Pre-diagnostic comorbidity pathways are associated with the timing of pulmonary hypertension outcomes: a case control study

Abstract Background Pulmonary hypertension (PH) is a life-threatening condition often diagnosed late due to non-specific symptoms and overlapping comorbidities with other cardiovascular and respiratory diseases. Although there have been advances in treatment, diagnostic delays are common, and patients are treated late in their disease course. Investigating the temporal ordering of recorded comorbidities may provide insight into diagnostic patterns associated with delayed PH recognition and adverse outcomes. Methods We identified 2,727 PH patients from the UK Biobank cohort of 440,014 participants with diseases diagnoses, and sampled two demographics-matched cohorts of chronic obstructive pulmonary disease (COPD) and heart failure (HF) with the same number of patients as the PH cohort but were not diagnosed with PH. Employing process mining, a methodology originating from business process analysis to extract temporal patterns from event logs, the temporal diagnostic sequences of comorbidity combinations preceding PH diagnosis were compared to those sequences in patients who received a final diagnosis COPD or HF. We also assessed the relative risk of mortality given different combinations of comorbidities and the genetic correlation between comorbidities. Results Patients diagnosed with PH had prior diagnoses of hypertension (58.1%), atrial fibrillation (31.2%), and heart failure (43.2%). Process mining revealed that the temporal order of cardiometabolic comorbidities influences the timing to diagnosis of PH, COPD and HF. Patients in which type 2 diabetes was diagnosed before systemic hypertension had received PH diagnosis 2 years and 9 months faster than the reverse sequence, where hypertension preceded diabetes. For patients with a final diagnosis of PH, HF and COPD, diagnostic sequences in which hypercholesterolemia was recorded before systemic hypertension and type 2 diabetes were associated with shorter median intervals to PH diagnosis. A sequence involving diabetes, hypertension, and hypercholesterolemia was associated with a significantly higher mortality risk (HR = 1.97; P = 0.001). Additional mortality risk factors included specific orderings of metabolic conditions and comorbidity clusters. Conclusion Temporal diagnostic sequences of comorbidities were associated with the timing of PH diagnosis. Our findings suggest that the presentation and ordering of metabolic diseases may serve as hypothesis-generating epidemiological indicators of PH risk. Given the limited PH subtype granularity available within UK Biobank, these findings should be interpreted within a heterogeneous PH population and require further validation in clinically characterized cohorts with detailed PH subtype information.