Failure to attend hospital appointments needlessly delays clinical care and consumes resources better spent on improving quality. The fact that attendance rates have remained relatively unchanged over the past 10 years suggests the problem is anything but simple. Two interacting factors arguably account for its difficulty.
Combining machine learning with large-scale data allows UCLH to create rich, complex, high-dimensional models able to operate within wider causal fields. Such models may not only predict attendance, enabling targeted intervention, but also prescribe it by matching detailed appointment and patient characteristics. By capturing individual variability better, they may also be used to infer systemic, modifiable hospital causes of non-attendance currently obscured by the many other factors in play.
Focusing on an important exemplar of hospital outpatient scheduling – magnetic resonance imaging (MRI) – we sought to answer two related questions: what is the relationship between the complexity of predictive models of attendance and their predictive performance, and can sufficient predictive performance be achieved to render targeting cost-effective? Models were trained and
evaluated on an unselected set of 22,318 consecutive scheduled MRI appointments at two of the Trust’s hospitals. Optimal predictive performance required 81 variables. Simulations showed net potential benefit across a wide range of attendance characteristics, peaking at £3.15 per appointment at current prevalence and call efficiency.