Design
Cost analysis of population data from the real world to calculate the total direct health care costs of patients with NVAF treated with oral anticoagulation, comparing the two groups of anticoagulants (VKAs and DOACs), analysed by intention-to-treat. The study was carried out from the perspective of the health system, over a 1-year period (2017) (Fig. 1).
Population
This study included all patients identified with a diagnosis of atrial fibrillation and treated with VKAs or DOACs in the population health database of Catalonia and who met the inclusion and exclusion criteria described below as of 1/1/2017.
In Spain, universal health care is paid for principally through taxation, and each autonomous community is responsible for the health care of its population. In the autonomous community of Catalonia, most residents are assigned a primary care centre that is managed by the Institut Català de la Salut, the largest public health service company in the community. It covers 6 million people, who represented 80% of the Catalan population in 2017.
All patients with a diagnosis of atrial fibrillation in the primary care of the Institut Català de la Salut were initially selected, applying the following inclusion and exclusion criteria:
-
Inclusion criteria: diagnosis of NVAF at least 1 year before and having been treated with a VKAs (warfarin or acenocoumarol) or DOACs (dabigatran, rivaroxaban, apixaban, or edoxaban) for at least 2 months before 1/1/2017.
-
Exclusion criteria: valvular atrial fibrillation (mitral stenosis or valve prosthesis), no anticoagulated patient, transfer of the patient to another autonomous community (different health system), and pregnancy (Fig. 1).
-
Definition of drug exposure: We identified the VKAs and DOACs available in Spain in 2017. Acenocoumarol and warfarin were included in the VKA group, and dabigatran, rivaroxaban, apixaban, and edoxaban were included in the DOAC group. Exposure to a drug was recognised for patients with an active prescription of one of the anticoagulants for the 2 months leading up to the start of the study (1/1/2017). If there was an overlap of different drugs, the one with the latest prescription end date was considered. Change of treatment group (VKAs to DOACs, or vice versa) was defined if, on 01/01/2018, a different type of anticoagulant from the initial one was detected.
The results were analysed within an intention to treat framework, although patients who switched anticoagulant treatment were determined. After applying the inclusion criteria, 128,915 patients were selected; applying the exclusion criteria reduced the sample to 82,034 patients.
Variables
Sociodemographic variables (age and sex), clinical variables (personal history of cardiovascular disease, cerebrovascular disease, arterial hypertension, diabetes mellitus, heart failure, kidney failure, and bleeding), CHADS2-VASC Score (risk of stroke in the next 12 months), prescription of oral anticoagulants by the primary care physician, as well as their pharmacy dispensing (we have included all commercialized anticoagulants, the VKAs acenocoumarol and warfarin, and the DOACs dabigatran, rivaroxaban, apixaban, and edoxaban), follow-up variables (medical and nursing visits in primary care and hospital), were considered in our statistical analyses data about hospital admissions and laboratory and complementary tests in addition to those performed during hospital admissions, that is, performed on an outpatient basis.. To assess the direct costs from the perspective of the health system, the variables of treatment costs, follow-up, and hospital admissions were analysed (Fig. 2).
Data sources
The population database used in the study was the primary care database of the Institut Català de la Salut for 2017 and was obtained from the Information System for the Development of Research in Primary Care, which contains information from different sources, such as the shared electronic medical record, which includes sociodemographic and geo-health variables, health problems categorized by the International Classification of Diseases 10 codes, laboratory determinations, complementary examinations, drugs recorded by their Anatomical Therapeutic Chemical code, and medical or nursing primary care and referrals from primary care to other hospital specialties.
The drugs prescribed by the primary-care professional came from real data contained in shared electronic medical records. Information about invoiced drugs came from the database of the Servei Català de la Salut Pharmacy Offices. Hospital data were also extracted from the Minimum Basic Dataset for Hospital Discharges, which includes diagnoses and hospital procedures linked to each admission to any of the hospitals in Catalonia, classified by hospital group based on structure and complexity.
