Information on all hospitalizations of patients with a diagnosis of NRAVD as the main or secondary diagnosis occurring in all Spanish hospitals between 2003 and 2018 was obtained from the national hospital registry.12 The completion of this record is mandatory and covers virtually all hospitalizations in public hospitals in Spain.13 We included patients with International Classification of Diseases (ICD)-9 code 424.1 (aortic valve disorders), used until 2015, and ICD-10 code I35 (nonrheumatic aortic valve disorders), used since 2016. These comprise aortic stenosis, aortic insufficiency and mixed aortic valve disease (stenosis + insufficiency). We excluded patients younger than 18 years or with unspecified sex, with Marfan syndrome (ICD9, 759.82; ICD-10, Q87), rheumatic valve disease (ICD-9, 395; ICD-10, I06, I08), congenital valve disease (ICD-9, 746.3; ICD-10, Q23) and aortic aneurysm or dissection (ICD-9, 441; ICD-10, I71), as well as those without proper codes. The matching between the ICD-9 and ICD-10 codes is shown in the table 1 supplementary data. We also excluded hospital stays of less than 2 days, as well as those with a discharge destination different from home or death.

This study was approved by the research ethics committee of Hospital Universitario 12 de Octubre. Due to the administrative nature of the data used, it was not necessary to obtain informed consent.

Hospitalizations were classified as those with NRAVD as the main diagnosis or as a secondary diagnosis. Age, sex and clinical characteristics (comorbidities) were registered. Age groups were categorized as: 18-44, 45-54, 55-64, 65-74, 75-84 and> 84 years. The degree of comorbidity was measured by the Charlson Index (Deyo version for ICD-9 codes), and the Quan Charlson Index for ICD-10, using the “Charlson” Stata and “comorbidity” R packages, respectively.14,15 Length of stay was measured in days.

The number of hospitalizations and specific NRAVD therapeutic procedures were counted for each year: aortic valve replacement (AVR), including SAVR and TAVI, and balloon aortic valvuloplasty (BAV). The codes used to identify each procedure are shown in table 2 of the supplementary, both for the ICD9 (2003-2015) and ICD10 (2016-2018) periods. In the case of TAVI, its codification began in 2014, showing no previous records. Age- and sex-standardized hospitalization and utilization rates per 100 000 population were obtained through the direct method for each procedure and subgroup. The population used for hospitalization and utilization standardized rates was established by the profile of the Spanish population during the study period (data obtained from the National Institute of Statistics, Spain).16 Both crude (cCFR) and age- and sex-standardized case fatality rates (sCFR) were used. The reference population for case fatality rates (sCFR, per 100 hospital admissions) was done using the whole population admitted during the study period.

Categorical variables are presented as numbers and percentages, and quantitative variables as mean±standard deviation. To analyze changes in categories during the study period, a chi square for trends test (Cochrane-Armitage) was performed, using the “ptrend” Stata command. For mean comparisons, a Kruskal-Wallis test was used, after ruling out normality. Trends in year-to-year standardized rates were analyzed with Poisson regression models with robust variance, with the incidence rate ratio (IRR) as the association measure (percentage of annual variation of trends).17 As changes in trend direction over time cannot be estimated with the Poisson model, an additional joinpoint model analysis was performed for each subgroup. This method provides the annual percentage change (APC). As the change in ICD coding between 2015 and 2016 should have an impact on the trend analysis, a joinpoint jump model analysis with 2015 as the jump location was performed. This correction for trends analysis after code changes has previously been evaluated in other diseases.18,19 For CFR, a standard joinpoint model was used. The trend in TAVI utilization rates was also analyzed by this method, due to the lack of sufficient observation units (years) to perform a joinpoint jump model. The joinpoint jump model and standard joinpoint model were performed with the Surveillance Research Program (National Cancer Institute of the United States) Software.20 All other analyses were performed with Stata IC 16.1 (Stata-Corp., College Station, United States) and R (R Core Team, 2019, Austria) software.

A total of 1 003 284 hospitalizations with a diagnosis of NRAVD were registered between 2003 and 2018. Of these, 163 150 were excluded (figure 1 of the supplementary data). NRAVD was a secondary diagnosis in most—716 115 (85.2%)—, and was the main diagnosis in 124 019 (14.8%). The total number of NRAVD hospitalizations increased from 32 957 in 2003 to 70 669 in 2018. The number of annual hospitalizations increased for total, main and secondary diagnoses (table 1). Total hospitalization rates increased from 103.6 per 100 000 in 2003 (95% confidence interval [95%CI], 102.5-104.7) to 166.4 in 2015 (95%CI, 165.2-167.6). After the change to ICD-10 codes, there was a drop in the rate to 108.0 in 2016 (95%CI, 107.0-109.0) with no change in the upward trend, reaching 121.1 hospitalizations per 100 000 in 2018 (95%CI, 120.1-122.1). The IRR for all hospitalizations during the whole period was 1.014 (95%CI, 0.998-1.029; P=.090). No changes in the direction of the trend for hospitalization rates were observed during the study period (APC, 3.9; 95%CI, 2.3-5.6; P <.05). Similar trends were observed for the groups with main and secondary diagnoses (figure 1).

