Pith. sign in

REVIEW 3 major objections 5 minor 74 references

A Novel Tool for Evaluating Effect Modification in Older Adults with ADRD Using Medicare Claims

T0 review · 3 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read This paper claims that its pseudo-data estimator PD-Robust yields consistent, asymptotically normal subgroup-effect estimates under truncation by death when any two of three nuisance models are correctly specified, and that the method…

desk verdict A genuinely new triply robust estimator for effect modification under truncation by death; the case study's headline subgroup is more assumption-dependent than the sensitivity analysis acknowledges. read the letter →

arxiv 2608.06654 v1 pith:6YX4LMLD submitted 2026-08-06 stat.ME

Yilin Zhang , Michelle Shardell , Jason Falvey , Rozalina McCoy , Elizabeth Stuart , Chixiang Chen This is my paper · ORCID
classification stat.ME MSC 62D2062P10
keywords heterogeneityoftreatmenteffectsprincipalstratificationtruncationbydeathtriplerobustnessdaysathomeMedicareclaimseffectmodificationpseudo-outcomeestimation
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 develops PD-Robust, a frequentist semiparametric estimator for effect modification when the outcome trajectory is truncated by death, and uses it to ask who is hurt most when hospitalized dementia patients acquire a hospital-acquired condition (HAC) after hip fracture. The central methodological claim is triple robustness: solving the PD-Robust estimating equation yields a consistent estimator of the projection parameter $\beta_{t0}$ if any two of the propensity score, principal score, and conditional outcome mean models are correctly specified, and the estimator is asymptotically normal. Applying the method to Medicare claims, the paper reports that males under 85 who would survive regardless of exposure lose up to 23 days at home over six months post-discharge due to HAC, exceeding the 8-day threshold considered clinically meaningful. The paper also provides model diagnostics, sensitivity analyses for the untestable principal ignorability assumption, and tools to characterize the patient profile of the unobserved always-survivor stratum. A reader would care because survivor-only analyses are biased and treating death as zero days at home conflates death with institutional care, whereas this approach offers an identifiable causal estimate for a clinically interpretable subpopulation.

What carries the argument

The load-bearing machinery is the PD-Robust pseudo-data construction. For each subject at each time $t$, the procedure builds a pseudo-outcome $\hat\phi_{1,i} - \hat\phi_{0,i}$ and a pseudo-scalar $\hat\psi_{S1,i}$ from the fitted propensity score, principal score, and conditional mean models, then solves the estimating equation $\sum_i g(t,t^*,\tilde X_i;\beta_t)\{\hat\phi_{1,i} - \hat\phi_{0,i} - \hat\psi_{S1,i}\, f(t,t^*,\tilde X_i;\beta_t)\} = 0$. The defining identities are $E(\phi_1 - \phi_0) = E[\{Y^{a=1}(t)-Y^{a=0}(t)\}I\{U(t^*)=(1,1)\}]$ and $E(\psi_{S1}) = P(U(t^*)=(1,1))$, which hold whenever at least one pair of the three nuisance models is correct; because the product of two estimation errors enters the influence function, any two correct models suffice, producing triple robustness. The estimand itself is the projection parameter $\beta_{t0} = \arg\min_{\beta_t} E[(\tau^{(1,1)}(t,t^*,\tilde X) - f(t,t^*,\tilde X;\beta_t))^2 \mid U(t^*)=(1,1)]$, which maps the possibly high-dimensional conditional effect onto an interpretable structural working model such as the linear form $\eta^T(\tilde X)\beta_t$.

What would settle it

A re-analysis that lets the principal-ignorability ratio $\epsilon_0(t,t^*,X)$ vary with covariates rather than stay a fixed scalar, and shows the 23-day estimate dropping below the 8-day threshold under mild violations, would falsify the clinical conclusion; more directly, a dataset with both potential survival outcomes observed, such as a fully followed randomized trial, could verify whether the principal ignorability equalities hold at all.

