CoChat
Sign Up
Cited by 6Health Systems, Economic Evaluations, Quality of LifeSimulation Techniques and ApplicationsStatistical Methods in Clinical Trials

A Tutorial on Time-Dependent Cohort State-Transition Models in R Using a Cost-Effectiveness Analysis Example

Open access
Alarid-Escudero, Fernando, Krijkamp, Eline, Enns, Eva A., Yang, Alan, Hunink, M. G. Myriam, Pechlivanoglou, Petros, Jalal, Hawre·2023·Medical Decision Making
DOI 10.1177/0272989x221121747PMC PMC9844995
Open this paper in CoChat — free account

Abstract

In an introductory tutorial, we illustrated building cohort state-transition models (cSTMs) in R, where the state transition probabilities were constant over time. However, in practice, many cSTMs require transitions, rewards, or both to vary over time (time dependent). This tutorial illustrates adding 2 types of time dependence using a previously published cost-effectiveness analysis of multiple strategies as an example. The first is simulation-time dependence, which allows for the transition probabilities to vary as a function of time as measured since the start of the simulation (e.g., varying probability of death as the cohort ages). The second is state-residence time dependence, allowing for history by tracking the time spent in any particular health state using tunnel states. We use these time-dependent cSTMs to conduct cost-effectiveness and probabilistic sensitivity analyses. We also obtain various epidemiological outcomes of interest from the outputs generated from the cSTM, such as survival probability and disease prevalence, often used for model calibration and validation. We present the mathematical notation first, followed by the R code to execute the calculations. The full R code is provided in a public code repository for broader implementation.

AI snapshot

Written from the paper’s full text.

TL;DR

This tutorial extends time-independent cohort state-transition models (cSTMs) in R to incorporate simulation-time dependence and state-residence time dependence using a cost-effectiveness analysis example. It provides mathematical notation, R code, and worked outputs for transition rewards, probabilistic sensitivity analysis, and epidemiological outcomes.

Key findings

  • The tutorial distinguishes simulation-time dependence, implemented with a 3-dimensional transition probability array where the third dimension is time since simulation start, from state-residence time dependence, implemented with tunnel states that expand the state space.
  • In the Sick-Sicker case study, the state-residence model uses 75 S1 tunnel states plus 3 other states, for a total of 78 states; the cohort starts at age 25 and is simulated over 75 one-year cycles.
  • The CEA found SoC least costly and effective, Strategy B with an ICER of $65,288/QALY, Strategy AB with an ICER of $104,461/QALY, and Strategy A dominated.
  • PSA used 1,000 parameter sets and 4,000 simulations; SoC had the highest probability of being cost-effective below $70,000/QALY, Strategy B between $70,000 and $105,000/QALY, and Strategy AB at or above $105,000/QALY; EVPI at $100,000/QALY was $7,836. Restricted life expectancy under SoC was 40.8 cycles (95% credible interval 38.4–43.6).

Methods

  • Tutorial design using a previously published 4-state Sick-Sicker cSTM expanded to include age-dependent background mortality from US life tables and state-residence dependence in the Sick state via a Weibull hazard function and tunnel states.
  • The case study evaluates four treatment strategies—SoC, Treatment A, Treatment B, and Treatment AB—with Treatment A increasing S1 utility from 0.75 to 0.95 at $12,000/year, Treatment B reducing S1-to-S2 progression with HR 0.6 at $13,000/year, and transition rewards including a 0.01 utility decrement and $1,000 cost for H-to-S1 and $2,000 cost at death.
  • Analyses include CEA with ICERs using dampack, probabilistic sensitivity analysis with 1,000 sampled parameter sets, and epidemiological outcomes such as survival probability, life expectancy, and prevalence generated from the cohort trace.

