Importance of dialysis specialists in early mortality in elderly hemodialysis patients: a multicenter retrospective cohort study

The early mortality rate in elderly patients undergoing hemodialysis is more than twice that in young patients, requiring more specialized healthcare. We investigated whether the number of professional dialysis specialists affected early mortality in elderly patients undergoing hemodialysis. This multicenter retrospective cohort study analyzed data from 1860 patients aged ≥ 70 years who started hemodialysis between January 2010 and December 2017. Study regions included Seoul, Gyeonggi-do, Gangwon-do, Daejeon/Chungcheong-do, Daegu/Gyeongsangbuk-do, and Busan/Ulsan/Gyeongsangnam-do. The number of patients undergoing hemodialysis per dialysis specialist was calculated using registered data from each hemodialysis center. Early mortality was defined as death within 6 months of hemodialysis initiation. Gangwon-do (28.3%) and Seoul (14.5%) showed the highest and lowest early mortality rate, respectively. Similarly, Gangwon-do (64.6) and Seoul (43.9) had the highest and lowest number of patients per dialysis specialist, respectively. Relatively consistent results were observed for the regional rankings of early mortality rate and number of patients per dialysis specialist. Multivariate Cox regression analysis—adjusted for previously known significant risk factors—revealed that the number of patients per dialysis specialist was an independent risk factor for early mortality (hazard ratio: 1.031, p < 0.001). This study underscores the growing need for dialysis specialists for elderly hemodialysis patients in Korea.


Calculation of dialysis specialist-related data
Study regions were classified according to the administrative division standards of the ROK as follows: Seoul, Gyeonggi-do, Gangwon-do, Daejeon/Chungcheong-do, Daegu/Gyeongsangbuk-do, and Busan/Ulsan/Gyeongsangnam-do.The number of patients undergoing hemodialysis per dialysis specialist by region was calculated using data from the Korean Renal Data System (KORDS), a dialysis patient registration program implemented by the KSN.The KSN issues a dialysis specialist license to internal medicine specialists who have undergone training in nephrology and have worked at a hemodialysis center for > 2 years after being recommended by two existing dialysis specialists.Physicians with a dialysis license were defined and counted as dialysis specialists.The number of dialysis patients was calculated based on the number of patients whose data were registered at each hemodialysis center.As for transferred patients, the record from the previous hemodialysis center was deleted in the KORDS, and patients newly transferred to a hemodialysis center were regarded as "transferred." Thus, transferred patients were calculated as the number of patients who were newly transferred to the hemodialysis center.
The fulfillment of recommendations for dialysis specialists at each hemodialysis center was assessed based on the standards of the KSN's hemodialysis center certification program.The recommended standards for dialysis specialists were ≤ 24, ≤ 26, and ≤ 36 dialysis sessions per dialysis specialist per day in general hospitals, hospitals, and clinics 14 .In the ROK, general hospitals, hospitals, and clinics are defined as hospitals with ≥ 100 beds, 30-99 beds, and ≤ 29 beds, respectively (for more details, see reference 16 ).Given that the number of recommended dialysis specialists differs depending on the size of medical institutions, the number of patients per dialysis specialist was calculated at the general hospital, hospital, and clinic levels.