Watch

Extended reading notes

Core claim

The paper's central discovery is that the causal effect of an exposure on a longitudinal outcome among the always-survivor principal stratum, the patients who would be alive at a fixed time $t^*$ regardless of exposure, can be recovered under principal ignorability through the projection-based estimand $\beta_{t0}$ defined in Equation (3), and that this parameter admits a triply robust estimator. Theorem 2.2 states that the estimator obtained by solving Equation (4) converges in probability to $\beta_{t0}$ if any two of the three nuisance working models, for the propensity score $\pi(X)$, the principal score $e^{(1,1)}(t^*,X)$, and the conditional outcome mean $\mu_{as}(t,t^*,X)$, are correctly specified, and that the estimator is asymptotically normal. On the applied side, the paper claims this identifies a clinically significant heterogeneity: within the always-survivor stratum, males under 85 years experience up to 23 fewer days at home in the six months after discharge when HAC occurs, compared with the 8-day minimal clinically meaningful difference, while the estimated effects for females are near zero.

Load-bearing premise

What everything rests on is principal ignorability, that once covariates are accounted for, patients who die under a given exposure have the same average outcome as patients who survive under that same exposure, an assumption no data can verify and one the paper checks only against a single scalar violation pattern.

Editorial extensions

If this is right

  • Estimating subgroup effects in studies with heavy post-treatment mortality no longer forces the choice between survivor-only analyses, which are biased, and coding death as zero, which distorts institutional-care utilization; PD-Robust targets the always-survivor stratum directly.
  • Consistency survives misspecification of any one of the three nuisance models: a wrong propensity score, principal score, or outcome regression alone does not break the estimate, so applied conclusions rest on a weaker modeling requirement than standard doubly robust procedures.
  • The reported 23-day reduction for males under 85 is a causal claim about patients who would survive to month 6 regardless of HAC, not an observed association, and it exceeds the 8-day threshold regarded as clinically meaningful.
  • Clinicians and payers can use the profiling tools to identify which observed patients most resemble the always-survivor stratum, making the causal estimate actionable rather than confined to a latent subpopulation.
  • Because the method ships with an R package, the same pipeline can be applied to other claims-based trajectory questions without re-deriving the influence function.

Reading between the lines

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

  • Because the sensitivity analysis perturbs only a scalar $\epsilon_0(t,t^*,X)$, the 23-day estimate has not been stress-tested against violations of principal ignorability that differ by age or sex; letting the ratio vary with covariates is a direct next stress test.
  • The near-agreement with the adapted DR-learner may be specific to this cohort's relatively low death rate and large always-survivor stratum; in populations where survival differs sharply by exposure, the two approaches would likely diverge, which is a testable prediction.
  • The same pseudo-outcome template could be carried over to composite endpoints such as days alive and at home, or to count outcomes with a log-link working model, generalizing the projection estimand beyond the linear working models used here.
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. This paper proposes PD-Robust, a frequentist projection-based method for estimating heterogeneous treatment effects on a longitudinal outcome that is truncated by death, within a principal stratum of patients who would survive regardless of exposure. The target parameter beta_{t0} is defined as the projection of tau(t, t*, Xtilde) = E[Y^1(t) - Y^0(t) | Xtilde, U(t*)=(1,1)] onto a parametric working model f(t, t*, Xtilde; beta_t). Under Assumptions 1-4, the authors construct pseudo-outcomes from propensity score, principal score, and conditional outcome models and estimate beta by solving Equation (4); Theorem 2.2 claims triple robustness and asymptotic normality. The case study applies the method to Medicare claims for older adults with ADRD after hip fracture, estimating HAC effects on six-month days-at-home and reporting that males under 85 lose up to 23 DAH, exceeding the 8-day clinically meaningful threshold. Simulations under various combinations of correctly and incorrectly specified nuisance models show low bias and near-nominal coverage, and the paper provides an R package and practical guidance.