Limitations

  • The paper states that time-dependent cSTMs can cause “state explosion” and require substantial RAM; a typical PC with 8GB RAM can handle about 1,000 states and 600 time-cycle slices, and high granularity may make individual-based state-transition models preferable.
  • The paper notes that alternative implementations exist, including updating transition matrices at each cycle to save memory, using a 3-dimensional array for state-residence time, or using R packages such as heemod and hesim; it also states that the case study assumes all parameters are known and discusses only a single tunnel.
  • The text makes evident that the tutorial is a methodological and coding demonstration rather than an empirical validation study; epidemiological outcomes are model-generated and described as useful for calibration and validation, but no external validation against observed data is reported.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1177/0272989x221121747

Cites

37 of 37 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

7 of 12 cited sources checked · 5 supported · 0 flagged · accuracy not yet measured

  • Interpretationreasoning

    Since all members of the cohort in the simulation age together, age-specific background mortality can be implemented by changing the transition to death as the simulation progresses.

    “Since all members of the cohort in the simulation age together, we can implement this dependency by changing the transition to death as the simulation progresses.”
    Quote in the paperSubstantial claim1 of 1 sources checked
    • cites [8] Snowsill T A New Method for Model-Based Health Economic Evaluation Utilizing and Extending Moment-Generating Functions. Medical Decision Making [ Internet] 2019; 39(5): 523– 39. Available from: http://journals.sagepub.com/doi/ 10.1177/0272989X19860119 31319752 source
  • Premisecitation

    Changing the core computation in a traditional cSTM from mtPt to diag(mt)Pt allows computation of the proportion of the cohort that transitions between any two states in a cycle t.

    “Briefly, this approach involves changing the core computation in a traditional cSTM, from mtPt to diag(mt)Pt. This simple change allows us to compute the proportion of the cohort that transitions between any two states in a cycle t.”
    Quote in the paperSubstantial claim1 of 1 sources checked
    • cites Krijkamp et al. (2020) source
  • Findingown data

    The non-dominated strategies SoC, B, and AB form the cost-effectiveness efficient frontier of the CEA based on the simulation-time-dependent Sick-Sicker model.

    “The non-dominated strategies, SoC, B, and AB, form the cost-effectiveness efficient frontier of the CEA based on the simulation-time-dependent Sick-Sicker model (Figure 5).”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: Rewards occur at the beginning of the cycle.
  • Findingown data

    Strategy B has the lowest expected loss for WTP thresholds greater than or equal to $70,000 and less than $105,000 per QALY.

    “The SoC Strategy has the lowest expected loss for WTP thresholds less than $65,000 per QALY, Strategy B has the lowest expected loss for WTP thresholds greater than or equal to $70,000 and less than $105,000.”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: Rewards occur at the beginning of the cycle.
  • Findingown data

    In the PSA, standard of care has the lowest expected cost and QALYs among the four strategies.

    “The CE scatter plot shows that Strategy AB has the highest expected costs and QALYs. Standard of care has the lowest expected cost and QALYs. Strategy B is more effective and least costly than Strategy A. Therefore, Strategy A is strongly dominated by Strategy B.”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: Rewards occur at the beginning of the cycle.
  • Classified as probably not a substantial claim. 20
  • May combine two claims. 12

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Other effect measures such as odds ratio, risk ratio, or risk differences can be used depending on the methods used to estimate them.

    “However, other effect measures can be used, such as odds ratio, risk ratio, or risk differences depending on the methods used to estimate them.30–32”
    Quote in the paperNot a substantial claim1 of 3 sources checked
    • cites [30] Kuntz KM, Weinstein MC. Modelling in economic evaluation. In: Drummond MF, McGuire A, editors. Economic evaluation in health care: Merging theory with practice. 2nd ed. New York, NY: Oxford University Press; 2001. p. 141– 71. source
    • cites [31] Ades AE, Lu G, Claxton K. Expected value of sample information calculations in medical decision modeling. Medical Decision Making [ Internet] 2004; 24(2): 207– 27. Available from: http://www.ncbi.nlm.nih.gov/pubmed/15090106 15090106 source
    • cites [32] Gidwani R, Russell LB. Estimating Transition Probabilities from Published Evidence: A Tutorial for Decision Modelers. PharmacoEconomics [ Internet] 2020; 38(11): 1153– 64. Available from: 10.1007/s40273-020-00937-z PMC7426391 32797380 source

      The “Relative Risks and Odds Ratios” section discusses using risk ratios and odds ratios to derive transition probabilities, including when an odds ratio can approximate a risk ratio. The passages found do not discuss risk differences as an effect measure, so the source supports only part of the cited list. Page numbers were unavailable in our copy. (LLM, conf=0.84)

      Experimental — accuracy not yet measured

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    R packages such as heemod and hesim can support modeling arbitrary patient populations or many treatment strategies.

