Objectives: Despite international debate over growing pressures on hospital beds, particularly from rises in emergency admissions, there are few published data on the characteristics of patients admitted to hospital. Most research has attempted to explore the appropriateness of general practitioner (GP) admissions, yet even the proportion of emergency admissions made via the GP is not known. This study aimed to investigate GP and patient factors associated with acute medical admissions to one hospital in Birmingham, UK. Methods: A study was performed on all acute medical admissions made over one week to a district general hospital in Birmingham and local general practices referring acutely to it. Admissions were identified from routine hospital dataset and the information validated. Factors perceived as linked to admission were explored by semi-structured interview or postal questionnaire with the admitting doctors and with a proportion of patients. The subjects were 158 patients and the 84 GPs involved in a proportion of the admissions. Outcome measures were admission routes, diagnoses, linked medical and social factors. Results: Only 84 (53%) patients were admitted via their GP (or deputy). 47% were self-referrals or admitted via outpatients. Data on 16% patients could not be validated because of GP refusal (3 doctors, 15 patients) or untraceable doctor (10 patients). Most admissions were for respiratory problems (47, 30%), mainly cases of infective illness (13% all admissions, 43% respiratory). Cardiovascular disease involved 16% cases, principally myocardial infarction or angina. Drug overdose was the most frequent single diagnosis, almost 1 in 10 acute admissions. 69% of the GP admissions were over 65 years old (30% over 75) against 36% of the non-GP (p<0.01). Additional influences on GP admissions included: exacerbations of long-standing medical problems (29%); clarification of uncertain diagnoses (61%); and requests for advice (5%). Some admissions also resulted from social pressures including: pressure from carers (15%); visit occurred at night (4%); and even pressure from patients to avoid admission (8%). In 11% of cases, GP admission was stated as solely related to non-acute medical problems. The average length of stay was 10.6 days (range 1 to 120 days). Mean length of stay for patients admitted by a doctor was 13 days compared to 8 days for non-GP admissions (p=0.025). 13% of patients died during or shortly after admission. Conclusions: Most patients sent in by the GPs or selfreferring, in this small sample of urban emergency admissions, were admitted with serious medical problems, on the basis of the coded discharge diagnoses, and required extended hospital stays. GP admissions in this sample appeared a sicker, or less coping, cohort. There was evidence that GPs were attempting to maintain patients at home and felt that around 15% of their admissions might have been managed in some alternative low technology environment.