Pith. sign in

REVIEW 3 major objections 5 minor 80 references

ExOSITO: Explainable Off-Policy Learning with Side Information for Intensive Care Unit Blood Test Orders

T0 review · 3 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read ExOSITO learns ICU lab orders that cut cost without missing vital tests.

desk verdict A useful, reproducible clinician-facing bandit for ICU lab ordering whose core GPS reliability claim is contradicted by its own Table 1. read the letter →

arxiv 2504.17277 v1 pith:H4UUNABM submitted 2025-04-24 cs.LG cs.AI

Zongliang Ji , Andre Carlos Kajdacsy-Balla Amaral , Anna Goldenberg , Rahul G. Krishnan This is my paper · ORCID
classification cs.LGcs.AI
keywords off-policylearningcontextualbanditsICUlabtestorderingclinicaldecisionsupportglobalpropensityscoretime-seriesforecastingMIMIC-IVHiRID
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper tries to establish that ICU lab test ordering can be improved without sacrificing essential tests by reframing it as an off-policy contextual bandit whose reward is built for clinicians rather than for mortality. It claims that a policy balancing how much a test is expected to change a patient's lab values, how well an order respects clinical rule-based bounds, and what each test costs beats the observed physician policy and prior RL methods on two large ICU datasets. The practical point is that such a tool could reduce over-ordering, with its costs, patient discomfort, and environmental waste, while keeping the clinician as the final decision maker and making each recommendation explainable.

What carries the argument

The load-bearing object is the lab order utility function $g(t,x)=\Delta X(t,x)-\beta_1 L_b(t,x)-\beta_2 C(t)$, where $\Delta X$ scores the mean and range change between observed past and predicted future lab values, $L_b$ penalizes deviations from rule-derived minimal and maximal order bounds, and $C$ charges test costs. It carries the argument by turning whether an order is useful to a clinician into a differentiable objective. Around it, the method builds a PatchTST forecasting model to supply predicted future patient status for both the reward and explainability, and a conditional normalizing flow estimate of the global propensity score (the conditional density of test orders given patient status) that enforces overlap through a Lagrangian constraint, keeping the learned policy inside actions the data can support.

What would settle it

Run a silent ICU deployment: log ExOSITO's daily recommended orders without acting on them, and check whether the tests it omits but physicians ordered are followed within 48 hours by a clinically significant event such as a new vasopressor start, transfusion, arrhythmia, or death. If omitted tests are systematically followed by such events, the paper's claim that no vital lab orders are omitted is contradicted.

Watch

Extended reading notes

Core claim

On its own terms, ExOSITO's central claim is that a reliable, explainable lab-ordering policy can be learned offline from routine EHR data, and that this policy dominates the observed physician policy: it yields higher lab-test informativeness (measured by mean and extreme-value changes between predicted future and observed past lab values), lower cost, and fewer out-of-bounds orders relative to clinician-derived minimal and maximal order sets. The paper further claims that adding a learned global propensity score constraint raises the average total outcome by about 12% compared with training without the constraint, and that prior RL methods using mortality as reward are ill-suited because lab tests help clinicians decide rather than directly changing patient outcomes.

Load-bearing premise

The pivotal assumption is that the paper's reward formula -- how much a test's values are predicted to change, how well an order matches the clinical rules, and what it costs -- actually captures what makes a lab order worth placing; if it does not, then outperforming the physician policy on this formula is not the same as improving care.

Editorial extensions

If this is right

  • Daily ICU lab ordering can be reduced in cost and volume without increasing the number of missed rule-required tests, because the learned policy holds orders inside clinician-derived bounds.
  • Mortality-based RL rewards are the wrong signal for lab-ordering decisions; an informativeness-plus-bounds-plus-cost objective is what lets a learned policy beat the physician policy.
  • Adding the GPS overlap constraint yields a more reliable policy; the paper reports roughly 12% higher average total outcome for GPS-constrained than unconstrained training.
  • Because each recommendation is tied to concrete patient data and a predicted future trajectory, clinicians can audit why a test was recommended, which is what makes clinician-facing deployment plausible.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Editorial inference: the same template of predicted change, rule bounds, and cost could transfer to other recurring diagnostic-ordering decisions such as imaging or microbiology where the clinician stays the decision maker and overuse is driven by precaution.
  • A testable extension the paper does not run: compare ExOSITO against a purely rule-based ordering policy under the same utility function, which would isolate how much of the gain comes from learned patient-specific adaptation rather than from the rules themselves.
  • The paper's admission that the data contain no true utility outcomes implies the decisive validation would be a silent trial in which recommendations are logged but not enacted, and omitted tests are checked against downstream deterioration.
  • Left implicit: since the reward includes predicted future lab values, the policy should improve as forecasting accuracy improves; the ablations point that way but do not map the sensitivity curve.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper proposes ExOSITO, a contextual-bandit method for daily ICU lab test ordering. It augments each patient's 48-hour history with a PatchTST forecast of the next 24 hours, derives per-stay lower and upper bounds on ten test orders from clinician-curated rules (Algorithm 1), and defines a lab-order utility g = ΔX − β1 Lb − β2 C (Eq. 5) that rewards predicted changes in lab values, penalizes deviation from the bounds, and penalizes cost. A conditional normalizing flow estimates the propensity of the logging policy, and Algorithm 2 trains a policy by a Lagrangian objective intended to constrain the learned policy to actions with estimated propensity above a threshold ε. The method is evaluated on MIMIC-IV and HiRID, comparing against random, physician, bound, and prior RL policies, and the paper claims lower cost, better information, and fewer out-of-bound orders than the physician, with a 12% improvement from the GPS constraint.

