Spatiotemporal trends and ecological determinants of cardiovascular mortality among 2844 counties in mainland China, 2006–2020: a Bayesian modeling study of national mortality registries

Table 2 Posterior estimates of associations between nonmedical ecological determinants and CVD mortality risk in China, 2006–2020: estimated from Bayesian multivariable regression model

Region	RR_GDP^a	RR_NTL	RR_NB	RR_PD
Nationwide	0.92 (0.89, 0.96) \({P}_{RR_{GDP<1\mid data}}=99.03\%\)^bc	0.89 (0.83, 0.94) \({P}_{RR_{NTL<1\mid data}}=99.99\%\)	0.98 (0.94, 1.02) \({P}_{RR_{NB<1\mid data}}=43.09\%\)	0.99 (0.95, 1.01) \({P}_{RR_{PD<1\mid data}}=51.22\%\)
Eastern areas	0.90 (0.88, 0.93) \({P}_{RR_{GDP<1\mid data}}=99.91\)%	0.83 (0.78, 0.87) \({P}_{RR_{NTL<1\mid data}}=99.99\%\)	0.89 (0.86, 0.92) \({P}_{RR_{NB<1\mid data}}=99.82\%\)	0.96 (0.90, 1.02) \({P}_{RR_{PD<1\mid data}}=54.93\%\)
Central areas	0.99 (0.95, 1.02) \({P}_{RR_{GDP<1\mid data}}=58.94\%\)	0.92 (0.85, 1.00) \({P}_{RR_{NTL<1\mid data}}=94.13\%\)	0.99 (0.95, 1.04) \({P}_{RR_{NB<1\mid data}}=49.13\%\)	0.98 (0.94, 1.02) \({P}_{RR_{PD<1\mid data}}=47.66\%\)
Western areas	0.96 (0.91, 1.01) \({P}_{RR_{GDP<1\mid data}}=87.51\%\)	0.84 (0.75, 0.92) \({P}_{RR_{NTL<1\mid data}}=99.98\%\)	0.98 (0.92, 1.04) \({P}_{RR_{NB<1\mid data}}=42.29\%\)	0.97 (0.92, 1.01) \({P}_{RR_{PD<1\mid data}}=41.63\%\)

^aThe Bayesian model was adjusted for annual average temperature (TEMP, °C), temperature variability (TV, °C), annual average relative humidity (HUMID, %), longitude (LT), altitude (AT), and concentration of PM_2.5 emission (PM_2.5, μg/m³), as confounders in main analysis
^bRRs represented the relative risk of posterior median and its 95% confidence interval (CI), equals to an exponential transformation of the regression coefficients of proxies and CVD mortality
^c\({P}_{RR_{proxies<1\mid data}}\) represented the posterior probability of negative association between different proxies and CVD mortality during 2006 and 2020. In a county in which the entire posterior probability of CVD mortality was negatively associated with those proxies, there was around a 100% posterior probability of a negative association and a 0% posterior probability of a positive association, and vice versa. Meanwhile, posterior probabilities more distant from 50% indicate more certainty

ISSN: 1741-7015