    “However, there are R packages for building and running cSTMs and conducting CEAs, such as heemod35 and hesim,36 that are made such that they can support modeling arbitrary patient populations or many treatment strategies.”
    Quote in the paperNot a substantial claim1 of 2 sources checked
    • cites [35] Filipović-Pierucci A, Zarca K, Durand-Zaleski I. Markov Models for Health Economic Evaluation: The R Package heemod. arXiv:170203252v1 [ Internet] 2017;April:30. Available from: http://arxiv.org/abs/1702.03252 source
    • cites [36] Incerti D, Jansen JP. hesim: Health Economic Simulation Modeling and Decision Analysis. arXiv:210209437v2 [ Internet] 2021 Feb;March:38. Available from: http://arxiv.org/abs/2102.09437 source

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationcitation

    Implementing transition rewards by calculating incurred rewards based on the proportion of the cohort making the relevant transition requires storing a cohort state-transition array in addition to the cohort trace.

    “For implementation, this requires storing not just the cohort trace, which reflects how many individuals are in each state at a given cycle, but also a cohort state-transition array, which records how many individuals are making each possible transition at a given cycle.9”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites [9] Krijkamp EM, Alarid-Escudero F, Enns E, Pechlivanoglou P, Hunink MM, Jalal H. A Multidimensional Array Representation of State-Transition Model Dynamics Medical Decision Making. 2019;In Press. PMC7065927 31989862 source
    • The authors state this as an assumption: Rewards occur at the beginning of the cycle.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Obtaining information on transitions between health states with origin and destination is relatively straightforward in a cSTM and is described in detail by Krijkamp et al. (2020).

    “However, obtaining this information is relatively straightforward in a cSTM and is described in detail by Krijkamp et al. (2020).”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites Krijkamp et al. (2020) source

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    In the case of hesim, all the memory pre-allocation and loops are made at the C++ level, making it more efficient than looping over base R but might require more RAM.

    “And in the case of hesim, all the memory pre-allocation and loops are made at the C++ level, making it more efficient than looping over base R but might require more RAM.”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites [36] Incerti D, Jansen JP. hesim: Health Economic Simulation Modeling and Decision Analysis. arXiv:210209437v2 [ Internet] 2021 Feb;March:38. Available from: http://arxiv.org/abs/2102.09437 source

      The Discussion says hesim’s simulations are performed at the C++ level and reports substantially faster runtimes than other R packages. The Conclusion also says its algorithms are written in C++. Neither passage establishes that all memory pre-allocation occurs in C++ or that hesim may require more RAM; the paper does not make a direct comparison with loops in base R. (LLM, conf=0.88)

      Experimental — accuracy not yet measured

  • May combine two claims.

    Findingown data

    In the PSA, Strategy B is more effective and less costly than Strategy A.

    “The CE scatter plot shows that Strategy AB has the highest expected costs and QALYs. Standard of care has the lowest expected cost and QALYs. Strategy B is more effective and least costly than Strategy A. Therefore, Strategy A is strongly dominated by Strategy B.”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: Rewards occur at the beginning of the cycle.

  • May combine two claims.

    Findingown data

    At WTP thresholds less than $70,000 per QALY, SoC has the highest probability of being cost-effective and the highest expected NMB.

    “At WTP thresholds less than $70,000 per QALY, SoC is the strategy with the highest probability of being cost-effective and the highest expected NMB.”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: Rewards occur at the beginning of the cycle.

  • May combine two claims.

    Findingown data

    Strategy B produces an expected benefit of 1.32 QALYs per individual for an additional expected cost of $86,162 compared with SoC, with an ICER of $65,288/QALY.

    “The SoC Strategy is the least costly and effective strategy, followed by Strategy B, producing an expected benefit of 1.32 QALYs per individual for an additional expected cost of $86,162 with an ICER of $65,288/QALY followed by Strategy AB with an ICER $104,461/QALY.”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: Rewards occur at the beginning of the cycle.