Significance. If Lemma 2.1 and Theorem 2.2 hold, the paper makes a substantive methodological contribution: a triply robust, finite-dimensional projection estimator for principal-stratum heterogeneous effects, together with model diagnostics, sensitivity tools, and principal-stratum profile characterization. The application addresses an important clinical question, and the concrete subgroup finding is falsifiable and clinically relevant. The paper also ships an R package and detailed implementation guidance, which is a practical strength. The main weakness is that the most prominent applied claim—males under 85 lose up to 23 DAH—relies on an untestable principal ignorability assumption, and the sensitivity analysis presented for this assumption is not stratified by the subgroup that drives the headline result.

major comments (3)
  1. [Section 2.5 and Figure 4B] The sensitivity analysis for Assumption 4 (principal ignorability) varies a single scalar epsilon0(t, t*, X) over (0,1] and does not report subgroup-specific sensitivity curves for the males-under-85 subgroup that drives the abstract's 'up to 23 fewer DAH' claim. Because Model 3 includes a sex-by-age interaction, a violation of principal ignorability concentrated in younger males (e.g., HAC selectively removes younger males with poor unexposed recovery) could change the subgroup projection parameter without affecting the overall sensitivity pattern shown in Figure 4B. Please report sensitivity results separately for the males-under-85 subgroup, or use X-dependent epsilon0 values that represent violations concentrated in that subgroup, and state whether the estimated effect remains above the 8-day threshold under those violations. If the current implementation already permits X-dependent epsilon0, the figure and text should describe how those covariate-specific values were chosen.
  2. [Theorem 2.2 and Lemma 2.1] Lemma 2.1 and Theorem 2.2 are the load-bearing theoretical results of the paper, but their proofs and all regularity conditions are deferred to the Supplementary Material, which is not included in the manuscript text. The main text should state the extra conditions needed for asymptotic normality or clearly identify the supplement statement and confirm that the supplement is part of the submission. As written, the triple robustness and asymptotic normality claims cannot be fully verified from the manuscript alone.
  3. [Section 4.2 and Table 2] The simulation evaluation focuses almost entirely on the first element of beta_t, while the case-study conclusion relies on interaction coefficients that distinguish males under 85 from other subgroups. To support the applied claim, please add simulation results for the interaction parameters (or at least for the subgroup-specific linear combination corresponding to males under 85), especially under partial model misspecification, so that the finite-sample behavior of the parameters driving the headline is assessed directly.
minor comments (5)
  1. [Abstract and Section 3] Please clarify whether 'up to 23 fewer DAH' is a six-month cumulative difference or a month-specific estimate for the male-under-85 subgroup, and identify the model and figure from which this number is taken.
  2. [Section 2.4, Eq. (4)] The definition of \hat\phi_{0,i} uses \hat\psi_{S1,i} before \hat\psi_{S1,i} is defined; please reorder the displayed definitions to avoid circular reading.
  3. [Figure 3B] The 'standardized t-statistic' diagnostic for the PPS model is informal; please state a threshold or reference distribution for judging when the t-statistics are small enough to indicate adequate PPS fit.
  4. [Table 2 and Figure 5 captions] There are typographical errors: 'devision' should be 'deviation' in Table 2, and 'confidence internals' should be 'confidence intervals' in Figure 5.
  5. [Section 2.5] The exposure ignorability sensitivity analysis is described only by reference to Section S1.6.2 of the Supplement; please provide at least a brief account of its identifying assumptions and interpretation in the main text.

Circularity Check

0 steps flagged · score 2.0 of 10

Low circularity: the estimator is built to solve the moment condition defining the projection parameter, which is standard semiparametric estimation rather than a reduction; minor self-citations are not load-bearing.

full rationale

No load-bearing circular step is demonstrated. The target parameter β_t0 is explicitly a projection parameter defined by the moment condition in Equation (3), and the PD-Robust estimator solves the sample moment condition in Equation (4) using EIF-based pseudo-outcomes. This is standard semiparametric estimation: consistency of the estimator for β_t0 is a theorem (Theorem 2.2) proved under different correct-specification combinations, and the pseudo-outcome identities E[φ1−φ0] = E[(Y^1−Y^0)I{U(t∗)=(1,1)}] and E[ψS1] = P(U(t∗)=(1,1)) are nontrivial identifying results, not definitions of the estimand. The applied headline (up to 23 fewer DAH in males under 85) is a fitted coefficient in the working structural model, not an independent prediction obtained from separately fitted inputs, so it is not a fitted-input-called-prediction. The paper's self-citations—Shen et al. for the DAH-mortality association supporting Assumption 4 and Chen et al. (2025) for a finite-sample caution—are not load-bearing; Assumption 4 is stated as an untestable assumption, and the method's identification does not reduce to those citations. The main weakness is a robustness gap rather than circularity: the principal-ignorability sensitivity analysis in Section 2.5 and Figure 4B varies a scalar ε0(t,t∗,X) with ε0≤1 and is not reported separately for the sex-by-age subgroup driving the 23-day claim, so a heterogeneous violation concentrated in that subgroup could alter the clinical conclusion. The paper itself flags that Assumption 4 cannot be directly verified from observed data. Because there is no demonstrated reduction of the derivation to its inputs, the circularity score is low.