Figure 1.

Temporal trends in hospitalization rates (top) and case fatality rates (bottom) among patients with nonrheumatic aortic valve disease as a main diagnosis (left) and secondary diagnosis (right) by joinpoint analysis. A and B: No inflection points in the jump model. C: Two inflection (2005 and 2013). D: One inflection (2012). Annual percent changes (APC) are shown in each graph for each stable time period (those within 2 inflection points).

(0.46MB).

Demographic and clinical characteristics of the study population are shown in table 1. The comorbidities analyzed are shown in table 3 of the supplementary data. Age and comorbidities increased over time in both groups with no change in the proportion of women and men. Only the proportion of cases with cardiogenic shock decreased. The mean age, proportion of women and comorbidities were higher in the secondary diagnosis group compared with the main diagnosis group (tables 3 and 4 of the supplementary data).

A total of 82 664 AVR procedures were registered, increasing from 3506 in year 2003 to 7118 in 2018. The standardized AVR rate increased from 10.5 procedures per 100 000 population (95%CI, 10.1-10.8) in 2003 to 17.1 (95%CI, 16.7-17.4) in 2018 (IRR, 1.033, 95%CI, 1.030-1.037; P <.001). This increase was similar for both the main and secondary diagnosis groups (table 2) and was stable throughout the observation time (2003-2018 APC, 3.7; 95%CI, 3.3-4.1; P <.05). While the main diagnosis group showed a similar trend (APC, 3.7, 95%CI, 3.3-4.1; P <.05), the joinpoint analysis showed an increase in the slope of AVR utilization in the secondary diagnosis group after 2015 (APC from 2003-2016, 3.8; 95%CI, 3.3-4.2, P <.05; APC vs 15.5; 95%CI, 6.6-25.0; P <.05 from 2016-2018) (figure 2A,B).

Figure 2.

Temporal trends in procedure utilization rates (top) and case fatality rates (bottom) among patients with nonrheumatic aortic valve disease as a main (left) and secondary diagnosis (right). A: Inflection points for SAVR and balloon aortic valvuloplasty (BAV) but not for transcatheter aortic valve implantation (TAVI). B: No inflection points for surgical aortic valve replacement (SAVR) or TAVI. C and D: A decrease in case fatality rate in SAVR. There are inflection points for SARV and balloon aortic valvuloplasty (BAV).

(0.55MB).

During the study period, 74 565 patients underwent SAVR (66 336 in the main diagnosis group, 89.0%) and 8138 TAVI (7361 in the main diagnosis group, 90.5%). The annual number of SAVR increased from 3506 in 2003 to 4431 in 2018, with a peak of 5682 procedures in 2013, the year prior to the introduction of TAVI. The SAVR utilization rate did not change significantly during the study period (from 10.5 procedures per 100 000 population; 95%CI, 10.1-10.8 in 2003 to 10.8; 95%CI, 10.4-11.1 in 2018; IRR, 1.008; 95%CI, 0.996-1.020; P=.201) (table 2). However, the joinpoint analysis showed a steady increase between 2003 and 2013 (APC, 3.3%; 95%CI, 2.6-4.0; P <.05), which reversed in 2013, inverting the annual utilization rate to −3.3% (95%CI, 1.3-5.1). A similar decrease in trends for SAVR utilization rates was observed when only patients with NRAVD as the main diagnosis were considered, while rates in patients with NRAVD as a secondary diagnosis increased consistently from 2003 to 2018 (figure 2A,B).

The absolute number of registered TAVI procedures increased from 734 in 2014 to 2690 in 2018, also increasing the proportion within all AVRs (from 11.7% in 2014 to 37.8% in 2018). TAVI utilization rates increased from 2.5 procedures per 100 000 population in 2014 (95%CI, 2.3-2.7) to 8.7 in 2018 (95%CI, 8.4-9.0), with an IRR of 1.341 (95%CI, 1.302-1.389; P <.001), which was similar in the main and secondary diagnosis groups (table 2). No joinpoints were detected in the TAVI utilization rate trend (APC, 35.8; 95%CI, 28.0-44.0; P <.05).

The absolute number of registered BAV procedures increased from 9 in 2006 to 577 in 2010, and its utilization rate increased from 0.03 procedures per 100 000 population to 1.53 in the same period (APC, 196; 95%CI, 20.2-629; P <.05), with an annual 21.2% (95%CI, 4.1%-35.3%; P <.05) drop since year 2010 onwards.

