The study protocol was approved by the Ethical Committee of Clinical Research of Euskadi (PI2014074).
This is a descriptive study including all T2DM patients with public health insurance in the Basque Country in the period between 1 September 2010 and 31 August 2011. Osakidetza (Basque Public Health Service) has an electronic health record and other computerised sources in which information concerning all contact between its inhabitants and the public health service is recorded. In this system, diagnoses are coded according to the International Classification of Diseases (ICD-9-CM) [9], whereas the coding system used for medicinal products is the Anatomical, Therapeutic, Chemical classification system [10].
For this study, information concerning demographic and clinical variables was obtained from the PREST stratification database. This database combines information from various sources (primary and specialised care, A&E and hospital admittance registries) to obtain diagnoses, prescriptions and procedures. With this information, all citizens registered with Osakidetza are classified every year using the Adjusted Clinical Group (ACG) case-mix system [11]. A more detailed description of the database can be found in previous publications [12].
The study population was considered to be all T2DM patients residing in the Basque Country. To identify this population, all patients who, according to the PREST database, had received a diagnosis corresponding to T2DM or unspecified DM (including their complications) at any point in their life or who had been prescribed antidiabetic medication, irrespective of whether they had made use of care services during the observation period, were included. Those patients who had a diagnosis relating to type 1 DM at any contact or for whom all diagnoses corresponded to unspecified DM, but the only medication prescribed was insulin, were excluded from this group. As T2DM is rare in young people, an age limit of 34 years was established and only people above this threshold were analysed. As a result, a total of 134,413 people were considered to be T2DM patients.
Aggregated Diagnosis Groups (ADGs) were used to detect the presence of diseases other than diabetes in our study population. ADGs are a component of the ACG system and comprise 32 diagnostic categories, which are groups with similar severity, expected duration of the disease and treatment needs.
In addition, the glycosylated haemoglobin (HbA1c) values for the study patients were collected. In the case of patients with multiple determinations in the study period, only the last value recorded was considered.
The deprivation index from the census tract of residence was used as a socioeconomic variable [13]. This index is usually categorised into quintiles, although for this study the subjects were grouped into three categories, with category one corresponding to residents in the least deprived areas and category 3 those in the most deprived regions.
Health care provision were calculated by cost-weighted health care utilisation. We consider the cost of the following services: prescribing, primary care (including visits to physician and nurse, laboratory and radiology), specialised outpatient care (doctor visits, rehabilitation, dialysis, radiotherapy and chemotherapy services), A&E attendances and inpatients stays.
In the case of prescribing, the cost was computed directly from primary care prescriptions recorded in the electronic health records. For the other cases (visits to A&E, rehabilitation sessions, outpatient specialty care and primary care doctors and nurses; lab tests and X-rays requested in primary care; some procedures such as dialysis, radiation therapy or chemotherapy performed in day hospitals) the number of services used by each patient was multiplied by their standard cost. The costs of hospital stays and major outpatient surgical procedures were calculated according to the weights of their corresponding Diagnosis Related Groups (DRGs) [14].
Some services for which no information was available, namely admission to psychiatric hospitals, hospital-at-home and day hospital services (except the above-mentioned procedures), medical transport, prostheses and other equipment delivered to patients at home, were excluded from the cost estimate.
For study of severe hypoglycaemic episodes, we considered those which required hospital-based care (either resolved by A&E services or required hospitalisation). The population was classified into two groups: persons with no hypoglycaemic episode and those who presented one or more.
As the hypoglycaemic episodes recorded upon discharge from hospital are often incorrectly coded, a modification of the algorithm proposed by Ginde et al. was used [15]. These authors considered hypoglycaemic episodes to be:
-
1.
Those episodes assigned the corresponding ICD-9-CM codes (251.00; 251.02; 251.10; 251.12; 251.20; 251.22).
-
2.
Those related to intoxication with antidiabetic drugs (code 962.3).
-
3.
Those in which the diagnosis was coded as 250.80 or 250.82 (diabetes with other specified complaints), excluding those which also presented one of the following diseases as co-diagnosis: cellulitis, lower limb ulcers, osteomyelitis, Oppenheim-Urban syndrome, diabetic lipidosis or secondary diabetic glycogenosis.
Although the positive predictive value (PPV) for the episodes indicated by this third method in the study by Ginde et al. exceeded 79 % of correctly identified hypoglycaemic episodes, in our population the PPV did not reach 40 %. Consequently, in this latter case we decided to add an additional criterion, namely a laboratory analysis in which a blood glucose level of <60 mg/dL was recorded. Although this algorithm (Fig. 1) did not allow us to include hypoglycaemic episodes diagnosed using a reagent strip, it was found to be a specific method for detecting these episodes.
The mean was calculated for all continual variables and the frequency for discrete variables, stratifying by sex, age group and deprivation index. A logistic regression analysis was performed to study the effect of the independent variables (sex, age group, deprivation index, HbA1c and ADG) on the probability of presenting a hypoglycaemic episode, and a linear regression analysis was performed to analyse the differences in annual healthcare costs per person. Values with p <0.05 were considered to be significant.
Statistical calculations were performed using Stata, Data Analysis and Statistical Software, Release 12 (StataCorp, LP, College Station, TX, USA).