Faced with a rapidly aging population and the rising prevalence of Alzheimer’s disease (AD) and related dementias, the field of dementia research needs to urgently consider screening tools that utilize widely accessible data modalities. We stratified relevant data based upon collection time, expertise needed, cost of administration, and data availability in a variety of clinical settings to form three distinct tiers: low-burden, medium-burden, and high-burden. Low-burden data, for instance, may include demographics, patient history, family history, and simple cognitive tests (e.g., Mini-Mental State Exam), information that could be reasonably obtained by a primary care physician, non-physician practitioner, or caregiver. Medium-burden information may include an in-depth neuropsychological testing battery or structural magnetic resonance imaging scan, both requiring more resources and time to administer. High-burden data would include information not routinely collected in clinical settings due to cost, time, and expertise required, such as the Clinical Dementia Rating. Using methods from statistical machine learning, we investigated disease prediction using low-burden data in three related projects. First, we developed a framework to evaluate whether inexpensive, accessible data can accurately classify AD clinical diagnosis and predict likelihood of progression. Second, we compared different data integration or “fusion” methods to determine optimal strategies for fusing low-burden data with additional neuroimaging data for predicting AD clinical diagnosis. Third, we developed models to infer dementia neuropathology using low-burden data. We found that using low-burden data yielded relatively high clinical diagnosis prediction performance and enabled us to identify subgroups with high- and low-risk of progression within 1.5-3 years. When integrating multimodal data, we observed that using a supervised encoder with intermediate fusion significantly improved performance. For neuropathology prediction, we identified three groups of subjects with significantly different proportions of positive neuropathology lesions, and we discovered that, with at least three clinical visits, using low-burden features yielded comparable prediction performance to using higher cost features. In conclusion, low-burden clinical data can effectively identify at-risk individuals for both AD clinical diagnosis and postmortem dementia neuropathology.