Introduction Health care providers in many countries are increasingly aware that quality of care can be improved by measures to reduce errors. Evidence suggests that the greatest opportunities for improvement may be in the management of the acutely ill hospitalized patient, where uncertainty, urgency and lack of integration substantially increase the risk of errors leadinf to adverse outcomes. Different approaches to this problem have evolved, including hospitalists in the USA, medical emergency teams in Australia, and outreach care in the UK. Critical care has a central role to play in all three developments, which share the common aim of improving the safe care of acutely ill patients as they travel through the health care system. We will review the background, methods, roles and benefits of these various systems which we group under the heading of 'outreach' care. Method We focused on identifying research publications that examined ways of improving the integration of critical care with acute care services as a means of improving the safe care of acutely ill hospitalized patients. Publications were initially identified by an electronic search of Medline and Cinahl, and the cited references provided additional material. The initial date range searched was 1995–2003 to ensure that current research and up to date literature was reviewed. However, this produced only a limited number of references, and the time period was therefore extended to 1990. Keywords searched singularly and in combination were 'acute pain team', 'suboptimal care', 'patient at risk', 'critical illness' and 'cardiopulmonary resuscitation'. Only English language references were included in the analysis. Background Trends in hospital care 1 2 3 4 5 Error and quality in health care 6 7 9 Identifying at-risk populations Acutely ill patients in general 7 8 7 10 11 12 13 14 15 Postoperative patients 16 17 This type of audit based on a large observational database is essential for identifying current practice and opportunities for improvement. However, recommendations were based on peer review and data from questionnaires provided by assessors who were unblinded to clinical outcomes, examining only those patients who died – there are no denominator data. Cardiopulmonary arrest 18 21 18 20 The interface with critical care 22 23 24 25 26 27 Treatment limitation decisions, futility, and end of life care 28 29 29 30 A more appropriate approach to management of acutely ill patients at risk for critical illness or cardiac arrest would be to prevent these complications, or agree treatment goals and limitations, by earlier recognition of simple warning signs. We consider the various approaches that are being adopted below. Systems for earlier recognition and management of patients at risk 31 33 34 35 36 37 38 Structures and processes for early intervention 39 40 Hospitalists 41 42 43 This model is suited to the system of care in the USA, where primary care physicians often have continuing responsibility for inpatient care. It might also be of value to other health care systems in which increasing specialization is creating a need for generalist acute care clinicians in hospital practice. Acute pain teams 44 45 44 46 47 16 48 49 46 50 Medical emergency teams 51 52 53 54 36 48 52 54 25 a priori  55 56 Critical care outreach 'Outreach' care is a systems approach for identifying and managing patients at risk of critical illness through collaborative care and education. Rather than providing a service through an external group, it aims to empower ward staff by offering them regular support, usually led by critical care trained nurses visiting the wards, with the facility to call on more expert assistance if required. Currently, critical care is seen as occurring within a defined environment (the ICU) and patients must achieve a certain level of severity of illness to merit admission. Outreach services facilitate a more flexible approach that is based on the needs of the patients and the skills and abilities of the ward staff. 57 55 56 58 Identifying the patient at risk – scoring systems for decision support 36 54 59 60 61 36 54 56 62 66 1 67 17 54 Does earlier intervention improve outcome? It may seem unnecessarily argumentative to suggest that something that is so evidently 'good' requires evaluation. Is it not self-evident that the earlier a life-threatening disease process is identified and treated, the better? This may be so, but complex systems – of which the acutely ill patient is one example – demonstrate sensitivity to initial conditions; that is, the outcome may be more difficult to predict with earlier application of an intervention. The outcome from cardiopulmonary resuscitation is well defined – a survival rate of around 10–15%, with death usually occurring when resuscitation attempts are discontinued. Is it not possible that earlier intervention that prevents cardiac arrest might result in delayed death following prolonged organ system support in the ICU, with the attendant suffering that this may cause? What is the cost–benefit of implementing the different models of care? Who will manage the process of discussing treatment goals or limitations with patients, and what do patients themselves feel about it? Physiological goals 68 72 73 74 75 77 Clinical outcomes 62 63 Autonomy and treatment limitation decisions 78 79 80 Changing behaviour through education and training 81 82 83 84 85 86 Conclusion There are substantial opportunities for improving the safety and quality of care delivered to acutely ill hospitalized patients. The methods which are adopted will vary according to local circumstance, but common elements include the need for better integration of care across disciplines and systems for earlier identification of patients at risk, and we refer to these as 'outreach' care. We do not yet know which of these various approaches will best improve patient outcomes, and there is a need for prospective studies in this area which take into account the difficulties of using randomization and controls, and which employ long-term follow-up. METs appear to reduce the incidence of cardiac arrests in ordinary wards, and consequential use of scarce intensive care resources. Outreach-based systems which support and educate ward-based staff in delivering clinical care appear to have achieved a high degree of acceptance in the UK. Health care managers need to work closely with clinicians to introduce these methods of team-working into hospital practice, while evaluating their effectiveness. Competing interests None declared. Abbreviations ICU = intensive care unit; MET = medical emergency team.