- Rating
- 3(+1)
- Confidence
- 2
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 2
Read review
Paper summary
This paper explores how to jointly model the high-order clinical relationships within a single visit in electronic health records and the long-term dependencies across visits. The authors first construct an ultra-knowledge graph using the semantic grouping of medical concepts, and based on this, generate a patient-level ultra-graph. Subsequently, the model extracts structural representations within the visit through multiple parallel ARMA ultra-graph neural networks, and then inputs the representations of visit nodes into a state space model based on HiPPO to form the patient's longitudinal representation, which is used for tasks such as death prediction, hospital stay duration prediction, and drug recommendation. The main value of this paper lies in integrating the encoding of ultra-graph structure and HiPPO temporal modeling into a framework applicable to longitudinal EHR. However, there are still some theoretical correspondence relationships that need to be further clarified between Sobolev projection, memory update under dynamic topology, and the final nonlinear ARMA encoder and the patient-level state.
Critical strengths
- The paper distinguishes between two complementary modeling problems: the joint relationships among multiple clinical concepts within a single visit, and the longitudinal dependencies between different visits of patients. Representing the former with hypergraphs and the latter with state space models is a reasonable and understandable design. The paper further focuses on the potential over-smoothing problem caused by conventional hypergraph convolutions, and attempts to retain the distinguishable node information through more flexible spatial encoding.
- The method has a certain degree of theoretical organization. Starting from the HiPPO projection problem with Sobolev regularity terms, a closed-form solution is provided to establish the connection between the hypergraph structure and the historical memory coefficients.
- The experiment covered multiple prediction tasks and achieved significant performance improvements. The main results are highly appealing.
- The paper not only compares the complete model with external baselines, but also includes comparisons between ordinary graphs and hypergraphs, hypergraph convolution and ARMA, SSM and LSTM, as well as analyses for longer medical history. The appendix further examines the medical ontology, co-occurrence clustering, and the hypergraph construction methods of different LLMs.
Critical weaknesses
- There is still a lack of clear boundary connection between the state update in the dynamic topology and the Sobolev regularized projection solution. The author has already explained in Remark that the derivative of the topological change boundary is not calculated, and a strategy of "first diffusing the current input and then updating the memory" is adopted. This explanation can define a reasonable piecewise state space model, but it has not fully explained whether this model still corresponds to the projection optimal solution of Theorem 3.1 at each moment.
- The evidence of long-term reliance on modeling is still relatively limited, and it cannot fully explain the gains on the overall dataset. The author compared SSM, LSTM, and the model without time-series aggregation on a subset of patients who had more than 10 visits. The results supported the role of SSM in this setting. However, this experiment did not fully report the subset size, uncertainty estimation, and the threshold settings corresponding to sensitivity, making it difficult to judge the improved stability. Additionally, the overall cohort contains a large number of short visit sequences. The larger overall improvement on MIMIC-IV cannot be attributed solely to the dataset containing more long-term patients, due to the long-term reliance on modeling. More direct evidence should include results stratified by historical length and the performance changes after gradually truncating the history.
- The relationship between the Sobolev filter and the final nonlinear ARMA-SSM implementation is not yet fully established. The direct correspondence established in the appendix mainly pertains to the linear recursion when , and relies on appropriate sharing coefficients and convergence conditions. However, the actual model uses finite-layer recursion, nonlinear activation, and learnable feature transformation matrices.
Questions for authors
- When , how does Equation (5) maintain ? Are there any implicit boundary state transformations?
- After adopting the readout at the treatment level and the finite-depth nonlinear ARMA, which Sobolev–HiPPO theoretical results still hold strictly?
- Under the same knowledge, residual structure and parameter budget, how much is the current ARMA module's advantage over other hypergraph encoders?
- Can the size of the long historical subset, confidence intervals, and historical truncation experiments be supplemented?
- Do all baselines use the same task-specific patient cohort, prediction time point, and input truncation rules?
If the author addresses my concerns and clarifies my doubts, I will consider upgrading to 3.
10 comments
W2 and Q4 Long range
Repeated-run robustness: We additionally report mean ± standard deviation across 10 runs on long-range mortality prediction task on MIMIC-IV dataset. Specifically, it includes 59,262 patients, 964 have >10 visits, yielding 5,912 eligible prediction samples with 208 positive outcomes. The patient-disjoint train/validation/test splits contain 670/153/141 patients, 3,970/977/965 samples, and 144/36/28 positive outcomes, respectively. Importantly, we included results on SSM variants for EHR data (EHRMAMBA), whereas graph based SSM results are already presented in Table 3 of the manuscript.
