Undoubtedly, the way we define MLTC matters. Comparing studies, or estimating prevalence over time becomes almost impossible if one measure includes 10 conditions while another includes 200. Yet there is no one-size-fits-all approach.
Early on in my thesis, I spent a lot of time worrying about which conditions to include. Discussions with my patient and public involvement group highlighted the complexity: many other conditions could be included, and greater granularity could be valuable, for example, specifying the subtypes of dementia, rather than grouping them together. As we also found in a paper published in BMJ Medicine, how we chose to define whether a condition is ‘chronic’ also has a major impact.
Involving a diverse set of voices – including patients, clinicians and researchers – in decisions on which conditions to include is vital. However, this will inevitably introduce a bias towards including more familiar diseases and overlooking rarer ones. Rare diseases, by definition, affect fewer than 1 in 2,000 people, yet cumulatively, their impact is substantial: 3.5 million people in the UK have a rare condition.
Excluding any condition is hard to justify. Even if a condition affects only one person, its impact may be large, warranting inclusion. This led me to consider a fully inclusive, ‘data-driven’ approach of including anything classed as a ‘disorder’ in medical taxonomies. But I soon realised this would be almost impossible to interpret and so limit any actionable insights it could generate.
As I argued in a recent piece in the BMJ, achieving universal consensus on the definition of MLTC may not be attainable, and may limit research if we stick too rigidly to it. The appropriate diseases to include will depend on what we are trying to achieve from the research. For example, exploratory studies investigating risk factors for diseases may benefit from a broader set of diseases, while studies examining how health services could work together may achieve more practical insights by focusing on a narrower set of common conditions.
So, yes, it matters which conditions we include in MLTC. But this isn’t just about justifying the choice to ourselves - as others have argued, it is about providing a clear rationale, documenting the decision-making process, and sharing code lists to ensure reproducibility for others.
Ultimately, MLTC is an inherently messy concept. Rather than seeking a definition to fit all cases, we should ensure our approach is aligned to the specific aims of our work.