Estimating duration in sex work from cross-sectional studies: a Bayesian global synthesis suggests shorter durations
- 11) MAP Centre for Urban Health Solutions, Unity Health Toronto
Abstract
Durations selling sex are key determinants of HIV/STI transmission dynamics and service coverage. Such durations are usually estimated from cross-sectional surveys of people actively selling sex. However, cross-sectional surveys are more likely to include individuals selling sex for longer, and survey-reported durations will be right-censored. Prior works have not adequately addressed these two competing biases. In this paper, we derive the mathematical relationship between the distribution of durations reported in cross-sectional surveys and the distribution of true durations in the population overall. These distributions will be identical if true durations are exponentially distributed; otherwise, reported durations will typically underestimate the variance in true durations. We develop a parametric Bayesian model of this relationship, and apply it to re-analyze aggregate data from a global systematic review. Our model-estimated average true durations selling sex are significantly shorter than previously published for all regions - e.g. in Africa: 2.1 (95% CI: 1.8-2.5) versus 5.5 years; in Asia: 1.4 (1.2-1.5) versus 2.8-4.0 years. Shorter average durations then suggest that HIV incidence in sex work could be higher than previously thought, to match the same HIV prevalence data. Our approach can be applied beyond sex work, to characterize duration distributions for any transient risk/disease state.
Related
- Estimating the Range and Duration of Post-influenza Symptoms, Conditions, and Diseases
- Bayesian Anchored Latent Learning for Estimating State Trajectories
- Prescription-Duration Patterns of Outpatient Macrolide Use in Participating Community Pharmacies in Tokyo, Japan: A 24-Month Repeated Cross-Sectional Study
- Explainable machine learning for identifying insufficient sleep duration among preschool children in cold-region China: a cross-sectional study
- Association between duration of smoking cessation and depressive symptoms among community-dwelling older adults: a cross-sectional study
BibTeX
@misc{knight2026estimating,
title = {{Estimating duration in sex work from cross-sectional studies: a Bayesian global synthesis suggests shorter durations}},
author = {J. Knight and S. Wang and L. Wang and S. Mishra},
howpublished = {medRxiv},
year = {2026},
month = nov,
doi = {10.64898/2026.11.03.26367114},
url = {https://www.medrxiv.org/content/10.64898/2026.11.03.26367114v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - Estimating duration in sex work from cross-sectional studies: a Bayesian global synthesis suggests shorter durations
AU - J. Knight
AU - S. Wang
AU - L. Wang
AU - S. Mishra
PY - 2026
DA - 2026/11/03
PB - medRxiv
DO - 10.64898/2026.11.03.26367114
UR - https://www.medrxiv.org/content/10.64898/2026.11.03.26367114v1
AB - Durations selling sex are key determinants of HIV/STI transmission dynamics and service coverage. Such durations are usually estimated from cross-sectional surveys of people actively selling sex. However, cross-sectional surveys are more likely to include individuals selling sex for longer, and survey-reported durations will be right-censored. Prior works have not adequately addressed these two competing biases. In this paper, we derive the mathematical relationship between the distribution of durations reported in cross-sectional surveys and the distribution of true durations in the population overall. These distributions will be identical if true durations are exponentially distributed; otherwise, reported durations will typically underestimate the variance in true durations. We develop a parametric Bayesian model of this relationship, and apply it to re-analyze aggregate data from a global systematic review. Our model-estimated average true durations selling sex are significantly shorter than previously published for all regions - e.g. in Africa: 2.1 (95% CI: 1.8-2.5) versus 5.5 years; in Asia: 1.4 (1.2-1.5) versus 2.8-4.0 years. Shorter average durations then suggest that HIV incidence in sex work could be higher than previously thought, to match the same HIV prevalence data. Our approach can be applied beyond sex work, to characterize duration distributions for any transient risk/disease state.
ER -