Pith. sign in

REVIEW 3 major objections 27 references

Reformulating irregular ICD diagnosis sequences as visit-bucketized directed graphs with time-decay message passing improves IBD risk detection while cutting model complexity.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · grok-4.3

2026-06-29 13:14 UTC pith:6X37UA4O

load-bearing objection The paper reformulates ICD sequences as bucketized directed graphs with time-decay message passing for IBD risk prediction, but the abstract supplies no numbers, baselines, or ablations to back the performance claims. the 3 major comments →

arxiv 2605.27799 v1 pith:6X37UA4O submitted 2026-05-27 cs.AI eess.SP

GraD-IBD: Graph Representation Learning from Diagnosis Trajectories for Early Detection of Inflammatory Bowel Disease

Leo Y. Li-Han , Ellen L. Larson , Elizabeth B. Habermann , Cornelius A. Thiels , Hojjat Salehinejad This is my paper
classification cs.AI eess.SP
keywords inflammatory bowel diseaseICD codesgraph representation learningtemporal message passingdisease risk predictiondiagnosis trajectoriesclinical data modelingearly detection
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper establishes that longitudinal ICD code trajectories, which are irregular and hierarchical, can be recast as visit-bucketized temporally directed graphs rather than processed by sequential models. A context-aware time-decay message passing step then propagates information across these graphs to predict inflammatory bowel disease risk. On a real-world clinical dataset the resulting model shows consistent gains in detection accuracy over prior methods together with lower computational cost than sequence-based alternatives. If the reformulation preserves the needed temporal structure, the approach offers a route to more scalable risk prediction from standard diagnosis records.

Core claim

GraD-IBD reformulates longitudinal ICD trajectories as visit-bucketized, temporally directed graphs and introduces a context-aware time-decay message passing mechanism that captures temporal dependencies; experiments on a real-world clinical dataset show consistent improvements in IBD detection over state-of-the-art methods together with significant reductions in computational complexity relative to sequential models.

What carries the argument

Visit-bucketized temporally directed graphs combined with context-aware time-decay message passing, which encodes diagnosis trajectories and propagates temporal information without the complexity of lattice-based sequential models.

Load-bearing premise

Bucketizing visits into directed graphs and applying time-decay message passing preserves all temporal dependencies needed for accurate downstream risk prediction.

What would settle it

A head-to-head comparison on the same clinical dataset in which a well-tuned sequential model (with explicit handling of irregular and hierarchical ICD sequences) matches or exceeds GraD-IBD's detection metrics while using comparable or lower compute.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • The graph formulation yields higher IBD detection performance than existing sequential approaches on the evaluated real-world dataset.
  • Computational complexity drops substantially compared with sequential models because message passing replaces lattice-based sequence handling.
  • The same graph construction and message-passing design can be applied to other longitudinal diagnosis-code tasks beyond IBD.
  • Risk scores become available from standard ICD records without requiring additional feature engineering for temporal irregularity.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same bucketized-graph construction could be tested on other chronic conditions whose diagnosis codes exhibit similar hierarchy and irregularity.
  • Integration with additional EHR streams such as medication or lab orders could be examined by extending the node and edge types of the directed graph.
  • Real-time deployment would require measuring how the time-decay parameter behaves when new visits arrive incrementally rather than in batch.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

3 major / 0 minor

Summary. The paper proposes GraD-IBD, a graph representation learning approach that reformulates irregular hierarchical longitudinal ICD diagnosis trajectories as visit-bucketized temporally directed graphs. It introduces a context-aware time-decay message passing mechanism to capture temporal dependencies for early IBD risk detection and reports consistent improvements over state-of-the-art methods along with reduced computational complexity on a real-world clinical dataset.

Significance. If the experimental claims are substantiated, the work would indicate that graph-based reformulations can provide efficiency advantages over sequential models when processing irregular medical event sequences, potentially supporting more scalable disease risk prediction from electronic health records.

major comments (3)
  1. [Abstract] Abstract: the claim of 'consistent and robust improvements in IBD detection over state-of-the-art methods' and 'significant reductions in computational complexity' is asserted without any metrics, baselines, statistical tests, dataset size, or ablation results, preventing evaluation of whether the data supports the central claim.
  2. [Method] Method description: no equations define the time-decay function, and no argument is supplied showing retention of intra-visit order or ICD hierarchy after bucketization; this directly bears on whether the graph construction preserves the temporal dependencies needed for risk prediction.
  3. [Experiments] Experiments: no ablation or comparison is described that tests whether the visit-bucketized graph construction degrades signal relative to the original sequences, which is required to substantiate the reported gains over sequential models.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for their thorough review and constructive comments on our manuscript. We address each of the major comments point by point below, indicating the revisions we plan to make to strengthen the paper.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the claim of 'consistent and robust improvements in IBD detection over state-of-the-art methods' and 'significant reductions in computational complexity' is asserted without any metrics, baselines, statistical tests, dataset size, or ablation results, preventing evaluation of whether the data supports the central claim.

