24	0.0039	0.13	0.25	0.44	1.0
ED Provider Notes (n=4156)	0.2	0.15	0.0042	0.092	0.16	0.26	0.96

Supplementary Table 4: Annotation by TRACE in 1000 patients for 30 most frequent note types

type	mean	std	min	25	50	75	max
Progress Notes (n=120467)	0.16	0.22	0.0	0.023	0.075	0.2	1.0
Telephone Encounter (n=34780)	0.27	0.29	0.0	0.061	0.13	0.4	1.0
Patient Instructions (n=15399)	0.62	0.33	0.0	0.35	0.72	0.91	1.0
Consults (n=12202)	0.14	0.15	0.0	0.045	0.094	0.18	1.0
H&P (n=4633)	0.098	0.13	0.0	0.028	0.063	0.12	1.0
ED Provider Notes (n=4156)	0.1	0.079	0.0	0.054	0.083	0.13	0.91

Supplementary Table 5: Template annotation by TRACE in 1000 patients for 30 most frequent note types

type	mean	std	min	25	50	75	max
Progress Notes (n=120467)	0.37	0.36	0.0	0.0	0.26	0.73	1.0
Telephone Encounter (n=34780)	0.11	0.26	0.0	0.0	0.0	0.0	1.0
Patient Instructions (n=15399)	0.13	0.3	0.0	0.0	0.0	0.0	1.0
Consults (n=12202)	0.16	0.24	0.0	0.0	0.031	0.22	1.0
H&P (n=4633)	0.22	0.25	0.0	0.021	0.14	0.35	1.0
ED Provider Notes (n=4156)	0.095	0.14	0.0	0.0	0.031	0.13	0.96

Supplementary Table 6: Copyforward annotation by TRACE in 1000 patients for 30 most frequent note types

Cohort	Metric	$\Delta$	95% CI for $\Delta$
OB/PPH	F1	$0.0180$	( $0.0023$ , $0.0341$ )
OB/PPH	AUROC	$0.0047$	( $-0.0034$ , $0.0127$ )
SHC-Inpatient	F1	$-0.0020$	( $-0.0141$ , $0.0103$ )
SHC-Inpatient	AUROC	$0.0003$	( $-0.0052$ , $0.0059$ )
MIMIC	F1	$0.0070$	( $-0.0106$ , $0.0244$ )
MIMIC	AUROC	$-0.0033$	( $-0.0147$ , $0.0080$ )

Supplementary Table 7: Outcome-prediction performance differences with 95% confidence intervals.

Cohort	Task	$\Delta$	95% CI for $\Delta$
Transplant	Average IE	$0.0040$	( $-0.0047$ , $0.0126$ )
Transplant	alcoholic fatty liver disease	$0.0210$	( $-0.0341$ , $0.0784$ )
Transplant	ascites	$0.0094$	( $-0.0100$ , $0.0308$ )
Transplant	cirrhosis	$-0.0037$	( $-0.0093$ , $0$ )
Transplant	cryptogenic cirrhosis	$0$	( $0$ , $0$ )
Transplant	hepatic encephalopathy	$0.0091$	( $0$ , $0.0241$ )

Supplementary Table 8: Aggregate and individual information-extraction F1 differences with 95% confidence intervals.

Supplementary Figure 6: TRACE 95% confidence interval for Information Extraction F1 compared to distribution of F1 for random removal

Term	$\bm{\beta}$ (SE)	OR (95% CI)	$\bm{p}$	$\bm{p_{\mathrm{Holm}}}$
Family 1: Information extraction (2 hypotheses)
Transplant: average IE
Intercept	$0.863$ ( $0.1565$ )	$2.37$ ( $1.74$ , $3.22$ )	$3.529\times 10^{-8}$	—
TRACE	$0.0496$ ( $0.0394$ )	$1.05$ ( $0.973$ , $1.14$ )	$0.209$	$0.4179$
Prevalence	$2.1238$ ( $0.2854$ )	$8.36$ ( $4.78$ , $14.6$ )	$9.869\times 10^{-14}$	—
Obstetrics: average IE

Supplementary Table 9: Logistic regression coefficients for TRACE versus Original. Endpoint: sensitivity. IE denotes information extraction; prevalence is included only in IE models.

