Aims: To define and identify ‘ambulatory care sensitive’ (ACS) and ‘ambulatory care insensitive' (ACI) conditions. To assess the quality of the NHS hospital episode statistics (HES) dataset. To assess the reasons for variation in hospitalisation for ACS conditions across small areas. To discuss whether variations in admissions for ACS conditions could be used to indicate equity of access to primary care. Main subjects: Hospital admissions for residents of North West Thames NHS region 1991/2, 1992/3, 1993/4. Methods: A modified nominal group technique was used to identify ACS and ACI conditions. The quality of HES was analysed by measuring shortfalls in key fields and the reproducability of clinical coding. Factors influencing variations in hospitalisation were assessed using multivariate analysis. The results were compared to similar studies from the US and Canada. Results: 30 ACS and 66 ‘weakly ACS' conditions were identified. Only S out of the 16 main acute providers serving the North West Thames region had adequate HES data across all three study years. In two providers, the exact clinical codes for main diagnosis could be reproduced in 43% and 60% respectively, and the first 3-digits of the ICD-9 codes in 33% and 72%. A significant positive relationship was found between admissions and socioecononomic deprivation independently of indicators for mortality and morbidity. This was weaker than that between admissions and area income found in the US and Canada. The relationship with deprivation was stronger for ACS than ACI conditions. A significant and positive relationship was found with an indicator of access to hospital care, but not access to primary care. The overall explanatory power of the model was weak at 10-20%. Conclusions: Admissions for ACS conditions are higher in deprived areas, suggesting that access to ambulatory care may be lower. Higher rates of ACS conditions in poor areas in North America suggest lower access to timely and effective ambulatory care than in the UK NHS.