It is well established in the literature that adverse drug reactions (ADRs) and drug non-compliance contribute substantially to the admissions at medical wards. Some important questions, however, remain unanswered. The purpose of this thesis was to characterise the drug-related hospital admissions (DRH) and to assess the magnitude of the problem seen in relation to the demographic parameters and drug use of the background population. In addition, an attempt was made to reduce the DRH incidence by an intervention program. The scope of the study program was adverse drug reactions, intended self-poisoning, non-compliance, underdosing and interactions. The material included 1999 admissions to six departments of internal medicine at Odense University Hospital. The patients were reviewed prospectively, while they were still in the wards, but use of standardised criteria fOR assessment of drug-ADR causality. With inclusion of a definite, probable and possible causal relationship, ADRs and toxic reactions were found as an important factor in 8.4% of all admissions. The incidense of ADR related admissions was 400 per 100,000 per year for the background population as a whole, but showing a strong increase with age. The drug-specific ADR incidences were generally small compared to the drug sales figures. Non-compliance contributed to 2.0% of admissions with diuretics and anti-asthmatics as the drugs most frequently involved. Two departments were re-investigated after an intervention program, primarily targetting general practitioners. The over-all incidence of DRHs was unaffected by the intervention, but the subset classified as avoidable DRHs showed a significant decline. The case material was subject to a blinded evaluation by an external peer group using the same criteria as the investigators. There was no indication that the observed decline in avoidable DRHs should be explained by a shift in the investigators' assessment of cases. It was concluded that the intervention probably had a specific effect on avoidable DRHs. Admissions caused by adverse ractions to over-the-counter remedies, usually salicylates, were characterised by a particularly inappropriate use of drugs, suggesting that the public may also be a suitable target for interventions. In 1990 the author established the Odense PharmacoEpidemiologic Database (OPED), a research registry with person-identifiable data on computerised refund claims from Funen County. The registry was validated by a cohort study on the association of non-steroidal antiinflammatory drugs (NSAID) and admission for severe upper gastrointestinal bleeding (UGB). All 31,503 recorded NSAID users in Odense during a 19 month period and a control cohort were studied. For both cohorts we obtained data on admissions caused by UGB and other diagnosis- and prescription-data that would be relevant for confounder control. The standardised incidence ratio was 5.5, controlled for age, sex and previous peptic ulcer episodes. A multivariate analyse revealed a number of other risk factors, but no other important confounders. The standardised incidence of UGB was particularly high during the first month of NSAID therapy. The size of the data set permitted an estimate of excess risk within various patient categories. NSAID users over 75 years of age or with a previous peptic ulcer episode had particularly high excess risks. The utilization pattern of NSAIDs was surprisingly sporadic and appeared to favour a high UGB rate for a given sales volume. It is pointed out that individualised dispensing data may be an important tool in future endeavors to control DRH incidences. The data may be used for generating or confirming hypotheses on unknown and suspected ADRs, for studies on risk factors of ADRs, for characterising the population's drug utilization, for identifying objects for interventions and for monitoring the effect of interventions.