Early hospital readmissions after ABO- or HLA- incompatible living donor kidney transplantation

Early hospital readmission (EHR) after kidney transplantation (KT) is associated with adverse outcomes and significant healthcare costs. Despite survival benefits, ABO- and HLA-incompatible (ABOi and HLAi) KTs require desensitization and potent immunosuppression that increase risk of EHR. However, little data exist regarding EHR after incompatible KT. We defined EHR as admission for any reason within 30 days of discharge from the index hospitalization. Patients who underwent living donor KT from 2010–2017 were classified into one of three groups (control, ABOi KT, or HLAi KT). Our study included 732 patients, 96 (13.1%) of who experienced EHR. HLAi KT patients had a significantly higher incidence of EHR than other groups (26.6%; P < 0.001). In addition, HLAi KT (HR, 2.26; 95% CI, 1.35–3.77; P = 0.002) and advanced age (≥60 years) (HR, 1.93; 95% CI, 1.20–3.12; P = 0.007) were independent risk factors for EHR. Patients with EHR showed 1.5 times and 3 times greater risk of late hospital readmission and death-censored graft loss, respectively, and consistently exhibited inferior renal function compared to those without EHR, regardless of immunologic incompatibilities. We recommend that KT recipients experiencing EHR or its risk factors be managed with extreme care due to their increased susceptibility to adverse outcomes.

Demographic and baseline clinical characteristics. Patient characteristics are summarized in Table 1.
We found no significant differences in body mass index (BMI), dialysis duration, and history of diabetes among the three KT groups. However, the proportion of female recipients and re-transplant cases were significantly higher in the HLAi KT group than in the other groups. HLAi KT patients were significantly older and had higher number of HLA mismatches than those in other groups. Of HLAi KT patients, 26 (32.9%) exhibited a positive complement-dependent cytotoxicity (CDC) crossmatch, and 53 (67.1%) had a positive flow cytometric crossmatch. Patients with CDC crossmatch positive had significantly higher mean fluorescence intensity than those with flow cytometric positive (9031 ± 5912 vs. 5867 ± 4694, P = 0.013). A total of 26 patients (33%) had DSA against both class I and class II. Twenty-one recipients (26.9%) had class I DSA and 31 (39.7%) had class II DSA. Of ABOi KT patients, 122 (77.2%) had low titer of anti-A/B (≤1:64) and 36 (22.8%) had high titer of anti-A/B (≥1:128). We observed no significant differences in donor characteristics among the groups, with the exception of a lower proportion of female donors in the HLAi KT group. The use of anti-thymocyte globulin for induction was significantly more common for patients in the HLAi KT group than for those in the ABOi KT and control groups. Mean serum tacrolimus concentrations during the entire study period were similar in the three groups (P > 0.05). The median duration of the follow-up period was 45 months.

Index hospitalization.
Clinical outcomes during index hospitalizations are shown in Table 2. During the pre-KT hospital stay, we performed thorough medical, immunological, and psychological evaluation. ABOi KT and HLAi KT patients had significantly longer pre-KT hospital stays than control group patients (P < 0.001),  www.nature.com/scientificreports www.nature.com/scientificreports/ while the length of post-KT hospital stays in the HLAi KT group was significantly longer than the stay lengths of both ABOi KT and control patients (P < 0.001).
The incidence of delayed graft function was significantly higher in the HLAi KT group than that of other groups (3.4% in controls, 1.9% in ABOi KTs, and 8.9% in HLAi KTs; P < 0.001). Fifty-four patients developed biopsy-proven acute rejection during their index hospitalization for KT [41 antibody-mediated rejection (AMR) and 13 T-cell medicated rejection (TCMR)]. The median time from KT to biopsy was 8 days [interquartile range (IQR), 5-14 days]. HLAi KT patients exhibited significantly higher rates of AMR than ABOi KT and control groups (19.0% in HLAi KTs, 4.4% in ABOi KTs, and 3.8% in controls, respectively; P < 0.001). We found no significant differences among the groups for TCMR incidence (P = 0.507). No patients died or experienced allograft loss during index hospitalization for KT.

