In this study, we focused on predicting the short-term survival probability and identifying influencing factors in hypothyroidism patients with heart failure using patient medical records from the MIMIC database. Our analysis included several factors such as age, RR, Cr, Glu, SOFA score, BUN, and CCI score. We successfully developed and validated a nomogram that exhibited excellent performance. This nomogram is the first visual model designed specifically for predicting the prognosis of hypothyroidism patients with heart failure. Its implementation can aid in identifying factors associated with higher survival risks and enable more targeted clinical management.
Among the factors we analyzed, age emerged as significant independent risk factors. In view of the high incidence rates of hypothyroidism and heart failure in the elderly14,15, there is a reasonable association with higher mortality among patients. Similar correlations have also been observed in other studies16. These associations can be attributed to various age-related changes in the cardiovascular system, including endothelial dysfunction, increased left ventricular stiffness, impaired baroreflex and autonomic nervous reflexes, and degenerative changes in the conduction system. Furthermore, chronic low-grade inflammation and excessive oxidative stress in aging individuals contribute to a reduced regenerative capacity of the heart17.
Respiratory rate is also an important factor in our study.Hypothyroidism also affects the respiratory system in various ways. For instance, it can alter the sensitivity of the respiratory center, alveolar ventilation, and respiratory-related muscles. When thyroid hormone levels decline, the sensitivity of the respiratory center may decrease, leading to a slower respiratory rate18,19. However, in patients with heart failure, the respiratory rate often increases due to the need for enhanced ventilation and oxygenation. Notably, a higher respiratory rate has been identified as a significant risk factor for mortality in patients with hypothyroidism and heart failure. This association may be attributed to the correlation between an elevated respiratory rate, severe hypoxia, and exacerbation of heart failure. This also suggests that clinical doctors need to pay attention to the patient's respiratory rate
Among the serological indicators, creatinine (Cr) as an evaluation marker for kidney function. It is a byproduct of normal muscle metabolism and is transported to the kidneys through the bloodstream, where it is filtered and eliminated. Elevated levels of creatinine can indicate impaired kidney function20. The relationship between thyroid hormones and renal function has been well-established. Chang et al.21 showed in their studies that hypothyroidism linked to decreased renal function and proteinuria, primarily affecting renal function regardless of whether it is subclinical or overt hypothyroidism. Levothyroxine replacement therapy has been shown to improve renal function in patients with hypothyroidism22. Another cross-sectional study found that serum creatinine levels were higher, and the glomerular filtration rate (eGFR) was significantly lower in the subclinical hypothyroidism group compared to the euthyroid group. The group with overt hypothyroidism displayed significantly higher levels of serum creatinine and lower eGFR compared to both groups. Additionally, thyroid-stimulating hormone (TSH) levels showed a negative correlation with eGFR and a positive correlation with serum creatinine23. In addition, Bjorklund et al.24 demonstrated renal dysfunction has an impact on mortality in patients with heart failure, which aligns with our study findings.
We found blood glucose (Glu) is an independent risk factors, consistent with other research findings.Among patients with heart failure, blood glucose concentrations are powerfully prognostic for short-term survival, independent of a diagnosis of diabetes mellitus or other clinical variables25. Research has demonstrated that a 2 mmol/L increase in admission glucose is associated with a 6% increase in annual readmissions. Thyroid hormone and TSH have also been found to be associated with blood sugar in various studies26,27. Because blood glucose is easily modifiable, it may represent a valid target for therapeutic intervention.
The SOFA score, or Sequential Organ Failure Assessment, is an index utilized to assess organ dysfunction in critically ill patients. Grooth et al.28 have shown a significant correlation between the SOFA score and patient mortality. Generally, higher SOFA scores are associated with an increased risk of death, which aligns with the findings of this study29. Similarly, the Charlson score is a widely used tool to evaluate disease severity and predict the risk of mortality30. This scoring system quantifies the number and severity of diseases a patient has to predict their risk of death. Single factor analysis revealed that it is a risk factor, but in multi factor analysis, it was found that its HR is 0.85, which needs further verification. We consider it to be related to the small number of testing set samples.
Blood urea nitrogen (BUN) is involved in changes in renal perfusion and is considered to reflect the progression of heart failure more accurately than traditional indicators like N-terminal pro-brain natriuretic peptide (NT-pro-BNP)31. Interestingly, in our COX regression model, the HR for BUN was 0.98, indicating a protective effect, contrary to findings from previous studies about heart failure32. And the predictive value of BUN for heart failure remains controversial in the literature. Sachdeva et al.33 found that BUN did not have any predictive value for heart failure prognosis, while Takaya et al. reported that BUN was predictive only when combined with renal failure. Therefore, further exploration is necessary to determine whether BUN affects the prognosis of heart failure. And we also need further confirmation that the reason our research results are inconsistent with those of other studies is that our study focuses on hypothyroidism combined with heart failure, rather than solely on heart failure patients.
Our model can provide effective predictive information and has also uncovered some interesting phenomena. However, it is important to note some limitations. Firstly, this study is retrospective and conducted at a single center. To enhance the study's robustness, future research should involve a larger dataset collected from multiple centers. Secondly, the data we incorporated did not include all information about patient comorbidities; hence, we cannot rule out the influence of underlying disease states on clinical outcomes. Moreover, some laboratory indicators might affect each other, but we could not detect interactions between covariates. Lastly, our study is based on a retrospective cohort, unknown confounding factors may affect our results, and nomograms obviously cannot provide completely accurate prognosis prediction, thus nomograms need further prospective validation before considering clinical application.