Purpose - Research on hospital system organization is dated and crosssectional. We analyze trends in system structure during 2000-2010 to ascertain whether they have become more centralized or decentralized. Design/Methodology/Approach - We test hypotheses drawn from organization theory and estimate empirical models to study the structural transitions that systems make between different "clusters" defined by the American Hospital Association. Findings - There is a clear trend toward system fragmentation during most of this period, with a small recent shift to centralization in some systems. Systems decentralize as they increase their members and geographic dispersion. This is particularly true for systems that span multiple states; it is less true for smaller regional systems and local systems that adopt a hub-and-spoke configuration around a teaching hospital. Research Limitations - Our time series ends in 2010 just as health care reform was implemented. We also rely on a single measure of system centralization. Research Implications - Systems that appear to be able to centrally coordinate their services are those that operate in local or regional markets. Larger systems that span several states are likely to decentralize or fragment. Practical Implications - System fragmentation may thwart policy aims pursued in health care reform. The potential of Accountable Care Organizations rests on their ability to coordinate multiple providers via centralized governance. Hospitals systems are likely to be central players in many ACOs, but may lack the necessary coherence to effectively play this governance role. Originality/Value - Not all hospital systems act in a systemic manner. Those systems that are centralized (and presumably capable of acting in concerted fashion) are in the minority and have declined in prevalence over most of the past decade.