    Authors: We agree with this observation. The abstract in the current version is indeed high-level and does not include quantitative details. In the revised manuscript, we will update the abstract to include specific metrics such as AUC improvements, F1 scores, comparisons to state-of-the-art baselines, statistical test results, the size of the clinical dataset used, and a brief mention of ablation studies to better support our central claims. revision: yes

  2. Referee: [Method] Method description: no equations define the time-decay function, and no argument is supplied showing retention of intra-visit order or ICD hierarchy after bucketization; this directly bears on whether the graph construction preserves the temporal dependencies needed for risk prediction.

    Authors: We acknowledge the need for more rigorous mathematical formalization. We will add the explicit equations defining the context-aware time-decay message passing mechanism. Furthermore, we will include a dedicated subsection or paragraph providing arguments and possibly illustrative examples showing that the visit-bucketization retains intra-visit order and the ICD code hierarchy, thereby preserving the necessary temporal dependencies for accurate risk prediction. revision: yes

  3. Referee: [Experiments] Experiments: no ablation or comparison is described that tests whether the visit-bucketized graph construction degrades signal relative to the original sequences, which is required to substantiate the reported gains over sequential models.

    Authors: This is a valid point for substantiating our claims. We will incorporate an additional ablation experiment that directly compares the performance of GraD-IBD on the visit-bucketized graphs against sequential models applied to the original unbucketized diagnosis sequences. This will help confirm that the graph reformulation does not degrade the predictive signal while providing efficiency benefits. revision: yes

Circularity Check

0 steps flagged

No circularity identified; derivation self-contained.

full rationale

The abstract and available text describe a graph reformulation of ICD sequences and a novel message-passing mechanism but supply no equations, parameter-fitting steps, self-citations, or uniqueness theorems. No load-bearing claim reduces by construction to its inputs, and the central performance claims rest on external real-world dataset experiments rather than internal redefinitions. This is the normal case of a self-contained empirical modeling paper with no detectable circularity.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract supplies no mathematical formulation, parameters, or explicit assumptions; all ledger entries are therefore unknown.

pith-pipeline@v0.9.1-grok · 5718 in / 987 out tokens · 30543 ms · 2026-06-29T13:14:32.609954+00:00 · methodology

0 comments
read the original abstract

International Classification of Diseases (ICD) is a globally recognized coding system that records diagnostic events during each patient encounter, providing a standardized data foundation for various clinical tasks. However, the irregular and hierarchical nature of ICD code sequences poses challenges for N-D lattice-based sequential modeling methods, leading to overly complex model designs. In this paper, we propose GraD-IBD, a graph diagnosis model that reformulates longitudinal ICD trajectories as visit-bucketized, temporally directed graphs to detect the risk of inflammatory bowel disease (IBD). A novel context-aware, time-decay message passing mechanism was developed to capture temporal dependencies while reducing model complexity. The experimental results using a real-world clinical dataset demonstrated consistent and robust improvements in IBD detection over state-of-the-art methods, with significant reductions in computational complexity compared to sequential models. These findings highlight the potential of graph representation learning to enable efficient, scalable, and accurate disease risk prediction from longitudinal ICD diagnosis codes.

Figures

Figures reproduced from arXiv: 2605.27799 by Cornelius A. Thiels, Elizabeth B. Habermann, Ellen L. Larson, Hojjat Salehinejad, Leo Y. Li-Han.

