Quantitative variables, are expressed as mean and standard deviation (SD) or median and interquartile ranges (ICR) in the case that the variable does not follow a normal distribution. Categorical variables are presented as percentages. Comparisons between quantitative variables are analyzed using Student's t if the data follows a normal distribution or with Mann Whitney's U if the distribution is not normal. Differences between categorical variables are compared using a χ2 (chi-square). The follow-up time has been calculated as the difference between the starting date and the end date of treatment (or 31 Dec. 2015 if the data is censored). In the case of a transplanted patients, the date of initiation of RRT is considered the day of discharge after the kidney transplant surgery. The admission rate has been estimated as the number of admissions per patient, and the admission rate for cardiovascular reasons, as the number of admissions per patient and year.

The time until the first admission has been estimated using Kaplan-Meier, and the differences between modalities have been established with the log-rank test.

The CMBDH database allocates a cost for each GRD on discharge. This cost is the approximate average value considering all the elements of: hospitalization, personnel, drugs, procedures, dialysis in the hospital and other interventions. From here we estimate the average cost per admission, and during the first year of follow-up in a patient incident on RRT. The cost for the first year is calculated by dividing the cost of all admissions by the number of patients who have initiated RRT, independently of whether they have be admitted to the hospital or have died before completion of one year follow-up.

All analyzes have been performed with Stata14 (StataCorp2015. Stata Statistical Software: Release 14. College Station, TX: StataCorp LP.X).

During the 2013-14 period, a total of 767 patients started RRT; 108 in PD (14.1%), 567 in HD (73.9%) and 92 were TX (12.0%). Mean age was 63.3 (SD: 16.4). Patients receiving a kidney graft before starting dialysis were younger than those initiating PD or HD (54.4 vs 60.0 and 65.4 years respectively, p < 0.001, Table 1). The main cause of CKD was diabetes mellitus (24.9%), followed by nephroangiosclerosis (NAE)/vascular disease (19.5%) and glomerulonephritis (16.7%).

At the end of the first year, 71.1% remain in the same modality of treatment, 16.8% had changed (12.3% have been transplanted, 8.3% have died, 2.2% have recovered renal function and in 1.6 % the follow-up was lost (Fig. 1). The cumulative follow up was 646.22 years (0.8 year / patient).

Fig. 1.

Diagram of patients one year after starting renal replacement therapy.

PD: peritoneal dialysis; HD: hemodialysis; TX: kidney transplant.

(0.08MB).

A 36.3% of patients incidents in dialysis initiate their RRT during a hospital admission. These patients were older and with more comorbidity, especially in those that were started on PD. As compared with the cohort of incident patients there was a lower percentage of patients with polycystic kidney disease (1.6 vs. 6.8%) and more patients with NAE / vascular disease (24.1 vs. 19.5%). A 37.6% had DM and 60% had a previous diagnosis of cardiovascular disease (Table 2).

The average period (AP) of the first admission was 24.1days (SD 24.8) and AP adjusted for dialysis technique was 19.1 days (11.7), with no difference between DP and HD (p = 0.6). The calculated complexity of the admission is 1.65 (SD: 2.39); it was higher in PD than in HD (2.15 vs. 1.64), and more than two thirds of the admissions were for severe or very severe clinical problems.

Most of the admission to initiate RRT are recommended by the emergency room doctors (79.2%) and only 20.8% had been programmed, with very similar percentages in both techniques (HD / DP). Most patients were discharged by the nephrology service (80.8%) and the reason for admission were related to diseases of the cardiovascular (38.8%) or genitourinary system (33.9%).

A 13.5% of patients who initiate RRT during a Hospital admission die during the first year of treatment. The main cause of mortality was infection (42.4%), followed by cardiovascular causes (27.3%) and cancer (12.2%).

Apart from the initial admission to start RRT and the admission to receive a renal TX, a 58.4% of patients are admitted to the hospital during the first year of RRT. A total of 903 admissions were recorded: 10.0% were PD patients, 78.0% HD and 11.0% TX. The admission rate during the first year was 1.18 admissions per patient, and the average number of days staying in the hospital per patient during the year of follow-up, was 12.05 days. The number of days in the hospital was 8.6 days (SD: 12.7) shorter in PD patients. Almost half of the admission were due to an emergency and only 3.2% die during an admission (Table 3).

The average age of admitted patients was 63.7 years (16.4), with a Charlson index of 5.3 (SD 1.8). The transplanted patients were younger than those who started dialysis (p < 0.001) and have a lower Charlson Index (Table 3). One third of the admissions were patients with DM (30.2%), followed by those with NAE/vascular disease (20%) and with glomerulonephritis (15.3%).