Assumptions & free parameters 3 free parameters · 7 assumptions · 0 invented entities

No new physical or latent entity is introduced by the paper; the always-survivor principal stratum is a standard latent construct from the principal-stratification literature, not a newly postulated object. The central claim rests on the four principal-stratification assumptions, the choice of t*, the projection working model, and correct specification of at least two nuisance models.

free parameters (3)
  • Working-model projection coefficients beta_t = estimated from data; e.g., male<85 HAC effect up to -23 DAH
    Equation (2) defines beta_{t0} as the projection of the principal-stratum causal effect onto the working model; these coefficients are the empirical output on which the headline effect sizes rest, rather than independently predicted constants.
  • Survival horizon t* = 6 months primary; 10 months sensitivity
    t* defines the principal stratum and the target estimand. Estimates shift with t*, as the paper acknowledges, so it is an analyst-chosen input rather than a derived quantity.
  • Sensitivity parameter epsilon_0(t,t*,X) = varied over (0,1]
    The principal-ignorability sensitivity analysis requires choosing the ratio of outcome means across strata. The robustness conclusion is conditional on this functional specification; it is not fitted from data but introduced by the method.
assumptions (7)
  • domain assumption Positivity: 0 < pi(X) < 1 for all X (Assumption 1)
    Used to justify inverse probability weighting in the pseudo-outcomes. Plausible, but not verifiable if some patients have zero probability of HAC.
  • domain assumption Exposure ignorability: A is independent of potential survival and outcomes given X (Assumption 2)
    Assumes no unmeasured confounding of HAC in Medicare claims. Section 2.5 provides a latent-variable sensitivity analysis, but the primary estimate depends on this assumption.
  • domain assumption Monotonicity: S^{a=1}(t) <= S^{a=0}(t) (Assumption 3)
    Rules out the (1,0) stratum and identifies the principal score e11 with p1. Argued plausible because HAC should not improve survival, but it is a structural assumption.
  • domain assumption Principal ignorability: E[Y^a|U=(.,.),X] depends only on the same-exposure survival status (Assumption 4)
    Untestable and central to identification of the target stratum outcome means. The paper's sensitivity analysis only partially relaxes it and does not cover fully general violations.
  • domain assumption Correct specification of at least two of the three nuisance models
    Triple robustness guarantees consistency only if any two of PS, PPS, and CM are correctly specified. In practice this is a statistical assumption about the GLM and pooled logistic models used in the application.
  • domain assumption Claims-based DAH and HAC coding are valid for the study population
    The case-study outcome and exposure come from Medicare claims. Measurement error in HAC coding or in days-at-home would bias the effect estimates.
  • standard math Standard large-sample asymptotic theory for Z-estimators and bootstrap inference
    Theorem 2.2's asymptotic normality and the bootstrap procedure rely on standard empirical-process conditions, which are deferred to the supplementary material.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Novel Tool for Evaluating Effect Modification in Older Adults with ADRD Using Medicare Claims." pith.science (2026). https://pith.science/paper/6YX4LMLD

@misc{pith2026260806654,
  author       = {Pith},
  title        = {Pith review of: A Novel Tool for Evaluating Effect Modification in Older Adults with ADRD Using Medicare Claims},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6YX4LMLD}},
  note         = {Machine review of arXiv:2608.06654}
}
read the original abstract

