Data source

We used the Acute Hospital Discharge Dataset (HDD), which includes records for all inpatient admissions from public and private hospitals in Catalonia. This database includes all episodes of healthcare in Catalonian health centers that led to patient admission into modalities including conventional hospitalization, major same-day surgery, outpatient day hospital and domiciliary hospitalization[11]. Data for the 2003–2012 time period were made available by the Catalonian Health Service, Health Department of the Catalonian Government.

From the HDD, BC cases were identified on the basis of the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) through the ASEDAT software[12]. ASEDAT was developed following the International Agency for Research on Cancer (IARC) recommendations[13]. Based on computer algorithms, it was designed to automatically detect and extract incident cancer cases from different databases, including the HDD in a given study period, with the purpose of increasing data reliability and reduce costs associated with a cancer registry.

We included women diagnosed with invasive breast cancer (i.e., malignant neoplasm of female breast; ICD-9-CM codes: 174.0-174.9)[14]. Data related to patients with in situ breast carcinoma (ICD-9-CM code 233.0) were excluded.

Identifying hospital episodes of care for breast cancer patients

Administrative hospital discharge abstracts from these BC cases were used to appropriately identify and construct the complete hospital episodes belonging to each patient. Patient abstracts were sorted in ascending order according to date of admission and discharge date, in that order.

Hospital episodes were defined by grouping together consecutive abstracts that were closely related in time (i.e., <24 hrs between them) and involved in a continuous healthcare process such as inter-hospital and intra-hospital transfers.

Patients were identified by their valid Personal Health Identification Numbers (the social security number), which were used to match hospital abstracts with individual patients. These data were pulled from the Catalonian HDD from 2003 onwards, enabling us to identify and compile all the different public hospital discharges where an individual had been attended. The result for each patient was one or more hospital episodes, each constructed from the corresponding hospital abstracts.

Episodes of inpatient care attended in private health centers were excluded from the analysis, because private hospitals do not use the Personal Health Identification Number system-wide, and we were unable to identify them within the public health system. Prevalent cases, defined as those with evidence of breast disease prior to the respective study periods, were also excluded by ASEDAT.

Defining the population and the breast cancer procedures carried out

In summary, we used the Catalonian HDD to identify all women with incident invasive BC diagnosed and attended during the years 2005, 2008 and 2011.

We only examined three one-year periods because of: 1) about 40% cases (1200 cases per year) remained unsolved after applying ASEDAT for the whole period, and they need to be manually checked by reviewing computer medical records, making it unfeasible; 2) the reported years are equidistants; and 3) data available for the time period 2003–2012 allow us to discard prevalent cases and ensure the minimum 12-month follow-up for incident BC cases.

For each woman, an index surgical procedure (BCS or mastectomy) was identified, as were subsequent surgical procedures related to breast reconstruction that took place within 12 months. Primary BCS and radical surgery procedures were identified based on the ICD-9-CM. Likewise, we also considered other related procedures such as removal of lymphatic structures (simple or radical), and reconstruction (immediate or delayed) (Additional file1: Table S1). Immediate breast reconstruction was defined as the presence of reconstructive surgery either as a primary or as a secondary procedure that took place simultaneously within the same first surgical episode as the mastectomy. Delayed reconstruction was limited to procedures performed within 12 months after the mastectomy. According to the ICD-9-CM codes used, our patients did not undergo any procedures of BCS or mastectomy during the three one-year study periods were classified as non-surgical.

Hospitals were categorized into three groups according to the Catalonian Health Service classification: i) high technology, including health centers providing highly specialized services, ii) reference hospitals and iii) county hospitals. In addition, the Charlson Comorbidity Index (CCI)[15], was calculated taking into account all the main and secondary diagnoses for the previous and reference principal episode.

