Based on experience gained from the registry, CUs are classified into 5 groups (Table 1). Type 1 CUs (without a structured cardiology unit) remain within the scope of the registry by providing valuable information on cardiology-related activity in hospitals serving small populations (clusters 1 and 2 of the MHSSE).

Because the RECALCALCAR survey did not classify all the CUs, MBDS data were used to develop a classification of all hospitals (Table 2).

The MHSSE provides the SEC with the MBDS database (MBDS_CAR) for the RECALCAR project. The MBDS_CAR comprises hospital discharges coded according to the IDC-9-CM (International Classification of Diseases, Ninth Revision, Clinical Modification) in NHS hospitals with a principal diagnosis of “cardiovascular disease” (CVD).6 In the absence of this diagnosis, it includes in the process fields coronary interventionists or, in the absence of a principal diagnosis or interventional procedures related to the 2 preceding points, discharges by a cardiology or cardiac surgery unit. The MBDS _CAR comprises approximately 400 000 hospital discharges per year of which approximately 350 000 have a principal diagnosis of CVD. Between 2007 and 2014, the MBDS_CAR provided data on 3 183 370 hospital discharges.

There is a high level of agreement between the RECALCAR survey (Table 1) and the MBDS (Table 2) on the classification of CUs (quadratric weighted kappa = 0.92; 95% confidence interval, 0.86-0.99; P < .0001).

The multivariable logistic regression risk adjustment model used by RECALCAR has been described in other studies,24 and is based on the methodology developed by Krumholz for the CMS,25–27 adapted to the characteristics of the MBDS_CAR. Unlike the CMS methodology, RECALCAR calculates hospital mortality instead of overall 30-day mortality; only CVD readmissions are estimated because the MHSSE does not provide the complete database. Discharges with a secondary diagnosis (such as shock) cannot be eliminated because the current NHS MBDS does not record them. Risk factors and their coefficients that are positively entered in the adjustment models are derived from those in the MBDS_CAR database.

Regarding the survey data presented in this article, qualitative variables are expressed as frequency distributions (number of cases and percentage) and quantitative variables are expressed as mean, median, and standard deviation. The chi-square test was used to analyze difference in proportions of qualitative variables. The Student t test was used analyze the normality of the distribution and homogeneity of variance of quantitative variables or, if needed, the Mann-Whitney U test for comparisons between 2 groups and the analysis of variance with the Bonferroni correction, and the Kruskal-Wallis test or the median test for comparisons of 3 or more groups. In all tests, a P value of < .05 was used as a cutoff for statistical significance. All data were analyzed using the SPSS 21.0 software package.

From 2011 to 2014, considerable stability was maintained in all the availability, utilization, and productivity indicators (Table 5). The median number of cardiologists per CU ranged from 13 to 15, with significant variations between CUs (standard deviation ± 9). There was a mean of 50 cardiologists/million population with significant variations between autonomous communities. The mean rate of CU discharges was 5/1000 population/y with a mean stay of 5.5 ± 1.1 days. No duty or on-call cardiologist was available in 30% of CUs with 24 or more beds; of these, no critical care beds were available in 45%. The mean rate of initial consultations was 16 ± 4/1000 population/y, with a ratio of 2:1 between successive and initial consultations. During the period analyzed, there was an increase in the number of heart failure (HF) units (42%-56%) and cardiac rehabilitation units (31%-45%). The mean rate of echocardiographic studies was 23 ± 4/1000 population/y, most of which were simple diagnostic imaging studies (92%). There was considerable variability in the productivity of echocardiography machines (1800 ± 650 studies/y), which in part may be due to variations in the amount of equipment (the survey asked respondents to only include equipment in the echocardiography laboratory, but some also included equipment in operating rooms, cardiac catheterization laboratories, etc.).

The number of cardiac catheterization laboratories varied by autonomous community. Although sufficient, the number was less than that recommended in the standards document (400 000 population per laboratory).6 These data did not take into account the age of the laboratories. There was marked variability between autonomous communities in the rate of catheterization studies, particularly in diagnostic studies (2750 ± 660/million population/y). There was also marked variability between units in productivity indicators, such as the number of studies by laboratory and cardiologist (1400 ± 550 studies per laboratory per year and 700 ± 330 studies per cardiologist attached to the unit). These figures suggest that there is room for improvement in resource use. The mean rate of percutaneous coronary interventions (PCI) per unit and cardiologist was more than the minimum recommended by the SEC Working Group on Cardiac Catheterization and Interventional Cardiology.30

In general, there was a sufficient number of electrophysiology laboratories, although less than that recommended in the standards document (600 000 inhabitants per laboratory).6 There was an occasional lack of laboratories in some autonomous communities. There was marked variability between autonomous communities in the rate of electrophysiological studies, particularly in diagnostic studies (2750 ± 660/million population/y).

