- Rating
- 2
- Confidence
- 2
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 2
Read review
Paper summary
The paper proposes HiSTOR for longitudinal potential-outcome prediction with high-dimensional covariates. It combines a current-covariate encoder, a causal temporal Transformer, and a four-part latent representation for shared, outcome-specific, treatment-specific, and residual-reconstruction information. Factual outcome prediction, auxiliary treatment prediction, reconstruction, adversarial distribution matching, and an optional rollout loss are used for training, with one-step, autoregressive, and direct multi-step prediction. Experiments cover CancerSim, CVSim, and a semi-synthetic MIMIC-III study. However, the main components are established causal representation-learning and longitudinal EHR techniques. The CVSim extension adds noisy measurements of an already observed low-dimensional factor rather than new confounding information.
Critical strengths
-
The paper introduces a structured representation framework for high-dimensional longitudinal causal inference, encoding the current patient state and prior history by their treatment and outcome roles, and supports both one-step and multi-step treatment-conditioned prediction.
-
The empirical evaluation covers varying confounding strengths, covariate dimensions, and prediction horizons, with HiSTOR reporting the lowest mean errors across the tested CancerSim and CVSim settings.
Critical weaknesses
-
The proposed method largely recombine established components: causal temporal Transformers, time embeddings, shared/treatment/outcome latent decomposition, task-specific heads, adversarial separation, and reconstruction. These ideas already appear in existing studies. The paper does not show that the four-branch routing outperforms a shared multi-task representation or introduces a new identification mechanism. The novelty is somewhat weak.
-
The four latent branches are slices of one MLP output and are separated only by which heads read them. There is no independence, orthogonality, balancing, or sufficiency constraint, so z_t^0 may omit confounders or duplicate information stored elsewhere. The method still depends on unmeasured-confounding assumptions.
-
The “w/o Reconstruction” ablation does not isolate the residual branch z_t^3; it removes the reconstruction objective itself. Since the direct decoder uses all four branches, the performance change cannot be attributed specifically to z_t^3. This ablation should preserve both the reconstruction loss and parameter count while removing only the residual branch.
-
The loss weighting hyperparameters appearing in the equations are fixed without sensitivity analysis. In particular, , , , , , and the gradient-penalty weight each use one setting, while only the adversarial schedule is specified.
Questions for authors
Can the authors provide a real-data off-policy evaluation using target-trial emulation or a doubly robust estimator, including overlap diagnostics and sensitivity analysis? If this is not feasible, please state explicitly that the real MIMIC-IV evaluation is factual forecasting only and does not establish treatment-effect validity.
4 comments
Reviewer3_part2of4
Response to Reviewer 3 (2/4)
W2. The branches are slices of one MLP output without independence, orthogonality or sufficiency constraints.
An independence or orthogonality penalty between the slots is deliberately not part of the method, for three reasons.
First, it is not what the estimand requires. Under sequential ignorability (Eq. 3), a representation supports valid adjustment if it is sufficient for the history with respect to the outcome (a prognostic score, Hansen, 2008) or with respect to treatment assignment (a balancing score, Rosenbaum and Rubin, 1983). Both are sufficiency conditions on what a slot contains, and neither involves independence between slots. In HiSTOR these conditions are targeted directly by the supervision: the outcome loss trains (, ) to be predictive of the outcome and the treatment loss trains (, ) to be predictive of treatment.
Second, statistical independence between latent blocks is not identifiable from observational data without further inductive bias or supervision (Locatello et al., 2019), and linear orthogonality can always be obtained by a rotation of the latent without changing any prediction. A penalty of this kind would therefore enforce a property that cannot be verified and that has no bearing on the estimand.
Third, such penalties can remove information that adjustment needs. Representations constrained to be invariant or balanced are in general not invertible, and the information they discard induces confounding bias (Johansson et al., 2019, Melnychuk et al., 2024). HiSTOR takes the opposite route and keeps the latent information-preserving: the reconstruction objective and the residual slot retain covariate information that neither head needs, and the plan decoder reads all four slots, so nothing is discarded from the latent.
The design is thus structure learning by routing: the outcome loss can only shape (, ), the treatment loss only (, ), and the reconstruction residual only , so the supervision signals are assigned to different subspaces by construction, which makes each head's problem easier to learn than with one shared vector. The adversarial term complements this at the level of the covariate code: it matches the compressed code to an isotropic Gaussian prior, which pushes its coordinates towards independence. On confounders missed by : is read by both heads and is the only slot trained by both losses, so information relevant to both treatment and outcome is what it is trained to carry. Like all methods compared, HiSTOR relies on sequential ignorability (Eq. 3). This is an assumption of the setting, not of our method. We will add this discussion to the appendix.