  • May combine two claims.

    Findingown data

    Strategy AB has the highest expected NMB for WTP thresholds greater than or equal to $105,000 and is the strategy with the highest probability of being cost-effective.

    “Strategy AB has the highest expected NMB for WTP thresholds greater than or equal to $105,000 and is the strategy with the highest probability of being cost-effective.”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: Rewards occur at the beginning of the cycle.

  • May combine two claims.

    Findingown data

    Strategy B has the highest probability of being cost-effective and the highest expected NMB for WTP thresholds greater than $70,000 and lower than $105,000 per QALY.

    “Strategy B has the highest probability of being cost-effective and the highest expected NMB for WTP thresholds greater than $70,000 and lower than $105,000 per QALY.”
    Quote in the paperNot a substantial claim
    • The authors state this as an assumption: Rewards occur at the beginning of the cycle.

  • May combine two claims.

    Findingown data

    SoC is the least costly and least effective strategy among the four strategies in the simulation-time-dependent Sick-Sicker CEA.

    “The SoC Strategy is the least costly and effective strategy, followed by Strategy B, producing an expected benefit of 1.32 QALYs per individual for an additional expected cost of $86,162 with an ICER of $65,288/QALY followed by Strategy AB with an ICER $104,461/QALY.”
    Quote in the paperNot a substantial claim
    • The authors state this as an assumption: Rewards occur at the beginning of the cycle.

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Interpretationnone given

    Any time-varying feature in a discrete-time model can most generally be implemented as tunnel states, with the extreme case of every state having a different tunnel state for each time step.

    “Any time-varying feature in a discrete-time model can most generally be implemented as tunnel states with, at the extreme, every state having a different tunnel state for each time step.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationreasoning

    Using time-varying transition probabilities is a shortcut when the cohort experiences time-varying processes simultaneously as a function of the time from the simulation start.

    “Using time-varying transition probabilities is a shortcut when the cohort experiences these time-varying processes simultaneously as a function of the time from the simulation start.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationreasoning

    The implementation of state-residence time dependence is more involved than simulation-time dependence.

    “The implementation of state-residence time dependence is more involved than simulation-time dependence.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Interpretationnone given

    The same principles for a single tunnel can be applied for adding more tunnels as needed.

    “Here we only discuss models with a single tunnel. The same principles can be applied for adding more tunnels as needed.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based). May combine two claims.

    Interpretationnone given

    Substantial state expansion can be cumbersome and potentially computationally expensive.

    “However, as we see below, substantial state expansion can be cumbersome and potentially computationally expensive.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Interpretationreasoning

    It is possible to build reasonably complex time-dependent cSTMs in R as long as there is sufficient RAM to store the transition probability array and outputs of interest.

    “It is possible to build reasonably complex time-dependent cSTMs in R as long as there is sufficient RAM to store the transition probability array and outputs of interest.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: A typical PC with 8GB of RAM can handle a transition probability array of about 1,000 states and 600 time-cycle slices.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationreasoning

    Each intermediate "Sick" state can have different transition probabilities, costs, or utilities.

    “Each intermediate “Sick” state can have different transition probabilities (e.g., mortality, risk of complications, etc.), costs, or utilities.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Interpretationnone given

    The state expansion approach used for state-residence time dependence can also be used to model simulation-time dependence.

    “We can also use this state expansion approach to model simulation-time dependence.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationreasoning

    The cohort trace matrix M can be used to compute epidemiological outcomes such as prevalence and survival probability over time, and economic outcomes such as cumulative resource use and costs.

    “Thus, M can be used to compute various epidemiological outcomes, such as prevalence and survival probability over time, and economic outcomes, such as cumulative resource use and costs.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Interpretationnone given

    There is typically some probability that the cohort remains in the last tunnel state, like for other non-transient states.

    “There is typically some probability that the cohort remains in that state in the last tunnel state, like for other non-transient states.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductown data

    The transition probability arrays were checked to be valid, ensuring transition probabilities are between 0 and 1 and that transition probabilities from each state sum to 1, using darthtools functions.

