Crash counts by crash type and severity are significantly correlated. The effects of contributing factors on crashes can vary by crash type and by crash severity. The Multivariate Poisson Lognormal (MVPLN) model outperforms the Univariate Poisson Lognormal (UPLN) model in crash prediction accuracy. The Integrated Nested Laplace Approximation (INLA) MVPLN model can decrease the computational time compared with the Markov Chain Monte Carlo (MCMC) MVPLN model.