<b><i>Introduction:</i></b> There is little research on factors predicting conversion to dementia in early-onset mild cognitive impairment (eoMCI), a transitional stage between healthy ageing and dementia in individuals below the age of 65. We aimed to examine whether sociodemographic and clinical factors at initial presentation predicted dementia progression in a cohort of eoMCI patients attending a memory service, at a university teaching hospital in the UK. <b><i>Methods:</i></b> This is a retrospective case note study of individuals diagnosed with eoMCI between 2000 and 2013 at the Younger Person’s Memory Service (YPMS) in Leicestershire, England. Data collected at assessment included social factors, demographic characteristics, and medical and psychiatric history, as well as standardized cognitive assessment scores. Variables were analysed using χ² or independent sample <i>t</i> tests to identify associations. A Cox regression survival analysis was done to identify predictive factors for dementia conversion. An ROC analysis for total CAMCOG was used to investigate sensitivity and specificity for dementia converters versus non-converters. <b><i>Results:</i></b> Out of 531 subjects who attended YPMS, 65 patients were given a diagnosis of eoMCI (47.7% female; mean age 56.4 ± 7.54 years). Of these, 21 (32.3%) converted to dementia during their course within the service. Comparison between subgroups revealed a significant association between dementia conversion and higher years of education and lower MMSE and CAMCOG (total and subscale) scores at baseline. Smoking history, alcohol use, or medical history such as diabetes or heart disease were not associated with conversion. Cox regression survival analysis showed higher education in years and lower total CAMCOG scores were significant predictors for conversion. Lower scores on the recent memory, remote memory, learning memory, and executive function subscales of the CAMCOG were also significant predictors for conversion. ROC curve analysis for total CAMCOG demonstrated that the best detection of dementia converters can be achieved with a cutoff score of 90.5/107 (sensitivity of 76.2% and specificity of 68.2%). Area under the curve was 0.808 (95% CI: 0.697–0.920). <b><i>Conclusion:</i></b> More years in education and lower cognitive scores on CAMCOG at initial assessment are associated with progression to dementia from eoMCI. Further research is required to explore these predictive factors more.