Study design
Data were obtained from Intego, a Flemish general practice-based morbidity registration network at the Department of General Practice of the University of Leuven [12]. Ninety-seven general practitioners (GPs), all using the medical software programme Medidoc (Corilus NV, Aalter, Belgium), collaborated in the Intego project. These 97 GPs work in 55 practices evenly spread over Flanders, Belgium. GPs applied for inclusion in the registry. Before acceptance of their data, registration performance was audited using a number of algorithms that compared their results with those of all other applicants. Only the data of the practices with an optimal registration performance were included in the database. The selection procedure has been described in detail previously [12]. The Intego GPs prospectively and routinely registered all new diagnoses together with new drug prescriptions, as well as laboratory test results and some background information (including gender and year of birth), using computer-generated keywords internally linked to codes. Using specially framed extraction software, new data were encrypted and collected from the GPs’ personal computers and entered into a central database. Registered data were continuously updated and historically accumulated for each patient. New diagnoses were classified according to a very detailed thesaurus automatically linked to the International Classification of Primary Care (ICPC-2) and International Statistical Classification of Diseases and Related Health Problems 10th Revision (ICD-10). Drugs were classified according to the WHO’s Anatomical Therapeutic Chemical (ATC) classification system.
Intego started to collect data in 1994, but for the present study, data over a 10-year time interval from January 1st, 2002, to December 31st, 2011, were used.
Study population
In 2002, 20,301 patients registered in the Intego database were aged 60 years and older and had at least had one contact with their general practitioner. Of these, 1068 had already been diagnosed with AF before 2002 and were therefore excluded from our study. The inclusion criterion for the cases was a reported first diagnosis of AF between January 1st, 2002 and December 31st, 2011. In total 1830 patients were diagnosed with AF. All participants were followed until the last contact date in the Intego registry or until December 31st, 2011, whichever came first.
Each patient with AF was matched with 3 to 4 control patients, bringing the number of controls up to 6622 [13]. These patients were still in the Intego database at the moment of the AF diagnosis in the case, belonged to the same age stratum in 2002 (i.e., 60–69, 70–79 or ≥80 years old), were of the same gender and originated from the same GP practice but had not been diagnosed with AF before or during the follow-up period.
Clinical characteristics
Outcome variables
The first outcome variable was the diagnosis of AF. The date of diagnosis served as the baseline date for the case and its matched controls. The association between AF and comorbidities was explored in the entire group of cases and controls (n = 8452). A second outcome variable was the diagnosis of a cerebrovascular event (i.e., transient ischaemic attack (TIA) (ICPC – code K89) or cerebrovascular accident (ICPC – code K90)) in patients with AF after baseline. The risk of cerebrovascular events was calculated in the group of cases (n = 1830).
The internationally used definition of TIA changed in 2009, during our follow-up period. Whereas the temporary nature of neurologic symptoms (<24 h) were previously emphasized, a TIA is currently defined as a transient episode of neurological dysfunction caused by focal brain, spinal cord, or retinal ischaemia without acute infarction [14]. It is unclear which of both definitions the GPs used. Therefore, we here used ‘cerebrovascular event’ for both TIA and CVA.
CHADS2 and CHA2DS2-VASc (congestive heart failure, hypertension, age ≥75 years, diabetes mellitus, prior cerebrovascular event or thromboembolism, vascular disease (myocardial infarction or peripheral arterial disease), age 65–74 years and female gender) scores at baseline were calculated to determine the risk of a cerebrovascular event in patients with AF [5, 6].
Comorbidity
The Charlson Comorbidity Index (CCI) includes 19 chronic diseases, which are weighted based on their association with mortality [15]. The presence of these chronic diseases was not assessed for this study, but the history of chronic diseases is registered by the general practitioner in the electronic health record. A one-time registration before the baseline date was considered positive for cases and controls. The following diagnoses were included: diabetes mellitus; a history of myocardial infarction; heart failure; a history of cerebrovascular event; peripheral arterial illness; chronic pulmonary disease; a history of peptic ulcer; dementia; liver disease; hemiplegia; a history of cancer; a history of leukaemia; a history of lymphoma; and AIDS. Connective tissue disease could not be reliably assessed from the registry. Furthermore, the differentiation between cancers with or without metastasis, diabetes with or without end organ failure, and mild or moderate to severe liver disease could not be made. Consequently, all patients with any cancer history were assigned the same score (=2), as were all patients with diabetes (=1) and all patients with liver disease (=1). Therefore, we used a modified CCI (mCCI) in all further analyses [16].