Figure 1
Figure 1. Figure 1: The overall architecture of the proposed Grad-IBD model. Unstruc [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Sequential representation of a patient’s ICD diagnostic history. The [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 4
Figure 4. Figure 4: Testing performance of different graph and sequential modeling methods with variable prediction intervals ranging from 1 to 6 months: (a) AUROC, [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

27 extracted references · 7 canonical work pages · 5 internal anchors

  1. [1]

    International classification of diseases (ICD),

    World Health Organization, “International classification of diseases (ICD),” accessed: 2025-07-25. [Online]. Available: https://www.who.int/standards/classifications/classification-of-diseases

    2025

  2. [2]

    BEHRT: transformer for electronic health records,

    Y . Li, S. Rao, J. R. A. Solares, A. Hassaine, R. Ramakrishnan, D. Canoy, Y . Zhu, K. Rahimi, and G. Salimi-Khorshidi, “BEHRT: transformer for electronic health records,”Scientific reports, vol. 10, no. 1, p. 7155, 2020

    2020

  3. [3]

    Scalable and accurate deep learning with electronic health records,

    A. Rajkomar, E. Oren, K. Chen, A. M. Dai, N. Hajaj, M. Hardt, P. J. Liu, X. Liu, J. Marcus, M. Sunet al., “Scalable and accurate deep learning with electronic health records,”NPJ digital medicine, vol. 1, no. 1, p. 18, 2018

    2018

  4. [4]

    arXiv preprint arXiv:1906.00346 , year=

    S. Junyuan, M. Tengfei, X. Cao, and S. Jimeng, “Pre-training of graph augmented transformers for medication recommendation,”arXiv preprint arXiv:1906.00346, 2019

    work page arXiv 1906

  5. [5]

    Inflammatory bowel disease: cause and immunobiology,

    D. C. Baumgart and S. R. Carding, “Inflammatory bowel disease: cause and immunobiology,”The Lancet, vol. 369, no. 9573, pp. 1627–1640, 2007

    2007

  6. [6]

    Presenting symptoms in inflammatory bowel disease: descriptive analysis of a community-based inception cohort,

    B. K. Perler, R. Ungaro, G. Baird, M. Mallette, R. Bright, S. Shah, J. Shapiro, and B. E. Sands, “Presenting symptoms in inflammatory bowel disease: descriptive analysis of a community-based inception cohort,”BMC gastroenterology, vol. 19, no. 1, p. 47, 2019

    2019

  7. [7]

    Systematic review with meta-analysis: time to diagnosis and the impact of delayed diagnosis on clinical outcomes in inflammatory bowel disease,

    N. Jayasooriya, S. Baillie, J. Blackwell, A. Bottle, I. Petersen, H. Creese, S. Saxena, R. C. Pollok, and P.-I. study group, “Systematic review with meta-analysis: time to diagnosis and the impact of delayed diagnosis on clinical outcomes in inflammatory bowel disease,”Alimentary pharma- cology & therapeutics, vol. 57, no. 6, pp. 635–652, 2023

    2023

  8. [8]

    Percentage of having a wellness visit in past 12 months for adults aged 18 and over, United States, 2019—2024,

    National Center for Health Statistics, “Percentage of having a wellness visit in past 12 months for adults aged 18 and over, United States, 2019—2024,” 2024, accessed: 2025-07-20. [Online]. Available: https://wwwn.cdc.gov/NHISDataQueryTool/SHS adult/index.html

    2019

  9. [9]

    No health insurance coverage among people under age 65, by selected characteristics: United States, selected years 1984–2019,

    Centers for Disease Control and Prevention, “No health insurance coverage among people under age 65, by selected characteristics: United States, selected years 1984–2019,” 2021, accessed: 2025- 07-20. [Online]. Available: https://www.cdc.gov/nchs/data/hus/2020- 2021/HINone.pdf

    1984

  10. [10]

    Con- trastive transfer learning for prediction of adverse events in hospitalized patients,

    H. Salehinejad, A. M. Meehan, P. J. Caraballo, and B. J. Borah, “Con- trastive transfer learning for prediction of adverse events in hospitalized patients,”IEEE Journal of Translational Engineering in Health and Medicine, vol. 12, pp. 215–224, 2023

    2023

  11. [11]

    Novel machine learning model to improve performance of an early warning system in hospitalized patients: a ret- rospective multisite cross-validation study,

    H. Salehinejad, A. M. Meehan, P. A. Rahman, M. A. Core, B. J. Borah, and P. J. Caraballo, “Novel machine learning model to improve performance of an early warning system in hospitalized patients: a ret- rospective multisite cross-validation study,”EClinicalMedicine, vol. 66, 2023

    2023

  12. [12]

    Deep learning on Hester Davis scores for inpatient fall prediction,

    H. Salehinejad, R. Rojas, K. Iheasirim, M. Yousufuddin, and B. Borah, “Deep learning on Hester Davis scores for inpatient fall prediction,” in 2025 IEEE Symposium on Computational Intelligence in Health and Medicine (CIHM). IEEE, 2025, pp. 1–6

