Article

Effects of Categorizing Continuous Variables in Decision-Analytic Models

Faculty of Arts and Sciences, Harvard University, Cambridge, Massachusetts, USA.
Medical Decision Making (Impact Factor: 2.27). 08/2009; 29(5):549-56. DOI: 10.1177/0272989X09340238
Source: PubMed

ABSTRACT When using continuous predictor variables in discrete-state Markov modeling, it is necessary to create categories of risk and assume homogeneous disease risk within categories, which may bias model outcomes. This analysis assessed the tradeoffs between model bias and complexity and/or data limitations when categorizing continuous risk factors in Markov models.
The authors developed a generic Markov cohort model of disease, defining bias as the percentage change in life expectancy gain from a hypothetical intervention when using 2 to 15 risk factor categories as compared with modeling the risk factor as a continuous variable. They evaluated the magnitude and sign of bias as a function of disease incidence, disease-specific mortality, and relative difference in risk among categories.
Bias was positive in the base case, indicating that categorization overestimated life expectancy gains. The bias approached zero as the number of risk factor categories increased and did not exceed 4% for any parameter combinations or numbers of categories considered. For any given disease-specific mortality and disease incidence, bias increased with relative risk of disease. For any given relative risk, the relationship between bias and parameters such as disease-specific mortality or disease incidence was not always monotonic.
Under the assumption of a normally distributed risk factor and reasonable assumption regarding disease risk and moderate values for the relative risk of disease given risk factor category, categorizing continuously valued risk factors in Markov models is associated with less than 4% absolute bias when at least 2 categories are used.

0 Followers
 · 
99 Views
  • [Show abstract] [Hide abstract]
    ABSTRACT: Markov models of disease progression are widely used to model transitions in patients' health state over time. Usually, patients' health status may be classified according to a set of ordered health states. Modelers lump together similar health states into a finite and usually small, number of health states that form the basis of a Markov chain disease-progression model. This increases the number of observations used to estimate each parameter in the transition probability matrix. However, lumping together observably distinct health states also obscures distinctions among them and may reduce the predictive power of the model. Moreover, as we demonstrate, precision in estimating the model parameters does not necessarily improve as the number of states in the model declines. This paper explores the tradeoff between lumping error introduced by grouping distinct health states and sampling error that arises when there are insufficient patient data to precisely estimate the transition probability matrix. Copyright © 2013 John Wiley & Sons, Ltd.
    Statistics in Medicine 09/2013; 32(22). DOI:10.1002/sim.5808 · 2.04 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: INTRODUCTION: Despite the known harmful effects of smoking during pregnancy, the highly addicted find it difficult to quit. Decreased smoking may be regarded as a means of harm reduction. There is limited information on the benefits of smoking reduction short of quitting. This study used salivary cotinine to assess the impact of change in smoking exposure on birth weight in full-term infants. METHODS: In a prenatal smoking cessation study, smoking status was validated by saliva cotinine at baseline and end of pregnancy (EOP). Salivary cotinine ≥15ng/ml defined active smoking. Based on salivary cotinine, women were grouped as nonsmoking/quit, light exposure (<150ng/ml), and heavy exposure (≥150ng/ml) at baseline and EOP. EOP and baseline smoking status were stratified to form smoking exposure change groups. Mean birth weight was compared among those who quit, reduced, maintained, and increased. RESULTS: Smoking cessation was associated with a 299g increase in birth weight compared with sustained heavy smoking, p = .021. Reduced exposure from heavy to light was associated with a 199g increase in birth weight compared with sustained heavy exposure, a 103g increase compared with increased exposure, and a 63g increase compared with sustained light exposure. Differences among continuing smokers were not statistically significant.Conclusions:Although not statistically significant, the increase in infant birth weight associated with reduction from heavy to light exposure suggests potential for benefit. The only statistically significant comparison was between quitters and sustained heavy smokers, confirming that smoking cessation should remain the goal for pregnant women.
    Nicotine & Tobacco Research 09/2012; 15(3). DOI:10.1093/ntr/nts184 · 2.81 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: The evaluation of the cost and health implications of agreeing to cover a new health technology is best accomplished using a model that mathematically combines inputs from various sources, together with assumptions about how these fit together and what might happen in reality. This need to make assumptions, the complexity of the resulting framework, the technical knowledge required, as well as funding by interested parties have led many decision makers to distrust the results of models. To assist stakeholders reviewing a model’s report, questions pertaining to the credibility of a model were developed. Because credibility is insufficient, questions regarding relevance of the model results were also created. The questions are formulated such that they are readily answered and they are supplemented by helper questions that provide additional detail. Some responses indicate strongly that a model should not be used for decision making: these trigger a “fatal flaw” indicator. It is hoped that the use of this questionnaire, along with the three others in the series, will help disseminate what to look for in comparative effectiveness evidence, improve practices by researchers supplying these data, and ultimately facilitate their use by health care decision makers.
    Value in Health 03/2014; 17(2):174–182. DOI:10.1016/j.jval.2014.01.003 · 2.89 Impact Factor