Background: Several risk factors have emerged for novel 2019 coronavirus disease (COVID-19) infection and severity. Yet, it is unknown to what degree these risk factors alone or in combination can accurately predict who is most at risk. It is also worthwhile to consider serological antibody titers to non COVID-19 infectious diseases, which may influence host immunity to COVID-19. Methods: In this retrospective study of multicenter UK Biobank participants, as of May 26th 2020, all COVID-19 testing data was collected by Public Health England for older adult in- and out-patients (69.6 ± 8.8 years). We used linear discriminant analysis with cross-validation and bootstrapping to determine the accuracy, specificity, and sensitivity of baseline data from 2006-2010 to predict COVID-19 infection and presumptive severity (i.e., testing at hospital). Receiver operating characteristic (ROC) curves were used to derive the area under the curve (AUC). Findings: This retrospective study included 4,510 unique participants and 7,539 testing instances (i.e., test cases). Testing resulted in 5,329 negative cases and 2,210 positive cases, split into 996 mild and 1,214 severe disease outcomes. Baseline data including demographics, bioimpedance-derived body composition, vitals, serum biochemistry, self-reported illness/disability, and complete blood count. A randomized subset of 80 participants with 124 test cases also had antibody titers for 20 common to rare infectious diseases. Among all test cases, accuracy was modest for final diagnostic models of COVID-19 infection (70.2%; AUC=0.570, CI=0.556-0.584) and severity (58.3%; AUC=0.592, CI=0.568-0.615). In the sub-group with serology, by contrast, final models predicted infection and severity with an accuracy of 93.5% (AUC=0.969, CI=0.934-1.000) and 74.4% (AUC=0.803, CI=0.663-0.943) respectively. Models included titers to common pathogens (e.g., human cytomegalovirus), age, blood cell counts, lipids, and other biochemical markers. Interpretation: Serological titers for infectious diseases and other risk factors could help policy makers and clinicians better identify who may get COVID-19 and require hospitalization.