Time-to-Event Estimation with Unreliably Reported Events in Medicare Health Plan Payment
Pith reviewed 2026-05-21 14:51 UTC · model grok-4.3
The pith
Novel time-to-event estimators track upcoding in Medicare Advantage while handling unreliable reporting.
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
Core claim
We propose several novel time-to-event estimators of incident coding intensity and possible upcoding in Medicare Advantage, including accounting for unreliable reporting. In simulations, our novel estimators recovered differences in upcoding within and across monitoring periods. Undercoding had a limited effect on our novel estimators while an existing estimator was more sensitive to undercoding.
What carries the argument
Time-to-event estimators that model the timing of reported health conditions while adjusting for reporting unreliability to estimate coding intensity.
If this is right
- These estimators can help track new coding behaviors earlier and at scale.
- They account for several real-world data considerations such as unreliable reporting.
- The open-source R package enables more reproducible methods development for coding intensity evaluation.
- Policymakers can monitor effects from updates to risk adjustment formulas using these tools.
Where Pith is reading between the lines
- If applied to real claims, the estimators could reveal upcoding trends that aggregate statistics miss.
- The simulation package could be used to test estimator performance under varying rates of reporting errors.
- Similar time-to-event approaches might extend to detecting incentive-driven reporting in other insurance programs.
Load-bearing premise
The simulated data accurately capture real-world patterns of incident coding, undercoding, and upcoding behavior including specific incentives and reporting reliability in Medicare claims.
What would settle it
Running the estimators on actual Medicare claims data from periods before and after known risk-adjustment formula updates and checking whether detected upcoding shifts match the expected direction and size would test the claim.
read the original abstract
OBJECTIVE: To propose time-to-event estimators that help evaluate incident diagnostic coding and possible upcoding in Medicare as well as introduce an open-source software package that enables more reproducible methods development relevant to Medicare billing behavior. STUDY SETTING AND DESIGN: Observational analysis of simulated upcoding based on coding by insurers or providers that may be incentivized by Medicare Advantage risk adjustment. DATA SOURCES AND ANALYTIC SAMPLE: Two years of separately simulated incident health condition coding data for a Medicare Advantage population and a Traditional Medicare population where coding patterns are aligned with known practices in each program. PRINCIPAL FINDINGS: We propose several novel time-to-event estimators of incident coding intensity and possible upcoding in Medicare Advantage, including accounting for unreliable reporting. We demonstrate estimator performance in simulated data leveraging the National Institutes of Health's All of Us study and also develop an open source R package to simulate longitudinal realistic labeled upcoding data, which were not previously available for researchers. In simulations, our novel estimators recovered differences in upcoding within and across monitoring periods. Undercoding had a limited effect on our novel estimators while an existing estimator was more sensitive to undercoding. CONCLUSIONS: Our proposed estimators can help researchers and policymakers track new coding behaviors (e.g., as may be incentivized by risk adjustment formula updates) earlier and at scale while accounting for several real-world data considerations. Further, the R package we provide can be used to improve the development, accessibility, and reproducible evaluation of coding intensity and upcoding methodology.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The manuscript proposes several novel time-to-event estimators for incident diagnostic coding intensity and possible upcoding in Medicare Advantage, explicitly incorporating unreliable reporting. These are evaluated exclusively via simulations constructed from high-level descriptions of coding patterns aligned with All of Us data and known MA/TM practices; an open-source R package is also released to generate such longitudinal labeled data. In the reported simulations the new estimators recover within- and across-period upcoding differences and exhibit lower sensitivity to undercoding than a comparator method.
Significance. If the recovery and robustness properties hold under real claims-generating processes, the estimators could enable earlier, scalable monitoring of coding responses to risk-adjustment formula changes. The accompanying simulation package fills a documented gap in reproducible methods development for Medicare billing studies and is a concrete strength.
major comments (2)
- [§4] §4 (Simulation Design and Results): all performance claims rest on data generated by the authors' own model whose hazard rates, reporting-reliability parameters, and incentive alignments are described only at a high level; without either (a) explicit sensitivity analyses that vary these parameters outside the base case or (b) an application to actual Medicare claims files, it is impossible to determine whether the reported recovery of upcoding differences is an artifact of the simulation assumptions rather than a property of the estimators.