    “We used the functions check_sum_of_transition_array and check_transition_probability in the darthtools package (https://github.com/DARTH-git/darthtools) to check if the transition probability arrays are valid (i.e., ensuring transition probabilities are between 0 and 1, and transition probabilities from each state sum to 1).”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationreasoning

    Simulation-time dependence can be used to incorporate seasonal or temporal variation in disease incidence, where risk varies by season or year in the simulation.

    “Simulation-time dependence can also be used to incorporate seasonal or temporal variation in disease incidence, where the risk of developing a disease can vary based on the current season or year in the simulation.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    In a real-world CEA, parameters are estimated from electronic health records, clinical trials, and observational studies, and when there is no individual-level data to inform them, through model calibration.

    “However, in a real-world CEA, parameters are estimated from electronic health records,33 clinical trials, and observational studies, and when there is no individual-level data to inform them, through model calibration.34”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [33] Rodriguez PJ, Ward ZJ, Long MW, Austin SB, Wright DR. Applied Methods for Estimating Transition Probabilities from Electronic Health Record Data. Medical Decision Making [ Internet] 2021; 41(2): 143– 52. Available from: https://journals.sagepub.com/doi/abs/ 10.1177/0272989X20985752?casa_token=SIhw-ApH8ogAAAAA:lxalsZ3tBihW8nVpK1WhSjkuVNR3Z7kKUe44gC5UJk4Rs8KfYoAQ6Y9efV8oVvR1GeLeTDwWdRXKtWk 33563111 source
    • cites [34] Welton NJ, Ades AE. Estimation of markov chain transition probabilities and rates from fully and partially observed data: uncertainty propagation, evidence synthesis, and model calibration. Medical Decision Making [ Internet] 2005; 25(6): 633– 45. Available from: http://www.ncbi.nlm.nih.gov/pubmed/16282214 16282214 source

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    The three-dimensional transition probability array with tunnels was checked to be valid using the darthtools package.

    “Again, we check the three-dimensional transition probability array with tunnels is valid using the darthtools package.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Interpretationreasoning

    Even if the time-varying process has a different periodicity than the cycle length, either tunnel states or time-varying transition probabilities can be used to capture these effects, but the process must be represented or approximated over an integer number of cycle lengths.

    “Even if the time-varying process has a different periodicity than the cycle length, either tunnel states or time-varying transition probabilities can be used to capture these effects. However, this time-varying process must be represented or approximated over an integer number of cycle lengths.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Interpretationreasoning

    As state-residence time dependence extends in more health states, the total number of health states will increase, causing a state explosion.

    “For example, as state-residence time dependence extends in more health states, the total number of health states will increase, causing a “state explosion’’.”
    Quote in the paperSubstantial claim
27 of 27

  • The paper relies on this without support.

    Many cSTMs require transitions, rewards, or both to vary over time.

    “However, in practice, many cSTMs require transitions, rewards, or both to vary over time (time-dependent).”

  • The paper relies on this without support.

    Epidemiological outcomes are used to calibrate and validate models against observed real-world data.

    “Such epidemiological outcomes can be informative, but they are also used to calibrate and validate the model against observed real-world data.”

  • The authors state this as an assumption.

    Healthy individuals face an age-specific background mortality.

    “In this tutorial, we assume that healthy individuals face an age-specific background mortality.”

  • The authors state this as an assumption.

    It is not clinically possible to distinguish between individuals in S1 and S2, so any treatment strategy involves treating individuals in both S1 and S2, regardless of whether they experience a benefit.

    “We assume that it is not clinically possible to distinguish between individuals in S1 and S2, so any treatment strategy involves treating individuals in both S1 and S2, regardless of whether they experience a benefit.”

  • The authors state this as an assumption.

    When Treatment A and B are combined, their effects are independent and treatment costs are additive.

    “When combined, the effects of Treatment A and B are assumed to be independent and treatment costs are additive.”

  • The authors state this as an assumption.

    The risk of progression from S1 to S2 increases as a function of time spent in S1 following a Weibull hazard function.

    “We assume the risk of progression from S1 to S2 increases as a function of the time τ = 1,…, ntunnels the cohort remains in the S1 state. This increase follows a Weibull hazard function”

  • The authors state this as an assumption.

    Transition rates are constant within each cycle interval (piecewise exponential transition times).

