4.1 Principal Results
This study explored the association between baseline comorbidities and 3-month frailty transitions and analyze the impact of different multimorbidity patterns on long-term survival outcomes in short-term frailty transition groups. The study found that frailty transition patterns and 2-year mortality risk are associated with age, gender, education level, and lifestyle.
The comorbidity rate in this study was 56.77%. The prevalence of multimorbidity in the persistent frailty group was 71.92%, which is higher than the 49.30% observed in the persistent non-frailty group. This finding is consistent with previous research. The results of the BLSA II study showed that multimorbidity increases the risk of frailty occurrence by 6.24 times[16]. Multimorbidity promotes the onset and progression of frailty by increasing physical burden and reducing physical function[17]. Although multimorbidity are associated with frailty, the results showed that even in the highest-risk persistent frailty group, 28.08% of patients had no comorbidities, indicating the need to identify specific multimorbidity states with frailty risk. In the group with CCI ≥ 2, the 2-year mortality risk increased for each frailty transition pattern, suggesting that the occurrence of comorbidities increases the mortality risk for each frailty transition group. Controlling the occurrence of comorbidities improves long-term survival outcomes for each frailty group.
In pattern 1, hypertension, diabetes, metabolic disorders, and coronary heart disease cluster together, known as cardiometabolic multimorbidity (CMM). Cardiometabolic diseases are a group of symptoms that include cardiovascular, renal, metabolic, prothrombotic, and inflammatory abnormalities[18]. These symptoms are typically characterized by insulin resistance, impaired glucose tolerance, dyslipidemia, hypertension, and central obesity. This is one of the most recurrent multimorbidity features[18]. Previous studies have shown that this multimorbidity pattern can be influenced by similar lifestyle and environmental factors[18–20]. Improving the lifestyle and environment of these patients can effectively intervene in the CMM population. Additionally, prior research using data from the UK Biobank found that frailty is an independent risk factor for all transitions in CMM progression trajectories[21]. Based on the bidirectional causal relationship between comorbidity and frailty, to our knowledge, no studies have explored the association of CMM with frailty development. Our study observed that in pattern 1, the risk of death in the worsening group has a cumulative effect on prognosis under this pattern. We speculate that CMM can increase the mortality risk in the frailty worsening population. In pattern 2, cognitive impairment and depression cluster together, and we similarly observed a cumulative effect on prognosis in the worsening group for frailty and pattern expression. A study on the elderly population in Italian nursing homes found that for the disease pattern of dementia and sensory impairment, the risk gradually increased with the expression of the pattern in the high frailty group (HR = 1.70, 95%CI: 1.44–2.01; HR = 1.92, 95%CI: 1.64–2.25; HR = 2.12, 95%CI: 1.83–2.47)[15]. This is similar to our findings. The mortality risk in the frailty worsening group (HR = 3.26, 95%CI: 2.43–4.37) is slightly lower than that in pattern 1 (HR = 3.40, 95%CI: 2.54–4.57), possibly because the factor loading of cancer in this pattern is < 0, indicating that patients susceptible to pattern 2 have a lower risk of cancer. This may offset part of the mortality risk of pattern 2. In this pattern, we still observed the presence of CMM (stroke and transient ischemic attack clustered together), but due to the lower association of TIA with pattern 2, and researchers focusing more on the combination of diabetes, stroke, and heart disease in CMM, we do not use CMM to explain the frailty mortality risk in pattern 2. In pattern 3 of kidney-hematologic diseases and pattern 4 of respiratory-skeletal diseases, we observed cumulative effects on prognosis for frailty and pattern expression in the persistent frailty group. For pattern 5, patients susceptible to this multimorbidity pattern had a lower risk of death. This is because the factor loadings of the three diseases in this pattern are all < 0, indicating that patients susceptible to this pattern have a lower risk of eye disease, arthritis, and atrial fibrillation. In other words, pattern 5 represents a healthy pattern.
In the Whitehall II study[22], it was found that after a 24-year follow-up, the mortality risk for patients with comorbidity and frailty increased compared to those without adverse health conditions. The association between comorbidity and mortality (HR = 4.12) was stronger than the association between frailty and mortality (HR = 2.38). Our findings confirm the additive effect of multimorbidity on the risk of death among frail individuals. Compared with the group with CCI = 0 and persistent non-frailty, in the group with CCI ≥ 2, the mortality risk for persistent frailty was as high as 14.27 times, and the mortality risk for worsening frailty was as high as 10.02 times. Even after adjusting for all confounding factors, this positive correlation remained significant. However, we did not observe such a high mortality risk in the multimorbidity patterns. The reason might be that, considering the excessive number of frailty transition groups, we divided each multimorbidity pattern only into high and low factor loading groups to avoid insufficient test power due to small group sizes. This also introduced a new issue: the differences between the high and low factor loading groups were not as significant, thereby reducing the mortality risk of the high factor loading group.
4.2 strengths and limitations
Our study has several advantages: First, previous studies have mostly focused on the impact of diseases on the entire frail population. In contrast, our research segmented the population, identifying which multimorbidity patterns pose greater harm to the worsening frailty group and which patterns pose greater harm to the persistently frail group. Second, to our knowledge, this is the first study to use PCA to explore the impact of multimorbidity patterns on different frailty transition survival outcomes in hospitalized elderly patients in China. While the dimensionality reduction technique used in this study does not have direct clinical applicability for individual patients as it requires a sample for analysis, it provides valuable insights into broader trends. Similarly, assigning individuals to specific multimorbidity patterns is based on probabilistic methods, making it challenging to identify unique correspondences between individuals and disease patterns. This approach reflects real-world situations where numerous conditions lead to a variety of disease combinations, some explained by pathophysiological principles and others occurring randomly. Third, our study not only examined the impact of multimorbidity clustering patterns on the long-term outcomes of frail patients but also investigated the effect of the CCI on the mortality risk of these patients. We used the CCI instead of the commonly used disease count method to weight the severity of chronic diseases, thereby enhancing the accuracy and applicability of our findings. Finally, our data come from a prospective cohort study, providing a higher level of causal evidence. This study still has limitations. Since our data were sourced from tertiary hospitals, the severity of patients' conditions may be higher, potentially leading to an overestimation of the multimorbidity rate. However, our findings are similar to previous studies that reported a multimorbidity rate of 69.3% among elderly hospitalized patients in China.