Clinical parameters and outcomes
Data on demographics, such as height, weight, unplanned dialysis, comorbidities (diabetes mellitus, hypertension, ischemic heart disease, congestive heart failure, cardiac arrhythmia, cerebrovascular disease, peripheral vascular disease, active malignancy, severe behavioral disorder, and liver cirrhosis), mobility status, nursing home residence, hospitalization history within 6 months before dialysis, and laboratory test results were collected at the time of hemodialysis initiation.Patient mortality data were requested from the National Statistical Office of the ROK and collected using the resident registration number.Using the collected data, early mortality prediction scores for elderly patients undergoing hemodialysis were calculated according to the French renal epidemiology and information network (REIN) study, and studies by Santos et al., and Thamer et al. (hereafter referred to as REIN, Santos, and Thamer scores) [17][18][19] .The REIN score is composed of age, sex, congestive heart failure, peripheral vascular disease, cardiac arrhythmia, active malignancy, severe behavioral disorder, mobility status, and serum albumin (for more details, see reference 17 ).The Santos score comprises age, ischemic heart disease, cerebrovascular disease, serum albumin, and prior nephrologist's care, but prior nephrologist's care scores were omitted in the present study as they were not collected (for more details, see reference 18 ).The Thamer score encompasses age, serum albumin, mobility status, nursing home residence, malignancy, congestive heart failure, and hospitalization history within 6 months before dialysis (for more details, see reference 19 ).The patients were followed up until December 31, 2020.The mortality rate of hemodialysis patients is high within the first 6 months of starting hemodialysis 7,20 .Furthermore, in previous studies on the development of a prediction model for early mortality in elderly hemodialysis patients, death within 6 months of starting hemodialysis was defined as early mortality [17][18][19]21 . Theefore, in this study, death within 6 months of starting hemodialysis was defined as early mortality and set as the primary outcome.

Statistical analysis
All continuous variables are expressed as mean ± standard deviation.Student's t-tests were used to compare continuous variables.Nominal variables are expressed as proportions.The chi-squared test or Fisher's exact test was used for the comparison of nominal variables, as appropriate.As this study was a retrospective cohort analysis, sample size calculation based on statistical power was not performed, and all patient data during the relevant period were collected from each research institution.Receiver operating characteristic (ROC) curve analysis was performed to set the cut-off value for the number of patients per dialysis specialist, and the value showing the highest sensitivity and specificity in predicting early mortality was set as the cut-off value 22 .Kaplan-Meier survival curve analysis was performed to compare patient survival between the low and high patients per dialysis specialist groups according to the cut-off value, which were compared using the log-rank test.Multivariate Cox proportional hazard regression analysis was performed to identify the risk factors for early mortality in elderly patients undergoing hemodialysis.We conducted 1:2 propensity score matching for factors showing significant differences between the early mortality and non-early mortality groups.Nearest neighbor matching was performed for propensity score matching using the Matchlt package in R 4.1.1(https:// www.R-proje ct.org/).Patients were matched to the closest pair based on the propensity score difference between the two groups.All statistical  www.nature.com/scientificreports/

Comparison of early mortality and number of dialysis specialists by region in the ROK
The early mortality rate was the highest in Gangwon-do (28.3%), followed by Daegu/Gyeongsangbuk-do (22.2%), and the lowest in Seoul (14.5%, Fig. 2a).The number of patients per dialysis specialist was also the highest in Gangwon-do (64.6), followed by Daegu/Gyeongsangbuk-do (63.9), and the lowest in Seoul (43.9, Fig. 2b).Gangwon-do had the highest proportion of centers that did not meet the recommended number of dialysis specialists (62.5%), followed by Daegu/Gyeongsangbuk-do (60.2%); conversely, Seoul had the lowest proportion of such centers (30.3%, Fig. 2c).Relatively consistent results were observed for the regional rankings of the early mortality rate, number of patients per dialysis specialist, and proportion of centers that did not meet the recommended number of dialysis specialists (Fig. 2).Overall, 57,167 patients were registered in the KORDS database, of whom 25,696 (44.9%) were in clinics, 7963 (13.9%) in hospitals, and 23,508 (41.1%) in general hospitals (Supplementary Table S1 online).Except for Busan/Ulsan/Gyeongsangnam-do, the number of patients per dialysis specialist in all regions was lower in general hospitals than at hospitals and clinics.Nevertheless, the proportion of centers that did not meet the recommended number of dialysis specialists varied by region.Notably, Busan/ Ulsan/Gyeongsangnam-do showed a low proportion of unmet hemodialysis centers with dialysis specialists at the clinic level (39.3% in clinics vs. 60.4% in hospitals and 63.6% in general hospitals).In contrast, Gangwon-do exhibited a high proportion of unmet hemodialysis centers with dialysis specialists at the general hospital level (72.7% at the general hospitals vs. 58.3% at the clinics and 55.6% at the hospitals).The median survival duration was the lowest in Gangwon-do (Table 2).