Term	$\bm{\beta}$ (SE)	OR (95% CI)	$\bm{p}$	$\bm{p_{\mathrm{Holm}}}$
Family 1: Information extraction (2 hypotheses)
Transplant: average IE
Intercept	$-1.0865$ ( $0.2572$ )	$0.337$ ( $0.204$ , $0.558$ )	$2.392\times 10^{-5}$	—
TRACE	$0.0048$ ( $0.0291$ )	$1$ ( $0.949$ , $1.06$ )	$0.8696$	$0.8696$
Prevalence	$4.6439$ ( $0.3721$ )	$104$ ( $50.1$ , $216$ )	$9.442\times 10^{-36}$	—
Obstetrics: average IE

Supplementary Table 10: Logistic regression coefficients for TRACE versus Original. Endpoint: specificity. IE denotes information extraction; prevalence is included only in IE models.

Cohort / task	df	Wald $\bm{\chi^{2}}$	$\bm{p}$	$\bm{p_{\mathrm{Holm}}}$
Family 1: Information extraction (2 hypotheses)
Transplant: average IE	4	$4.1295$	$0.3888$	$0.3888$
Obstetrics: average IE	4	$9.5479$	$0.04877$	$0.09755$
Family 2: Zero-shot outcome prediction (1 hypothesis)
Transplant: survival*	1	$3.2908$	$0.06967$	$0.06967$
Family 3: Embedding-based outcome prediction (3 hypotheses)

Supplementary Table 11: Wald tests of text-removal associations among TRACE predictions. Endpoint: sensitivity. *Linear term for text removal (1 df) because matrix was singular.

Cohort / task	df	Wald $\bm{\chi^{2}}$	$\bm{p}$	$\bm{p_{\mathrm{Holm}}}$
Family 1: Information extraction (2 hypotheses)
Transplant: average IE	4	$3.1419$	$0.5344$	$1$
Obstetrics: average IE	4	$0.7973$	$0.9388$	$1$
Family 2: Zero-shot outcome prediction (1 hypothesis)
Transplant: survival	4	$2.506$	$0.6436$	$0.6436$
Family 3: Embedding-based outcome prediction (3 hypotheses)

Supplementary Table 12: Wald tests of text-removal associations among TRACE predictions. Endpoint: specificity.

Supplementary Figure 7: Association between text removal and information extraction accuracy. Adjusted predicted probability of a correct prediction among TRACE-processed inputs as a function of the percentage of characters removed by TRACE, estimated from restricted cubic spline logistic regression models (Equation 3) fit separately among reference-positive (sensitivity) and reference-negative (specificity) observations. Original input length was held at its mean and feature prevalence at the task-specific prevalence. Thick lines show the average across IE tasks; thin lines show individual tasks. Shading denotes pointwise 95% confidence intervals from the patient-cluster-robust covariance matrix.

Supplementary Figure 8: Association between text removal and clinical outcome prediction accuracy. Adjusted predicted probability of a correct prediction among TRACE-processed inputs as a function of the percentage of characters removed by TRACE, estimated from restricted cubic spline logistic regression models (Equation 3) fit separately among reference-positive (sensitivity) and reference-negative (specificity) observations, with original input length held at its mean. Shading denotes pointwise 95% confidence intervals from the patient-cluster-robust covariance matrix. Overall survival sensitivity uses a linear term for text removal because the spline model was singular.

Cohort / task	Metric	$\bm{\Delta}$	Lower 97.5% bound	$\bm{p}$	$\bm{p_{\mathrm{Holm}}}$	Non- inferior
Family 1: Information extraction F1 (2 hypotheses)
Transplant: average IE	F1	$0.0040$	$-0.0046$	$4.882\times 10^{-35}$	$9.764\times 10^{-35}$	Yes
Obstetrics: average IE	F1	$-0.0086$	$-0.0222$	$1.277\times 10^{-9}$	$1.277\times 10^{-9}$	Yes
Family 2: Outcome-prediction AUROC (3 hypotheses)
Obstetrics: PPH	AUROC	$0.0047$	$-0.0033$	$6.451\times 10^{-41}$	$1.290\times 10^{-40}$	Yes
Inpatient-SHC: readmission	AUROC	$0.0003$	$-0.0053$	$6.640\times 10^{-71}$	$1.992\times 10^{-70}$	Yes

Supplementary Table 13: Exploratory Wald non-inferiority tests of TRACE versus Original at an absolute margin of 0.05.

