Severe Hypoglycemia: Ascertainment, Surveillance and Pharmacovigilance

修改的项目摘要/摘要 大多数2型糖尿病患者都会服用降糖药物，以降低长期并发症的风险。严重低血糖(SH)是药物的意外后果(即医源性效应)，定义为患者需要帮助的低血糖水平。自闭症与较差的生活质量、严重的跌倒、车祸、室性心律失常、痴呆症、住院以及数倍增加的死亡风险有关。尽管药物治疗取得了进展，但导致急诊科就诊或住院的重型肝炎已成为糖尿病治疗中最常见的并发症之一，也是一个严重的公共卫生问题。老年人因不良药物事件而紧急住院的四分之一是由SH引起的。然而，我们缺乏足够强大的确定工具来估计SH的总发病率。尽管治疗的复杂性和普遍的多药联用，我们仍然缺乏对患者开始联合降糖治疗或它们与非糖尿病药物相互作用时SH风险如何变化的可靠了解。我们建议从一个集成的医疗保健提供系统(Kaiser Permanente Northern California或KPNC)中对大规模(n~229,000)不同的2型糖尿病成人队列中的SH进行研究，具体目标有以下三个：在目标1中，我们将开发和验证从EMR数据(ED和医院)中识别SH的新算法。使用这些EMR编码算法，我们将对临床识别的SH的趋势进行全面的监测。我们还将估计未被临床识别的SH的比例(基于与先前调查中自我报告的SH的关联)，这将有助于计算SH的总发生率。在目标2中，我们将使用严格的因果建模技术(例如，差异分析、边际结构模型、有向无环图)，量化与糖尿病患者常用药物(即检查糖尿病药物、非糖尿病药物或它们之间的相互作用)有关的SH风险的变化。在目标3中，我们将评估与开始用药相关的SH风险的变化(与目标2相比)是否在特定的风险亚组(例如，既往的SH、CKD、长期糖尿病、老年人)以及基线和HbA1c的变化之间存在实质性差异。这项研究将1)提供新的SH确定工具，以改进全国范围内的监测，从而使人们能够更全面地了解SH流行病学；以及2)提供与开始糖尿病患者常用药物相关的SH风险变化的有效估计，从而帮助提供者和患者个体化管理糖尿病，同时将SH风险降至最低。

MODIFIED PROJECT SUMMARY/ABSTRACT Most patients with type 2 diabetes are prescribed medications to lower glucose levels and reduce the risk of long-term complications. Severe hypoglycemia (SH) occurs as an unintended consequence of medications (i.e., an iatrogenic effect) and is defined as a low blood glucose level for which the patient requires assistance. SH is associated with poorer quality of life, serious falls, car accidents, ventricular arrhythmia, dementia, hospitalizations, and a several-fold increased risk of death. Despite advances in pharmacotherapy, SH leading to emergency department (ED) visits or hospitalization has emerged as one of the most prevalent complications of diabetes treatment and is a critical public health concern. One in four emergency hospitalizations for adverse drug events among older adults is due to SH. Yet, we lack sufficiently robust ascertainment tools to estimate the total incidence of SH. Despite increasing treatment complexity and prevalent polypharmacy, we also lack a reliable understanding of how SH risk changes when patients initiate combinations of glucose-lowering therapies or their interactions with non-diabetic drugs. We propose to study SH in a large (n~229,000), diverse cohort of adults with type 2 diabetes from an integrated healthcare delivery system (Kaiser Permanente Northern California or KPNC) with these 3 specific aims: In Aim 1, we will develop and validate novel algorithms for identifying SH from EMR data (ED and hospital). Using these EMR coding algorithms, we will conduct comprehensive surveillance of trends in clinically recognized SH. We will also estimate the proportion of SH not clinically recognized (based on linkage with self-reported SH from a previous survey), which will facilitate calculation of the total rates of SH. In Aim 2, we will quantify the change in risk of SH associated with initiation of medications commonly used by people with diabetes (i.e., examining the effect of diabetes medications, non-diabetes medications, or their interactions), using rigorous, causal modeling techniques (e.g., difference-in-difference analysis, marginal structural models, directed acyclic graphs). In Aim 3, we will estimate whether the change in SH risk associated with initiation of medications (from Aim 2) differs substantively across specific risk subgroups (e.g., prior SH, CKD, long duration diabetes, the elderly), and by baseline and changes in HbA1c. This study will 1) provide new SH ascertainment tools to improve surveillance nationwide and thereby enable a more comprehensive understanding of SH epidemiology; and 2) provide valid estimates of change in SH risk associated with initiation of medications commonly used by people with diabetes, thus helping providers and patients individualize diabetes management while minimizing the risk of SH.