Kaplan-Meier survival curve analysis of the low and high patients per dialysis specialist groups
The cut-off value for the number of patients per dialysis specialist set by ROC curve analysis was 57.2.Based on this, the patients were classified into the low and high patients per dialysis specialist groups, and Kaplan-Meier survival curve analysis was performed.There was a clear difference in survival between the two groups with a log-rank test p value of 0.008 (Fig. 3).

Multivariate Cox proportional hazards regression model analysis for early mortality
Table 3 shows the results of the multivariate Cox proportional hazards regression model analysis conducted to identify the significant factors affecting early mortality.Baseline characteristics showing significant differences between the early mortality and non-early mortality groups (age, BMI, hypertension, cardiac arrhythmia, active malignancy, severe behavioral disorder, mobility status, nursing home residence, hospitalization history within 6 months before dialysis, unplanned dialysis, BUN, creatinine, albumin, and inorganic phosphorus levels), factors that were observed to be important in previous prediction models for early mortality in elderly patients undergoing hemodialysis (sex, congestive heart failure, and peripheral vascular disease), and the number of patients per dialysis specialist were included in the analysis.The early mortality rate was the highest in Gangwon-do, followed by Daegu/Gyeongsangbuk-do, Gyeonggi-do, Busan/Ulsan/Gyeongsangnam-do, Daejeon/Chungcheong-do, and Seoul.Gangwon-do had the highest number of patients per dialysis specialist, followed by Daegu/Gyeongsangbuk-do, Daejeon/Chungcheong-do, Gyeonggi-do, Busan/Ulsan/Gyeongsangnam-do, and Seoul.Additionally, Gangwon-do had the highest proportion of centers that did not meet the recommended number of dialysis specialists, followed by Daegu/ Gyeongsangbuk-do, Daejeon/Chungcheong-do, Busan/Ulsan/Gyeongsangnam-do, Gyeonggi-do, and Seoul.The regional rankings for the early mortality rate, number of patients per dialysis specialist, and proportion of centers that did not meet the recommended number of dialysis specialists were relatively consistent (created with Bing, Geonames Tomtom, Microsoft, produced by Microsoft Excel 2016).
Age, active malignancy, mobility status (totally dependent on transfers), unplanned dialysis, creatinine levels, albumin levels, and number of patients per dialysis specialist were identified as significant factors influencing early mortality in the multivariate model.The risk of early mortality increased by 3.1% when the number of patients per dialysis specialist increased by one.
Table 4 presents the results of the Cox proportional hazards regression model analysis conducted to examine the effect of the number of patients per dialysis specialist at the clinic, hospital, and general hospital levels.The multivariate model included factors reported by previous studies to be important and showing significant differences in baseline characteristics (i.e., the same as the factors included in Table 3; Supplementary Tables S2, S3, and S4 online).The multivariate hazard ratio for the number of patients per dialysis specialist was the highest at the general hospital level (1.025), as compared with the clinic level (1.013) and hospital level (1.009).www.nature.com/scientificreports/

Multivariate Cox proportional hazards regression model analysis for early mortality in the propensity score matched cohort
Table 5 shows the results of the multivariate Cox proportional hazards regression model analysis conducted to identify the significant factors affecting early mortality in the propensity score matched cohort.Baseline characteristics showing significant differences between the early mortality and non-early mortality groups in the propensity score matched cohort (hypertension, cardiac arrhythmia, peripheral vascular disease, active malignancy, mobility status, nursing home residence, hospitalization history within 6 months before dialysis, unplanned dialysis, calcium levels), and the number of patients per dialysis specialist were included in the analysis (Supplementary Table S5).The number of patients per dialysis specialist remained an independent risk factor for early mortality.