\Delta

denotes TRACE minus Original performance. For each endpoint, we tested

H_{0}:\Delta\leq-0.05

against

H_{1}:\Delta>-0.05

. Holm correction was applied separately within each family shown; non-inferiority was concluded when the adjusted one-sided

p

-value was below 0.025. Family-wise error was controlled within each family, rather than jointly across all nine hypotheses. *Survival prediction used zero-shot inference; the other outcome-prediction tasks used embedding-based classifiers.

feature	arm	precision	precision_lower	precision_upper
Overall Survival (Liver Transplant)	Original	0.22	0.17	0.28
Overall Survival (Liver Transplant)	TRACE	0.22	0.17	0.27
Postpartum Hemorrhage (Obstetrics)	Original	0.30	0.28	0.32
Postpartum Hemorrhage (Obstetrics)	TRACE	0.34	0.32	0.37
30-Day Readmission (Inpatient-SHC)	Original	0.52	0.49	0.54
30-Day Readmission (Inpatient-SHC)	TRACE	0.48	0.46	0.5

Supplementary Table 14: Clinical outcome prediction precision with 95% confidence intervals. OS prediction was evaluated using zero-shot inference.

feature	arm	recall	recall_lower	recall_upper
Overall Survival (Liver Transplant)	Original	0.94	0.87	1.000
Overall Survival (Liver Transplant)	TRACE	0.91	0.82	0.98
Postpartum Hemorrhage (Obstetrics)	Original	0.59	0.56	0.62
Postpartum Hemorrhage (Obstetrics)	TRACE	0.53	0.50	0.56
30-Day Readmission (Inpatient-SHC)	Original	0.63	0.61	0.66
30-Day Readmission (Inpatient-SHC)	TRACE	0.68	0.66	0.7

Supplementary Table 15: Clinical outcome prediction recall with 95% confidence intervals. OS prediction was evaluated using zero-shot inference.

feature	arm	f1	f1_lower	f1_upper
Overall Survival (Liver Transplant)	Original	0.36	0.29	0.43
Overall Survival (Liver Transplant)	TRACE	0.35	0.28	0.42
Postpartum Hemorrhage (Obstetrics)	Original	0.40	0.38	0.42
Postpartum Hemorrhage (Obstetrics)	TRACE	0.42	0.39	0.44
30-Day Readmission (Inpatient-SHC)	Original	0.57	0.55	0.59
30-Day Readmission (Inpatient-SHC)	TRACE	0.57	0.55	0.58

Supplementary Table 16: Clinical outcome prediction F1 with 95% confidence intervals. OS prediction was evaluated using zero-shot inference.

feature	arm	auroc	auroc_lower	auroc_upper
Postpartum Hemorrhage (Obstetrics)	Original	0.747	0.729	0.763
Postpartum Hemorrhage (Obstetrics)	TRACE	0.751	0.734	0.768
30-Day Readmission (Inpatient-SHC)	Original	0.802	0.791	0.814
30-Day Readmission (Inpatient-SHC)	TRACE	0.803	0.791	0.814
30-Day Readmission (MIMIC-III)	Original	0.712	0.687	0.735
30-Day Readmission (MIMIC-III)	TRACE	0.708	0.685	0.731

Supplementary Table 17: Clinical outcome prediction AUROC with two-sided 95% percentile-bootstrap confidence intervals from 10,000 paired resampling replicates.

Outcome	Prevalence	Original	TRACE
Overall Survival (Liver Transplant)	0.19	0.222	0.216
Postpartum Hemorrhage (Obstetrics)	0.150	0.380	0.379
30-day Readmission (Inpatient-SHC)	0.247	0.598	0.599
30-day Readmission (MIMIC-III)	0.066	0.157	0.162

Supplementary Table 18: Clinical Outcome Prediction AUPRC

Input	Minimum span (characters)	AUROC	Precision	Recall	F1
Original	–	0.747 (0.729, 0.764)	0.299 (0.278, 0.321)	0.593 (0.561, 0.624)	0.398 (0.374, 0.421)
TRACE	25	0.748 (0.731, 0.765)	0.303 (0.282, 0.323)	0.619 (0.588, 0.650)	0.407 (0.383, 0.429)
TRACE	50 *	0.751 (0.734, 0.768)	0.342 (0.317, 0.367)	0.531 (0.499, 0.563)	0.416 (0.391, 0.441)
TRACE	75	0.745 (0.728, 0.762)	0.341 (0.315, 0.367)	0.480 (0.448, 0.511)	0.399 (0.373, 0.424)

Supplementary Table 19: Embedding-based postpartum hemorrhage (PPH) prediction across TRACE minimum span thresholds. Values are estimates (95% confidence intervals). Original denotes unprocessed notes.

