The objective of the study is to estimate the frequency of multimorbidity in type 2 diabetes patients classified by health statuses in a European region and to determine the impact on pharmaceutical expenditure.
The study was carried out in the Comunidad Valenciana (CV) in eastern Spain, working with the population for 2012. The total resident population for that year was 5,150,054, from which information was taken on 491,854 patients diagnosed with diabetes, being treated with antidiabetics and/or blood glucose reagent strips. Patients with juvenile diabetes, gestational diabetes and those not receiving pharmacological treatment were excluded, reducing the group to 350,015 patients.
The study was a cross sectional observational study. The data were obtained retrospectivity from the Electronic Health Record (EHR) and the Electronic Prescriber system (GAIA) corresponding to patients attended during the period January to December of 2012. Each patient has a healthcare identification number in the Population information System (PIS). Cases were identified from the population morbidity database (SCP-cv) of the Directorate General of Pharmacy of the Valencian Community Government Health Department (Conselleria de Sanitat). The population was organized and grouped using CRG. This model makes an initial classification which groups International Classification of Disease, 9th revision, clinical modification (ICD-9-MC) diagnostic codes in Major Diagnostic Categories (MDC). 350,015 people were selected as classified in MDC 101 Diabetes Mellitus (DM) from the ICD-9-MC codes of 250.x0 assigned in their medical encounters in 2012 from a CV total of 5,150,054.
Each MDC group’s different episode disease categories (EDC) also made up from CIE-9-MC codes. It is considered that EDC 427 Diabetes identifies patients with T1D and EDC 424 Diabetes those with T2D. The chronic conditions (EDCs) analyzed in the study as comorbid were: hypertension, dyslipidemia, mental disorders, osteoarticular disease, cardiovascular and cerebrovascular disease (CVD), thyroid disorders, severity obesity, osteoporosis, COPD, cancer, renal disease, Alzheimer, asthma and pulmonary embolism. CVD and Renal failure were included as comorbid still being an event produced by diabetes complications however not being the only cause. For example, a patient diagnosed with diabetes would be in different levels of severity according to comorbidities and complications. A patient with diabetes but without any other associated chronic pathology would be in status 5. Another patient with add on dominant chronic condition, such as hypertension, would be in status 6.
To determine which comorbidities should be considered in this analysis, a review of previous literature was conducted. In addition, we established a consensus with specialists in internal medicine.
Other microvascular diabetes related complications were identified from EHR using codes 205.x2 except for codes 250.10 and 250.12 (indicative of TD2 with ketoacidosis and a condition close associate with T1D) similar to the algorithm proposed by Kho et al. . Nephropathy results are not included here due to them having been insufficiently recorded.
Our algorithm was developed with primarily data warehouse, easily accessed through Structured Query Language (SQL).
From the data obtained, we estimated the frequency of episodes of the chronic diseases in the cohort of patients with T2D selected and classified in CRGs. To study the prevalence of other chronic diseases, we adapt the methodology previously reported by other authors . Following Barnet et al., we display frequencies, percentages and cross tabulations for descriptive analysis. Furthermore, we conduct a binary logistic regression to examine associations between physical and mental health comorbidities. We further add a factorial analysis to study prevalence of common chronic diseases and co-occurrence of diseases and include CRG groups in the descriptive analysis.
An exploratory factor analysis was performed with a varimax rotation applied with the aim of identifying the multimorbidity clusters. Factor analysis is one of the most widely used techniques in building models of multimorbidity [19, 20], although clinical criteria such as grouping by affected organ systems are also pertinent [21, 22].
The EDCs were coded in binary format (i.e., 0 = no disease and 1 = presence of the disease).
The factor analysis was based on a correlation matrix to determine which diagnostic variables comprised each factor . The factors obtained from this analysis were interpreted as multimorbidity patterns. Pharmaceutical use and expenditure were also analyzed bearing in mind the resulting multimorbidity groups. Specifically, we analyzed the use of insulins, oral antidiabetics (OAD) – secretegogues and non-secretegogues – and diabetes test strips, as well as the average cost of these medicines and the total primary health care pharmaceutical expenditure, taking into account the multimorbidity factors obtained.
The incidence of diabetic related microvascular complications (retinopathy, nephropathy and neuropathy) were estimated grouped according to age and gender. Logistic regression identified the effect of gender and age groups (independent variables) on the coexistence of such complications. Pearson chi 2 test was performed to observe the differences between gender and we reported P value under 0.05 were considered statistically significant.
The data used was analyzed using the statistics packet SPSS version 20.0. To comply with the data protection law, an anonymous and irreversibly dissociated number replaced the patient’s Healthcare Identification Number (PIS). This observational study was also approved by the Behavioural Research Ethics Board of the Generalitat Valenciana.