    “we assume constant rates within each cycle interval (i.e., piecewise exponential transition times)”

  • The authors state this as an assumption.

    State-residence dependence affects the cohort in the S1 state throughout the whole simulation (ntunnels = nT).

    “We assume that state-residence dependence affects the cohort in the S1 state throughout the whole simulation (i.e., ntunnels = nT )”

  • The authors state this as an assumption.

    Rewards occur at the beginning of the cycle.

    “and assumes that rewards occur at the beginning of the cycle.9”

  • The authors state this as an assumption.

    All parameters are known for the case study.

    “We assume that all parameters are known for the case study.”

  • The paper relies on this without support.

    A typical PC with 8GB of RAM can handle a transition probability array of about 1,000 states and 600 time-cycle slices.

    “For example, a typical PC with 8GB of RAM can handle a transition probability array of about 1,000 states and 600 time-cycle slices.”
11 of 11

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

  • Transition-array cycle count differs by one

    Minor

    The transition probability array is specified to have nT time slices, but the text later says a constant transition probability is replicated nT + 1 times. Both cannot describe the same array.

1 of 1

Cites

Flagged first, then by citations

Research works printed but not linked: 3

  1. Arias E, 2017, Natl Vital Stat Rep, 66, 63
  2. Rothman KJ, 2008, Modern Epidemiology
  3. Gidwani R, 2020, Pharmacoeconomics, 38

Cited by

Sorted by citations

Related

Authors
Alarid-Escudero, F., Krijkamp, E., Enns, E. A., Yang, A., Hunink, M. G. M., Pechlivanoglou, P., & Jalal, H.
Affiliations
  • Department of Health Policy, School of Medicine, and Stanford Health Policy, Freeman-Spogli Institute for International Studies, Stanford University, Stanford, California, USA
  • Erasmus MC (NL)
  • University of Minnesota (US)
  • Hospital for Sick Children (CA)
  • Harvard University (US)
  • University of Ottawa (CA)
ORCID
Venue
Medical Decision Making
Publisher
SAGE Publications
Published
2022-09-16
Volume
43
Issue
1
Pages
21-41
ISSN
  • 0272-989X (ISSN-L)
  • 1552-681X
DOI
10.1177/0272989x221121747
PMC
PMC9844995
OpenAlex
W3198103700
Cited by
5
References
33
Citations by year
2025: 1 · 2022: 3 · 2021: 1
FWCI
0.79
Type
journal article
Language
en
Open access
Licence
Locations
  • Medical Decision Making (publishedVersion)
  • EUR Research Repository (Erasmus University Rotterdam) (publishedVersion, other-oa)
  • PubMed (publishedVersion)
  • RePEc: Research Papers in Economics (submittedVersion)
  • PubMed Central (submittedVersion)
  • arXiv (Cornell University) (submittedVersion)
  • EUR Research Repository (Erasmus University Rotterdam) (submittedVersion)
  • EUR Research Repository (Erasmus University Rotterdam) (submittedVersion, cc-by)
  • arXiv (Cornell University)
  • Sage Journals Data (cc-by)
Funders
  • Society for Medical Decision Making
  • National Cancer Institute (U01-CA199335, U01-CA253913, U01-CA265729, U01-CA265750)
  • National Institute of Allergy and Infectious Diseases (K25AI118476)
  • National Institute on Drug Abuse (K01DA048985)
  • Gordon and Betty Moore Foundation (GBMF7853)
Topics
  • Social Sciences › Economics, Econometrics and Finance › Economics and Econometrics › Health Systems, Economic Evaluations, Quality of Life
  • Social Sciences › Decision Sciences › Management Science and Operations Research › Simulation Techniques and Applications
  • Physical Sciences › Mathematics › Statistics and Probability › Statistical Methods in Clinical Trials
Keywords
cost-effectiveness analysis, probabilistic sensitivity analysis, survival probability, disease prevalence, model calibration, health economic modelling, R implementation
MeSH terms
Cost-Effectiveness Analysis, Computer Simulation, Cost-Benefit Analysis, Humans, Markov Chains, Probability

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-16.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-16.

Cited sources are checked by gpt-6-sol.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review