Studying consequences following baseline exposures has become increasingly important for advancing comparative effectiveness research using real-world data. This case study evaluates the impact of hospital-acquired conditions (HAC) during hospitalization for hip fracture on post-discharge recovery trajectories among older adults living with Alzheimer Disease and Related Dementia, a population particularly vulnerable to high post-hospital mortality. To appropriately account for truncation of recovery trajectory due to death and to explore heterogeneity in effect modification by patient demographics, we introduce a novel pseudo data-based robust (PD-Robust) analysis strategy, accompanied by an R package and detailed usage guidance to inform real data analysis. Grounded in an interpretable estimand via principal stratification under principal ignorability and a structural working model, PD-Robust accommodates truncation by death, provides model diagnosis and robustness check against assumption violation, and facilitates the characterization of patient profiles among the principal stratum. Applied to Medicare claims data, where better recovery is defined as more days at home (DAH) over six months post-discharge, PD-Robust reveals heterogeneity in HAC effects, with males under the age of 85 years as a high-risk subgroup experiencing up to 23 fewer DAH, comparing HAC to no HAC. This exceeds the 8-day threshold regarded as clinically meaningful difference in DAH due to any exposure. Moreover, simulation studies further demonstrate that PD-Robust achieves low estimation bias and accurate statistical inference, supporting its utility in real-world data applications.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

74 extracted references · 45 canonical work pages

  1. [1]

    Journal of the Royal Statistical Society Series B: Statistical Methodology , volume=

    Multiply robust estimation of causal effects under principal ignorability , author=. Journal of the Royal Statistical Society Series B: Statistical Methodology , volume=. 2022 , publisher=

    work page 2022

  2. [2]

    A Joint Learning for Analyzing a National Geriatric Centralized Networks: A New Toolbox Deciphering a Real-World Complexity , author=

    work page

  3. [3]

    Journal of the Royal Statistical Society Series B: Statistical Methodology , volume=

    Principal stratification with continuous post-treatment variables: Nonparametric identification and semiparametric estimation , author=. Journal of the Royal Statistical Society Series B: Statistical Methodology , volume=. 2026 , publisher=

    work page 2026

  4. [4]

    BMJ open , volume=

    Scoping review exploring the impact of hip fracture in older adults with cognitive impairment or dementia , author=. BMJ open , volume=. 2025 , publisher=

    work page 2025

  5. [5]

    2018 , publisher=

    Double/debiased machine learning for treatment and structural parameters , author=. 2018 , publisher=

    2018

  6. [6]

    arXiv preprint arXiv:2303.03520 , year=

    The Effect of Substance Use on Brain Ageing: A New Causal Inference Framework for Incomplete and Massive Phenomic Data , author=. arXiv preprint arXiv:2303.03520 , year=

    work page arXiv

  7. [7]

    Randomization analysis of experimental data in the Fisher randomization test

    Discussion of "Randomization analysis of experimental data in the Fisher randomization test" by D. Basu , author=. Journal of the American statistical association , volume=

    work page

  8. [8]

    censoring

    Causal inference through potential outcomes and principal stratification: application to studies with" censoring" due to death , author=. Statistical Science , number =. 2006 , publisher=

    work page 2006

Show all 74 references
  1. [9]

    2015 , publisher=

    Causal inference in statistics, social, and biomedical sciences , author=. 2015 , publisher=

    2015

  2. [10]

    Developing a protocol for observational comparative effectiveness research: A user's guide , year=

    Estimation and reporting of heterogeneity of treatment effects , author=. Developing a protocol for observational comparative effectiveness research: A user's guide , year=

  3. [11]

    Annual review of statistics and its application , volume=

    Precision medicine , author=. Annual review of statistics and its application , volume=. 2019 , publisher=

    2019

  4. [12]

    Biometrics , volume=

    Principal stratification in causal inference , author=. Biometrics , volume=. 2002 , publisher=

    2002

  5. [13]

    2010 , publisher=

    Causal inference , author=. 2010 , publisher=

    2010

  6. [14]

    The Econometrics Journal , volume=

    Debiased machine learning of conditional average treatment effects and other causal functions , author=. The Econometrics Journal , volume=. 2021 , publisher=

    2021

  7. [15]

    Journal of the Royal Statistical Society Series B: Statistical Methodology , volume=

    Principal stratification analysis using principal scores , author=. Journal of the Royal Statistical Society Series B: Statistical Methodology , volume=. 2017 , publisher=