Mean and standard deviation results across multiple seed runs.
| Model | AUROC (%) | Macro-F1 (%) |
|---|---|---|
| Hyper-GNN | 60.6 ± 0.3 | 62.8 ± 1.2 |
| SoAR-LSTM | 65.1 ± 0.5 | 68.9 ± 0.9 |
| HypEHR | 63.2 ± 0.7 | 67.1 ± 1.4 |
| EHRMamba | 67.6 ± 1.3 | 72.3 ± 1.7 |
| SoAR-SSM | 69.3 ± 1.2 | 76.1 ± 1.4 |
SoAR-SSM outperforms SoAR-LSTM by +4.2 AUROC/+7.2 Macro-F1 points and EHRMamba by +1.7 AUROC/+3.8 Macro-F1 points, while requiring substantially fewer parameters. The improvements are consistent with the reported run-to-run variability, with particularly clear gains in Macro-F1.
Overall, these controlled, independently tuned comparisons demonstrate that SoAR-SSM achieves superior long-range prediction performance with lower model capacity, strengthening the empirical evidence for effective and parameter-efficient temporal modeling.
To directly investigate reliance on historical information, we conduct a progressive history-truncation experiment on the MIMIC-IV long-history subset (>10 visits), retaining only the most recent () visits or the full available history.
| Model | K=1 | K=5 | K=10 | Full history |
|---|---|---|---|---|
| SoAR-SSM (AUROC) | 0.586 | 0.603 | 0.632 | 0.693 |
| SoAR-LSTM (AUROC) | 0.523 | 0.574 | 0.592 | 0.651 |
AUROC improves monotonically with increasing historical context, from 58.6% (K=1) to 69.3% (full history), an absolute improvement of 10.7 percentage points. Notably, retaining visits beyond the most recent 10 provides an additional 6.1-point improvement, indicating that earlier clinical information contributes predictive value beyond recent observations. These findings complement the SSM-versus-LSTM ablation in above table and provide direct empirical evidence supporting the model's utilization of long-range patient history.
W3, Generalization to
We thank the Reviewer for the comment. The manuscript provides the connection of Sobolev filter to ARMA-1 filter and remarks the connection to ARMA-K filters. However, for completeness we present the result for any integer . Let be a symmetric positive semidefinite hypergraph Laplacian, and define
Recall, for an eigenvalue of , its spectral response is
Now we express the relation of Sobolev filter to generalized ARMA filter.
Proposition C.1 (Integer-order Sobolev–ARMA decomposition). For every positive integer , the operator admits an exact decomposition into first-order rational branches i.e,
where , with and
Proof. Consider the polynomial as . Observe the roots of i.e., are distinct and nonzero, and , so every root is simple, as and
The partial-fraction expansion gives
Let . Each is either nonreal or negative, so every is invertible. Applying (C.2) yields
Since , we obtain and in particular is invertible. Substituting into (C.3) gives , which proves (C.1).
For the roots occur in conjugate pairs (, hence , ), so each pair sums to a real second-order rational branch and, for odd , the real root gives one real first-order branch; thus is realizable with real weights.
Connection to the rational ARMA family. For integer ,
Dividing the numerator and denominator by obtains the rational ARMA family.
Proposition C.2 (Recursive realization and truncation error). For an input , initialize consider
Further if for every , then
Whereas the error between the representation are bounded as
Proof. Unrolling (C.4) gives
Since , the Neumann series converges to , and satisfies
.
Summing these bounds and applying Proposition C.1 proves the result.
Recovery of the first-order construction. For , , which gives
Thus , recovering the original first-order Sobolev–ARMA connection. Here , so the recursion converges for every .
For completeness, detailed connection of to generalized ARMA filter is included in Appendix C
Q1 and W1, On memory update at topology change
We thank the reviewer for the encouraging and insightful review. Theorem 3.1 characterizes instantaneous Sobolev-optimal coefficients; Eqs. (5) and (9) define causal memory of topology-filtered visits. Their state conventions differ at topology changes.
