Abstract Background and Purpose Over half of the elderly population experiences multiple comorbidities. By analyzing the chronic disease trajectory patterns within a large population, it is possible to predict future disease risks for individuals. Methods We collected data from 37.38 million patients in northern China between 2016 and 2024. Using the ICD-10 level 3 codes and de-duplicating by patient, we obtained a total of 207.08 million diagnoses from 2,272 ICD-10 codes. Figure 1 shows how diagnosis numbers vary with age. We excluded acute diseases, external cause diseases, pregnancy-related diagnoses, and non-disease diagnoses, leaving 777 chronic disease nodes (DN). The weight of the edge from DN1 to DN2 was defined by the number of individuals diagnosed with DN2 after DN1 within a 5-year interval, resulting in 265,831 bidirectional edges between nodes. The relative risk (RR) of progression from DN1 to DN2 was calculated, and Chi-squared test p-values were adjusted using Benjamini-Hochberg method to reduce false discovery rate. Edges with RR ≤ 1, non-significant RR, or weight 1000 were removed. The final temporal comorbidity network consisted of 351 chronic diseases and 8,672 edges. For each initial DN, we used 1 million Monte Carlo random walk to simulate individual disease trajectories. Assuming a patient currently has multiple chronic diseases, we randomly selected a starting DN based on the sum of the outgoing edge weights of each DN, and weighted-randomly selected an outgoing edge from the selected DN as a step, adding the arrival DN to the patient’s chronic diseases. This process was repeated until the number of chronic diseases reached a threshold, set at 7 based on our data and previous studies. Results In our data, 9 of the top 10 common diseases belonged to cardiovascular-kidney-metabolic (CKM) diseases (Figure 2). The most common downstream diseases of CKM diseases were also CKM diseases. For hypertension (15/20), chronic ischaemic heart disease (16/20), cerebral infarction (16/20), type 2 diabetes mellitus (15/20), heart failure (13/20), other cerebrovascular diseases (14/20), angina pectoris (15/20), transient cerebral ischaemic attacks (14/20), and dyslipidemia (14/20), more than half of the 20 common downstream diseases were CKM diseases. Common non-CKM diseases downstream of CKM diseases were other diseases of liver (9/9), hyperplasia of prostate (6/9), other intervertebral disc disorders (6/9), emphysema (6/9), other chronic obstructive pulmonary disease (5/9), and other nontoxic goitre (5/9). Conclusions Using population-based temporal disease trajectories can help predict the risk of future chronic diseases. CKM diseases often serve as downstream diseases for each other. These predictive results require confirmation through lifelong follow-up studies.Diagnosis Numbers by Age Chronic Diseases and Downstream Diseases
Fei et al. (Sat,) studied this question.