Supplementary Figure 9: Estimated net system-wide cost savings with TRACE. Gross annual savings were estimated from the mean reduction of 220,167 tokens per patient timeline (based on the 1,000-patient timeline sample), 2,058,497 FY2024 Stanford Health Care encounters (from published utilization data), and the model prices and query frequencies shown. Each query was assumed to process a full timeline. Processing 435,797 notes from 1,000 patients required 130.746 billable vCPU-hours. Assuming 8 GiB per vCPU gives 1,045.968 GiB-hours and a sample processing cost of

4.47 at Google Cloud E2 virtual machine 1-year committed-use rates under an existing institutional commitment. Extrapolation to 3 million SHC patients gives 1.307 billion baseline notes and an initial cost of

13,403. New notes equal to a fixed 9% of baseline volume (approximately 117 million notes) adds a fixed cost of $1,206 annually, assuming stable model pricing. Net savings subtract initial and annual processing costs beginning in year 1. Projections assume representative sample timelines, constant average processing cost per note, incremental processing without rerunning the historical corpus, and constant query volume and token reduction. Storage and other operational costs are excluded.

Explore similar work

How much of an LLM-generated clinical corpus is actually new? A production-scale measurement of content redundancy for provenance classification
Jun 28, 2026Ali H. Lazem, William J. TeahanLarge Language Model Pipelines Text Corpora
School of Computer Science and Engineering, Bangor University, Bangor, LL57 2DG, Gwynedd, United Kingdom. · University of Thi-Qar, Nasiriyah, 64001, Iraq.
Systematic Evaluation of the Quality of Synthetic Clinical Notes Rephrased by LLMs at Million-Note Scale
May 18, 2026Jinghui Liu, Sarvesh Soni, Anthony NguyenInternational Classification Of Disease Coding Notes
Australian e-Health Research Centre, CSIRO, Australia · National Library of Medicine, National Institutes of Health, USA
ClinTraceBench: Source-Verifiable Longitudinal Clinical Reasoning over EHR-Derived Dialogues
Sep 1, 2026Huimin Wang, Zhengyi Zhao, Yutian ZhaoPatient Trajectories Mimic-Iv Datasets
Shenzhen University · The Chinese University of Hong Kong · Dealism

Jun 28, 2026, cs.CLJ/K move · Enter open · S save

How much of an LLM-generated clinical corpus is actually new? A production-scale measurement of content redundancy for provenance classification

Ali H. Lazem, William J. Teahan

School of Computer Science and Engineering, Bangor University, Bangor, LL57 2DG, Gwynedd, United Kingdom. · University of Thi-Qar, Nasiriyah, 64001, Iraq.

Large Language Model Pipelines Text Corpora Redundancy Provenance Volume

Open paper AskPDF arXiv

Clinical machine learning increasingly relies on training corpora generated by large language models (LLMs) rather than annotated by clinicians, and such corpora are described and reused largely on the basis of their reported scale. We test whether volume reflects information content. Analysing the complete output of a multi-agent clinical extraction pipeline applied to 167,034 patient narratives, 2.51 billion generated tokens across the ten text-bearing channels of an eleven-channel pipeline, we introduce Provenance-based Redundancy Decomposition, a token-level classification of the entire output by source. Only 10.9% of the output is trainable-unique content while 79.4% is redundant; raw token count overstates information content by roughly ninefold. The redundancy arises through two distinct mechanisms, verbatim copying of source context into per-item fields, and duplication of generated text across records, of which only the former is losslessly removable. An independent, model-free analysis based on lossless compression confirms the redundancy, recovering the two mechanisms without reference to the provenance labels. One pipeline channel carries almost no redundancy, showing that the level of redundancy depends on how each channel is structured rather than being a fixed property of LLM extraction. Because uncorrected redundancy up-weights the longer, more complex presentations that generate the most items, it skews the token-level training distribution of the corpus, a property we measure directly. In a controlled downstream test, de-duplicating the corpus before adaptation improved a clinical encoder on external disease-recognition benchmarks at equal token budget, robustly across adaptation depths and replicated on a second benchmark, confirming that the redundancy carries a measurable cost beyond storage. The classification tool is released openly.

Abstract

Explore similar work

How much of an LLM-generated clinical corpus is actually new? A production-scale measurement of content redundancy for provenance classification

Systematic Evaluation of the Quality of Synthetic Clinical Notes Rephrased by LLMs at Million-Note Scale

ClinTraceBench: Source-Verifiable Longitudinal Clinical Reasoning over EHR-Derived Dialogues