A literature review on the analysis of symptom-based clinical pathways: Time for a different approach?

Breathlessness is a common clinical presentation, accounting for a quarter of all emergency hospital attendances. As a complex undifferentiated symptom, it may be caused by dysfunction in multiple body systems. Electronic health records are rich with activity data to inform clinical pathways from undifferentiated breathlessness to specific disease diagnoses. These data may be amenable to process mining, a computational technique that uses event logs to identify common patterns of activity. We reviewed use of process mining and related techniques to understand clinical pathways for patients with breathlessness. We searched the literature from two perspectives: studies of clinical pathways for breathlessness as a symptom, and those focussed on pathways for respiratory and cardiovascular diseases that are commonly associated with breathlessness. The primary search included PubMed, IEEE Xplore and ACM Digital Library. We included studies if breathlessness or a relevant disease was present in combination with a process mining concept. We excluded non-English publications, and those focussed on biomarkers, investigations, prognosis, or disease progression rather than symptoms. Eligible articles were screened before full-text review. Of 1,400 identified studies, 1,332 studies were excluded through screening and removal of duplicates. Following full-text review of 68 studies, 13 were included in qualitative synthesis, of which two (15%) were symptom and 11 (85%) disease focused. While studies reported highly varied methodologies, only one included true process mining, using multiple techniques to explore Emergency Department clinical pathways. Most included studies trained and internally validated within single-centre datasets, limiting evidence for wider generalisability. Our review has highlighted a lack of clinical pathway analyses for breathlessness as a symptom, compared to disease-focussed approaches. Process mining has potential application in this area, but has been under-utilised in part due to data interoperability challenges. There is an unmet research need for larger, prospective multicentre studies of patient pathways following presentation with undifferentiated breathlessness.Author summary: Breathlessness is a common symptom for patients attending hospital. This can be caused by many conditions and getting a diagnosis may be complex. The journey from attending hospital to a diagnosis is called a ‘clinical pathway’. Process mining is a way of understanding the order and timing of events within a clinical pathway. We have reviewed the evidence for using process mining to better understand the clinical pathways for patients with breathlessness. We found 13 relevant studies and of those, only two focussed on the symptom and the rest on related diseases such as Covid-19. Most studies were also of small-scale, and the results were generally not replicated in more than one hospital. Only one study used true process mining to look at symptoms in patients attending a single Emergency Department. We believe that considering symptoms and using process mining could help clinicians understand how different patients interact with healthcare. This may improve the effectiveness of clinical pathways in the future, such as by reducing delays in diagnosis. Process mining may also help explain differences in response to treatments between individuals and help guide clinicians on the best time to organise tests such as specialist scans.