Clinical Significance Of CT Imaging Screening For Adrenal Disease,Cost-Performance Analysis And Construction Of A Predictive Model For High-risk Groups

ObjectiveSome adrenal diseases are insidious and have no significant symptoms,and because of the technical limitations of ultrasound examination of the adrenal glands,CT examination is currently the most ideal adjunct to detect adrenal diseases.The incidence of adrenal disease is not low,but it is often underdiagnosed or underestimated in clinical practice,especially in China,where there is no large sample to elucidate the incidence of adrenal disease in detail,and some cases with subclinical status are mostly ignored.The purpose of this study is to investigate the detection rate of abnormal CT images of the adrenal gland,to explore and analyze the cost-effectiveness of CT image screening for adrenal disease,and to further construct a predictive model for adrenal disease through machine learning algorithms.The purpose of this study is to investigate the detection rate of adrenal CT imaging abnormalities and the cost-effectiveness of CT imaging screening for adrenal disease,and to further construct a prediction model for screening of high-risk groups for adrenal disease through machine learning algorithms to achieve personalized screening for adrenal disease.Methods1.A total of 82,646 patients who visited our hospital between March 2011 and February 2021 were retrospectively collected to analyze the detection rate of abnormal adrenal CT imaging and its imaging characteristics,and further analyze the relationship between diabetes,hypertension and obesity.subgroups and adrenal diseases.2.Statistics of the examination cost of patients who underwent adrenal CT examination in our hospital in the past 10 years,and analysis of adrenal CT screening3.To collect clinical and imaging data of 824 patients with 349 cases of abnormal adrenal gland and 475 cases of normal adrenal gland in our hospital,and to construct a prediction model for the high-risk group of adrenal gland disease by machine learning.Results1.The proportion of adrenal abnormalities in the population was about 4.1%,with an average age of 57.09±16.20 years,more women than men(4.32%:3.94%),with left-sided lesions predominating(68%on the left,17%on the right,and 15%bilaterally),and the proportion of adrenal abnormalities in the hypertensive and diabetic subgroups was 10.59%and 9.52%,respectively,higher than in the general population.2.Adrenal-related screening cost a total of 123,936,600 yuan,of which patients with adrenal abnormalities spent a total of 508,200yuan,accounting for 4.1%of the overall cost,and only 0.25%required intervention.Patients in the hypertension subgroup spent a total of $193,350 for adrenal CT screening,of which $204,750 was spent by patients with adrenal abnormalities,accounting for 10.59%of the overall cost,of which only 1.09%required intervention.Patients in the diabetes subgroup spent a total of $943,500,of which patients with adrenal gland abnormalities spent a total of $90,300,accounting for 9.57%of the overall cost,of which only 0.7%required intervention.3.The results of the machine learning prediction model suggested that patients with abnormal adrenal imaging were more likely to have adrenal abnormalities at grade 2 hypertension and above,and the significance of performing adrenal-related CT examinations was greater.Patients with only grade 1 hypertension or no significant hypertension and no significant abnormalities in renal function or blood chloride levels were less likely to have abnormal adrenal imaging findings(AUC=0.72).Conclusion1.The positive CT detection rate of adrenal abnormalities was 4.1%,with female patients and left-sided lesions predominating;the detection rate of adrenal abnormalities was higher in the subgroup of patients with hypertension and diabetes mellitus than the normal population.2.the cost-effectiveness of imaging screening for adrenal disease in the overall population is low,and the cost-effectiveness of adrenal screening in the hypertensive and diabetic subgroups is higher than that in the normal population.3.A decision tree model based on three risk indicators,namely blood pressure,glomerular filtration rate and blood chloride,can effectively help clinical selection of screening strategy(AUC=0.72).