1. Topology switches and Eq. (5). At visit k, Theorem 3.1 gives the exact optimizer of Eq. (2):
Skinst=RkCk,Rk=[I+λ(LˉH,k+ϵI)β]−1.
On aligned nodes, maintaining this identity across a switch with continuous C requires
Sinst(tk−1+)=RkRk−1−1Sinst(tk−1−).
SoAR-SSM instead carries accumulated memory forward unchanged and uses the current topology only to filter new information. Under zero-order hold, R(t)=Rk and X(t)=Xk on (tk−1,tk], with
S˙(t)=S(t)A⊤+RkXkB⊤,S(tk−1+)=S(tk−1−).
Exact integration yields Eq. (5):
Sk=Sk−1Aˉk+RkXkB⊤Gk,
where Aˉk=eΔkA⊤ and Gk=∫0ΔkeuA⊤,du.
This matches S=RkC dynamics within constant-topology intervals but preserves memory across switches. Thus Eq. (5) is an exact ZOH discretization of the specified filtered-input dynamics, not a claim that Sk=RkCk continues to hold under arbitrary topology changes. This also follows from the diffuse-then-update recurrence structure of (Li et., al., 2024, ref line no:599), however adapted to our modelling strategy. Additionally, we also provide a variational characterization supporting the diffusion then update.
2. Sobolev/HiPPO variational meaning. Let Qk=(LˉH,k+ϵI)β⪰0. Then each visit has the exact spatial Sobolev optimizer as
Zk⋆=argminZ[21∥Z−Xk∥F2+2λTr(Z⊤QkZ)]=RkXk.
The compressed memory follows zq(τ)=R(τ)xq(τ), where R(τ) is the topology at observation time. For an orthonormal HiPPO basis ptK and its measure μtK,
Mq,K⋆=argminM21∫0tK∥zq(τ)−MptK(τ)∥22,dμtK(τ),
which, by orthonormality, equals Mq,K⋆=∫0tKzq(τ)ptK(τ)⊤,dμtK(τ).
A matching HiPPO state-space realization has the filtered-input structure of Eq. (5). Hence, in the idealized linear setting, the two-stage interpretation is Sobolev-regularize each visit, then temporally compress it. Theorem 3.1 asserts the latter's instantaneous optimality, not equality of the two histories after topology changes.
3. Exact discrepancy characterization. On a common aligned node space with matching transitions, zero initialization and Uk=XkB⊤Gk, unrolling gives
SK−RKCK=∑k=1K(Rk−RK)Uk(∏j=k+1KAˉj),
and the resolvent identity gives
Rk−RK=λRk(QK−Qk)RK.
Since Qk⪰0 implies ∥Rk∥2≤1,
∥SK−RKCK∥F≤λ∑k=1K∥QK−Qk∥2,∥Uk∥F,∏j=k+1KAˉj2.
Thus the two states coincide for static topology; otherwise their difference is controlled by operator variation, input strength and temporal propagation.
4. Eq. (9) and actual implementation. Each visit is processed independently in its own hypergraph:
Zk=fθ(Lk,Xk),uk=ev,k⊤Zk∈R1×d, where fθ is the ARMA encoder and
sk=sk−1Aˉk+ukB⊤Gk, The model stores one fixed-dimensional patient state, not persistent states for individual hypergraph nodes. Variable node identities/counts do not require cross-topology basis transformations: all topology-specific reasoning occurs before the visit embedding enters patient memory. With aligned nodes, exact Rk and a fixed selector ev, Eq. (9) is the left projection sk=ev⊤Sk of Eq. (5); with changing node sets/selectors it directly defines a fixed-dimensional patient model rather than a lossless projection of node-wise memory. As mentioned, maintaining per node wise statistics is difficult due to changing concept sets. The design is motivated from downstream tasks where all our tasks are patient-level and would require a pooling step. As Linear pooling commutes with propagation, given the readout is fixed, performing pooling through a fixed visit coordinate is lossless (detailed derivation below) with respect to the propagated dynamics.
We made this distinction clear in Secs. 3.2, 3.4 of the revised draft. To summarize, Theorem 3.1 gives exact instantaneous spatial regularization; Eq. (5) gives exact linear ZOH carried-memory evolution; Eq. (9) implements the corresponding causal diffuse-then-update principle with shared patient-level memory.