Significance. Strengths: the problem framing as a clinician-facing assistive bandit is appropriate; the use of clinical rules as side information and forecasts as explainability is a useful idea; the manuscript includes code, two large real-world datasets, and extensive ablations. If the GPS constraint actually improved reliability and if the reward captured clinical utility, this would be a valuable contribution. As it stands, the numerical evidence in the paper's own tables contradicts the main GPS claim, the evaluation is circular with respect to the hand-crafted reward, and the paper acknowledges there are no true outcome measures, so the clinical significance is not yet established.

major comments (3)
  1. [§5.3, Table 1, Appendix F] The claim that 'the average total outcome of a reliable policy is approximately 12% higher than that of a policy trained without GPS constraints' is not supported by Tables 1 and 11 under the evaluation specified in Appendix F. With β1 = β2 = 1, the test metric is g = ΔX − Lb − Cost. On MIMIC-IV, Ours(w/o GPS) gives 0.44 − 1.06 − 0.80 = −1.42 and Ours(w GPS) gives 0.42 − 1.16 − 0.66 = −1.40, a 1.4% change, not 12%. On HiRID, the corresponding values are 1.08 − 0.62 − 0.57 = −0.11 and 1.01 − 0.89 − 0.52 = −0.40, a large degradation. Moreover, the out-of-bounds metric Lb, which the GPS constraint is supposed to reduce, increases on both datasets (MIMIC: 1.06 → 1.16; HiRID: 0.62 → 0.89). Either the 12% figure was computed under different hyperparameters, which must be reported and used consistently, or it is a miscalculation; as written, the paper's central reliability claim is contradicted by its own results.
  2. [§4.4, Eq. (7), Algorithm 2] Equation (7) as printed does not enforce the overlap constraint f̂ ≥ ε. The objective is min_θ max_{λ_i ≥ 0} −(1/n)Σ [g − λ_i(f̂ − ε)]. Expanding the inner term gives −g + λ_i(f̂ − ε), so when f̂ > ε the inner maximization is unbounded above as λ_i → ∞, and when f̂ < ε it is maximized at λ_i = 0. The λ update in Algorithm 2, λ ← λ + η_λ(f̂ − ε), likewise increases the multiplier exactly when the constraint is satisfied. The printed optimization therefore encourages f̂ ≤ ε, the opposite of the intended feasible set Π_r. Only the final validation selection (the indicator 1[f̂ > ε]) applies the constraint as stated. This needs to be corrected, and the GPS experiments re-run or re-explained, before the overlap-guaranteeing contribution can be assessed.
  3. [§4.3, §5.2, §6] The evaluation is circular with respect to the definition of clinical benefit. The policy is trained to maximize g(t,x) in Eq. (5), and the reported comparisons use the same g with β1 = β2 = 1 (Appendix F); the Lb and C terms are simultaneously training objectives and evaluation metrics, so the learned policy is being compared with the physician on its own reward. The paper states in Section 6 that the dataset has no 'true' outcome measures quantifying lab test utility. Consequently, statements such as 'reduces costs without omitting any vital lab orders' and 'outperforming both a physician's policy and prior approaches' are only claims about the hand-crafted proxy. Table 1 additionally shows nonzero Llow for both learned policies (MIMIC w GPS: 0.67; HiRID w GPS: 0.5), which contradicts the abstract's 'without omitting any vital lab orders.' The authors should either provide expert/outcome validation or substantially temper the clinical claims.
minor comments (5)
  1. [Eq. (3), §4.3] The bound loss only penalizes tests where tmax_j = tmin_j; for tests with tmax_j > tmin_j there is no penalty at all, so the sentence 'The indicator function ensures the policy includes all necessary labs suggested by rules while avoiding redundant ones' overstates the constraint that Lb actually imposes.
  2. [§5.2, Appendix F] The notation for random restarts is inconsistent: Section 5.2 says m = 5 random restarts, while Appendix F refers to k = 5 experimental runs; please unify the notation.
  3. [Appendix C] The rule 'If patient Phosphate is greater than 0.6 or greater than 1.8' is likely a typo for 'less than 0.6 or greater than 1.8'; please correct it.
  4. [Appendix A] The text says the MIMIC-IV database follows 'HIPPA'; the correct acronym is HIPAA.
  5. [Appendix G, Table 7] The row '10, 5 (Main result)' is ambiguous: please clarify whether these numbers refer to tests per 24-hour window or per 48-hour window, and align them with the description in Section 5.1.

Circularity Check

2 steps flagged · score 6.0 of 10

The central evaluation is self-referential: the lab-order utility g is both the training objective and the reported performance metric, and the ablation 'benefit' of clinical rules is measured by a penalty built from those same rules.

  1. self definitional [Section 4.3, Eq. (5); Section 5.2, Policy Training and Evaluation; Appendix F]
    "Synthesizing these desirable qualities, we define the lab test order utility function (conditional outcome function) as: g(t,x ) = ∆X(t,x )−β1Lb(t,x )−β2C(t) ... We then evaluate the policy with the best validation set performance on test set with our outcome function g(t,x)."

    The policy is obtained by maximizing g in Eqs. (6)-(7), and the paper's central claim of outperforming the physician policy is a comparison of values of the same g on the test set. The Lb and C terms are exactly the quantities the policy is trained to reduce, so 'reduces costs without omitting any vital lab orders' and 'outperforms physician' restate the optimization objective rather than an independently measured clinical benefit. Section 6 concedes that the dataset has no 'true' outcome measures quantifying lab test utility, leaving g as the only target.

  2. self definitional [Appendix H.6, Ablation Study on the Use of Privileged Information (Clinical Rules)]
    "Llower increased by about 56%: The learned policy’s behavior diverges greatly from the clinical standard, indicating poor adherence to necessary tests."

