Al-Lawati Najla A, Alfonso Helman, Al-Lawati Jawad
Department of Non-Communicable Diseases, Directorate General of Primary Health Care, Ministry of Health, Muscat, Oman.
Faculty of Health Sciences, School of Public Health, Curtin University, Perth, Australia.
Oman Med J. 2022 Jan 31;37(1):e340. doi: 10.5001/omj.2021.123. eCollection 2022 Jan.
We sought to develop and validate a diabetic risk score model as a non-invasive and self-administered screening tool to be used in the general Omani population.
The 2008 World Health Survey (WHS) data from Oman (n = 2720) was used to develop the risk score model. Multivariable logistic regression with the backward stepwise method was implemented to obtain risk factors regression coefficients for sex, age, educational attainment, marital status, place of residence, hypertension, body mass index (BMI), waist circumference, tobacco use, daily fruit and vegetable intake, and weekly physical activity. The model coefficients were multiplied by a factor of five to allocate each variable category a risk score. The total score was calculated as the sum of these individual scores. The score was validated using another Omani cohort (Sur Survey 2006 dataset, n = 1355) by calculating the area under the receiver-operating characteristic (ROC) curve (AUC), and optimal score sensitivity and specificity were determined.
A robust diabetes risk score model was produced composed of eight variables (age, sex, education level, marital status, place of residence, hypertension, smoking status, and BMI) with an optimal cutoff point of ≥ 15 to classify persons with possible prevalent type 2 diabetes mellitus (T2DM). At this cutoff point, the model had a sensitivity of 71.1%, specificity of 74.4%, and AUC of 0.80 (95% confidence interval (CI): 0.78-0.82), when internally validated (in the WHS 2008 cohort). When the model was externally validated (using the Sur 2006 cohort), the optimal cutoff point for the score was ≥ 13, with a lower sensitivity (54.0%), higher specificity (79.0%), and an AUC of 0.74 (95% CI: 0.70-0.78). In contrast, the test of the old Omani, Kuwaiti, Saudi, and Finnish diabetes risk scores in our study populations showed poor performance of these models among Omanis with poor sensitivity (29% to 63.5%) and reasonable specificity (70% to 80%).
The developed diabetes risk score for screening prevalent T2DM, provides an easy-to-use self-administered tool to identify most individuals at risk of this condition in Oman. The score incorporates eight diabetes-associated risk factors that can also act as a tool to increase people's awareness about the importance of diabetes-related risk factors and provide information for policymakers to establish diabetes prevention programs.
我们试图开发并验证一种糖尿病风险评分模型，作为一种非侵入性的自我管理筛查工具，用于阿曼普通人群。
使用来自阿曼的2008年世界卫生调查（WHS）数据（n = 2720）来开发风险评分模型。采用向后逐步法进行多变量逻辑回归，以获得性别、年龄、教育程度、婚姻状况、居住地点、高血压、体重指数（BMI）、腰围、烟草使用、每日水果和蔬菜摄入量以及每周体育活动等风险因素的回归系数。将模型系数乘以5，为每个变量类别分配一个风险评分。总评分计算为这些个体评分的总和。通过计算受试者工作特征（ROC）曲线下面积（AUC），使用另一个阿曼队列（2006年苏尔调查数据集，n = 1355）对该评分进行验证，并确定最佳评分的敏感性和特异性。
生成了一个强大的糖尿病风险评分模型，该模型由八个变量（年龄、性别、教育水平、婚姻状况、居住地点、高血压、吸烟状况和BMI）组成，最佳截断点为≥15，用于对可能患有2型糖尿病（T2DM）的人群进行分类。在这个截断点，该模型在内部验证（在2008年WHS队列中）时，敏感性为71.1%，特异性为74.4%，AUC为0.80（95%置信区间（CI）：0.78 - 0.82）。当该模型进行外部验证（使用2006年苏尔队列）时，评分的最佳截断点为≥13，敏感性较低（54.0%），特异性较高（79.0%），AUC为0.74（95%CI：0.70 - 0.78）。相比之下，在我们的研究人群中对旧的阿曼、科威特、沙特和芬兰糖尿病风险评分进行测试，结果显示这些模型在阿曼人中表现不佳，敏感性较差（29%至63.5%），特异性合理（70%至80%）。
所开发的用于筛查T2DM患病率的糖尿病风险评分，提供了一种易于使用的自我管理工具，可识别阿曼大多数有患此病风险的个体。该评分纳入了八个与糖尿病相关的风险因素，也可作为一种工具，提高人们对糖尿病相关风险因素重要性的认识，并为政策制定者建立糖尿病预防计划提供信息。