    2017

  8. [16]

    Journal of the Royal Statistical Society Series B: Statistical Methodology , pages=

    Principal stratification with continuous post-treatment variables: Nonparametric identification and semiparametric estimation , author=. Journal of the Royal Statistical Society Series B: Statistical Methodology , pages=. 2025 , publisher=

    2025

  9. [17]

    Proceedings of the national academy of sciences , volume=

    Metalearners for estimating heterogeneous treatment effects using machine learning , author=. Proceedings of the national academy of sciences , volume=. 2019 , publisher=

    2019

  10. [18]

    JAMA surgery , volume=

    Effect of delirium and other major complications on outcomes after elective surgery in older adults , author=. JAMA surgery , volume=. 2015 , publisher=

    2015

  11. [19]

    Journal of Orthopaedic Surgery and Research , volume=

    A retrospective study: exploring preoperative hyponatremia in elderly patients with hip fractures , author=. Journal of Orthopaedic Surgery and Research , volume=. 2024 , publisher=

    2024

  12. [20]

    BMC psychiatry , volume=

    Electrolyte disorders and aging: risk factors for delirium in patients undergoing orthopedic surgeries , author=. BMC psychiatry , volume=. 2016 , publisher=

    2016

  13. [21]

    The Journals of Gerontology: Series A , volume=

    Sex differences in recovery across multiple domains among older adults with hip fracture , author=. The Journals of Gerontology: Series A , volume=. 2022 , publisher=

    2022

  14. [22]

    Journal of Alzheimer's Disease , volume=

    Evaluation of Medicare claims data as a tool to identify dementia , author=. Journal of Alzheimer's Disease , volume=. 2019 , publisher=

    2019

  15. [23]

    Curse of dimensionality and related issues in nonparametric functional regression , volume =

    Geenens, Gery , year=. Curse of dimensionality and related issues in nonparametric functional regression , volume =. Statistics Surveys , number =

  16. [24]

    Journal of the American Statistical Association , volume=

    Identifiability and estimation of causal effects by principal stratification with outcomes truncated by death , author=. Journal of the American Statistical Association , volume=. 2011 , publisher=

    2011

  17. [25]

    Journal of the American Statistical Association , volume=

    Causal effects of treatments for informative missing data due to progression/death , author=. Journal of the American Statistical Association , volume=. 2010 , publisher=

    2010

  18. [26]

    Journal of the Royal Statistical Society Series B: Statistical Methodology , volume=

    Principal causal effect identification and surrogate end point evaluation by multiple trials , author=. Journal of the Royal Statistical Society Series B: Statistical Methodology , volume=. 2016 , publisher=

    2016

  19. [27]

    Biometrika , volume=

    Quasi-oracle estimation of heterogeneous treatment effects , author=. Biometrika , volume=. 2021 , publisher=

    2021

  20. [28]

    Electronic Journal of Statistics , volume=

    Towards optimal doubly robust estimation of heterogeneous causal effects , author=. Electronic Journal of Statistics , volume=. 2023 , publisher=

    2023

  21. [29]

    arXiv preprint arXiv:2306.16297 , year=

    A meta-learning method for estimation of causal excursion effects to assess time-varying moderation , author=. arXiv preprint arXiv:2306.16297 , year=

    arXiv

  22. [30]

    Biometrics , volume=

    Identification of causal effects in the presence of nonignorable missing outcome values , author=. Biometrics , volume=. 2014 , publisher=

    2014

  23. [31]

    Biometrika , volume=

    Addressing complications of intention-to-treat analysis in the combined presence of all-or-none treatment-noncompliance and subsequent missing outcomes , author=. Biometrika , volume=. 1999 , publisher=

    1999

  24. [32]

    Biostatistics , volume=

    Doubly robust estimation and causal inference in longitudinal studies with dropout and truncation by death , author=. Biostatistics , volume=. 2015 , publisher=

    2015

  25. [33]

    Annals of translational medicine , volume=

    Balance diagnostics after propensity score matching , author=. Annals of translational medicine , volume=

  26. [34]