    The 'clinical standard' and 'necessary tests' are operationalized by the tmin rules from Algorithm 1, and Lb in Eq. (3) is defined as deviation from those bounds. The ablation removes tlower by setting it to zero and then reports that Llower increases; that increase is simply the removed penalty term. The conclusion that privileged clinical rules are critical is therefore a restatement of how the metric was constructed, not an independent empirical finding about clinical value.

full rationale

The paper's derivation chain is internally consistent as an optimization exercise, but its central evaluation is self-referential. Eq. (5) defines the lab-order utility g = ΔX − β1 Lb − β2 C; Eq. (6)-(7) train the policy to maximize that same g; Section 5.2 then evaluates the policy with g; and Appendix F fixes β1=β2=1 for testing. Thus the headline 'outperforms the physician policy' is a statement that a policy trained to maximize g scores higher on g than the logging policy. This is not an independent clinical validation, and the paper explicitly acknowledges that no ground-truth lab-utilization outcomes exist. The ΔX term does carry some external content through the forecasting model trained on observed data, and the held-out test split means the result is not purely a training-set artifact, which prevents the score from being higher. The rule-ablation analysis is more directly circular: removing the clinical rules removes the definition of 'necessary tests' from the metric, so the observed increase in Llower is a computation of the penalty that was removed. Separately, the specific claim that GPS raises the average total outcome by about 12% is not reproducible from Table 1 under the stated β1=β2=1 setting: recomputing g gives only about 1.4% improvement on MIMIC and a large degradation on HiRID, and GPS increases Lb on both datasets. That is a falsifiability and correctness concern rather than a circularity concern, and it is not scored here as circularity. No load-bearing self-citation or imported uniqueness theorem was found; citations to prior work by the authors are not central to the derivation.

Assumptions & free parameters 4 free parameters · 6 assumptions · 0 invented entities

The central claim rests on a hand-crafted utility function and hand-authored clinical rules, introduced without external outcome validation. Free parameters β1, β2, ε, and cost weights shape the learned policy; the clinical rules constrain the action space and the evaluation metric. The causal assumptions (ignorability, overlap, MAR, temporal independence) are standard but acknowledged as imperfect in the ICU setting.

free parameters (4)
  • beta_1 (bound penalty weight) = 1 at evaluation; searched over {0,1,10,100}
    Weights the Lb bound-adherence term in the lab order utility function g. Chosen by validation; changing it drastically alters policy behavior (Appendix H.3).
  • beta_2 (cost penalty weight) = 1 at evaluation; searched over {0,1,10,100}
    Weights the cost term C in g. Chosen by validation; large values drive the policy toward ordering nothing.
  • epsilon (GPS reliability threshold) = 5th percentile of estimated GPS on training set
    Minimum allowed propensity under the logging policy; set as a data-dependent quantile and used in Eq (7) and validation selection.
  • alpha_j cost weights = normalized [12,5,12.36,18,9.1,10,18.62,1.5,18,1.5] from literature
    Relative clinical costs of the ten tests; taken from cited literature (Kandalam 2020; Spoyalo 2023) and normalized to sum to 1. The policy's cost minimization depends on these values.
assumptions (6)
  • domain assumption Ignorability: no unobserved confounders for test ordering decisions.
    Stated in Section 3 and Appendix A.1; the authors acknowledge unobserved confounding is 'inevitable given the complexity of the ICU environment' (Discussion). If violated, the GPS-based support constraint and policy value estimates are biased.
  • standard math Overlap: f(t,x) > epsilon for all relevant (t,x).
    Standard overlap condition in causal inference, Section 3; estimated via conditional normalizing flows. Strong overlap is assumed to hold.
  • domain assumption Missing data are Missing At Random (MAR).
    Stated in Section 3 and Discussion; used to justify mean imputation in the forecasting pipeline. Questionable for EHR data where missingness often depends on clinical severity.
  • domain assumption Temporal independence of test ordering decisions.
    Bandit formulation treats each ICU stay as one time step; the paper acknowledges in Discussion that this is a strong assumption and plans semi-Markov extensions.
  • ad hoc to paper The clinical rule set CR is conservative and clinically correct.
    Rules in Appendix C are hand-crafted with expert clinicians. They define tmin and tmax (Algorithm 1), which constrain the policy action space and the Lb evaluation metric. Errors in the rules would invalidate both the training bounds and the evaluation.
  • ad hoc to paper The lab order utility function g is a valid proxy for clinical utility of test orders.
    Eq (5). g replaces any ground-truth outcome; the paper states no true outcome measures are available (Section 6). All physician-comparison claims are evaluated under g.

how reviews work

0 comments
Cite this review

Pith. "Pith review of ExOSITO: Explainable Off-Policy Learning with Side Information for Intensive Care Unit Blood Test Orders." pith.science (2026). https://pith.science/paper/H4UUNABM

@misc{pith2026250417277,
  author       = {Pith},
  title        = {Pith review of: ExOSITO: Explainable Off-Policy Learning with Side Information for Intensive Care Unit Blood Test Orders},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/H4UUNABM}},
  note         = {Machine review of arXiv:2504.17277}
}
read the original abstract

Ordering a minimal subset of lab tests for patients in the intensive care unit (ICU) can be challenging. Care teams must balance between ensuring the availability of the right information and reducing the clinical burden and costs associated with each lab test order. Most in-patient settings experience frequent over-ordering of lab tests, but are now aiming to reduce this burden on both hospital resources and the environment. This paper develops a novel method that combines off-policy learning with privileged information to identify the optimal set of ICU lab tests to order. Our approach, EXplainable Off-policy learning with Side Information for ICU blood Test Orders (ExOSITO) creates an interpretable assistive tool for clinicians to order lab tests by considering both the observed and predicted future status of each patient. We pose this problem as a causal bandit trained using offline data and a reward function derived from clinically-approved rules; we introduce a novel learning framework that integrates clinical knowledge with observational data to bridge the gap between the optimal and logging policies. The learned policy function provides interpretable clinical information and reduces costs without omitting any vital lab orders, outperforming both a physician's policy and prior approaches to this practical problem.