Costs
The costs are presented in euros for the year 2017. The invoiced drugs come from real data from the Servei Català de la Salut, which contains the record of the medication dispensed for each patient to a National Health System pharmacy. We obtained the public sale price for each drug (that billed to the health system).
The costs of medical and nursing visits and complementary tests were obtained from the Official Gazette of the Generalitat de Catalunya for 2012 (Fig. 2). The costs were updated to 2017 according to the interannual variation of the Consumer Price Index published by the National Institute of Statistics [23]. To calculate the cost of hospital admissions, we analysed the variation in the price of the overall hospital discharge and of each hospital in the Public Use Hospital Network (XHUP) by analysing the variation in the price at hospital discharge with respect to hospital structure and complexity, and two indices, the relative intensities of structure (RIS) and of resources (RIR), for each hospital using the following formula [24]:
Hospital admission price i
(65% × high price RIS XHUP × RIS hospital i ) + (35% × high price RIR XHUP × RIR hospital i )
Data common to all hospitals.
Individual data of each hospital.
RIS/RIR hospital
these are individual values for each hospital. The RIS and RIR are indicators of the structural capacity of the hospital and of resource consumption, respectively.
High price RIS/RIR XHUP
these are prices approved and published annually in the Official Gazette of the Generalitat de Catalunya by Servei Català de la Salut. These rates are common to all centres.
Weights of the RIS and RIR
The two indices are differentially weighted, with 65% for the RIS and 35% for the RIR. These weights have not changed since 2000.
Since 1997, Servei Català de la Salut has used this formula to calculate payment for the purchase of health services, assigning a weight of 65% to the unit price of hospital discharge modulated by structure and a weight of 35% to the unit price of discharge modulated by hospital complexity (resources). For each hospital centre, the price of discharge modulated by structure was adjusted by the RIS of the centre, and the price of discharge modulated by complexity was adjusted by the RIR of the centre.
Statistical analysis
Categorical variables are summarised as the frequency and percentage within each treatment group; continuous variables are summarised as the median and interquartile range. To detect statistically significant group differences, the contrast of differences in proportions was calculated with a Z test for categorical variables and the Mann‒Whitney U test for continuous variables.
Total cost was imputed according to the treatment group, and the per capita cost was calculated for each variable.
Propensity score matching was performed by randomly selecting 1,500 patients with DOACs. A couple with the same characteristics was assigned according to age, sex, years since diagnosis of NVAF, and their history of ischaemic stroke, indeterminate stroke, intracranial haemorrhage, heart failure, and kidney failure. The propensity score matching model was 1:1 without replacement, the amplitude of the calliper was ≤ 0.2 standard deviations of the logit propensity score, and the nearest neighbour method was used for matching. The standardized difference of the means and the difference between proportions were used to evaluate group differences for continuous and categorical variables, respectively. Standardized differences < 0.2 were considered acceptable values. The same analyses were carried out on this sample as on the whole population. The incidence rate ratio was calculated taking the DOAC group as a reference, and the confidence interval was calculated. The level of significance was set at 0.05. Statistical analyses were carried out with IBM SPSS® Statistics 20.0, R® 4.0.2 and EDIDAT® 4.2.
Sensitivity analysis
The costs were updated to 2022, based on the interannual variation of the Consumer Price Index published by the National Institute of Statistics [23]. The costs for 2023 were capitalized using discount rates of 1%, 5.6% and 8.9% with the formula:
Where r is the discount rate and t is the number of years
The rate of 5.6% was applied because it was the Bank of Spain's most recent forecast for 2023 [25], and that of 8.9% was used because it was the interannual variation of the Consumer Price Index for 2022 published by the Spanish National Institute of Statistics [23]. The price of the medications did not change until 2023. 2017 was considered the standard year, and the follow-up variables were not modified.