GIANTT background and population
GIANTT contains anonymized healthcare data of a dynamic, cumulative cohort of about 60,000 patients with type 2 diabetes, treated in primary care in the province of Groningen (The Netherlands). In particular, this entails patients with a type 2 diabetes diagnosis that is confirmed by their general practitioner (GP) and all their routine care data. These are primarily data about diagnostic tests, (co)morbidity and prescribed medication (beyond diabetes medication alone). From 2007 onwards, the covered population slightly changed due to new diabetes bundled care agreements between health insurers and general practitioners in the region (1). Potential changes in data characteristics over time could also be attributed to revised clinical guidelines or reimbursement rules. From 2013 onwards, there are some changes in the population and the data due to new ways of data extraction methods.
Currently, full clean data are available for the period 2007-2013. This entails an expanding cohort of GP practices, all members of the Groninger Huisartsen Coöperatie (GHC), including all their type 2 diabetes patients . The GIANTT patient population from 2007 has been compared with the ZODIAC diabetes cohort from the Zwolle region (2). This analysis revealed that both populations are comparable regarding their general and clinical characteristics. In 2007, a total of 95 GP practices took part in GIANTT (3). This number increased to about 180 GP practices in 2013 (80% of all practices in Groningen). After 2013, only data from diabetes patients that are included in bundled care plans by their GP are available. Reasons not to include patients in bundled care plans include a short life expectancy, high complexity or refusal of care in general.
Extraction method and validation
Until 2007 most data were extracted from GP’s notes and from structured tables in the GP Information System using validated methods (4). From 2007 onwards, GP practices started using more structured care data registration, yet GIANTT extraction methods remains the same. From 2013 onwards, data are only extracted from structured tables using extraction methods as provided by Proigia/Calculus, a service provider that takes care of GPs care data analysis and benchmarking.
GIANTT contains 4 main tables with the following data:
• Patient-table: date of entry in GIANTT, (estimated) date of diagnosis, leave from GIANTT, reason of leave, month/year of birth, sex, responsible physician (GP or specialist)
• Measurement-table: value and date of diagnostic measurements from GP systems including a WCIA code and GIANTT codes (combinations of WCIA codes). Measurements can be requested as first or last before a certain (index) date or as a list of all measurements in a defined period. It is possible to distinguish by source (notes and/or structured tables). Regarding dates, the latter is more reliable. Yet, some measures (e.g. blood pressure, weigth) are sometimes only registered in GP’s notes.
• Comorbidity-table: dates of diagnosis/complaint or selected diabetes-complication treatments (e.g. amputation, angioplasty, bypass) are registered in the diagnosis/episode table of the GP system using an ICPC code and GIANTT codes (free text fields are coded by medical students). Comorbidity can be requested as first or last data before a certain (index) date or as the presence during a certain period.
• Medication-table: date, amount, use (free text), estimated duration of prescription with ATC code. Medication can be requested by individual ATC code or ATC class as first or last prescription after a certain (index) date or presence within a defined period (i.e. start and/or estimated end date and a given duration of run-in or follow-up).
Measurements that are part of routine quartile/yearly checks of diabetes patients are increasingly available from 2007 onwards (in particular HbA1c, SBP/DBP, TC, HDL, LDL, creatinine/eGRF, albuminuria/ACR, weigth/length/BMI, smoking status). Presence of other measurements depends on patient’s individual comorbidity (indicating certain measurements) and protocols/habits of the GP practice and prevailing guidelines (that can vary over time).
Since 2007, comorbidity data are increasingly available yet there is profound variation across GP practices. Generally, ‘chronic’ diagnoses are rarely repeated and the date of the first record can be slightly later than the true date of the diagnosis, for example due to manual entry after a relocation of the patient. Comorbidity data provide a good overview of the medical history of the patient without the need for an exact start date of the diagnosis. From 2007 onwards, it may be possible to assess occurrence of comorbidity prospectively yet some delay in diagnosis should be taken into account. Pouwels et al (2016) linked GIANTT data to Dutch hospital data to assess how reliable GIANTT can identify cardiovascular events (5). For hospitalizations due to ischemic heart disease/cerebrovascular diseases/CABG/PTCA based on GIANTT data, this revealed a sensitivity of 43% and a specificity of 97%. Adding drug codes as proxy could increase the sensitivity to 94%.
All medication prescribed in the GP practice are included in GIANTT, inclusive of repeat prescriptions of medication initiated by a medical specialist. To estimate the duration of the prescription (in days), the free text field was transformed into numeric values. Of note, when using this parameter, caution is required given some prescriptions have missing data regarding exact usage (for example, in cases of “as needed/pro re nata” medication).
(1) de Vries ST, Voorham J, Haaijer-Ruskamp FM, Denig P. Potential overtreatment and undertreatment of diabetes in different patient age groups in primary care after the introduction of performance measures. Diabetes Care. 2014; 37(5):1312-20.
(2) Alberts S, Denig P. Validity and representativeness of GIANTT. [Unpublished]
(3) Voorham J, Haaijer-Ruskamp FM, van der Meer K, de Zeeuw D, Wolffenbuttel BH, Hoogenberg K, Denig P; GIANTT-Group. Identifying targets to improve treatment in type 2 diabetes; the Groningen Initiative to aNalyse Type 2 diabetes Treatment (GIANTT) observational study. Pharmacoepidemiol Drug Saf. 2010; 19(10):1078-86.
(4) Voorham J, Denig P. Computerized extraction of information on the quality of diabetes care from free text in electronic patient records of general practitioners. J Am Med Inform Assoc. 2007; 14(3):349-54
(5) Pouwels KB, Voorham J, Hak E, Denig P. Identification of major cardiovascular events in patients with diabetes using primary care data. BMC Health Serv Res. 2016; 16:110.