W3. "w/o Reconstruction" removes the whole reconstruction objective and cannot isolate . The decoder reads all four branches.
We ran both controls the reviewer suggests: (i) the same frozen backbone with a decoder that reads only
$z_0$, $z_1$, $z_2$, and (ii) a backbone without the residual slot, with the 64 dimensions reallocated so that the total width is unchanged. For (i), the five-horizon error rises by 0.7% on CancerSim ( = 4) and by 4.5% and 1.5% on CVSim Static and Temporal ( = 2000), none significant (five paired seeds). For (ii), it changes by -4.0% on CancerSim (paired interval excludes zero) and by +3.9% and -5.3% on CVSim Static and Temporal (not significant). The residual slot therefore matters little where the covariates are low-dimensional, as on CancerSim, and we will report both controls next to the existing "w/o Reconstruction" row so that the three effects can be read separately.
| Model (one-step / five-horizon) | CancerSim = 4 | CVSim Static = 2000 | CVSim Temporal = 2000 |
|---|---|---|---|
| HiSTOR | 0.666 / 0.904 | 0.876 / 1.150 | 0.845 / 1.158 |
| Decoder reads |
only | 0.666 / 0.911 (+0.7%) | 0.876 / 1.201 (+4.5%) | 0.845 / 1.176 (+1.5%) | | Backbone without the residual slot | 0.653 (-1.9%) / 0.868 (-4.0%*) | 0.889 (+1.5%) / 1.195 (+3.9%) | 0.857 (+1.4%) / 1.097 (-5.3%) |
In (i) the backbone is frozen and shared with HiSTOR, so the one-step error is identical. In (i) on CVSim Static the increase grows with the horizon (-4.2%, -0.1%, +6.6%, +6.5% and +10.7% at = 1, …, 5).
Reviewer3_part3of4
Response to Reviewer 3 (3/4)
W4. No sensitivity analysis of the loss weights.
We agree that the paper did not list all loss weights and reported no sensitivity analysis. we list all weights below, including = 1,together with a sensitivity study. = 10 with three critic updates is the WGAN-GP default (Gulrajani et al., 2017), and = = 0.01 are used unchanged on both simulators. We ran a one-at-a-time study over 0.1$\times$ and 10$\times$ of each weight and three alternative branch-width allocations, on CancerSim = 4 and CVSim = 2000 Static with five paired seeds. Branch widths (16/16/32/96, 32/32/64/32, 40/40/40/40) change the five-horizon error by -4.1% to +3.3%. Across all loss weights the five-horizon error changes by -6.0% to +13.0% (CancerSim) and -2.1% to +12.8% (CVSim). The largest increases (11-13%) come from the tenfold reconstruction and adversarial weights, = 1 and = 1 on CVSim, and the only increase outside its paired interval is the tenfold adversarial weight on CancerSim (+4.2%). Every configuration keeps a lower five-horizon error than the best baseline in the paper's tables (at most 1.02 vs. 1.07 on CancerSim and 1.30 vs. 1.67 on CVSim). At one step the same holds except for the two tenfold weights on CVSim (1.01 and 1.05 vs. 0.95 for IGC-Net). The full table will be added to the appendix.
| Changed setting | CancerSim = 4, one-step | CancerSim = 4, five-horizon | CVSim Static = 2000, one-step | CVSim Static = 2000, five-horizon |
|---|---|---|---|---|
| Submitted | 0.666 | 0.904 | 0.876 | 1.150 |
| = 0.001 | 0.660 (-0.9%) | 0.878 (-2.8%) | 0.891 (+1.7%) | 1.232 (+7.1%) |
| = 0.1 | 0.733 (+10.0%*) | 0.942 (+4.2%*) | 1.049 (+19.8%) | 1.280 (+11.3%) |
| = 0.001 | 0.687 (+3.1%) | 0.920 (+1.7%) | 0.833 (-0.5%) | 1.220 (+8.9%) |
| = 0.1 | 0.668 (+0.2%) | 0.923 (+2.1%) | 0.902 (+2.9%) | 1.239 (+7.8%) |
| = 0.1 | 0.685 (+2.8%) | 0.875 (-3.3%) | 0.880 (+0.5%) | 1.125 (-2.1%) |
| = 10 | 0.710 (+6.6%) | 1.022 (+13.0%) | 1.010 (+15.3%) | 1.297 (+12.8%) |
| = 1 | 0.668 (+0.3%) | 0.906 (+0.2%) | 0.846 (+1.0%) | 1.253 (+11.8%) |
| = 0 | 0.647 (-2.9%) | 0.893 (-1.2%) | 0.897 (+2.4%) | 1.164 (+1.3%) |
| = 0.001 | n/a | n/a | 0.883 (+0.8%) | 1.238 (+7.6%) |
| = 0.01 | 0.642 (-3.6%*) | 0.850 (-6.0%*) | submitted | submitted |
| = 0.1 | 0.662 (-0.6%) | 0.869 (-3.9%) | 0.884 (+0.9%) | 1.141 (-0.7%) |
| = 1 | submitted | submitted | 0.943 (+7.7%) | 1.279 (+11.3%) |
| = 0 | submitted | submitted | 0.853 (-2.6%*) | 1.171 (+1.8%) |
| = 0.1 | n/a | n/a | 0.873 (-0.4%) | 1.235 (+7.4%) |
| = 0.5 | n/a | n/a | 0.920 (+5.1%*) | 1.238 (+7.7%) |