    JAMA Internal Medicine , volume=

    Stakeholder perspectives on minimum clinically important difference and noninferiority margin for hospital-free days to assess interventions , author=. JAMA Internal Medicine , volume=. 2023 , publisher=

    2023

  27. [36]

    Journal of the American Statistical Association , volume=

    Estimation and inference of heterogeneous treatment effects using random forests , author=. Journal of the American Statistical Association , volume=. 2018 , publisher=

    2018

  28. [37]

    Econometrica , volume=

    The sorted effects method: Discovering heterogeneous effects beyond their averages , author=. Econometrica , volume=. 2018 , publisher=

    2018

  29. [38]

    2018 , institution=

    Generic machine learning inference on heterogeneous treatment effects in randomized experiments, with an application to immunization in India , author=. 2018 , institution=

    2018

  30. [39]

    Journal of the American Statistical Association , volume=

    Using secondary outcomes to sharpen inference in randomized experiments with noncompliance , author=. Journal of the American Statistical Association , volume=. 2013 , publisher=

    2013

  31. [40]

    Journal of the American Statistical Association , volume=

    Evaluating the effect of training on wages in the presence of noncompliance, nonemployment, and missing outcome data , author=. Journal of the American Statistical Association , volume=. 2012 , publisher=

    2012

  32. [41]

    The New England journal of medicine , volume=

    Days spent at home—a patient-centered goal and outcome , author=. The New England journal of medicine , volume=. 2016 , publisher=

    2016

  33. [42]

    The lancet , volume=

    Dementia prevention, intervention, and care: 2020 report of the Lancet Commission , author=. The lancet , volume=. 2020 , publisher=

    2020

  34. [43]

    Gait & posture , volume=

    Gait parameter risk factors for falls under simple and dual task conditions in cognitively impaired older people , author=. Gait & posture , volume=. 2013 , publisher=

    2013

  35. [44]

    BMC Medical Research Methodology , volume=

    Real-world data: a brief review of the methods, applications, challenges and opportunities , author=. BMC Medical Research Methodology , volume=. 2022 , publisher=

    2022

  36. [45]

    International orthopaedics , volume=

    Trajectories and prognostic factors for recovery after hip fracture: a longitudinal cohort study , author=. International orthopaedics , volume=. 2022 , publisher=

    2022

  37. [46]

    Aging and disease , volume=

    Emerging anti-aging strategies-scientific basis and efficacy , author=. Aging and disease , volume=. 2018 , publisher=

    2018

  38. [47]

    The lancet Diabetes & endocrinology , volume=

    The association between smoking cessation and glycaemic control in patients with type 2 diabetes: a THIN database cohort study , author=. The lancet Diabetes & endocrinology , volume=. 2015 , publisher=

    2015

  39. [48]

    2000 , publisher=

    Asymptotic statistics , author=. 2000 , publisher=

    2000

  40. [49]

    , author=

    Patient experience over time in patient-centered medical homes. , author=. The American journal of managed care , volume=

  41. [50]

    Proceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining , pages=

    Xgboost: A scalable tree boosting system , author=. Proceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining , pages=

  42. [51]

    arXiv preprint arXiv:2303.12687 , year=

    On weighted orthogonal learners for heterogeneous treatment effects , author=. arXiv preprint arXiv:2303.12687 , year=

    arXiv

  43. [52]

    Journal of the American Geriatrics Society , volume=

    Associations of sex, Alzheimer's disease and related dementias, and days alive and at home among older Medicare beneficiaries recovering from hip fracture , author=. Journal of the American Geriatrics Society , volume=. 2023 , publisher=

    2023

  44. [53]

    Journal of the American Geriatrics Society , volume=

    Association between pre-hip fracture depression and days at home after fracture and assessing sex differences , author=. Journal of the American Geriatrics Society , volume=. 2024 , publisher=

    2024

  45. [54]

    Journal of the American Statistical Association , pages=

    The Effect of Alcohol Intake on Brain White Matter Microstructural Integrity: A New Causal Inference Framework for Incomplete Phenomic Data , author=. Journal of the American Statistical Association , pages=. 2025 , publisher=

    2025

  46. [55]