The RECALCAR registry sometimes included questions on cardiovascular surgery units. According to the data collected, only 17% of cardiovascular surgery units performed more than 600 major surgical interventions, which is the minimum number recommended by the Spanish Society of Thoracic and Cardiovascular Surgery.6

In relation to “good practice” recommended in the standards document, there was poor implementation of process management (only 14% of the units) and scarce development of CVD health care networks (17%).

The most relevant finding was the marked variability in structure, activity, and management indicators between CUs and between the autonomous community health services. In general, there was no shortage of equipment. However, there were organizational and administrative issues: no duty or on-call cardiologist was available in 28% of CUs with 24 or more beds; of these no critical care beds were available in 50%; there was poor implementation of process management; and scarce development of CVD health care networks.

The results of comparisons between different types of units are shown in Table 6. More complex units had significantly more cardiologists, assigned beds, discharges, and external consultations. However, it is striking that no significant differences were found in relation to mean stay or the ratio of successive to first consultations.

In relation to cardiac catheterization activity, type 4 units had a significantly larger catchment population and more cardiologists attached to CUs, laboratories, and activity than type 3 units, with a similar per-laboratory performance and higher productivity per cardiologist attached to the CU (665 ± 267 in type 3 units vs 883 ± 295 in type 4 units; P = .002).

Regarding electrophysiology laboratory activity, the more complex units tended to have a significantly larger catchment population and more cardiologists attached to CUs, laboratories, and activity than less complex units. Type 4 units had significantly higher performance per laboratory and productivity per cardiologist attached to the CU.

Most type 3 and 4 units (78% and 97%) were accredited to teach the MIR program (Spanish acronym for the intern and resident physician program). More complex units conducted significantly more research activity (research projects, clinical trials, publications).

Regarding the “best practice” recommended in the standards document,6 there was poor implementation of process management, especially in type 3 units, and scarce development of CVD health care networks.

Associations between structure and activity data for 2014 (obtained from the 2015 survey of CUs) and MBDS indicators for 2013 were analyzed. As expected (Table 7), hospitals with the most complex units had more discharges for CVD, with a longer mean stay in hospitals with type 4 units, which was probably due to the higher complexity of the disease processes treated. The more complex units had better mortality and readmission rates with type 4 units obtaining the best outcomes (Table 7).

RECALCAR focused its first outcomes research on the 2 most prevalent disease processes in CU discharges: acute myocardial infarction (AMI) and HF. The first article published in this line showed that, within the Spanish NHS, the characteristics of the hospital, treatment in a CU, and PCI were associated with survival in patients hospitalized for myocardial infarction. The study recommended that the Spanish NHS establish health care networks that promote PCI and the participation of CUs in the care of patients with AMI.24 Based on RECALCAR data, Worner et al.31 found that in hospitals accredited to teach the MIR cardiology program with CUs that incorporated critical care and acute cardiovascular care units, risk-adjusted in-hospital mortality due to AMI was significantly lower than in those without these units. This finding supports the position statement of the SEC. Cequier et al.,32 found an association between the development of primary PCI health care networks, an increase in the PCI rate, and a regional decrease in the AMI mortality rate. Rodríguez-Padial et al.33 found similar AMI mortality rates, with lower readmission, mortality, and HF after AMI in patients with PCI treated in a CU and in hospitals with higher volume of activity.33 The 2 main recommendations of this study were the regionalization of units and the development of regional primary PCI networks (still incomplete in the Spanish NHS).

Similar preliminary results have been obtained in relation to HF (the second line of health outcomes research in RECALCAR).34–36

The main limitation of the data obtained is that their provision was voluntary and therefore cannot be extrapolated to all CUs. However, the participation rate was 72%, which is high for surveys of this kind. Despite this rate, the information may have been biased in favor of the participating CUs to the extent that these CUs showed better performance indicators than nonparticipating CUs. The reliability of the data collected in the survey has been verified by audits and compared with those obtained from other sources, such as the MBDS or the registries of the scientific working groups. There was a high level of agreement between the different sources.

Some of the limitations of the MBDS_CAR have beenmentioned. The MHSSE does not provide the complete database (with all discharge figures), and secondary diagnoses (such as shock) at admission cannot be eliminated. The MHSSE does not provide the SEC with the identity of the hospital, which is therefore deduced by indirect methods. This approach may lead to errors when analyzing associations between structure and activity data obtained from the survey and outcome indicators calculated using the MBDS. Furthermore, the MBDS has its own limitations because it is an administrative database. However, the use of administrative records to assess health service outcomes has been validated by comparison with data from clinical registries.37,38 In addition, the Spanish NHS MBDS undergoes regular audits and is a reliable information source.24