Results were analyzed according to (a) age groups (≤49, 50–69 and ≥70 years old); (b) stage of disease at diagnosis (using ICD-9-CM coding as a proxy)[16, 17] (Additional file1: Table S2a); (c) hospital complexity; and (d) comorbidity. Three levels of comorbidity were defined according to Quan’s update of the CCI, as applied to ICD-9-CM[18]: no comorbidity = 0, low comorbidity = from 1 to median truncated to 1, and high comorbidity = more than the median truncated to 1.

Tumor stage was classified as local (confined within the breast), regional (affecting the lymph nodes, primarily those in the armpit and/or upper arm) or distant (the cancer is found in other parts of the body as well). Hospital discharge ICD-9 codes were mapped using the American Joint Committee on Cancer (AJCC) stages[17] (Additional file1: Tables S2a and S2b).

Statistical analysis

For each variable considered in the analysis, trends in proportions during the years 2005, 2008 and 2011 were assessed using a log-binomial regression model[19]. The sign of the slope of this model (β) indicates if there was an increasing (positive) or decreasing (negative) trend in the proportions[19]. Statistical analyses were carried out through the statistical software R[20] assuming a significance level of α = 0.05.

Ethics

The study uses retrospective data from administrative databases in where patients are anonymus to the researchers and did not require informed consent nor Ethics Committee approval.

Conservative or radical surgery

In 2011 in Catalonia, BCS accounted for 74% of BC surgery, a significant increase in comparison to 2005 and 2008. BCS use was increasing in the 50–69 age group (49.8%, 52.7% and 54.5% in 2005, 2008 and 2011, respectively) and among women with presenting with regional tumor stage (7.6%, 10.7% and 16.2%).

Likewise, a significant drop in radical surgery was observed in the same period (from 32.1% in 2005 to 26.0% in 2011), mainly in patients over 50 years old. On the other hand, we detected a significant rise in mastectomies in younger women (from 28.0% to 33.8%). According to the disease stage, there is a clear decline over time in the practice of mastectomy in cases of local disease (from 90.7% to 74.5%) and a significant increase for regional (from 7.7% to 21.2%) and metastatic tumors (from 1.7% to 4.3%). The most common modality of radical breast surgery was, by far, the simple mastectomy (around 85% of all mastectomies throughout the study period); radical mastectomy and subcutaneous mastectomy followed at a considerable distance (Table 2).

Lymph node surgery in BCS significantly increased (from 77.1% in 2005 to 86.2% in 2011), with important advances in simple lymph node dissection (from 48.8% to 71.4%). In mastectomized women, removal of lymph nodes remained stable throughout time at around 88%, and we detected a significant increase of simple lymph node excisions.

Mastectomy and immediate reconstruction

Breast-conserving surgery versus mastectomy

The percentage of women with BC treated by conservative surgery in Catalonia rose significantly between 2005 and 2011, mainly in the 50–69 year-old age group. These figures (74% in the year 2011) are consistent with other studies in Spain[26, 27] and worldwide[23, 28–30], where reports ranged from 55% to 69%. On the other hand, the United States (US) and Japan show a certain underutilization of this procedure, with percentages ranging from 51% to 59% in the former[31, 32] and just 36% in the latter[33]. Differences in the survey periods, the patient selection criteria, the methods used, the screening programs and the national health policies could all play a role in explaining the differences in these results.

The rise of conservative surgery among regional stage BC patients in Catalonia could be explained by the administration of adjuvant treatment, mainly post-surgical radiation therapy. Other relevant data found in this study shows a generalization of conservative surgery in most Catalonian health centers, with a rising preference for this treatment modality in hospitals with medium and low complexity. The spread of conservative treatment favors equity, regardless of the geographic area where patients live.

As noted by other authors[34, 35], removal of lymphatic structures has increased significantly among patients treated with BCS, with a rise in simple lymph node excision. This finding is closely related to the implementation of sentinel lymph node biopsy (SLNB), which spread in Catalonia over the course of the study period.