The admission rate due cardiovascular causes was 0.29 admissions / patient-year, and it was lower in renal TX patients than in dialysis patients (HD / DP). The admission rate due to infection of any origin was 0.14 admissions / patient-year and admission due to peritoneal infection in PD was 0.13 admissions / patient-year. The 11 cases admitted because peritonitis had an average hospital stay of 4.6 days (3.2), ranging from1 to 10 days, all with severity 3-4, but no deaths. In HD patients, the admission rate for infection was 0.14 admissions / patient-year; The causes were: 28.3% respiratory infection, 19.8% vascular access infection and 10.4% urinary infection; none of them was a cause of death.

The average cost of admissions during the first year (Table 4 ), estimated by incident patient and according to the GRD and the degree of severity is 5,251€ for PD patient, 10,999 € for HD patients and 27,531€ for TX patients ( including the cost of the transplant procedure: 21,255 €).

The urgent readmission for the same diagnostic category only represents 8% of the admissions (6.8% in HD, 10.2% in PD, 15.3% in TX). The most frequent DRGs in readmissions were "malfunction, reaction or complication of device, graft or genitourinary transplant" (10.0%) and "renal failure" (9.1%).

By Kaplan-Meier it was estimated that the time elapsed from the initiation of RRT to the first admission was 147 days (95% CI: 128-176). According to modalities of RRT, HD had the earliest admissions : 131 days HD (95% CI: 116-151), 260 days for DP ( 95% CI: 177-363) and 197 days for TX ( 95 % CI: 88-435) (log-rank 14.57; p < 0.001) (Fig. 2).

Fig. 2.

Time (in years) until the first admission according to starting technique of RRT, estimated by Kaplan-Meier (log-rank test). Admissions to start renal replacement therapy or receive a transplant is not included.

(0.1MB).

As in other registries 4, the average age of incident patients in RRT, is increasing progressively, being even higher than the reference published by the PD center group (DPCG) 24. The etiology of renal disease is similar to that described by the REER at the national level 5, with a high prevalence of DM. However, a higher percentage of glomerulonephritis and PKD in both PD and TX stands out. PKD patients have an earlier diagnosis and are frequently followed in advanced CKD clinics, so it is easier for them to start on PD or be transplanted without going through dialysis 25.

Number of days per admission is lower in incidents patients in PD than HD and the TX spent the longest period of admission. As compared with reports from North America the admission are longer in PD and much shorter in the transplant patient. The explanation could be the differences of the health model, in Spain transplant recipients are older with more comorbidity, a majority of donors with expanded criteria versus more living donors more favorable selection profile of donor and recipient in the United States model 17. Data from American show a shorter admission period as compared with Spain, but it should be taken into consideration that the data from America do not include prevalent patients who die within the first 3 months.

Many regional registries do not collect information about admission on RRT (for example, REMER), and if they do, they do not publish the analysis of data in this field. Only the registry from the Basque Country communicates data on hospital admissions, showing an average admission period of 11 days for PD, 13 days for TX and 14.5 days for HD this is information from 2008 from a concerted center 26.

Most patients are admitted in the nephrology service, even in cases where the records indicates a DRG from surgery. Possibly this is due to the complexity of managing our patients and the dependence on their nephrology doctors, thus for the surgeon it is easier to attend the patient while admitted in the nephrology service.

The Department of Health of the Community of Madrid considers inadequate an early discharge with urgent readmissions (within 30 days), due to the same diagnostic category. In our study these readmissions represent less than 10% of the admissions, and the percent is higher in TX than in PD or HD, possibly because part of the clinical incidences of PD and HD can be treated / followed within the HD unit, as an informal model of post-discharge continual medical care. This data is better than in Canadian studies (17.2% in HD) 27 or in North American studies (34% overall, 37% in HD) 17. Another study including only PD patients 28 estimated 14.6% of readmissions at 30 days; the most relevant reasons for admission were myocardial infarction, complications with the procedure and DM; in our series reasons for admission were problems with the device, graft or transplant.

A recent feature of the analysis allows estimating the risks of complications and/or mortality based on the diagnoses present at the admission for a given DRG, classifying them into 4 categories. The comparison with the real mortality in each category provides us with a tool to analyze the quality of care during admission 10 ; In our study, for each DRG the mortality obtained that was lower than expected. We have not found any publication that analyzes this aspect in patients with CKD and / or RRT.