    The Journals of Gerontology: Series A , volume=

    Associations of days spent at home before hip fracture with Postfracture days at home and 1-year mortality among Medicare beneficiaries living with Alzheimer’s disease or related dementias , author=. The Journals of Gerontology: Series A , volume=. 2023 , publisher=

    2023

  47. [56]

    Statistics in medicine , volume=

    Analyzing risk factors for post-acute recovery in older adults with Alzheimer's disease and related dementia: A new semi-parametric model for large-scale medicare claims , author=. Statistics in medicine , volume=. 2024 , publisher=

    2024

  48. [57]

    Biometrics , volume=

    Bayesian inference for a principal stratum estimand on recurrent events truncated by death , author=. Biometrics , volume=. 2023 , publisher=

    2023

  49. [58]

    2019 , eprint=

    Survivor average causal effects for continuous time: a principal stratification approach to causal inference with semicompeting risks , author=. 2019 , eprint=

    2019

  50. [59]

    Biometrical Journal , volume=

    Survivor average causal effects for continuous time: a principal stratification approach to causal inference with semicompeting risks , author=. Biometrical Journal , volume=. 2025 , publisher=

    2025

  51. [60]

    American Journal of Epidemiology , volume=

    Survivor average causal effects and long-term outcomes after severe infections , author=. American Journal of Epidemiology , volume=. 2025 , publisher=

    2025

  52. [61]

    Biometrics , volume=

    On estimation of the survivor average causal effect in observational studies when important confounders are missing due to death , author=. Biometrics , volume=. 2009 , publisher=

    2009

  53. [62]

    BMJ Open

    Validation of days at home as an outcome measure after surgery: a prospective cohort study in Australia. BMJ Open. 2017; 7 (8): e015828 , author=

    2017

  54. [63]

    Statistics in medicine , volume=

    Identification and estimation of survivor average causal effects , author=. Statistics in medicine , volume=. 2014 , publisher=

    2014

  55. [64]

    2005 , publisher=

    Quantile regression , author=. 2005 , publisher=

    2005

  56. [65]

    2015 , publisher=

    Mathematical statistics: basic ideas and selected topics, volumes I-II package , author=. 2015 , publisher=

    2015

  57. [66]

    High dimensional probability II , pages=

    Preservation theorems for Glivenko-Cantelli and uniform Glivenko-Cantelli classes , author=. High dimensional probability II , pages=. 2000 , publisher=

    2000

  58. [67]

    Journal of the Royal Statistical Society Series B: Statistical Methodology , volume=

    Semiparametric localized principal stratification analysis with continuous strata , author=. Journal of the Royal Statistical Society Series B: Statistical Methodology , volume=. 2025 , publisher=

    2025

  59. [68]

    The annals of applied statistics , volume=

    A Bayesian machine learning approach for estimating heterogeneous survivor causal effects: applications to a critical care trial , author=. The annals of applied statistics , volume=

  60. [69]

    The Annals of Applied Statistics , volume=

    Multiply robust estimation for causal survival analysis with treatment noncompliance , author=. The Annals of Applied Statistics , volume=. 2026 , publisher=

    2026

  61. [70]

    Biometrika , volume=

    Semiparametric counterfactual density estimation , author=. Biometrika , volume=. 2023 , publisher=

    2023

  62. [71]

    Biometrics , volume=

    Instrumented difference-in-differences , author=. Biometrics , volume=. 2023 , publisher=

    2023

  63. [72]

    arXiv preprint arXiv:2510.19243 , year=

    A New Targeted-Federated Learning Framework for Estimating Heterogeneity of Treatment Effects: A Robust Framework with Applications in Aging Cohorts , author=. arXiv preprint arXiv:2510.19243 , year=

    arXiv

  64. [73]

    arXiv preprint arXiv:2501.17514 , year=

    Semiparametric principal stratification analysis beyond monotonicity , author=. arXiv preprint arXiv:2501.17514 , year=

  65. [74]

    JAMA Network Open , volume=

    Neighborhood deprivation and days spent at home after fall-related hip fracture , author=. JAMA Network Open , volume=. 2025 , publisher=

    2025

  66. [75]

    1993 , publisher=

    Efficient and adaptive estimation for semiparametric models , author=. 1993 , publisher=

    1993

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.