Statistical Modeling of Semicontinuous Health Outcomes

Semicontinuous health outcomes arise when an outcome variable combines a mass of exact zeros with a continuously distributed positive component. Common examples include healthcare costs, utilisation measures and biomarker readings where non-use or absence of an event yields zero, and usage or measurement yields a skewed distribution of positive values. Traditional linear models fail to capture this dual nature, prompting the development of specialised frameworks. Two-part or hurdle models separate the process generating zeros from that governing positive values, while zero-inflated alternatives embed a point mass at zero within a single distribution. Tweedie models offer a one-step solution by treating the outcome as a compound distribution with a discrete and a continuous component. For longitudinal or clustered data, multilevel and mixed-effects extensions account for within-subject correlations and variability across units. Bayesian approaches have further enriched the toolkit by enabling flexible distributional assumptions and coherent uncertainty quantification. These methods underpin cost-effectiveness analyses, risk stratification, nutritional and behavioural studies and trials involving semicontinuous biomarkers, thereby informing policy, resource allocation and personalised interventions globally.

Research from Nature Portfolio

No recent Nature Portfolio content available.

Statistical Modeling of Semicontinuous Health Outcomes publication trend

The graph below shows the total number of articles in statistical modeling of semicontinuous health outcomes across all publications each year (not limited to Nature Index journals).

Technical terms

Semicontinuous data: Data comprising a point mass at zero and a continuous positive distribution, common in cost and utilisation measures.

Two-part model: A framework that separately models the occurrence of a positive value and the magnitude given positivity, typically via a binary and a continuous submodel.

Zero-inflation: An excess frequency of zeros beyond that expected under standard count or continuous distributions, accommodated via mixture or hurdle components.

Tweedie distribution: A compound Poisson–Gamma family that unifies a discrete point mass at zero with a continuous right-skewed component in a single model.

Multilevel (mixed-effects) model: A regression approach incorporating fixed effects for population-level predictors and random effects to capture correlation within clusters or subjects.

References

1. Estimating Costs Associated with Disease Model States Using Generalized Linear Models: A Tutorial. PharmacoEconomics (2023).
2. Studying dietary intake in daily life through multilevel two-part modelling: a novel analytical approach and its practical application. International Journal of Behavioral Nutrition and Physical Activity (2021).
3. Tweedie distributions for fitting semicontinuous health care utilization cost data. BMC Medical Research Methodology (2017).
4. Zero-augmented beta-prime model for multilevel semi-continuous data: a Bayesian inference. BMC Medical Research Methodology (2022).