Use of Latent Profile Analysis to Model the Translation of University Research into Health Practice and Policy: Exploration of Proposed Metrics

The aim of this study is to profile academic institutions (n = 127) based on publications, citations in the top 10% of journals, patent citations in Food and Drug Administration (FDA) approvals, clinical trials with uploaded results, contributions to clinical practice guidelines, awarded patents, start-ups, and licenses generating income in response to the Association of University Technology Managers (AUTM) Licensing Activity Survey: Fiscal Years 2011–2015. Latent variable modeling (LVM) was conducted in Mplus v.8.1, specifically latent profile analysis (LPA) was utilized to predict institutional profiles of research, which were compared with the 2015 Carnegie Classification System ranks. Multivariate regression of profile assignment on research expenditure and income generated by licensure was used to show concurrent validity. The LPA resulted in three profiles as the most parsimonious model. Mantel-Haenszel test of trend to the Carnegie Classification found a positive and significant association among institution rankings (r = 0.492, χ 2 (1) = 26.69, p < 0.001). Profile assignment significantly predicted differences in research expenditure and income generated by licensure. By classifying academic institutions into improving, mobilizing and thriving translational research profiles allows for a universal metric of translation of science from basic or bench to practice or policy.