Discussion
Herein, we observed that the number of patients per dialysis specialist was significantly higher in the early mortality group than in the non-early mortality group.Similarly, the proportion of centers that did not meet the recommended number of dialysis specialists was higher in the early mortality group.Notably, the number of patients per dialysis specialist was an independent risk factor for early mortality in elderly patients undergoing hemodialysis, even in the propensity score matched cohort.Regarding baseline characteristics, the early mortality group had significantly lower BMI and albumin levels than the non-early mortality group.Additionally, the proportions of cardiac arrhythmia, active malignancy, severe behavioral disorder, total dependence on transfers, hospitalization history within 6 months before dialysis, and unplanned dialysis were significantly higher in the early mortality group than in the non-early mortality group.These results are consistent with those of many previous studies on early mortality in elderly patients 18,19,21,23,24 .BUN, creatinine, and inorganic phosphorus levels were significantly lower in the early mortality group than in the non-early mortality group, which may be a result of the decreased nutritional status in the early mortality group [25][26][27] .In addition, the prevalence of hypertension was significantly lower in the early mortality group; however, the reason for this was unclear.It is possible that the use of renin-angiotensin-aldosterone system inhibitor (RAASI) reflects a reduction in mortality, and the low RAASI use in the early mortality group (36.1% in the early mortality group vs. 51.6% in the non-early mortality group) was associated with a low prevalence of hypertension during the retrospective data collection process [28][29][30] .
Several studies have been conducted to develop prediction models for early mortality in elderly patients undergoing hemodialysis.Although there are slight differences between the models, they commonly include patient age, functional status, hospitalization history, unplanned dialysis, and nutritional status as major factors [17][18][19]21 . Preious studies have focused on determining whether conservative management should be performed in patients with a high probability of early mortality as they may not benefit sufficiently from hemodialysis 17 .However, while it is important to determine the appropriateness of conservative management in elderly patients undergoing hemodialysis at high risk of early mortality, efforts to reduce early mortality in these patients are also important.
Previous studies have analyzed the early mortality in elderly patients undergoing hemodialysis, focusing only on patient factors.However, early mortality in elderly patients undergoing hemodialysis may also be affected by higher-quality medical care, such as more specialized and meticulous management by dialysis specialists.Therefore, this study focused on physician factors rather than patient factors.In addition to dialysis, the management of patients undergoing hemodialysis also requires integrated management such as anemia, mineral bone disorder, blood pressure control, nutritional status improvement, dialysis adequacy, and vascular access management [31][32][33] .For meticulous management, the physician in charge of hemodialysis should be trained through a specialized educational program.
The mortality rate in elderly patients undergoing hemodialysis was the highest in Gangwon-do and lowest in Seoul.Similarly, the number of patients per dialysis specialist and the proportion of centers that did not meet the KSN recommendations were the highest in Gangwon-do and lowest in Seoul.When analyzed by region, the rankings for the early mortality rate and number of patients per dialysis specialist were relatively consistent.In addition, based on a cut-off of 57.2 patients per dialysis specialist, there was a clear difference in the Kaplan-Meier survival curve between the low and high patients per dialysis specialist groups.The early mortality www.nature.com/scientificreports/rate was 16.4% in the low group and 23.6% in the high group.Furthermore, after adjustment for various factors influencing early mortality in elderly patients undergoing hemodialysis, the number of patients per dialysis specialist was observed to be an independent risk factor for early mortality.In the multivariate Cox regression analysis that included the REIN score previously developed for predicting early mortality in elderly patients undergoing hemodialysis, the number of patients per dialysis specialist was also identified as an independent risk factor (hazard ratio: 1.034, p < 0.001).In the propensity score matched cohort analysis conducted to minimize selection bias, the number of patients per dialysis specialist remained an independent risk factor.This suggests that physician factors, represented by dialysis specialists, are important in patient