Figures

Figures reproduced from arXiv: 2504.17277 by the authors.

Figure 1
Figure 1. Overview of Proposed Method. Top: Development of an ICU patient status forecasting model ϕ for future predictions. Both observed past and predicted future of patient are inputs of policy πθ (green arrow), which determines the next-day lab test order for the patient. Bottom: πθ is learned by maximizing the reward function g, which evaluates each test order (red arrow) and incorporates three main components (dashed ar… view at source ↗
Figure 2
Figure 2. Integration of Medical Knowledge into Explainable Policy Learning for real patient data. Top: At de￾ployment, our policy transparently justifies actions, such as ordering a CBC test, based on predictions like a decrease in future Hemoglobin levels. Bottom: The policy incorporates clinical guidelines for lab test ordering. At test time, the policy increases test orders for deteriorating patients due to changes in tra… view at source ↗
Figure 3
Figure 3. Graphical representation of two prior works (Cheng et al., 2018; Chang et al., 2019) and our proposed methods. ity classification, we have to check whether each ICU stay’s end time is before the record time of death of the patients. In order for our model to learn meaning￾ful representation, we also eliminated the ICU stays with duration less than 12 hours and stays that has less than 5 lab tests ordered. After prep… view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Bar plot that shows the distribution of guideline-generated tests and observed tests [PITH_FULL_IMAGE:figures/full_fig_p022_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

80 extracted references · 57 canonical work pages

  1. [1]

    Systematic comparison of routine laboratory measurements with in-hospital mortality: Icu-labome, a large cohort study of critically ill patients

    Edris M Alkozai, Bakhtawar K Mahmoodi, Johan Decruyenaere, Robert J Porte, Annemieke Oude Lansink-Hartgring, Ton Lisman, and Maarten W Nijsten. Systematic comparison of routine laboratory measurements with in-hospital mortality: Icu-labome, a large cohort study of critically ill patients. Clinical Chemistry and Laboratory Medicine (CCLM), 56 0 (7): 0 1140...

    work page 2018

  2. [2]

    Evidence-based laboratory medicine

    Tony Badrick. Evidence-based laboratory medicine. The Clinical Biochemist Reviews, 34 0 (2): 0 43, 2013

    work page 2013

  3. [3]

    Eliminating catheter-related bloodstream infections in the intensive care unit

    Sean M Berenholtz, Peter J Pronovost, Pamela A Lipsett, Deborah Hobson, Karen Earsing, Jason E Farley, Shelley Milanovich, Elizabeth Garrett-Mayer, Bradford D Winters, Haya R Rubin, et al. Eliminating catheter-related bloodstream infections in the intensive care unit. Critical care medicine, 32 0 (10): 0 2014--2020, 2004

    work page 2014

  4. [4]

    Reducing test utilization in hospital settings: a narrative review

    Renuka S Bindraban, Maarten J Ten Berg, Christiana A Naaktgeboren, Mark HH Kramer, Wouter W Van Solinge, and Prabath WB Nanayakkara. Reducing test utilization in hospital settings: a narrative review. Annals of laboratory medicine, 38 0 (5): 0 402--412, 2018

    work page 2018

  5. [5]

    o khan C ak rca, Ay s e Torun, Ahmet Bindal, Murat \

    Tuba Damar C ak rca, G \"o khan C ak rca, Ay s e Torun, Ahmet Bindal, Murat \"U st \"u nel, and Ahmet Kaya. Comparing the predictive values of procalcitonin/albumin ratio and other inflammatory markers in determining covid-19 severity. Pakistan Journal of Medical Sciences, 39 0 (2): 0 450, 2023

    work page 2023

  6. [6]

    Dynamic measurement scheduling for event forecasting using deep rl

    Chun-Hao Chang, Mingjie Mai, and Anna Goldenberg. Dynamic measurement scheduling for event forecasting using deep rl. In International Conference on Machine Learning, pages 951--960. PMLR, 2019

    work page 2019

  7. [7]

    An optimal policy for patient laboratory tests in intensive care units

    Li-Fang Cheng, Niranjani Prasad, and Barbara E Engelhardt. An optimal policy for patient laboratory tests in intensive care units. In BIOCOMPUTING 2019: Proceedings of the Pacific Symposium, pages 320--331. World Scientific, 2018

    work page 2019

  8. [8]

    Reducing unnecessary lab testing in the icu with artificial intelligence

    Federico Cismondi, Leo A Celi, Andr \'e S Fialho, Susana M Vieira, Shane R Reti, Joao MC Sousa, and Stan N Finkelstein. Reducing unnecessary lab testing in the icu with artificial intelligence. International journal of medical informatics, 82 0 (5): 0 345--358, 2013

    work page 2013

Show all 80 references
  1. [9]

    G \'a lvez, Tracey Polsky, Genna Kreher, Blair Kraus, Luis M

    Maya Dewan, Jorge A. G \'a lvez, Tracey Polsky, Genna Kreher, Blair Kraus, Luis M. Ahumada, John J. McCloskey, and Heather Wolfe. Reducing unnecessary postoperative complete blood count testing in the pediatric intensive care unit. The Permanente Journal, 2016

    2016

  2. [10]

    Reducing unnecessary postoperative complete blood count testing in the pediatric intensive care unit

    Maya Dewan, Jorge Galvez, Tracey Polsky, Genna Kreher, Blair Kraus, Luis Ahumada, John McCloskey, and Heather Wolfe. Reducing unnecessary postoperative complete blood count testing in the pediatric intensive care unit. The Permanente journal, 21, 2017

    2017

  3. [11]