The mCCI at the time of diagnosis of AF was calculated for each case and its controls. An mCCI was not available for 87 patients with AF because no creatinine levels were available in the database. In addition to the mCCI, diagnoses of hypertension and valvular heart disease were extracted at baseline.
Pharmacotherapy
The prescription of cardiovascular medication (ATC-coded) including beta-blockers, angiotensin-converting enzyme (ACE) inhibitors, angiotensin receptor blockers, calcium antagonists, diuretics, digitalis and antiarrhythmic drugs was extracted for every study participant at baseline and six months after the diagnosis of AF.
The prescription of antithrombotic therapy was registered for cases in the first six months after the diagnosis of AF (ATC-coded: heparin, acetylsalicylic acid, thienopyridine, dipyridamol, vitamin K antagonists). Throughout this paper we used the term ‘anticoagulants’ for both oral anticoagulants (vitamin K antagonists) and subcutaneous heparin treatment. When assessing whether patients were receiving anticoagulation treatment after the diagnosis of AF, we used a six-month time frame as there could be a doctor-related delay in registering treatment changes.
Data analysis
Continuous data are presented as the mean and standard deviation (SD). Categorical data are presented as numbers and frequencies. Comparisons between cases and controls were performed using the independent samples t-test or the χ2 test for categorical data.
The prevalence and incidence of AF for each year were calculated in the yearly contact group. These included all the Intego patients of 60 years and older who had had at least one contact with their GP that year. Further analyses were made using the data from patients included in the case–control study.
The association between the presentation of novel AF and comorbidity was explored in the entire study population (n = 8452) by calculating odds ratios (ORs) with the corresponding 95 % confidence intervals (CIs) using bivariate and multivariable binary conditional logistic regression analyses adjusting for cardiovascular medication at the moment of diagnosis. Two different models were used; one with the mCCI and one with the different comorbidities of the mCCI without the overall index. These analyses were also performed in different age strata to see whether the pattern of multimorbidity in relation to AF changes between age groups.
The risk of cerebrovascular event after baseline in patients with AF (n = 1830) was calculated using a Cox proportional hazards model and adjusted for age and gender. Two different models were used: one with the mCCI and CHADS2 or CHA2DS2-VASc score and one with the different comorbidities without the overall index.
The association between the prescription of anticoagulants in the first six months after baseline and multimorbidity was investigated in patients with AF (n = 1830) using bivariate and multivariable binary logistic regression analyses adjusting for age, gender and cardiovascular medication prescribed in the six months after baseline. Two models were used: one with the mCCI and CHADS2 or CHA2DS2-VASc score and one with the different comorbidities without the overall index.
Subsequently, in patients with AF who developed a cerebrovascular event, individual patient profiles were investigated to explore further the relationship between the prescription of anticoagulation and the incidence of cerebrovascular events. For those who had received oral anticoagulation treatment (vitamin K antagonists), the last INR values before the cerebrovascular event were checked to see whether they were within a therapeutic range (INR 2–3.5) (cross section method for time-in-therapeutic range). Patients who had received a first diagnosis of AF at the same time as when they had had a cerebrovascular event were excluded from certain analyses because there had been no time to administer anticoagulants before the event.
To avoid co-linearity, the correlation coefficients between all covariates were calculated. In the case of co-linearity (Pearson’s r >0.80), only one of the two covariates was considered in the multivariable model. When clinically relevant, interaction was assessed between the variables used in the analyses. If the interaction term was statistically significant (P <0.05), separate models were run to assess the direction of association in different strata.
Statistical analyses were performed using SPSS 22.0 (SPSS Inc., Chicago, IL, USA).