prognosis even after correcting for important patient factors in the early mortality of elderly patients undergoing hemodialysis.After hemodialysis initiation, patients are usually transferred to the community; in particular, when their risk of early mortality is deemed to be high, they are often transferred to nursing hospitals.Therefore, we additionally analyzed the number of patients per dialysis specialist and centers that did not meet the recommended number of dialysis specialists in clinics, hospitals, and general hospitals.In most regions, there were more dialysis specialists in general hospitals than in hospitals or clinics.In Gangwon-do, the number of patients per dialysis specialist was lower at the general hospital level than at the hospital and clinic levels; nonetheless, the proportion of centers that did not meet the recommended number of dialysis specialists was abnormally high at the general hospital level, which might be attributable to an imbalance in the number of dialysis specialists among general hospitals in this region.General hospitals managing 41.1% of all patients undergoing hemodialysis and the number of patients per dialysis specialist in general hospitals showed the greatest impact on early mortality (Table 4).In other words, both the difference in the supply of dialysis specialists by region and lack of dialysis specialists in general hospital-sized medical institutions have some impact on early mortality.In particular, looking at Gangwon-do, which has the highest early mortality rate, the imbalance in the supply of dialysis specialists among general hospitals in the region can be assumed to be also important.
This study has several limitations.First, it did not include data from Jeolla-do and Jeju-do regions.This is because, at the time of the study, medical institutions in the Jeolla-do and Jeju-do regions were not included in the Korean Society of Geriatric Nephrology; as such, it was not possible to collect data from these regions.More objective results can be obtained by analyzing data from all regions of the ROK.
Second, the number of patients per dialysis specialist was calculated using data from the KORDS database; however, since the data in the KORDS database is voluntarily deposited by hemodialysis centers, accurate reflection of each region using the data is limited.Moreover, hemodialysis centers that deposit data in the KORDS database are likely to have dialysis specialists or physicians interested in the KSN.In other words, it is possible that the number of dialysis specialists in non-metropolitan areas was overestimated because the database contains data deposited by physicians who are interested in the KSN, and in non-metropolitan areas, the number of dialysis specialists is expected to be small.Nevertheless, the results of this study showed that there were fewer dialysis specialists in non-metropolitan areas, which had a significant effect on early mortality in elderly patients undergoing hemodialysis.
Third, the degree of medical access in each region could not be analyzed.Professional care provided by dialysis specialists is important in the context of early mortality in elderly patients undergoing hemodialysis; however, easy-to-access hemodialysis centers are also important.The number of hemodialysis centers registered in the KORDS data was significantly lower for Gangwon-do than for Seoul and Gyeonggi-do.Accordingly, differences in access to medical services among regions might also possibly have an impact on early mortality.
However, our study has strength in that this multicenter study recruited a relatively large number of elderly patients undergoing hemodialysis distributed nationwide.In the analysis using the previously described REIN, Thamer and Santos scores, the results were relatively consistent with the validation cohort analysis results of previous studies; thus, our cohort can be considered reliable for elderly patients undergoing hemodialysis (Supplementary Table S6 online).In addition, to our knowledge, this is the first study to show that the quality of medical services, represented by the number of patients per dialysis specialist, has a significant effect on early mortality in elderly patients undergoing hemodialysis.
In conclusion, the number of patients per dialysis specialist is an important factor for early mortality in elderly patients undergoing hemodialysis.Additionally, the imbalance between regions with dialysis specialists and among medical institutions can affect early mortality in elderly patients undergoing hemodialysis.This study demonstrates the importance of professionally trained dialysis specialists for the rapidly increasing number of elderly patients undergoing hemodialysis in the super-aging society of the ROK.Active introduction of the dialysis specialist system and active training of dialysis specialists are crucial for improving the prognosis of elderly patients undergoing hemodialysis in the future.