    Invertible generative modeling using linear rational splines

    Hadi Mohaghegh Dolatabadi, Sarah Erfani, and Christopher Leckie. Invertible generative modeling using linear rational splines. In AISTATS, 2020

    2020

  4. [12]

    Neural spline flows

    Conor Durkan, Artur Bekasov, Iain Murray, and George Papamakarios. Neural spline flows. In NeurIPS, 2019

    2019

  5. [13]

    Overlap in observational studies with high-dimensional covariates

    Alexander D’Amour, Peng Ding, Avi Feller, Lihua Lei, and Jasjeet Sekhon. Overlap in observational studies with high-dimensional covariates. Journal of Econometrics, 221 0 (2): 0 644--654, 2021

    2021

  6. [14]

    Evidence-based guidelines to eliminate repetitive laboratory testing

    Kevin P Eaton, Kathryn Levy, Christine Soong, Amit K Pahwa, Christopher Petrilli, Justin B Ziemba, Hyung J Cho, Rodrigo Alban, Jaime F Blanck, and Andrew S Parsons. Evidence-based guidelines to eliminate repetitive laboratory testing. JAMA internal medicine, 177 0 (12): 0 1833...

    2017

  7. [15]

    Tsmixer: Lightweight mlp-mixer model for multivariate time series forecasting

    Vijay Ekambaram, Arindam Jati, Nam Nguyen, Phanwadee Sinthong, and Jayant Kalagnanam. Tsmixer: Lightweight mlp-mixer model for multivariate time series forecasting. arXiv preprint arXiv:2306.09364, 2023

    2023 arXiv

  8. [16]

    Offline reinforcement learning for safer blood glucose control in people with type 1 diabetes

    Harry Emerson, Matthew Guy, and Ryan McConville. Offline reinforcement learning for safer blood glucose control in people with type 1 diabetes. Journal of Biomedical Informatics, 142: 0 104376, 2023

    2023

  9. [17]

    Managing the cost of diagnosis

    Lynn Feldman. Managing the cost of diagnosis. Manag Care, 5: 0 43--45, 2009

    2009

  10. [18]

    Off-policy evaluation with deficient support using side information

    Nicol \`o Felicioni, Maurizio Ferrari Dacrema, Marcello Restelli, and Paolo Cremonesi. Off-policy evaluation with deficient support using side information. Advances in Neural Information Processing Systems, 35: 0 30250--30264, 2022

    2022

  11. [19]

    Off-policy deep reinforcement learning without exploration

    Scott Fujimoto, David Meger, and Doina Precup. Off-policy deep reinforcement learning without exploration. In International conference on machine learning, pages 2052--2062. PMLR, 2019

    2019

  12. [20]

    Emaq: Expected-max q-learning operator for simple yet effective offline and online rl

    Seyed Kamyar Seyed Ghasemipour, Dale Schuurmans, and Shixiang Shane Gu. Emaq: Expected-max q-learning operator for simple yet effective offline and online rl. In International Conference on Machine Learning, pages 3682--3691. PMLR, 2021

    2021

  13. [21]

    Does pediatric housestaff experience influence tests ordered for infants in the neonatal intensive care unit? Critical care medicine, 25 0 (4): 0 704--709, 1997

    Charles H Griffith, John F Wilson, Nirmala S Desai, and Eugene C Rich. Does pediatric housestaff experience influence tests ordered for infants in the neonatal intensive care unit? Critical care medicine, 25 0 (4): 0 704--709, 1997

    1997

  14. [22]

    Kale, Greg Ver Steeg, and Aram Galstyan

    Hrayr Harutyunyan, Hrant Khachatrian, David C. Kale, Greg Ver Steeg, and Aram Galstyan. Multitask learning and benchmarking with clinical time series data. Scientific Data, 6 0 (1): 0 96, 2019. ISSN 2052-4463. doi:10.1038/s41597-019-0103-9. URL https://doi.org/10.1038/s41597-0...

    2019 doi

  15. [23]

    C. J. Hjortsø, M. Møller, A. Perner, and A. C. Brøchner. Routine versus on-demand blood sampling in critically ill patients: A systematic review*, 2023

    2023

  16. [24]

    Long short-term memory

    Sepp Hochreiter and J \"u rgen Schmidhuber. Long short-term memory. Neural computation, 9 0 (8): 0 1735--1780, 1997

    1997

  17. [25]

    Early prediction of circulatory failure in the intensive care unit using machine learning

    Stephanie L Hyland, Martin Faltys, Matthias H \"u ser, Xinrui Lyu, Thomas Gumbsch, Crist \'o bal Esteban, Christian Bock, Max Horn, Michael Moor, Bastian Rieck, et al. Early prediction of circulatory failure in the intensive care unit using machine learning. Nature medicine, 2...

    2020

  18. [26]

    Implementation of an on-demand strategy for routine blood testing in icu patients

    Ioulia Iosfina, Hayley Merkeley, Tara Cessford, Georgia Geller, Neda Amiri, Nazli Baradaran, Monica Norena, Najib Ayas, and Peter M Dodek. Implementation of an on-demand strategy for routine blood testing in icu patients. In D23. QUALITY IMPROVEMENT IN CRITICAL CARE, pages A53...

