Purpose <p>Assess the utility of access to healthcare, clinical conditions, and social determinants of health (SDoH) variables in population-level suicide prediction models.</p> Methods <p>Negative binomial regression models were constructed using data from population-level surveys, state death certificates, federal records of behavioral health services, and U.S. Census data. Outcomes of interest were suicidal ideation and suicide attempt (SISA), inpatient psychiatric hospitalization (IPH), and suicide death. The relative changes in pseudo R² were used to assess the impact of variable categories (i.e., clinical conditions, access to healthcare, and geo-derived SDoH) when added to a demographic-only baseline suicide prediction model.</p> Results <p>Clinical data showed a significant impact, with the largest percent increase in pseudo R² compared to the demographic-only baseline model (321.9% for SISA; 736.9% for IPH; 18.9% for suicide death). Access to healthcare and geo-derived SDoH also improved model performances for all outcomes, but considerably lower than clinical variables. Models with all variable categories had the highest pseudo R², with .68, .58, and .46 for SISA, IPH, and suicide, respectively. Availability of emergency mental health services was found to be protective against IPH (IRR .90; 95% CI .84—.96) and suicide death (IRR .91; 95% CI .84—.97).</p> Conclusions <p>Clinical data proved to have the most effective variables in predicting a continuum of suicidal behaviors. While the impacts of access to healthcare and SDoH factors were comparatively limited, these variables also contributed to additional model improvements. These findings show the utility of population-level healthcare services and SDoH for ecological suicide behavior risk prediction.</p>