Bias in the estimation of adult body mass index from general practice records
Bias in the estimation of adult body mass index from general practice records
复制标题
根据一般实践记录估计成人体重指数的偏差
DOI：
10.1093/eurpub/cku163.080
复制
发表时间：
2014
影响因子：
4.4
通讯作者：
Badrick E
中科院分区：
文献类型：
--
作者：
Badrick E
背景：在英国，来自全科医生（GP）记录的数据通常用于监测慢性疾病患者。我们评估了这些数据对成人体重指数（BMI）趋势产生假设的适用性，因为常规收集的数据被广泛用于推断人口健康。方法回顾性队列研究，使用2001年至2011年英国索尔福德成人全科医生系统的数据（n= 248,917例患者），可获得慢性病和所有BMI测量的Readcoded信息。结果：2001年至2011年间对男性和女性的BMI测量显示，平均BMI相对平稳。女性平均BMI (kg/m2) 2001年为28.3(0.06)，2006年上升至28.6(0.03)，2011年下降至28.0(0.03)，年测量频率从13895次增加到67,554次。我们调查了患有慢性疾病（冠心病、癌症或糖尿病）的患者测量BMI的几率，2001年男性的BMI为9.12(8.59-9.68)，2011年降至4.32（4.16-4.48）。在2型糖尿病（T2D）患者中，n= 14 002在诊断前几年观察到较高的BMI，随后下降并变得相对稳定（女性诊断前一年的测量次数为2271，诊断后一年为8325）。我们还观察到在引入质量结果框架（QOF）前后记录BMI频率的变化。结论：来自GP记录的发现可能表明索尔福德人口的平均BMI正在下降，这与其他来源相矛盾。随着时间的推移，这种趋势似乎更多地与测量的原因有关，而不是与实际的人口身高体重有关。特别是，在qof之前的时代，BMI可能只会因为疾病或患者的体重而被记录。为了避免偏差，我们主张捕获有关测量背景的详细元数据，并相应地制作纵向统计模型，即使这样，数据的价值也可能仅限于考虑个体内BMI随时间的变化。
BackgroundData from general practice (GP) records in the UK is routinely used for monitoring patients with chronic conditions. We assessed the suitability of these data for generating hypotheses about trends in adult body mass index (BMI), due to the expanding use of routinely collected data for making inferences about population health.MethodsRetrospective cohort study, using data from GP systems on adults in Salford, UK between 2001 and 2011 (n= 248,917 patients) Readcoded information was available on chronic conditions and all BMI measures.ResultsBMI measures for men and women between 2001-2011 show a relatively flat mean BMI. In women the mean BMI (kg/m2) in 2001 was 28.3 (0.06), rising to 28.6 (0.03) in 2006 then decreasing to 28.0 (0.03) in 2011, the annual frequency of measures increased from 13,895 to 67,554. We investigated the odds of having BMI measured if patients had a chronic condition (CHD, cancer or diabetes), for men in 2001 this was 9.12 (8.59-9.68) this had decreased to 4.32 (4.16-4.48) in 2011. In patients with Type 2 Diabetes (T2D), n= 14 002 a higher BMI is observed in the years preceding diagnosis which subsequently decreases and becomes relatively stable (number of measures in the year before diagnosis in women was 2271, in the year after 8325). We also observed changes in the frequency of recording BMI around the time the Quality Outcomes Framework (QOF) was introduced.ConclusionsThe findings from GP records could indicate that the mean BMI of the Salford population is decreasing, which contradicts other sources. The trends over time seem more related to why the measurement was made than to actual population weight for height. In particular, in the pre-QOF era, BMI may have only been recorded due to illness or if the patient’s weight was of concern. In order to avoid biases we advocate capturing detailed metadata about the context of measurement and crafting longitudinal statistical models accordingly, even then the value of the data may be restricted to considering within-individual changes in BMI over time.