In terms of mastectomy, our results for 2005–2011 showed a significant decrease among women treated surgically, in line with data from the European eusomaDB database[36], but in contrast to the trend in the US between 2004 and 2008[37, 38]. Moreover, mastectomy rates in the Catalonian population were slightly lower than those reported in most European countries[30, 32], Canada[39], the US[32, 37] and Japan[33]. This could suggest a slight underestimation in our study, due to a failure to consider those patients initially treated with conservative surgery who received a mastectomy during a 12-month follow-up period. It should also be noted that (as in our study) most of the cases involved in all these studies were in early breast cancer stages (0, I and II).

An upward trend in mastectomy rates among women under 50 years of age is consistent with other studies[38, 39], and may be explained by several factors, including the higher stage disease at diagnosis, the presence of more biologically aggressive tumors, the higher risk of recurrence over the patients’ lifetime, the availability of immediate breast reconstructive surgery, increasing the aesthetic requirements, patient preference and other intangible factors.

In accordance with other series, patients treated with radical surgery are older and have greater comorbidity compared to those treated conservatively[27, 32].

Reconstructive surgery after mastectomy

Worldwide, post-mastectomy reconstruction rates have increased over the last decade due to a number of factors, but the procedure is still performed only in a minority of patients. The relatively young age of patients, the local stage of their tumors, the null or low comorbidity, and patient access to more specialized health centers are all recognized as limiting factors in breast reconstruction[40, 41].

In our case, there was a considerable and general increase in reconstructive surgery during the seven-year period, similar to the rise observed in England[42], and in contrast to the trends reported in Canada[39, 40] and Australia[41] (historically low), and in the US[43, 44] (where rates are higher than ours). The rise of breast reconstruction in Catalonia occurred basically due to immediate reconstructive surgery, because delayed breast reconstruction dropped significantly in the study period.

Methodological strengths and limitations

Our study has several strengths that should be considered. First, the starting point was, the HDD, a population-based information source that has been proved valuable and reliable to carry out clinical and epidemiological studies. Second, the use of ASEDAT software (previously validated)[12], allowed us to identify patients from the first episode of breast cancer care, essentially equivalent to the incidence date. The number of incident breast cancer cases in 2008 in Catalonia (n = 3849) was similar to the estimated number from the Catalonian population-based cancer registries (n = 3907)[8], reinforcing the validity of the methodology used. Third, breast cancer is basically a surgical disease that requires hospitalization, a fact which favors the availability of comprehensive information for primary and reconstructive surgery, because cases are followed for a minimum of 12 and a maximum of 24 months. Finally, information about the application of primary and secondary ICD-9-CM diagnosis and procedures codes allowed us to determine comorbidity and estimate cancer stage, in addition to knowing the surgical techniques implemented.

As for the limitations, we should mention firstly the intrinsic weaknesses of the study design. This is a descriptive study in a database used by Catalonian hospitals, so there could be some variability in coding practices between centers and professionals and the ICD-9-CM changes that happened over the study. For instance, we found some specific deficiencies, such as the lack of any specific ICD-9-CM codes for the most recent techniques, such as SLNB, which prevented us from assessing the impact of its implementation simultaneously with the lymph node dissection.

Secondly, the underestimation of surgical cases by the lack of information on discharges in private hospitals. Thirdly, the lack of relevant information for tumor stage in the HDD, although we tried to mitigate this problem by creating a proxy for the disease stage based on the ICD-9-CM (Additional file1: Table S2a)[16, 17]. Despite not having been validated in a sample of patient’s medical records, as we mentioned previously, this method seems to have properly delimited the advanced neoplastic disease but overestimated local disease by about 10%. Our results, agree with some validation studies[17, 45, 46], suggests caution when interpreting the results, given the likelihood of overestimating local stages versus advanced stages.

Despite all the aforementioned limitations, this study may be considered an interesting methodological contribution to automated studies of quality indicators from clinical and administrative databases in Catalonia, with regard to surgical procedures for invasive breast cancer.