Adherence to nature-based social prescribing: a target trial emulation of RECETAS Prague study
Abstract
Background: In pragmatic intervention studies, adherence often varies over time and depends on participant characteristics that also influence outcomes, potentially biasing conventional per-protocol analyses. Target trial emulation(TTE) provides a structured framework for estimating causal effects of adherence strategies, while avoiding selection bias during follow-up.
Methods: We analyzed data from the RECETAS Prague trial(n = 131), evaluating the 9-week nature-based social prescribing intervention Friends in Nature. Adherence strategies were defined by sessions attended and grouped (0-1, 2–5, 6–7, and 8–9). Outcomes included quality of life(EQ-5D-5L) and loneliness(De Jong Gierveld Score) at 3-months. A TTE framework with cloning, artificial censoring, and inverse-probability-of-censoring weighting estimated causal per-protocol effects. Missing data were handled using MICE(n = 5). Weighted generalized estimating equation models were fitted in each imputed dataset, and estimates were pooled using Rubin’s rules. Overall influence was assessed using a test for trend.
Results: Compared with attendance of 0-1 sessions, EQ-5D scores at 3 months were similar for 2–5(0.008; 95%CI −0.02 to 0.04), 6–7(0.014; 95%CI −0.02 to 0.05), and 8–9(0.008; 95%CI −0.03 to 0.05). Loneliness scores were also similar (2–5: −0.587; 95%CI −2.10 to 0.92; 6–7: 0.157; 95%CI −1.19 to 1.50; 8–9: −0.090; 95%CI −1.73 to 1.55) (test for trend: p = 0.303).
Conclusions: This study shows that TTE can provide a transparent framework for estimating adherence-related effects under time-varying participation in pragmatic trials. No differences in quality of life or loneliness were observed across attendance levels at 3 months, suggesting that longer intervention or follow-up periods may be needed to detect potential effects. However, residual confounding from unmeasured attitudes and differential compliance, particularly lower comparator adherence due to the non-blinded study design, may limit causal interpretation.
Key messages: • TTE offers a transparent framework to estimate adherence-related effects under time-varying participation in pragmatic public health trials.
• No 3-month differences in quality of life or loneliness were observed across attendance levels, suggesting longer intervention or follow-up may be needed.
Related
- Slice regular functions on alternative *-algebras : prescribing zeroes and values on discrete sets and related extension problems
- The Physics Is the Network: Fast, Training-Free Neural Emulation
- Multifidelity Computer Model Emulation Via Diffusion Model Steering and Targeted Maximum Likelihood
- Bayesian Treed Vecchia Gaussian Process Emulation via Sequential Monte Carlo
- Target Trial Emulation with the R Package TTE: A Tutorial and Methodological Guide
BibTeX
@article{arvandi2026adherence,
title = {{Adherence to nature-based social prescribing: a target trial emulation of RECETAS Prague study}},
author = {M Arvandi and S Puntscher and JS Litt and A Bártová and B Novotná and V Papon and LR Hallsson and U Rochau and U Siebert},
journal = {The European Journal of Public Health},
year = {2026},
month = oct,
doi = {10.1093/eurpub/ckag169.1560},
url = {https://doi.org/10.1093/eurpub/ckag169.1560},
note = {CC BY-NC 4.0},
}
RIS
TY - JOUR
TI - Adherence to nature-based social prescribing: a target trial emulation of RECETAS Prague study
AU - M Arvandi
AU - S Puntscher
AU - JS Litt
AU - A Bártová
AU - B Novotná
AU - V Papon
AU - LR Hallsson
AU - U Rochau
AU - U Siebert
PY - 2026
DA - 2026/10/01
JO - The European Journal of Public Health
DO - 10.1093/eurpub/ckag169.1560
UR - https://doi.org/10.1093/eurpub/ckag169.1560
AB - Background: In pragmatic intervention studies, adherence often varies over time and depends on participant characteristics that also influence outcomes, potentially biasing conventional per-protocol analyses. Target trial emulation(TTE) provides a structured framework for estimating causal effects of adherence strategies, while avoiding selection bias during follow-up. Methods: We analyzed data from the RECETAS Prague trial(n = 131), evaluating the 9-week nature-based social prescribing intervention Friends in Nature. Adherence strategies were defined by sessions attended and grouped (0-1, 2–5, 6–7, and 8–9). Outcomes included quality of life(EQ-5D-5L) and loneliness(De Jong Gierveld Score) at 3-months. A TTE framework with cloning, artificial censoring, and inverse-probability-of-censoring weighting estimated causal per-protocol effects. Missing data were handled using MICE(n = 5). Weighted generalized estimating equation models were fitted in each imputed dataset, and estimates were pooled using Rubin’s rules. Overall influence was assessed using a test for trend. Results: Compared with attendance of 0-1 sessions, EQ-5D scores at 3 months were similar for 2–5(0.008; 95%CI −0.02 to 0.04), 6–7(0.014; 95%CI −0.02 to 0.05), and 8–9(0.008; 95%CI −0.03 to 0.05). Loneliness scores were also similar (2–5: −0.587; 95%CI −2.10 to 0.92; 6–7: 0.157; 95%CI −1.19 to 1.50; 8–9: −0.090; 95%CI −1.73 to 1.55) (test for trend: p = 0.303). Conclusions: This study shows that TTE can provide a transparent framework for estimating adherence-related effects under time-varying participation in pragmatic trials. No differences in quality of life or loneliness were observed across attendance levels at 3 months, suggesting that longer intervention or follow-up periods may be needed to detect potential effects. However, residual confounding from unmeasured attitudes and differential compliance, particularly lower comparator adherence due to the non-blinded study design, may limit causal interpretation. Key messages: • TTE offers a transparent framework to estimate adherence-related effects under time-varying participation in pragmatic public health trials. • No 3-month differences in quality of life or loneliness were observed across attendance levels, suggesting longer intervention or follow-up may be needed.
SN - 1101-1262
SN - 1464-360X
ER -