    2013

  19. [27]

    Way off-policy batch deep reinforcement learning of implicit human preferences in dialog

    Natasha Jaques, Asma Ghandeharioun, Judy Hanwen Shen, Craig Ferguson, Agata Lapedriza, Noah Jones, Shixiang Gu, and Rosalind Picard. Way off-policy batch deep reinforcement learning of implicit human preferences in dialog. arXiv preprint arXiv:1907.00456, 2019

    1907 arXiv

  20. [28]

    Measurement scheduling for icu patients with offline reinforcement learning

    Zongliang Ji, Anna Goldenberg, and Rahul G Krishnan. Measurement scheduling for icu patients with offline reinforcement learning. arXiv preprint arXiv:2402.07344, 2024

    2024 arXiv

  21. [29]

    Mimic-iii, a freely accessible critical care database

    Alistair EW Johnson, Tom J Pollard, Lu Shen, Li-wei H Lehman, Mengling Feng, Mohammad Ghassemi, Benjamin Moody, Peter Szolovits, Leo Anthony Celi, and Roger G Mark. Mimic-iii, a freely accessible critical care database. Scientific data, 3 0 (1): 0 1--9, 2016

    2016

  22. [30]

    Mimic-iv, a freely accessible electronic health record dataset

    Alistair EW Johnson, Lucas Bulgarelli, Lu Shen, Alvin Gayles, Ayad Shammout, Steven Horng, Tom J Pollard, Sicheng Hao, Benjamin Moody, Brian Gow, et al. Mimic-iv, a freely accessible electronic health record dataset. Scientific data, 10 0 (1): 0 1, 2023

    2023

  23. [31]

    Inappropriate repeat testing of complete blood count (cbc) and electrolyte panels in inpatients from alberta, canada

    Vijay Kandalam, Cheryl K Lau, Maggie Guo, Irene Ma, and Christopher Naugler. Inappropriate repeat testing of complete blood count (cbc) and electrolyte panels in inpatients from alberta, canada. Clinical Biochemistry, 77: 0 32--35, 2020

    2020

  24. [32]

    Reducing unnecessary testing in the intensive care unit by choosing wisely

    Ruth M Kleinpell, J Christopher Farmer, and Stephen M Pastores. Reducing unnecessary testing in the intensive care unit by choosing wisely. Acute and Critical Care, 33 0 (1): 0 1, 2018

    2018

  25. [33]

    Influence of educational, audit and feedback, system based, and incentive and penalty interventions to reduce laboratory test utilization: a systematic review

    Daniel M Kobewka, Paul E Ronksley, Jennifer A McKay, Alan J Forster, and Carl van Walraven. Influence of educational, audit and feedback, system based, and incentive and penalty interventions to reduce laboratory test utilization: a systematic review. Clinical Chemistry and La...

    2015

  26. [34]

    The artificial intelligence clinician learns optimal treatment strategies for sepsis in intensive care

    Matthieu Komorowski, Leo A Celi, Omar Badawi, Anthony C Gordon, and A Aldo Faisal. The artificial intelligence clinician learns optimal treatment strategies for sepsis in intensive care. Nature medicine, 24 0 (11): 0 1716--1720, 2018

    2018

  27. [35]

    Towards safe mechanical ventilation treatment using deep offline reinforcement learning

    Flemming Kondrup, Thomas Jiralerspong, Elaine Lau, Nathan de Lara, Jacob Shkrob, My Duc Tran, Doina Precup, and Sumana Basu. Towards safe mechanical ventilation treatment using deep offline reinforcement learning. In Proceedings of the AAAI Conference on Artificial Intelligenc...

    2023

  28. [36]

    Impact of clinical specialty on attitudes regarding overuse of inpatient laboratory testing

    Deborah Korenstein, Solomon Husain, Renee L Gennarelli, Cilian White, James N Masciale, and Benjamin R Roman. Impact of clinical specialty on attitudes regarding overuse of inpatient laboratory testing. Journal of hospital medicine, 13 0 (12): 0 844--847, 2018

    2018

  29. [37]

    Reducing unnecessary laboratory testing in the medical icu

    Nisha Kotecha, Janet M Shapiro, John Cardasis, and Gopal Narayanswami. Reducing unnecessary laboratory testing in the medical icu. The American journal of medicine, 130 0 (6): 0 648--651, 2017

    2017

  30. [38]

    Stabilizing off-policy q-learning via bootstrapping error reduction

    Aviral Kumar, Justin Fu, Matthew Soh, George Tucker, and Sergey Levine. Stabilizing off-policy q-learning via bootstrapping error reduction. Advances in Neural Information Processing Systems, 32, 2019

    2019

  31. [39]

    Effect of laboratory testing guidelines on the utilization of tests and order entries in a surgical intensive care unit

    Kanya Kumwilaisak, Alberto Noto, Ulrich H Schmidt, Clare I Beck, Claudia Crimi, Kent Lewandrowski, and Luca M Bigatello. Effect of laboratory testing guidelines on the utilization of tests and order entries in a surgical intensive care unit. Critical care medicine, 36 0 (11): ...

    2008

  32. [40]

    Batch reinforcement learning

    Sascha Lange, Thomas Gabel, and Martin Riedmiller. Batch reinforcement learning. In Reinforcement learning: State-of-the-art, pages 45--73. Springer, 2012

    2012

  33. [41]

    Batch policy learning under constraints

    Hoang Le, Cameron Voloshin, and Yisong Yue. Batch policy learning under constraints. In International Conference on Machine Learning, pages 3703--3712. PMLR, 2019

    2019

  34. [42]

    Bayesian optimization for policy search via online-offline experimentation

    Benjamin Letham and Eytan Bakshy. Bayesian optimization for policy search via online-offline experimentation. Journal of Machine Learning Research, 20 0 (145): 0 1--30, 2019

    2019

  35. [43]

    Reducing repetitive and reflexive diagnostic phlebotomy in an intensive care unit: a quality improvement project

    Olga Levi, Maverick Chan, Thomas Bodley, Smith Orla, Michelle Sholzberg, Shannon Swift, Hina Chaudhry, Jan O Friedrich, and Lisa K Hicks. Reducing repetitive and reflexive diagnostic phlebotomy in an intensive care unit: a quality improvement project. Blood, 134: 0 3406, 2019

    2019

  36. [44]

    Offline reinforcement learning: Tutorial, review, and perspectives on open problems

    Sergey Levine, Aviral Kumar, George Tucker, and Justin Fu. Offline reinforcement learning: Tutorial, review, and perspectives on open problems. arXiv preprint arXiv:2005.01643, 2020

    2005 arXiv

  37. [45]

    Tianyi Lin, Chi Jin, and Michael I. Jordan. On gradient descent ascent for nonconvex-concave minimax problems . In ICML , 2020

    2020

  38. [46]

    Is deep reinforcement learning ready for practical applications in healthcare? a sensitivity analysis of duel-ddqn for hemodynamic management in sepsis patients

    MingYu Lu, Zachary Shahn, Daby Sow, Finale Doshi-Velez, and H Lehman Li-wei. Is deep reinforcement learning ready for practical applications in healthcare? a sensitivity analysis of duel-ddqn for hemodynamic management in sepsis patients. In AMIA Annual Symposium Proceedings, ...