    2025

  13. [13]

    Graph representation of postoperative patients for opioids refill prediction: A real-world case study,

    A. Choudhary, C. A. Thiels, and H. Salehinejad, “Graph representation of postoperative patients for opioids refill prediction: A real-world case study,” in2024 46th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC). IEEE, 2024, pp. 1–4

    2024

  14. [14]

    GRAM: graph-based attention model for healthcare representation learning,

    E. Choi, M. T. Bahadori, L. Song, W. F. Stewart, and J. Sun, “GRAM: graph-based attention model for healthcare representation learning,” in Proceedings of the 23rd ACM SIGKDD international conference on knowledge discovery and data mining, 2017, pp. 787–795

    2017

  15. [15]

    Graphcare: Enhancing health- care predictions with personalized knowledge graphs,

    P. Jiang, C. Xiao, A. Cross, and J. Sun, “Graphcare: Enhancing health- care predictions with personalized knowledge graphs,”arXiv preprint arXiv:2305.12788, 2023

    work page arXiv 2023

  16. [16]

    Learning the graphical structure of electronic health records with graph convolutional transformer,

    E. Choi, Z. Xu, Y . Li, M. Dusenberry, G. Flores, E. Xue, and A. Dai, “Learning the graphical structure of electronic health records with graph convolutional transformer,” inProceedings of the AAAI conference on artificial intelligence, vol. 34, 2020, pp. 606–613

    2020

  17. [17]

    Dis- entangled dynamic heterogeneous graph learning for opioid overdose prediction,

    Q. Wen, Z. Ouyang, J. Zhang, Y . Qian, Y . Ye, and C. Zhang, “Dis- entangled dynamic heterogeneous graph learning for opioid overdose prediction,” inProceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 2022, pp. 2009–2019

    2022

  18. [18]

    Predictive modeling with tem- poral graphical representation on electronic health records,

    J. Chen, C. Yin, Y . Wang, and P. Zhang, “Predictive modeling with tem- poral graphical representation on electronic health records,” inIJCAI: proceedings of the conference, vol. 2024, 2024, p. 5763

    2024

  19. [19]

    Inductive representation learning on large graphs,

    W. Hamilton, Z. Ying, and J. Leskovec, “Inductive representation learning on large graphs,”Advances in neural information processing systems, vol. 30, 2017

    2017

  20. [20]

    Deep learning,

    Y . LeCun, Y . Bengio, and G. Hinton, “Deep learning,”Nature, vol. 521, no. 7553, pp. 436–444, 2015

    2015

  21. [21]

    Layer Normalization

    J. L. Ba, J. R. Kiros, and G. E. Hinton, “Layer normalization,”arXiv preprint arXiv:1607.06450, 2016

    work page internal anchor Pith review Pith/arXiv arXiv 2016

  22. [22]

    Adam: A Method for Stochastic Optimization

    D. P. Kingma and J. Ba, “Adam: A method for stochastic optimization,” arXiv preprint arXiv:1412.6980, 2014

    work page internal anchor Pith review Pith/arXiv arXiv 2014

  23. [23]

    Graph Attention Networks

    P. Veli ˇckovi´c, G. Cucurull, A. Casanova, A. Romero, P. Lio, and Y . Ben- gio, “Graph attention networks,”arXiv preprint arXiv:1710.10903, 2017

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  24. [24]

    Semi-Supervised Classification with Graph Convolutional Networks

    T. N. Kipf and M. Welling, “Semi-supervised classification with graph convolutional networks,”arXiv preprint arXiv:1609.02907, 2016

    work page internal anchor Pith review Pith/arXiv arXiv 2016

  25. [25]

    Attention is all you need,

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, and I. Polosukhin, “Attention is all you need,”Advances in neural information processing systems, vol. 30, 2017

    2017

  26. [26]

    Long short-term memory,

    S. Hochreiter and J. Schmidhuber, “Long short-term memory,”Neural computation, vol. 9, no. 8, pp. 1735–1780, 1997

    1997

  27. [27]

    BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding

    J. Devlin, M.-W. Chang, K. Lee, and K. Toutanova, “BERT: Pre-training of deep bidirectional transformers for language understanding,”arXiv preprint arXiv:1810.04805, 2018

    work page internal anchor Pith review Pith/arXiv arXiv 2018