| Widths 16/16/32/96 | 0.649 (-2.5%) | 0.873 (-3.5%) | 0.833 (-0.5%) | 1.133 (+1.1%) |
| Widths 32/32/64/32 | 0.652 (-2.1%) | 0.867 (-4.1%) | 0.844 (+0.8%) | 1.157 (+3.3%) |
| Widths 40/40/40/40 | 0.671 (+0.7%) | 0.880 (-2.6%) | 0.845 (+0.8%) | 1.132 (+1.0%) |
Submitted values: = = 0.01, = 1, = 10, widths 24/24/48/64. = 1, = 0 on CancerSim and = 0.01, = 0.25 on CVSim. Relative changes are paired by seed. On CVSim the rows for = 0.001, = 1 and the three width allocations use four seeds, so their means are not directly comparable with the five-seed submitted row, while the relative changes are.
Reviewer3_part4of4
Response to Reviewer 3 (4/4)
Q1. Off-policy evaluation on real data (MIMIC-IV).
Our clinical study uses MIMIC-III covariates with semi-synthetic treatments and outcomes (the IGC-Net generator), not MIMIC-IV. Appendix I states that it "is a semi-synthetic counterfactual benchmark, not a real-outcome clinical prediction experiment". This is the standard protocol of CT, Causal CPC and IGC-Net, and it exists because counterfactual accuracy cannot be scored on real outcomes: off-policy estimates on real data are themselves model-based and would add unverifiable assumptions to the comparison. The semi-synthetic design keeps real covariate complexity while retaining ground truth. We will make the semi-synthetic nature of this study more prominent in the main text, and list target-trial emulation with overlap diagnostics as future work.
We will update the revised paper with all of the clarifications and new results above.
Reviewer3_part1of4
Response to Reviewer 3 (1/4)
We thank the reviewer for the careful reading. We clarify each point below. All new experiments reported below use the submitted model, and the main result is unchanged by them: HiSTOR has the lowest five-horizon error in all 23 simulation settings of the paper (18 CVSim, 5 CancerSim). Unless stated otherwise, every new number is NRMSE (%) averaged over five seeds, every relative change is paired by seed against the submitted model trained in the same batch of runs, and * marks a change whose 95% paired t-interval excludes zero. "One-step" is the one-step-ahead error and "five-horizon" is the mean over the five multi-step horizons, as in the paper. In these reruns the submitted model obtains 0.666 / 0.904 on CancerSim γ = 4, 0.876 / 1.150 on CVSim Static dhd = 2000 and 0.845 / 1.158 on CVSim Temporal dhd = 2000 (one-step / five-horizon), against 0.670 / 0.911, 0.882 / 1.197 and 0.845 / 1.167 in the paper's tables.
W1. The method recombines existing components. No evidence that four-branch routing beats a shared representation.
None of CRN, CT, Causal CPC or IGC-Net compresses the current high-dimensional state separately from the history, and none is competitive on both tasks at dhd ≥ 200. HiSTOR's combination (separate compression of the current state, reconstruction-retained compression with a residual slot, and a frozen-representation, plan-conditioned direct decoder) is what yields the lowest five-horizon error in all 23 simulation settings, with margins of 22-28% over the best baseline for dhd ≥ 200.
The question of routing versus a shared representation is answered by the equal-width comparison in which every head reads the whole latent, run in eight data settings with three paired seeds. Under the paper's loss weights the two are statistically indistinguishable, so routing costs no accuracy. With a strong treatment term (λX = 1 instead of the submitted 0.01) on CVSim Static at dhd = 2000 with the patient factor withheld from all models, routing is 10.0% better over five horizons and 4.7% better one-step (paired intervals exclude zero). The components that the new controls isolate account for the accuracy. At dhd = 2000, removing the history encoder raises the five-horizon error by 44.7% (Static) and 15.1% (Temporal), removing the current-state path raises the one-step error by 27.5% and 30.8% and the five-horizon error by 15.8% and 12.4%, and autoregressive decoding is 4.8% and 5.5% worse than direct decoding over five horizons.