Volume–outcome relationship in transcatheter aortic valve implantations in Germany 2008–2014: a secondary data analysis of electronic health records

OBJECTIVES: We examine the volume–outcome relationship in isolated transcatheter aortic valve implantations (TAVI). Our interest was whether the volume–outcome relationship for TAVI exists on the centre level, whether it occurs equally for different outcomes and how it develops over time. DESIGN: Secondary data analysis of electronic health records. The comprehensive German Federal Bureau of Statistics Diagnosis Related Groups database was queried for data on all isolated TAVI procedures performed in Germany between 2008 and 2014. Logistic and linear regression analyses were carried out. Risk adjustment was applied using a predefined set of patient characteristics to account for differences in the risk factor composition of the patient populations between centres and over time. Centres performing TAVI were stratified into groups performing <50, 50–99 and ≥100 procedures per year. SETTING: Germany 2008–2014. PARTICIPANTS: All patients undergoing isolated TAVI in the observation period. INTERVENTIONS: None. PRIMARY AND SECONDARY OUTCOME MEASURES: In-hospital mortality, bleeding, stroke, probability of ventilation >48 hours, length of hospital stay and reimbursement. RESULTS: Between 2008 and 2014, a total of 43 996 TAVI procedures were performed in 113 different centres in Germany with a total of 2532 cases of in-hospital mortality. Risk-adjusted in-hospital mortality decreases over the years and is lower the higher the annual procedure volume at the centre is. The magnitude of the latter effect declines over the observation period. Our results indicate a ceiling effect in the volume–outcome relationship: the volume–outcome relationship is eminent in circumstances of relatively unfavourable outcomes. Alongside improving outcomes, however, the volume–outcome relationship decreases. Also, a volume–outcome relationship seems to be absent in circumstances of constantly low event rates. CONCLUSIONS: The hypothesised volume–outcome relationship for TAVI exists but diminishes and may disappear over time. This should be taken into account when considering mandatory minimum thresholds.