Figure 1 .
Figure 1.Study design and population.Among 2588 patients aged ≥ 70 years who started hemodialysis at sixteen medical institutions between January 2010 and December 2017, 728 patients with missing or erroneous data were excluded.Finally, 1860 patients were included in the analysis.The early mortality group comprised 321 patients, whereas the non-early mortality group consisted of 1539 patients, corresponding to an early mortality rate of 17.3%.

Figure 2 .
Figure 2. (a) Early mortality, (b) number of patients per dialysis specialist, and (c) proportion of centers that did not meet the recommended number of dialysis specialists according to the regions of the Republic of Korea.The early mortality rate was the highest in Gangwon-do, followed by Daegu/Gyeongsangbuk-do, Gyeonggi-do, Busan/Ulsan/Gyeongsangnam-do, Daejeon/Chungcheong-do, and Seoul.Gangwon-do had the highest number of patients per dialysis specialist, followed by Daegu/Gyeongsangbuk-do, Daejeon/Chungcheong-do, Gyeonggi-do, Busan/Ulsan/Gyeongsangnam-do, and Seoul.Additionally, Gangwon-do had the highest proportion of centers that did not meet the recommended number of dialysis specialists, followed by Daegu/ Gyeongsangbuk-do, Daejeon/Chungcheong-do, Busan/Ulsan/Gyeongsangnam-do, Gyeonggi-do, and Seoul.The regional rankings for the early mortality rate, number of patients per dialysis specialist, and proportion of centers that did not meet the recommended number of dialysis specialists were relatively consistent (created with Bing, Geonames Tomtom, Microsoft, produced by Microsoft Excel 2016).

Figure 3 .
Figure 3. Kaplan-Meier survival curve analysis of the low and high patients per dialysis specialist groups.There was a clear difference in survival between the two groups.The early mortality rate was 16.4% in the low group and 23.6% in the high group.

Table 1 .
Comparison of baseline characteristics between early mortality and non-early mortality groups.Continuous and categorical variables are presented as mean ± standard deviation and as number (%), respectively.REIN, renal epidemiology and information network cohort.a The Santos score consisted of age, ischemic heart disease, cerebrovascular disease, albumin, and prior nephrologist's care; however, the prior nephrologist's care score was omitted as this was not collected in the study.

Table 2 .
Comparison of early mortality and dialysis specialists in the Republic of Korea according to region.Continuous and categorical variables are presented as median (interquartile range) and proportions, respectively.IQR interquartile range.

Table 3 .
Multivariate Cox proportional hazards regression model analysis for early mortality.The multivariate model was adjusted for known significant factors and those showing statistical differences between the early mortality and non-early mortality groups.The following parameters were used: age, sex, body mass index, hypertension, congestive heart failure, cardiac arrhythmia, peripheral vascular disease, active malignancy, severe behavioral disorder, mobility status, nursing home residence, hospitalization history within 6 months before dialysis, unplanned dialysis, blood urea nitrogen, creatinine, albumin, inorganic phosphorus levels, and number of dialysis specialists.After excluding patients with missing values, 1805 (97.0%) participants were included in the multivariate model.CI confidence interval, HR hazard ratio, Ref. reference.a p value < 0.05.

Table 4 .
Univariate and multivariate hazard ratios for early mortality in relation to the number of patients per dialysis specialist at the clinic, hospital, and general hospital levels.The multivariate model was adjusted for known significant factors and those showing statistical differences between the early mand non-early mortality groups (age, sex, body mass index, hypertension, congestive heart failure, cardiac arrhythmia, peripheral vascular disease, active malignancy, severe behavioral disorder, mobility status, nursing home residence, hospitalization history within 6 months before dialysis, unplanned dialysis, blood urea nitrogen, creatinine, albumin, inorganic phosphorus levels).HR hazard ratio, CI confidence interval.a Number of patients per dialysis specialist at the clinic level was used.b Number of patients per dialysis specialist at the hospital level was used.c Number of patients per dialysis specialist at the general hospital level was used.

Table 5 .
Multivariate Cox proportional hazards regression model analysis for early mortality in the propensity score matched cohort.The multivariate model was adjusted for known significant factors and those showing statistical differences between the early mortality and non-early mortality groups in the propensity score matched cohort.The following parameters were used: hypertension, cardiac arrhythmia, peripheral vascular disease, active malignancy, mobility status, nursing home residence, hospitalization history within 6 months before dialysis, unplanned dialysis, calcium levels, and number of dialysis specialists.After excluding patients with missing values, 963 (97.8%) participants were included in the multivariate model.CI confidence interval, HR hazard ratio, Ref. reference.a p value < 0.05.