Objective: To assess the impact of specialization on processes and outcomes of care for cancer patients.
Data source: Papers published in English between 1980 and 1995 and identified through MEDLINE and Embase (MeSH terms: NEOPLASM (exploded), and PHYSICIAN PRACTICE PATTERNS (or DECISION MAKING, ATTITUDE OF HEALTH PERSONNEL, QUALITY OF HEALTH CARE, DELIVERY OF HEALTH CARE, HEALTH EDUCATION or OUTCOME ASSESSMENT HEALTH CARE), or through the reference lists of review articles.
Study selection: Studies providing information on the association between quality of care indicators for cancer patients and clinician/centre degree of specialization. A total of 47 papers concerning 46 empirical studies were considered.
Data extraction: For studies using process of care indicators, the proportion of specific procedures performed by specialists and non-specialists was abstracted. For studies using outcome indicators (e.g., mortality), the effect of specialization was quantified in terms of odds ratio (OR) expressing relative reduction in risk of death. The quality of individual studies using process or outcome indicators was assessed according to study design, avoidance of selection bias in patient identification and data analysis, degree of adjustment of the comparison between clinicians/centres with different levels of specialization.
Data synthesis: Specialized centres/clinicians fared better both when process and outcome indicators were used. While the former varied widely in different studies and their clinical relevance was often questionable, mortality was consistently lower when care was provided by specialized centres/clinicians, with the effect size being greater in smaller studies. For breast cancer, where all the studies were of sufficiently good quality, a pooled estimate of the effect of specialization was performed which showed that specialized cancer care was associated with an 18% (95% CI: 12%-23%) reduction in mortality.
Conclusions: Despite the fact that care provided by specialized centres/clinicians appeared to be better both when assessed in relation to process indicators and to mortality, this evidence should be considered far from conclusive because of major methodological flaws in these studies. Relative to current efforts to promote evidence-based policy-making, this review underscores the limited capability of scientific information to provide reliable guidelines for structuring better health care systems.