[Association between levels of thyroid-stimulating hormone and prognosis of patients admitted to intensive care units: an analysis using the MIMIC-III].

OBJECTIVE

To explore the association between levels of thyroid-stimulating hormone (TSH) on admission and prognosis of patients admitted to intensive care unit (ICU).

METHODS

The data were collected from patients who were admitted to the ICU of the Beth Israel Deaconess Medical Center in the United States from 2001 to 2012 with available TSH test records within 24 hours after the ICU admission via the Medical Information Mart for Intensive Care-III v1.4 (MIMIC-III v1.4). Information including gender, age, ethnicity, type of admission, mechanical ventilation (MV) or renal replacement therapy (RRT) received on admission, comorbidities, and TSH test records within 24 hours after the ICU admission were collected. The sequential organ failure assessment (SOFA) score, simplified acute physiology score II (SAPS II) and the comorbidities index Elixhauser (SID30) score were calculated according to the parameters. The primary outcome was hospital mortality. Differences in baseline characteristics and prognosis were examined between patients with normal TSH levels and abnormal TSH levels which was determined according to a dichotomous variable provided by the data. Multivariable Logistic regression was used to analyze the association between TSH levels and prognosis after adjusting for confounding factors. A sensitivity analysis was conducted which categorized the study population as three groups (i.e., decreased, normal, and elevated TSH levels) using the range of 0.30-3.00 mU/L as the normal range of TSH.

RESULTS

A total of 3 425 ICU patients were enrolled in the study, of which 2 692 (78.60%) were with normal TSH and 733 (21.40%) were with abnormal TSH. There was no statistically significant difference in gender, age, ethnicity, type of admission and the ratio of MV between the normal TSH and abnormal TSH groups. Compared with normal TSH group, the patients in abnormal TSH had a higher SOFA, SAPS II and SID30 scores as well as the ratio of RRT [SOFA score: 4 (2, 7) vs. 4 (2, 6), SAPS II score: 38.02±13.76 vs. 36.53±13.75, SID30 score: 11 (4, 22) vs. 11 (0, 20), RRT ratio: 5.32% (39/733) vs. 3.49% (94/2 692), all P < 0.05]. The hospital mortality of patients in normal TSH was significantly higher than that of those in abnormal TSH [9.82% (72/733) vs. 5.94% (160/2 692), P < 0.01]. After adjusting for confounding factors, abnormal TSH was significantly associated with hospital mortality [odds ratio (OR) = 1.71, 95% confidence interval (95%CI) was 1.24-2.35, P = 0.001]. In the sensitivity analysis in which the range of 0.30-3.00 mU/L was used as the normal range of TSH, compared with normal TSH, decreased TSH (OR = 2.36, 95%CI was 1.40-3.97, P = 0.001) and elevated TSH (OR = 1.44, 95%CI was 1.05-1.98, P = 0.023) were both significantly associated with increased hospital mortality.

CONCLUSIONS

An abnormal level of TSH within 24 hours after admitted to ICU is an independent risk factor for hospital mortality among ICU patients.