A primary objective of the Nephrology Strategic Plan of our community of Madrid is a programmed initiation of the RRT after a regular follow up in the advanced chronic kidney failure (ACKD) out patients clinic29. However, more than a third of patients initiate RRT during a Hospital admission which is not considered optimal. Starting RRT while in a hospital admission is considered as a suboptimal situation 18,19. In older series Initiation of RRT during Hospital admission was as high as 50% 18,30 or even 75% of the cases when ACKD clinics were not fully stablished.

Although other countries have published experiences of using PD as an emergency to initiate RRT31, the unscheduled start of RRT is almost always performed by HD in our country, which explains the different rates found in our study for both techniques. If the patient had chosen PD, an early transfer to PD can be proposed after stabilizing their situation with an urgent onset in HD 18. Patients starting dialysis during a hospital admission are older than the rest of the incident cohort, especially in PD, where the difference in age is almost 10 years. Unfortunately, we cannot analyze the impact of comorbidity on the rate of admissions, because we only have that information in those patients that have been admitted.

In our study, the admission for initiation of dialysis entails a duration of more than 2 weeks which is in contrast to the rest of subsequent admissions; this is similar to what was observed in the previously referenced study on initiation of unscheduled dialysis 19.

A non-programed initiation of HD determines a more frequent use of vascular catheters 19 which is associated with a higher rate of hospitalization due to infections 8 and, therefore, at a higher cost 18, in addition to the reduction of patient survival during the first year of RRT 8,18. A13.5% of the patients who started dialysis during a hospital admission died within the first year; this is almost twice than those who did not initate RRT during a hospital admission, which is consistent with studies previously published 18,19.

The main objective of this study was not an analysis of resources consumption, but it is a great opportunity to estimate cost, which is often undervalued. In fact, the cost of the dialysis technique, pharmacy and even transportation has been estimated quite accurately in our country and has served to define a strategy to support the development of home techniques 7,32. The annual cost is considered 45,170€ for a HD center, 61,280 € for hospital HD, 33,255 € for the PD and 11,766 for TX, averaging the impact of the first year on the half-life of graft and patient 6. However, in these studies the budget of hospitalized care is not considered, or if they do, it is applied a standard cost 33 instead of using the costs associated with GRD 22,23.

Our calculations are very accurate, since we used the cost per GRD rather than a fixed price per day of admission and the estimates average admission days. To the cost of RRT should add more than 7,000€ per incident patient the first year, plus another almost 5,000 € if the initiation of dialysis is performed during admission. In the results section it is described the costs for each of the RRT techniques and the extra cost of initiating non programed RRT while in the hospital admission. These references are of great value for planning and the development of predictive models of the economic impact of interventions that enhance household techniques 34 or the programed start of dialysis with the creation of ERCA consultations and the application of quality models.

But while we wait for the expected revolution of real clinical management and health policy planning based on costs and results of interventions a, we can estimate that, today and in our setting, between 9 and 17% of the cost of TSR corresponds to hospital expenses 14,23,33.

We must not forget that our study focuses on the first year of RRT. Admissions may decrease during the following years and this would explain why admissions are lower in other studies on prevalent patients 23.

In order to guarantee the inclusion of all hospital admissions of a given patient and analyze the integrated treatment, only the hospitals from the Community of Madrid with a complete portfolio of services have been included in the study. The inclusion of less complex hospitals, without transplant patients or on call services, would increase the statistical power but would make difficult the analysis of cross-hospital admissions. This aspect should be considered in the extrapolation of results to hospitals of a lower level of complexity or with a partial service portfolio. To consider the impact of RRT at the hospital level in a comprehensive manner, the emergency episodes should be included, which cannot be extracted from the MBDS. But one of the advantages of this analysis is that by its nature it encompasses 100% of patients incident to RRT, as well as all their admissions in the Community of Madrid.

Our article evaluates available information that allows the generation of hypotheses with the corresponding new area of research. So far the limitations have been organizational and never technical.

In recent years, health-related data has increased exponentially with an associated increase in medical publications including the term "big data". However, before analyzing the unstructured information and "diving in a sea of data ", we should take advantage of the already available structured databases to obtain relevant information on the real effect of our health interventions. We have databases of public domain, of mandatory implementation, legally protected and often dependent on the same service within the department of Health. By integrating national, regional, hospital records, including clinical, laboratory, etc. databases, and analyzing them together, the information available will increase enormously. The information obtained from structured data will provide the knowledge to design and deploy integrated health care plans and monitor the effects on our population health to ultimately improve management, results and care of our patients.