Background Antipsychotics are widely used to manage behavioral disorders in patients with dementia. Recently, serious concerns have been raised about the stroke and mortality risk of atypical antipsychotics when administered to patients with dementia. Aim The aim of this study was to examine the impact of atypical and conventional antipsychotics on mortality and hospital admissions among Finnish elderly institutionalized patients with dementia in a two-year follow up and to compare their prognosis with that of nonusers. Patients and Methods The authors examined 254 very frail patients with dementia, mean age 86 years, from seven Finnish nursing homes and two hospitals in1999–2000. Medical records provided information on the use of daily antipsychotic medication; central registers confirmed mortality for up to two years. Results Nearly one-half (48.4%) of the patients used antipsychotic medication: 37.4% received conventional neuroleptics (N = 95) and 11.0% received atypical antipsychotics (N = 28). The mean number of hospital admissions was higher among the nonusers than among the users of conventional or atypical antipsychotics. Of the users of atypical antipsychotics (risperidone, olanzapine), 32.1% died within 2 years. The respective figures for users of conventional neuroleptics were 45.3%, and for the nonusers, 49.6%. In the Cox proportional hazard model, a high number of medications and the use of physical restraint predicted higher mortality at two years. The use of atypical antipsychotics showed lower risk of mortality, if any. The respective test for conventional antipsychotics was nonsignificant. Conclusion Among these frail and very old patients with dementia, neither the use of atypical antipsychotics nor the use of conventional neuroleptics increased mortality or hospital admissions. The use of restraints, however, doubled the risk of mortality.