    2020

  39. [47]

    Learning optimal group-structured individualized treatment rules with many treatments

    Haixu Ma, Donglin Zeng, and Yufeng Liu. Learning optimal group-structured individualized treatment rules with many treatments. Journal of Machine Learning Research, 24 0 (102): 0 1--48, 2023

    2023

  40. [48]

    Hematological findings in covid-19 and their correlation with severity of disease

    Namra Mahmood, Zahra Riaz, Arooj Sattar, and Mehwish Kiran. Hematological findings in covid-19 and their correlation with severity of disease. Pakistan Journal of Medical Sciences, 39 0 (3): 0 795, 2023

    2023

  41. [49]

    Prediction of blood lactate values in critically ill patients: a retrospective multi-center cohort study

    Behrooz Mamandipoor, Wesley Yeung, Louis Agha-Mir-Salim, David J Stone, Venet Osmani, and Leo Anthony Celi. Prediction of blood lactate values in critically ill patients: a retrospective multi-center cohort study. Journal of clinical monitoring and computing, pages 1--11, 2022

    2022

  42. [50]

    Supported trust region optimization for offline reinforcement learning

    Yixiu Mao, Hongchang Zhang, Chen Chen, Yi Xu, and Xiangyang Ji. Supported trust region optimization for offline reinforcement learning. In International Conference on Machine Learning, pages 23829--23851. PMLR, 2023

    2023

  43. [51]

    Normalizing flows for interventional density estimation

    Valentyn Melnychuk, Dennis Frauen, and Stefan Feuerriegel. Normalizing flows for interventional density estimation . In ICML , 2023

    2023

  44. [52]

    Local policy search with bayesian optimization

    Sarah M \"u ller, Alexander von Rohr, and Sebastian Trimpe. Local policy search with bayesian optimization. Advances in Neural Information Processing Systems, 34: 0 20708--20720, 2021

    2021

  45. [53]

    Deep offline reinforcement learning for real-world treatment optimization applications

    Mila Nambiar, Supriyo Ghosh, Priscilla Ong, Yu En Chan, Yong Mong Bee, and Pavitra Krishnaswamy. Deep offline reinforcement learning for real-world treatment optimization applications. In Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, pag...

    2023

  46. [54]

    A time series is worth 64 words: Long-term forecasting with transformers

    Yuqi Nie, Nam H Nguyen, Phanwadee Sinthong, and Jayant Kalagnanam. A time series is worth 64 words: Long-term forecasting with transformers. arXiv preprint arXiv:2211.14730, 2022

    2022 arXiv

  47. [55]

    Embedding time-limited laboratory orders within computerized provider order entry reduces laboratory utilization

    Natalie M Pageler, Deborah Franzon, Christopher A Longhurst, Matthew Wood, Andrew Y Shin, Eloa S Adams, Eric Widen, and David N Cornfield. Embedding time-limited laboratory orders within computerized provider order entry reduces laboratory utilization. Pediatric Critical Care ...

    2013

  48. [56]

    Variational inference with normalizing flows

    Danilo Rezende and Shakir Mohamed. Variational inference with normalizing flows. In ICML, 2015

    2015

  49. [57]

    Estimating causal effects of treatments in randomized and nonrandomized studies

    Donald B Rubin. Estimating causal effects of treatments in randomized and nonrandomized studies. Journal of educational Psychology, 66 0 (5): 0 688, 1974

    1974

  50. [58]

    Off-policy bandits with deficient support

    Noveen Sachdeva, Yi Su, and Thorsten Joachims. Off-policy bandits with deficient support. In Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, pages 965--975, 2020

    2020

  51. [59]

    Diagnostic blood loss from phlebotomy and hospital-acquired anemia during acute myocardial infarction

    Adam C Salisbury, Kimberly J Reid, Karen P Alexander, Frederick A Masoudi, Sue-Min Lai, Paul S Chan, Richard G Bach, Tracy Y Wang, John A Spertus, and Mikhail Kosiborod. Diagnostic blood loss from phlebotomy and hospital-acquired anemia during acute myocardial infarction. Arch...

    2011

  52. [60]

    Reliable off-policy learning for dosage combinations

    Jonas Schweisthal, Dennis Frauen, Valentyn Melnychuk, and Stefan Feuerriegel. Reliable off-policy learning for dosage combinations. arXiv preprint arXiv:2305.19742, 2023

    2023 arXiv

  53. [61]

    Residents' self-report on why they order perceived unnecessary inpatient laboratory tests

    Mina S Sedrak, Mitesh S Patel, Justin B Ziemba, Dana Murray, Esther J Kim, C Jessica Dine, and Jennifer S Myers. Residents' self-report on why they order perceived unnecessary inpatient laboratory tests. Journal of hospital medicine, 11 0 (12): 0 869--872, 2016

    2016

  54. [62]

    Scalable joint models for reliable uncertainty-aware event prediction

    Hossein Soleimani, James Hensman, and Suchi Saria. Scalable joint models for reliable uncertainty-aware event prediction. IEEE transactions on pattern analysis and machine intelligence, 40 0 (8): 0 1948--1963, 2017

    1948

  55. [63]

    Patient, hospital and environmental costs of unnecessary bloodwork: capturing the triple bottom line of inappropriate care in general surgery patients

    Karina Spoyalo, Annie Lalande, Chantelle Rizan, Sophia Park, Janet Simons, Philip Dawe, Carl J Brown, Robert Lillywhite, and Andrea J MacNeill. Patient, hospital and environmental costs of unnecessary bloodwork: capturing the triple bottom line of inappropriate care in general...

