[Volume and health outcomes: an overview of systematic reviews]

Improving quality and effectiveness of health care is one of the priorities of health policies. Hospital or physician volume of activity may be a measurable variable with a relevant impact on effectiveness of health care. There are several studies and systematic reviews evaluating the association between volume and outcome of health care. The aim of this review is to identify: areas, clinical conditions or interventions (prevention, diagnostic, therapeutic, surgical or clinical) for which an association between volume and outcome has been investigated; those for which an association between volume and outcome has been proved

Overview of systematic reviews and Health Technology Assessment reports; search of MEDLINE, EMBASE, The Cochrane Library, Web sites of Health Technology Assessment, other HTA Agencies, National guideline Clearinghouse, National Health Care quality tools (1995-february 2005). For each studied area results are described separately for each review due to the heterogeneity of outcomes, volume thresholds and results reported. No metanalysis has been conducted. Completeness of reporting of the systematic reviews has been evaluated using the QUOROM statement. For each review we evaluated the number of studies included and the proportion of studies with statistically significant results (p < 0,05). As far as in-hospital mortality is concerned, the different areas have been classified in the following groups: Strong evidence ofpositive association: areas with > or =10 studies included in the reviews, and high prevalence (> or =50%) of positive studies (p <0. 05) in the majority of reviews. Weak evidence of association: areas with 5 to 9 studies included in the reviews and high prevalence (> or =50%) of positive studies (p <0.05) in the majority of reviews. Weak evidence of lack of association: areas with 5 to 9 studies included in the reviews and high prevalence (>50%) of not statistically significant studies (p >0.05) in the majority of reviews. No suficient evidence of association: areas with less than 5 studies included in the reviews. No evidence of association: areas with > or =10 studies included in the reviews, and high prevalence (>50%) of not statistically significant studies (p >0.05) in the majority of reviews. The same literature search was then applied to identify primary studies published in each considered area following the most recent systematic review published.

We identified 21 systematic reviews and included 11 of them analysing 46 different areas. The majority of studies evaluate the effect of specific surgical procedures; the main outcomes considered are hospital mortality and 5 year survival for cancers. Considering in-hospital mortalilty as outcome, in 11 areas there is strong evidence ofassociation between volume of activity and outcome: abdominalaortic aneurysm (unruptured), percutaneous transluminal coronary angioplasty knee arthroplasty coronary artery bypass, surgery for oesophageal and pancreatic cancer, surgery for prostate cancer, colecistectomy, carotid endarterectomy, myocardial infarction, neonatal intensive care. It is never possible however to identify a unique volume threshold. For some of these areas, particularly coronary angioplasty and coronary artery bypass, there are many new studies published following the last systematic review; some specific aspects are being investigated such as the role of temporal changes in the association, the effect of different risk adjustment procedures and the separate role of physician or hospital volume. In some cases, for example knee arthroplasty in-hospital mortality could be an inadequate outcome on which judging the strength of association, in fact, the few studies evaluating other outcomes such as complications provide inconsistent results. For a range of areas the evidence of association is weak: AIDS, appendicectomy, cardiac catheterization, surgery for breast, lung, stomach cancer, hernia repair, hip fracture, hysterectomy and injuries. As far as AIDS is concerned, the few number of studies found is probably due to the lack of studies published after the introduction of effective therapies. All the included studies show an evidence of association between volume and in-hospital mortality. In no case we found weak evidence of lack of association while we identified three conditions for which the number of studies included in the reviews together with the prevalence of non significant studies do suggest lack of association; these are abdominal aortic aneurysm (ruptured), hip arthroplasty and surgery for colorectal cancer. In the case of hip arthroplasty as well, inhospital mortality could be an inadequate outcome, but only one old study found a positive association with risk of complications. Eventually there is a group of areas (n=22) for which there is not enough evidence to draw conclusions about the association between volume and outcome due to a small number of studies. In some cases, such as transplants, this could be due to the low rate of events; in this case all the few published studies show positive results. There are some limitations which should be taken into account in the interpretation of these results: despite the overall good completeness of reporting of the included reviews, the majority of studies included in the reviews themselves are cross-sectional studies representing a very weak study design to evaluate causality of the investigated association. Moreover the methodology of risk adjustment applied is heterogenous among studies and it is difficult to know the extent to which this can affect the observed results. It is eventually necessary to consider the possible occurrence of publication bias which could lead to an overestimation of the positive effect of volume on health care outcomes attributable to the lack of publication of negative studies.

In some areas the evidence seems strong enough to guide health care organizational choices, although it is not possible to identify well defined volume thresholds. In other areas, particularly for non surgical conditions, where there is not enough evidence, it seems necessary to conduct proper epidemiological studies. Also the evaluation of effectiveness of using volume as an instrument of health policy requires further research. Taking into account the rapid and continuing process of technology development, the definition of standard and prerequisite volumes of care should be specific of each temporal period and health care system. It is therefore a dynamic process requiring a continuous review of the available evidence. In the area of evidence based public health, the limited available evidence should not impair the choice of actions based on limited evidence, but rather it should lead to the application of thefew available evidence on one side and to the planning of proper research in the areas of lack of evidence.