Diabetes

Patient consultations recorded as voice dictations are frequently stored electronically as transcriptions in free text format. The information stored in free text is not computer tractable. Advances in artificial intelligence permit the conversion of free text into structured information that allows statistical analysis.

Objective

This paper describes DMReporter, a medical language processing system that automatically extracts information pertaining to diabetes (demography, numerical measurement values, medication list, and diagnoses) from the free text in physicians’ notes and stores it in a structured format in a MYSQL database.

Public health departments have a strong interest in monitoring the incidence, care, and complications of gestational diabetes, as it is associated with poor outcomes for infants and increased risk of diabetes type II for mothers. Gestational diabetes rates are also a possible early marker for changes in the incidence of diabetes type II in the general population. However, diabetes is not generally a reportable condition and therefore, public health surveillance is limited to periodic telephone surveys (subject to self-report inaccuracies), sponsored clinical examinations (expensive, small sample size, no information about processes of care), and occasional research studies. Automated analysis of electronic health record data is a promising method to complement existing surveillance tools with longitudinal, continually updated, clinically rich data derived from large populations. We describe a pilot project to automatically survey electronic health record data in order to identify cases of gestational diabetes, describe their patterns of care and complications, and report summary data to the state health department.

Objective

To develop an electronic, prospective surveillance system to describe the incidence, care, and complications of gestational diabetes using live electronic health record data from a large defined population.

The research reported in this paper is part of a larger effort to achieve better signal-to-noise ratio, hence accuracy, in pharmacovigilance applications. The relatively low frequency of occurrence of adverse drug reactions leads to weak causal relations between the reaction and any measured signal. We hypothesize that by grouping related signals, we can enhance detection rate and suppress false alarm rate.

Objective

ICD-9 codes are commonly used to identify disease cohorts and are often found to be less than adequate. Data available in structured databasesFlab test results, medications etc.Fcan supplement the diagnosis codes. In this study, we describe an automated method that uses these related data items, and no additional manual annotations to more accurately identify patient cohorts.

International Diabetes Federation (IDF) estimates that 21.4 million women in 2013 had some form of hyperglycaemia in pregnancy and in India alone an estimated 4 million women have GDM. Recognizing the shortfall of trained manpower; Certificate Course in Gestational Diabetes Mellitus (CCGDM) was launched in 2012; since then it has spread across 17 states and 39 cities across 55 regional training centers and trained 2400 Primary care physicians (PCP) all across India.

Objective With implementation of program on all India level aim is to develop a robust monitoring and evaluation system to ensure quality assurance and standardized course delivery on all India level.

Traditional surveillance methods, such as registries that require manual validation of every diabetes case or questionnaires, are resource intensive and associated with considerable delay in reporting results. An EHR-based surveillance system may be more efficient for sustained monitoring of the incidence and prevalence of childhood diabetes, so as to inform health care needs for this growing population.

Objective

The study goal was to develop an efficient surveillance approach for childhood diabetes across two large Southeastern US public academic health care systems, using electronic health record (EHR) data.