With an increasing incidence of orthopaedic procedures performed worldwide, the quantity of data collected, including “Big Data”, is also rising. Widening indications for surgery, a growing number of implant options and variety of operative techniques, as well as an increasing need to demonstrate cost effectiveness, necessitate the use of robust analysis techniques to assess outcomes.
Traditionally, analysis of outcomes in orthopaedic surgery involves survival methods, where the outcome of interest is ‘time to event’, which is usually revision or re-operation. For arthroplasty, this represents the time from the date of insertion of the implant until the date on which the revision is performed and patients whose outcomes are not known or have died are censored. Revision is generally taken as the primary indicator of failure of a joint replacement. Although revision/re-operation is dependent on many factors, including the fitness for surgery of the patient, it provides a firm endpoint for analysis, particularly in epidemiological studies.
One of the strengths of survival analysis is the handling of incomplete data or follow-up. If an event is not seen within the timeframe observed or reported, there would be incomplete observations, known as censored events. ‘Right’ censoring is the most common and occurs either if a subject does not experience the event during the study period, is lost to follow-up or withdraws from the study. Death is another reason for censoring.
The ‘risk set’ at a specific time point is defined as the individuals/implants that at that time are at risk of experiencing the event (e.g. revision). These are the individuals that have survived up to …