The mean length of stay in the whole group decreased from 11.3±11.4 days in 2003 to 9.4±9.5 in 2018 (P <.001), with a fall in all groups, except for patients treated with TAVI for NRAVD as a secondary diagnosis (table 5 of the supplementary). For the total study population, the cCFR increased from 6.7% in 2003 (95%CI, 6.5-7.0) to 8.7% in 2018 (95%CI, 8.5-8.9), with an IRR of 1.015 (95%CI, 1.012-1.018; P <.001). After age- and sex-standardization, no significant change was observed (7.6%; 95%CI, 7.3%-7.9% in 2003 and 8.0%; 95%CI, 7.8%-8.2% in 2018; IRR, 1.000; 95%CI, 0.997-1.004; P=.891). However, a 0.7% annual decrease in the sCFR was observed between 2003 and 2012 (95%CI, 0.0-1.3; P <.05), with a subsequent 1.3% increase (95%CI, 0.0-2.5; P <.05) until 2018. Opposite trends in CFR among the main and secondary diagnoses groups was evident both in crude and standardized CFR (table 2, figure 1C,D).

In patients undergoing AVR, the crude cCFR decreased by 5.3% annually (IRR, 0.948; 95%CI, 0.939-0.957), from 7.8% in 2003 (95%CI, 7.0-8.8) to 3.9% in 2018 (95%CI, 3.5-4.4). Age- and sex-adjusted sCFR decreased from 8.8% in 2003 (95%CI, 7.6-10.1) to 4.0% in 2018 (95%CI, 3.6-4.5), with a mean IRR of 0.935 (95%CI, 0.930-0.947; P <.001), which was more pronounced between 2006 and 2018 (APC, −7.2; 95%CI, −6.3 to −8.2; P <.05). Trends were different for the main diagnosis group (2006-2018 APC, −8.2; 95%CI, −7.3 to −9.1; P <.05) and the secondary diagnosis group (2003-2018 APC, −3.4; 95%CI, −2.2 to −4.5; P <.05) (figure 2C,D). The cCFR decreased for SAVR from 7.8% (95%CI, 7.0-8.8) in 2003 to 4.6% (95%CI, 4.0-5.3) in 2018 (IRR, 0.954, 95%CI, 0.946-0.962) and for TAVI from 4.9% in 2014 (95%CI, 3.6-6.7) to 2.8% in 2018 (95%CI, 2.2-3.5) (IRR, 0.855; 95%CI, 0.776-0.941; P=.001). sCFR for SAVR fell from 8.3% (95%CI, 7.3-9.3) to 3.7% (95%CI, 4.0-5.3) (IRR, 0.949, 95%CI, 0.940-0.957; P <.001), with a significant intensification in the percentage of decrease between 2008 and 2013 (APC, 9.6%; 95%CI, 4.9%-14.1%). The adjusted CFR for TAVI fell from 4.8% (95%CI, 3.3-6.4) to 2.8 (95%CI, 2.2-3.4) (IRR, 0.860, 95%CI, 0.781-0.946), with no inflection point in the trend. CFR trends by type of diagnosis group are shown in table 2 and figure 2C,D.

Our study shows the importance of differentiating patients with NRAVD as the main or secondary diagnosis—that is, in those admitted for conditions not related to their valve disease—in epidemiological studies as these represent different clinical situations, patient risk profiles, and outcomes. While outcomes improved for patients with a main NRAVD diagnosis, probably related to an earlier diagnosis and the improvement in treatment—better results with surgery and the introduction of TAVI—, in patients with a secondary diagnosis, NRAVD probably contributed more as a comorbidity and a poor prognostic factor.38,39 Interestingly, the improvement in CFR among patients undergoing AVR, both SAVR and TAVI, was observed regardless of whether it was performed in patients with a main or secondary diagnosis. Previous studies reported improvements in survival for SAVR and TAVI in patients with a main diagnosis of NRAVD2 or a main and secondary diagnosis together24,26 but no stratified analysis by type of diagnosis has been published so far. As an increase in NRAVD diagnosis secondary to changes in the population2,22,40 is expected in the coming years, specific studies addressing the management and outcomes of patients with NRAVD as the cause of admission or comorbidity will be needed.

This work has clinical implications with an impact on daily practice. Patients with NRAVD as a secondary diagnosis would benefit less from specific treatment for their aortic valve disease, since we have observed a worse prognosis related to their comorbidities, which may determine their outcome in more decisively than the aortic valve disease itself. Therefore, in patients with a secondary diagnosis of NRAVD, the benefit of valve replacement is more controversial. Future studies are needed to improve the selection of candidates for AVR, including TAVI.

The national hospital registry covers more than 90% of hospitalizations in public centers in Spain. However, the inherent limitations of the use of secondary administrative databases must be acknowledged, including different bias risks. Due to the use of anonymized data, a data quality analysis could not be performed by the researchers. Outcomes of patients with a discharge destination other than home or exitus are not available, so they have been eliminated from the analysis.

From 2003 to 2015, diagnoses were coded according to the ICD-9, which has no specific code for aortic stenosis. Instead, NRAVD codes have been used as a proxy, as it accounts for the large majority of patients with NRAVD.21 The switch to ICD-10 coding by health care facilities has led to undercodification in Spain, mainly in 2016.41 The observed bump in the time series, deeper for NRAVD secondary diagnosis, may be explained by this fact. To attenuate this artefact, we used the joinpoint jump model, which allowed us to minimize the effect of this sharp fall and allow an evaluation of the whole trend. Finally, results stratified by NRAVD severity could not be determined since no echocardiographic records are included in the national hospital registry database.