Worth a second look Advances in hospital outcomes research Linda Aiken, Douglas Sloane Center for Health Services and Policy Research, University of Pennsylvania, USA Hospital outcomes research has a long history, dating back to Florence Nightingale's statistical analysis docu- menting that more British soldiers in the Crimea died as a result of hospital-related conditions than of wounds received in battle.' Her findings led to organizational and staffing innovations, including the introduction of trained nurses, which produced declines in hospital mortality. Over time, hospitals have become labour- intensive, complex organizations and the most expen- sive setting for the delivery of health care. Industrialized countries, struggling to bring rising health care costs under control, have increasingly sought to ration hospital care by admitting only the sickest patients and permitting them to stay for the shortest possible time. The average severity of illness of hospitalized patients has increased, as has per capita expenditure. Conse- quently, managerial ideologies and strategies developed in manufacturing industries to streamline production have increasingly been adopted by hospitals to contain rising operating costs. Because nursing and related activities constitute the main work in hospitals, the reorganization of nursing is a primary focus of wide- spread hospital re-engineering schemes.F There has been little systematic study of the conse- quences of hospital restructuring on patient outcomes, although it has been consistently shown that two of the principal structural characteristics that have been altered through hospital re-engineering - nurse to patient ratios and nursing skill mix - are important determinants of hospital performance. Additionally, despite an extensive literature on correlates of hospital mortality, relatively little is known about how manipul- able organizational attributes of hospitals affect mortality and other outcomes. Especially lacking is research on the effects of the organization of nursing care in hospitals. The number of nurses may be less strongly related to patient outcomes than how the work of nurses is organized: how much autonomy they are accorded to exercise their professional judgment, how Linda H. Aiken PhD, Claire M. Fagin, Leadership Professor of Nursing, Professor of Sociology, Director, Center for Health Services and Policy Research, University of Pennsylvania, 420 Guardian Drive, Philadel- phia, PA 19104-U096, USA. Douglas M. Sloane PhD, Associate Professor of Sociology, The Catholic University of America, Washington DC and Adjunct Associate Professor of Nursing, University of Pennsylvania, Philadelphia, PA, USA. Correspondence to: LA. much control they have over their practice, their relationships with physicians, and so on. Our thinking on this has been shaped, in part, by the work of Ann Barry Flood, who provides an informative critique and synthesis of research on the impact of organizational and managerial factors on the quality of health care. 3 She concludes that too often studies focus on identifying structural characteristics associated empirically with outcomes rather than on developing models that explain the processes by which structure affects quality of care and outcomes. Without under- standing how organization affects outcomes, she argues, we risk reproducing the structural characteristic but not its intended effect. Perhaps at no time since Night- ingale's studies has hospital outcomes research been more important. The purpose of this brief essay is to call attention to a few articles in the research literature - one somewhat dated now but the others quite recent - which, taken together, have the potential to contribute to our understanding of how the structure and organization of hospitals affect patient outcomes. The prohibitive cost of primary data collection across large numbers of institutions has resulted in increased interest among researchers in the use of routine administrative data to conduct hospital outcomes research. The use of these databases to rank the performance of hospitals has resulted in acrimonious debate on both sides of the Atlantic." Researchers, however, continue to exploit the potential and minimize the limitations of administrative data in the pursuit of knowledge about the relationships between allocation of resources, organizational structure, processes of care and patient outcomes. The primary problems associated with the use of such data are that they fail to include information on important organizational attributes (such as nurse autonomy, control and relations with physicians) and, in most countries, they provide few useful outcome measures except for inpatient mortality, which can be affected by selective admission and discharge practices by hospitals. One of the most useful strategies for deriving measures of organizational characteristics of hospitals was identified, nearly three decades ago, in a brief essay by Aiken (not the current author) and Hage. 5 In their research on organizational interdependence and organ- izational behaviour (such properties as complexity and innovativeness) among health and welfare organiza- tions, they obtained measures of the properties of 16 organizations by aggregating individual level data J Health Serv Res Policy Volume 3 Number4 October1998 249