- [§3.1–3.2] §3.1–3.2 (Estimator Derivation): the likelihood or survival-function adjustments for unreliable reporting are not shown in sufficient algebraic detail to verify that the claimed robustness to undercoding follows directly from the model rather than from post-hoc tuning; a concrete derivation or pseudocode step that isolates the effect of the reporting-error term on the estimator would be required to support the central robustness claim.
minor comments (2)
- [Table 1, Figure 2] Table 1 and Figure 2: axis labels and legend entries use inconsistent abbreviations for the proposed estimators; a single consistent naming convention would improve readability.
- [Abstract and §5] The abstract states that the package 'enables more reproducible methods development' but does not include a DOI or GitHub link in the main text; adding the permanent repository reference would strengthen the reproducibility contribution.
Simulated Author's Rebuttal
We appreciate the referee's constructive feedback on our manuscript. We address each major comment below, indicating revisions made to strengthen the presentation and evaluation of the proposed estimators.
read point-by-point responses
-
Referee: [§4] §4 (Simulation Design and Results): all performance claims rest on data generated by the authors' own model whose hazard rates, reporting-reliability parameters, and incentive alignments are described only at a high level; without either (a) explicit sensitivity analyses that vary these parameters outside the base case or (b) an application to actual Medicare claims files, it is impossible to determine whether the reported recovery of upcoding differences is an artifact of the simulation assumptions rather than a property of the estimators.
Authors: We agree that reliance on a single base-case simulation leaves open the possibility that reported performance is tied to specific modeling assumptions. In the revised manuscript we have added a dedicated sensitivity analysis section that systematically varies hazard rates, reporting-reliability parameters, and incentive alignments over ranges informed by the All of Us and MA/TM literature. These additional results show that the relative robustness of the new estimators to undercoding persists across the tested parameter space. Application to restricted Medicare claims files is not possible in the present study because of data-use agreements and the methodological focus of the work; we have expanded the limitations section to acknowledge this gap and to outline a concrete plan for future external validation. revision: partial
-
Referee: [§3.1–3.2] §3.1–3.2 (Estimator Derivation): the likelihood or survival-function adjustments for unreliable reporting are not shown in sufficient algebraic detail to verify that the claimed robustness to undercoding follows directly from the model rather than from post-hoc tuning; a concrete derivation or pseudocode step that isolates the effect of the reporting-error term on the estimator would be required to support the central robustness claim.
Authors: We thank the referee for this observation. The revised manuscript now contains expanded algebraic derivations in §§3.1–3.2 that explicitly incorporate the reporting-error probability into the likelihood and survival functions. We isolate the contribution of the error term by showing the difference between the adjusted and unadjusted estimators and demonstrate algebraically why the adjustment reduces bias under undercoding. Step-by-step pseudocode that implements the adjusted estimator has also been added to the supplementary materials. revision: yes
Circularity Check
No load-bearing circularity; validation uses author-developed simulations as standard methodological practice
full rationale
The paper proposes novel time-to-event estimators for incident coding intensity and upcoding (including under unreliable reporting) and introduces an R package for generating simulated longitudinal data. Performance is demonstrated by applying the estimators to data simulated under the authors' own model of hazard rates, reporting reliability, and incentive structures, where the estimators recover known differences. This is a conventional simulation-based validation for new statistical methods rather than a circular derivation: the estimators are not defined in terms of the simulation outputs, no fitted parameter is relabeled as a prediction, and no self-citation chain or uniqueness theorem is invoked to force the result. The central contribution remains the independent proposal of the estimators and the reproducible simulation tool. No specific equations or sections in the provided text exhibit self-definitional or fitted-input reductions.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.