Forty-one studies (n=1,030,853) were included. Of these, 18 were based on clinical data and 24 on administrative data.
All studies showed either an inverse relationship between provider volume and mortality or no volume-outcome effect. Sixteen of the 18 clinical reports showed a statistically significant positive relationship between provider volume and outcome; no study showed a significant negative relationship.
Pancreatic resection (11 studies): the quality scores ranged from 5 to 10 (median 7). Definitions of low-volume hospitals ranged from 1 to less than 22 cases per year, and high-volume from more than 13 to more than 81 cases per year. All studies reported a statistically significant inverse relationship between hospital volume and mortality. The NNT to prevent one death was 10 to 15.
Oesophagectomy (10 studies): the quality scores ranged from 6 to 10 (median 7). Definitions of low-volume hospitals ranged from less than 2 to 32 cases per year, and high-volume from more than 6 to 83 cases per year. Nine studies reported a statistically significant inverse relationship between provider volume and mortality. The NNT to prevent one death was 7 to 9.
Surgery for gastric cancer (5 studies): the quality scores ranged from 4 to 10 (median 7). Definitions of low-volume hospitals ranged from less than 5 to 26 cases per year, and high-volume from more than 10 to 67 cases per year. Three studies reported a statistically significant inverse relationship between provider volume and mortality. One study reported a non-statistically significant reduction in RAMR for high-volume hospitals. One study reported no significant relationship. The NNT to prevent one death was 20 to 100.
Surgery for lung cancer (10 studies): the quality scores ranged from 6 to 10 (median 7). Definitions of low-volume hospitals ranged from less than 6 to less than 38 cases per year, and high-volume from 20 to 66 cases per year. Four studies found a statistically significant inverse relationship between hospital volume and mortality, with a reduction in RAMR ranging from 1.65 to 5.4%. The NNT to prevent one death was 20 to 50.
Surgery for breast cancer (4 studies): the quality scores ranged from 7 to 11 (median 9). Definitions of low-volume hospitals ranged from less than 10 to less than 25 cases per year, and high-volume from 30 to more than 149 cases per year. Three studies reported a statistically significant inverse relationship between provider volume and mortality, with the smallest study not reporting a significant relationship. The NNTs were not reported. Surgery for colorectal cancer (16 studies): the quality scores ranged from 5 to 11 (median 9). Definitions ranged from 1 to fewer than 12 cases for low-volume surgeons and from one to fewer than 84 cases per year for low-volume hospitals. Ten of 15 studies assessing hospital volume reported a statistically significant inverse relationship between volume and outcome, as did three of 7 studies assessing surgeon volume. The NNT to prevent one death was 50 to 100.
Surgery for miscellaneous cancers: the results were conflicting for cystectomy (3 studies) and nephrectomy (2 studies). High-volume providers were associated with lower mortality for radical prostatectomy (1 study), liver resection (3 studies) and intracranial tumour (1 study).