Comparison of survival analysis approaches to modelling age at first sex among youth in Kisesa Tanzania

PLOS ONE RESEARCH ARTICLE Comparison of survival analysis approaches to modelling age at first sex among youth in Kisesa Tanzania Jacqueline Materu ID1,2*, Eveline T. Konje2, Mark Urassa1, Milly Marston3, Ties Boerma4, Jim Todd1,3 1 Program of Sexual and Reproductive Health, National Institute for Medical Research, Mwanza Centre, Mwanza, Tanzania, 2 Department of Biostatistics, Epidemiology and Behavioral Sciences, School of Public Health, Catholic University of Health, and Allied Sciences, Mwanza, Tanzania, 3 Department of Population Health, London School of Hygiene and Tropical Medicine, London, United Kingdom, 4 Institute for Global Public Health, University of Manitoba, Manitoba, Canada a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 * Abstract Background OPEN ACCESS Citation: Materu J, Konje ET, Urassa M, Marston M, Boerma T, Todd J (2023) Comparison of survival analysis approaches to modelling age at first sex among youth in Kisesa Tanzania. PLoS ONE 18(9): e0289942. https://doi.org/10.1371/ journal.pone.0289942 Editor: José Antonio Ortega, University of Salamanca, SPAIN Received: January 20, 2023 Accepted: July 30, 2023 Published: September 7, 2023 Copyright: © 2023 Materu et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Data Availability Statement: All relevant data are within the paper and its Supporting Information files. Funding: Fogarty International Center provided support for JM’s training to conduct this work (D43 TW011826). Competing interests: The authors have declared that no competing interests exist. Many studies analyze sexual and reproductive event data using descriptive life tables. Survival analysis has better power to estimate factors associated with age at first sex (AFS), but proportional hazards models may not be right model to use. This study used accelerated failure time (AFT) models, restricted Mean Survival time model (RMST) models, with semi and non-parametric methods to assess age at first sex (AFS), factors associated with AFS, and verify underlying assumptions for each analysis. Methods Self-reported sexual debut data was used from respondents 15–24 years in eight cross-sectional surveys between 1994–2016, and from adolescents’ survey in an observational community study (2019–2020) in northwest Tanzania. Median AFS was estimated in each survey using non-parametric and parametric models. Cox regression, AFT parametric models (exponential, gamma, generalized gamma, Gompertz, Weibull, log-normal and log-logistic), and RMST were used to estimate and identify factors associated with AFS. The models were compared using Akaike information criterion (AIC) and Bayesian information criterion (BIC), where lower values represent a better model fit. Results The results showed that in every survey, the Cox regression model had higher AIC and BIC compared to the other models. Overall, AFT had the best fit in every survey round. The estimated median AFS using the parametric and non-parametric methods were close. In the adolescent survey, log-logistic AFT showed that females and those attending secondary and higher education level had a longer time to first sex (Time ratio (TR) = 1.03; 95% CI: 1.01–1.06, TR = 1.05; 95% CI: 1.02–1.08, respectively) compared to males and those who PLOS ONE | https://doi.org/10.1371/journal.pone.0289942 September 7, 2023 1 / 18 PLOS ONE Comparison of survival analysis approaches to modelling age at first sex reported not being in school. Cell phone ownership (TR = 0.94, 95% CI: 0.91–0.96), alcohol consumption (TR = 0.88; 95% CI: 0.84–0.93), and employed adolescents (TR = 0.95, 95% CI: 0.92–0.98) shortened time to first sex. Conclusion The AFT model is better than Cox PH model in estimating AFS among the young population. Introduction Age at first sex (AFS) is a critical indicator for measuring the onset of an adolescent’s sexual and reproductive life. The onset of sex is a normative step in adolescent sexual development [1]. However, early sex is associated with negative outcomes, including unwanted pregnancies and sexually transmitted infections (STIs) [1–3]. Once young people become sexually active, they are at greater risk of having multiple, usually consecutive, short-term sexual relationships, and inconsistent use of condoms, putting them at higher risk of contracting HIV and other STIs [4, 5]. Accurate monitoring and estimation of AFS has become increasingly important in measuring behavioral changes in HIV prevention and family planning programs [5]. Data on AFS are often collected through self-reports in nationally representative household surveys to track health and population indicators such as the Demographic and Health Surveys (DHS) [6, 7]. Many challenges have been identified in the measurement and modelling of AFS, as studies have shown inconsistent trends that were difficult to interpret [8, 9]. Measurement challenges encompass recall biases, social desirability responses, and a lack of accurate information [10, 11]. Modelling challenges involve failing to account for age censoring [12– 14], and the use of inappropriate analysis methods such as logistic regression when dealing with AFS outcome. Some studies in the existing literature [8, 9] have provided arguments highlighting the distinct advantages of survival analysis in assessing the initiation of sexual and reproductive events because of the distinctive characteristics of the data and its population. AFS is most often estimated using time to event methods, and frequently, a Cox’s proportional hazards (PH) model is applied to estimate factors associated with AFS [8, 9, 15–18]. The Cox PH model necessitates the fulfillment of the assumption of hazard function proportionality. In the case of AFS, the Cox PH assumption is improbable to be satisfied since all individuals will eventually initiate sex, making it impossible for one group to consistently possess a greater risk (or hazard) than another group. When the Cox PH assumption is violated, the utilization of the standard Cox PH model becomes inappropriate, as it can introduce significant bias and result in diminished statistical power when estimating or inferring the impact of a specific risk factor on desired outcomes [19]. According to a review of survival analysis in cancer journals, it was reported that only 5% of all studies using the Cox PH model examined the underlying PH assumption [20]. Similarly, some studies have used Cox’s PH model to find factors associated with AFS without clearly stating whether they examined the PH assumption and if the assumption was met or not [8, 9, 15–18]. While time-to-event methods have been optimal for modeling and estimating factors related to AFS, other studies have classified AFS into predefined time intervals and used standard logistic re (...truncated)