    2023

  56. [64]

    Reinforcement learning: An introduction

    Richard S Sutton and Andrew G Barto. Reinforcement learning: An introduction. MIT press, 2018

    2018

  57. [65]

    Tabak and Eric Vanden-Eijnden

    Esteban G. Tabak and Eric Vanden-Eijnden. Density estimation by dual ascent of the log-likelihood. Communications in Mathematical Sciences, 8 0 (1): 0 217--233, 2010

    2010

  58. [66]

    Leveraging factored action spaces for efficient offline reinforcement learning in healthcare

    Shengpu Tang, Maggie Makar, Michael Sjoding, Finale Doshi-Velez, and Jenna Wiens. Leveraging factored action spaces for efficient offline reinforcement learning in healthcare. Advances in Neural Information Processing Systems, 35: 0 34272--34286, 2022

    2022

  59. [67]

    Trippe and Richard E

    Brian L. Trippe and Richard E. Turner. Conditional density estimation with Bayesian normalising flows. arXiv preprint arXiv:1802.04908, 2018

    2018 arXiv

  60. [68]

    Do we know what inappropriate laboratory utilization is?: A systematic review of laboratory clinical audits

    Carl Van Walraven and C David Naylor. Do we know what inappropriate laboratory utilization is?: A systematic review of laboratory clinical audits. Jama, 280 0 (6): 0 550--558, 1998

    1998

  61. [69]

    Changing resident test ordering behavior: a multilevel intervention to decrease laboratory utilization at an academic medical center

    Arpana R Vidyarthi, Timothy Hamill, Adrienne L Green, Glenn Rosenbluth, and Robert B Baron. Changing resident test ordering behavior: a multilevel intervention to decrease laboratory utilization at an academic medical center. American Journal of Medical Quality, 30 0 (1): 0 81...

    2015

  62. [70]

    Optimal and adaptive off-policy evaluation in contextual bandits

    Yu-Xiang Wang, Alekh Agarwal, and Miroslav Dud k. Optimal and adaptive off-policy evaluation in contextual bandits. In International Conference on Machine Learning, pages 3589--3597. PMLR, 2017

    2017

  63. [71]

    Learning from demonstrations with high-level side information

    Min Wen, Ivan Papusha, and Ufuk Topcu. Learning from demonstrations with high-level side information. In Proceedings of the Twenty-Sixth International Joint Conference on Artificial Intelligence, 2017

    2017

  64. [72]

    Using trajectory data to improve bayesian optimization for reinforcement learning

    Aaron Wilson, Alan Fern, and Prasad Tadepalli. Using trajectory data to improve bayesian optimization for reinforcement learning. The Journal of Machine Learning Research, 15 0 (1): 0 253--282, 2014

    2014

  65. [73]

    Learning likelihoods with conditional normalizing flows

    Christina Winkler, Daniel Worrall, Emiel Hoogeboom, and Max Welling. Learning likelihoods with conditional normalizing flows. arXiv preprint arXiv:1912.00042, 2019

    1912 arXiv

  66. [74]

    Supported policy optimization for offline reinforcement learning

    Jialong Wu, Haixu Wu, Zihan Qiu, Jianmin Wang, and Mingsheng Long. Supported policy optimization for offline reinforcement learning. Advances in Neural Information Processing Systems, 35: 0 31278--31291, 2022

    2022

  67. [75]

    Behavior regularized offline reinforcement learning

    Yifan Wu, George Tucker, and Ofir Nachum. Behavior regularized offline reinforcement learning. arXiv preprint arXiv:1911.11361, 2019

    1911 arXiv

  68. [76]

    Haoran Xu, Louis Agha-Mir-Salim, Zachary O’Brien, Dora C Huang, Peiyao Li, Josep G \'o mez, Xiaoli Liu, Tongbo Liu, Wesley Yeung, Patrick Thoral, et al. Varying association of laboratory values with reference ranges and outcomes in critically ill patients: an analysis of data ...

    2021

  69. [77]

    Offline policy selection under uncertainty

    Mengjiao Yang, Bo Dai, Ofir Nachum, George Tucker, and Dale Schuurmans. Offline policy selection under uncertainty. In International Conference on Artificial Intelligence and Statistics, pages 4376--4396. PMLR, 2022

    2022

  70. [78]

    The landscape of inappropriate laboratory testing: a 15-year meta-analysis

    Ming Zhi, Eric L Ding, Jesse Theisen-Toupal, Julia Whelan, and Ramy Arnaout. The landscape of inappropriate laboratory testing: a 15-year meta-analysis. PloS one, 8 0 (11): 0 e78962, 2013

    2013

  71. [79]

    Plas: Latent action space for offline reinforcement learning

    Wenxuan Zhou, Sujay Bajracharya, and David Held. Plas: Latent action space for offline reinforcement learning. In Conference on Robot Learning, pages 1719--1735. PMLR, 2021

    2021

  72. [80]

    Zimmerman, M

    J. Zimmerman, M. Seneff, Xiaolu Sun, D. Wagner, and W. Knaus. Evaluating laboratory usage in the intensive care unit: patient and institutional characteristics that influence frequency of blood sampling., 1997

    1997

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.