Faraji Gavgani Leili, Sarbakhsh Parvin, Asghari Jafarabadi Mohamad, Shamshirgaran Seyed Morteza, Jahangiry Leila
Department of Statistics and Epidemiology, Faculty of Health, Tabriz University of Medical Sciences, Tabriz, Iran.
Road Traffic Injury Research Center, Tabriz University of Medical Sciences, Tabriz, Iran.
Int J Endocrinol Metab. 2018 Apr 25;16(2):e12757. doi: 10.5812/ijem.12757. eCollection 2018 Apr.
Functional limitation is one of the most important health - related concerns of diabetic patients. This study aimed to identify the factors associated with functional limitation among diabetic patients using generalized additive model (GAM) as a flexible technique to reveal the non - linear and non - monotonic association between the response and a set of independent variables.
The source data belonged to two cross - sectional studies conducted in 2014. A total of 694 people with type 2 diabetes in the age range of 31 - 70 years were selected via convenience sampling from diabetes clinics in Ardabil and Tabriz. The data were collected by interviewers using structured questionnaires and checklists. The functional capacity was measured using the physical functioning subscale of the Medical Outcomes Study Short Form 36 - Item Health Survey (SF36). Participants with a total functional capacity of less than 90 were considered to have "moderate or high level of functional limitation." To identify the factors associated with functional limitation and reveal the shape of associations, the GAM procedure with "logit" link function was applied to the dataset of 378 diabetic patients without any missing data by smoothening of the effect of underlying factors. The Akaike information criterion (AIC) as the relative quality of the model's criterion was computed for GAM and compared with AIC of the simple logistic regression.
Sex (P = 0.029), age (P < 0.001), BMI (P = 0.029), and SBP (P = 0.04) were significant in the GAM. Moreover, age with a linear function (df = 0.98), BMI with quadratic function (df = 1.75), and SBP with the degree 1.33 were significantly related to functional capacity. AIC of the GAM was lower than that of the logistic model.
In our sample, GAM could identify some linear and nonlinear associations between underlying factors and functional limitation in diabetic patients. These complex associations could relatively increase the fit quality of the GAM when compared to logistic regression.
功能受限是糖尿病患者最重要的健康相关问题之一。本研究旨在使用广义相加模型（GAM）作为一种灵活的技术，来识别糖尿病患者中与功能受限相关的因素，以揭示响应变量与一组自变量之间的非线性和非单调关联。
源数据来自2014年进行的两项横断面研究。通过便利抽样，从阿尔达比勒和大不里士的糖尿病诊所选取了694名年龄在31 - 70岁之间的2型糖尿病患者。数据由访员使用结构化问卷和检查表收集。使用医学结局研究简明健康调查问卷（SF - 36）的身体功能分量表来测量功能能力。总功能能力低于90的参与者被认为具有“中度或高度功能受限”。为了识别与功能受限相关的因素并揭示关联的形式，通过平滑潜在因素的影响，将具有“logit”链接函数的GAM程序应用于378名无任何缺失数据的糖尿病患者数据集。计算GAM的赤池信息准则（AIC）作为模型准则的相对质量，并与简单逻辑回归的AIC进行比较。
在GAM中，性别（P = 0.029）、年龄（P < 0.001）、体重指数（BMI）（P = 0.029）和收缩压（SBP）（P = 0.04）具有显著意义。此外，具有线性函数的年龄（自由度 = 0.98）、具有二次函数的BMI（自由度 = 1.75）和度数为1.33的SBP与功能能力显著相关。GAM的AIC低于逻辑模型。
在我们的样本中，GAM可以识别糖尿病患者潜在因素与功能受限之间的一些线性和非线性关联。与逻辑回归相比，这些复杂关联可以相对提高GAM的拟合质量。
