[Volume and health outcomes: evidence from systematic reviews and from evaluation of Italian hospital data].

BACKGROUND: Improving quality and effectiveness of health care is one of the priorities of health policies. Hospital or physician volume represents a measurable variable with a relevant impact on effectiveness of health care. A recent Italian law, the "spending review", calls for the definition of "qualitative, structural, technological and quantitative standards of hospital care". There is a need for an accurate evaluation of the available scientific evidence in order to identify these standards, including the volume of care above or below which the public and private hospitals may be accredited (or not) to provide specific health care interventions. Since 2009, the National Outcomes Programme evaluates outcomes of care of the Italian hospitals; nowadays it represents an official tool to assess the National Health System (NHS). In addition to outcome indicators, the last edition of the Programme (2013) includes a set of volume indicators for the conditions with available evidence of an association between volume and outcome. The assessment of factors, such as volume, that may affect the outcomes of care is one of its objectives.

OBJECTIVES: To identify clinical conditions or interventions for which an association between volume and outcome has been investigated. To identify clinical conditions or interventions for which an association between volume and outcome has been proved. To analyse the distribution of Italian health providers by volume of activity. To measure the association between volume of care and outcomes of the health providers of the Italian NHS.

METHODS: Systematic review. An overview of systematic reviews and Health Technology Assessment (HTA) reports performed searching electronic databases (PubMed, EMBASE, Cochrane Library), websites of HTA Agencies, National Guideline Clearinghouse up to February 2012. Studies were evaluated for inclusion by two researchers independently; the quality assessment of included reviews was performed using the AMSTAR checklist. For each health condition and for each outcome considered, total number of studies, participants, high volume cut-off values (range, average and median) have been reported, where presented. Number of studies (and participants) with statistically significant positive association and metanalysis performed were also reported, if available. Analysis of the distribution of Italian hospitals by volume of activity and the association between volume of activity and outcomes. Outcomes National Programme 2011 The analyses were performed using the Hospital Information System and the National Tax Register pertaining the year 2011. For each condition, the number of hospitals by volume of activity was calculated. Hospitals with a volume of activity lower than 3-5 cases/year for the condition under study were excluded from the analysis. For conditions with more than 1,500 cases per year and frequency of outcome ≥ 3%, the association between volume of care and outcome was analysed. For these conditions, risk-adjusted outcomes were estimated according to the selection criteria and the statistical methodology of the National Outcome Programme.

RESULTS: The systematic reviews identified were 107, of which 47, evaluating 38 clinical areas, were included. Many outcomes were assessed according to the clinical condition/procedure considered. The main outcome common to all clinical condition/procedures was intrahospital/30-day mortality. Health topics were classified in the following groups according to this outcome: Positive association: a statistically significant positive association was demonstrated in the majority of studies/participants and/or a pooled measure (metanalysis) with positive results was reported. Lack of association: no association was demonstrated in the majority of studies/participants and/or no metanalysis with positive results was reported. No sufficient evidence of association: both results of single studies and metanalysis do not allow to draw firm conclusions on the association between volume and outcome. Evidence of a positive association between volumes and intrahospital/ 30-day mortality was demonstrated for 26 clinical areas: AIDS, abdominal aortic aneurysm (ruptured and unruptured), coronary angioplasty, myocardial infarction, knee arthroplasty, coronary artery bypass, cancer surgery (breast, lung, colon, colon rectum, kidney, liver, stomach, bladder, oesophagus, pancreas, prostate); cholecystectomy, brain aneurysm, carotid endarterectomy, hip fracture, lower extremity bypass surgery, subarachnoid haemorrhage, neonatal intensive care, paediatric heart surgery. For 2 clinical conditions (hip arthroplasty and rectal cancer surgery) no association has been reported. Due to a lack of evidence, it was not possible to draw firm conclusion for 10 clinical areas (appendectomy, colectomy, aortofemoral bypass, testicle cancer surgery, cardiac catheterization, trauma, hysterectomy, inguinal hernia, paediatric oncology). The relationship between volume of clinician and outcomes has been assessed only through the literature review; to date, it is not possible to analyse this association for Italian health providers hospitals, since information on the clinician/surgeon on the hospital discharge chart is missing. The literature found a positive association for: AIDS, coronary angioplasty, unruptured abdominal aortic aneurysm, hip arthroplasty, coronary artery bypass, cancer surgery (colon, stomach, bladder, breast, oesophagus), lower extremity bypass surgery. The analysis of the distribution of Italian hospitals per volume of activity concerned the 26 conditions for which the systematic review has shown a positive association between volume of activity and intrahospital/30-day mortality. For the following conditions it was possible to conduct the analysis of the association between volume and outcome of treatment using national data: unruptured abdominal aortic aneurysm, coronary angioplasty, knee arthroplasty, coronary artery bypass, cancer surgery (colon, pancreas, lung, prostate, stomach, bladder), laparoscopic cholecystectomy, endarterectomy, hip fracture and acute myocardial infarction. For them, the association between volume and outcome of care has been observed. The shape of the relationship is variable among different conditions, with heterogeneous "slope" of the curves. DISCUSSION For many conditions, the systematic review of the literature has shown a strong evidence of association between higher volumes and better outcomes. Due to the difficulty to test such an association in randomized controlled studies, the studies included in the reviews were mainly observational studies: however, the quality of the available evidence can be considered good both for the consistency of the results between the studies and for the strength of the association. Where national data had sufficient statistical power, this association has been observed by the empirical analysis conducted on the health providers of the NHS in 2011. Analysing national data, potential confounders, including age and the presence of comorbidities in the admission under study and in the admissions of the two previous years, have been considered. The systematic review of the literature does not permit to identify predefined volume thresholds. The analysis of national data shows a strong improvement in outcomes in the first part of the curve (from very low volumes to higher volumes) for the majority of the studied conditions. In some cases the improvement in outcomes remains gradual or constant with the increasing volume of care, in other the analysis could allow the identification of threshold values beyond which the outcome does not improve further. However, a good knowledge of the relationship between effectiveness of treatments and their costs, the geographical distribution and the accessibility to health care services are necessary to choose the minimum volumes of care, under which specific health procedures in the NHS should not be provided. Some potential biases due to the use of information systems data should also be taken into account. In particular, it is necessary to consider possible selection bias due to the different way of coding among hospitals that could lead to a different selection of cases for some conditions (e.g. acute myocardial infarction), less likely to occur in the selection of cases for oncologic, orthopaedic, vascular, abdominal, and cardiac surgery. Regarding the definition of the exposure (volume of care), a possible bias could result from misclassification of health providers with high volume of activity. In fact, performing the intervention in different departments/units of the same hospital would result in an overestimation of the volume of care measured for hospital rather than for department/unit. A similar bias could occur if the main determinant of the outcome of treatment was the case load of each surgeon: the results of the analysis may be biased when the same procedure was carried out by different operators in the same hospital/unit. In any case, the observed association between volumes of care and outcome is very strong, and it is unlikely to be attributable to biases of the study design. However, the foreseen bias is likely to be non-differential, and, therefore, it would eventually lead to an underestimation of the true association between volume of care and outcome. Health systems operate, by definition, in a context of limited resources, especially when societies and governments choose to reduce the amount of resources to allocate to the health system. In such conditions, the rationalisation of the organization of health services based on the volume of care may make resources available to improve the effectiveness of interventions. The identification and certification of services and provider with high volume of activity can help to reduce differences in the access to no effective procedures.