Parameters of EHR.
During the study period, 96 recipients experienced EHR (13.1%). As shown in Fig. 2, 55 patients (11.1%) in the control group, 20 patients (12.7%) in the ABOi KT group, and 21 patients (26.6%) in the HLAi KT group experienced EHR (P = 0.002). The median time from transplant discharge to EHR was 13 days (IQR, 6-22 days) with no significant differences seen among groups (P = 0.154). The median LOS for EHR was 8 days (IQR, 4-15 days) with a maximum of 149 days. Median LOS were similar in the control (8 days) and ABOi KT groups (7.5 days) but significantly longer in the HLAi KT group (12 days, P = 0.034).
Thirty-three patients underwent biopsies for acute allograft dysfunction during their EHR (n = 13, TCMR and n = 10, AMR). Most cases of rejection in the control group were diagnosed as TCMR. In contrast, AMR was found in 12.5%, 27.3%, and 83.3% of control, ABOi KT, and HLAi KT biopsied patients, respectively. There was no mixed rejection during EHR.
Mean estimated glomerular filtration rates (eGFR) of the three KT groups were similar for 12 months after KT, while patients with EHR had consistently lower eGFR during the same period compared to those without EHR (Fig. 4).
Cox regression analysis revealed advanced age (≥60 years) and HLAi KT as significant independent risk factors for EHR (Table 4). History of diabetes and cardiovascular disease, BMI, index hospitalization LOS, and ABOi KT status were not significantly associated with EHR.  Table 3. Causes of early hospital readmission.  www.nature.com/scientificreports www.nature.com/scientificreports/

Discussion
In this study, HLAi KT significantly increases the risk of EHR compared to ABOi KT and immunologically compatible KT. Additional risk factor for EHR was advanced age. Patients who experienced EHR experienced greater risk of late hospital readmission and death-censored graft loss and exhibited consistently lower eGFR compared to those without EHR during the first post-operative year.
KT recipients are vulnerable to EHR due to the large burden of comorbidities and lifelong immunosuppression. Among them, patients with high immunologic risks (ABOi or HLAi KT) receive desensitization and potent immunosuppression. Recent multi-center study reported that HLAi KT is associated with increased risk of hospital readmission compared to compatible KT 16 . Unlike prior studies, we included ABOi KT patients in order to evaluate the impact of desensitization to EHR. Interestingly, ABOi KT had a similar risk of EHR compared to compatible KT despite desensitization therapy. However, HLAi KT was associated with increased risk of EHR. The significantly higher risk of AMR in HLAi KT is a possible explanation for this result. In addition, intensified immunosuppressive therapy for AMR may further increase the risk of infectious complications, the most common cause of EHR in HLAi KT patients 21 . Our data suggest that the risk of EHR was associated with the high immunogenicity of HLAi KT patients, rather than desensitization therapy.
The incidence of EHR in our study (13.1%) was considerably lower than that found in a large population-based United States study (31%) 7 . A recent Canadian population-based study also showed a higher EHR incidence (20.7%) than in our patients 8 ; however we analyzed living donor KT patients, who typically have a lower risk for EHR. In addition, differences in healthcare systems and LOS across countries could account for variations in EHR incidence rates 22,23 . Because patients in Korea are in the hospital for approximately 2 weeks following KT, adverse events occurring during this period can be treated without affecting incidence of EHR. Long LOS is common in the Korean medical environment, which has a sufficient number of beds and a national health insurance system 24 . According to OECD reports, the average LOS for all causes in Korea is 2.1 times longer than the OECD average, 2.6 times longer than the US 25 .
Similar to prior studies, we found that advanced age is associated with EHR 7 , as elderly recipients are more likely to have comorbidities at the time of KT than younger patients 14,26 . In addition, older patients are more susceptible to infections due to progressive decline in immune function 27 . Therefore, elderly recipients should be managed especially carefully, as they are more vulnerable to adverse events requiring EHR.
Our study extended the previous work on EHR and poor outcomes including late readmission and death-censored graft loss. To our knowledge, our study was first to describe the outcomes following EHR in incompatible KT patients. McAdams-DeMarco et al. demonstrated that EHR was associated with late readmission, graft loss, and mortality 9 . In this study, EHR was associated with an increased risk of late readmission   www.nature.com/scientificreports www.nature.com/scientificreports/ and death-censored graft loss but not mortality. A noteworthy finding in our study involves the impact of EHR on outcomes with respect to immunological compatibility status. In addition, this is the first study to compare renal function according to EHR. Although population-based data contain highly accurate information about admission, graft failure, and mortality, they lack granular detail about immunologic risk, biopsy results, and renal function, which are vital when determining care of KT recipients.
EHR often represents an adverse event, results in increased cost of care, and has become an indicator of healthcare quality 6 . However, emerging evidence suggests that EHR is largely driven by patient characteristics 28 . In the transplant field, immunologic risks exert a meaningful effect on risk of EHR 17 , and advanced age increases risk of readmission because of comorbidities and compromised immunity. Despite the higher risk of EHR, incompatible KT continues to demonstrate a significant survival benefit compared with remaining on dialysis 5,26 . Consequently, EHR does not necessarily result from poor quality of care, but rather represents the best possible care for these most vulnerable patients 16,29 .
Our present study has some limitations that warrant consideration. First, our single-center design limits the generalizability of this study. However, this made it possible to maintain homogeneity in the desensitization protocol and immunosuppressive regimen. Second, certain recipient risk factors previously found to be predictors of EHR (e.g., frailty, chronic obstructive pulmonary disease) were not included in our analysis, necessitating further studies to understand why these patients in particular require hospital readmission.
In conclusion, we found that HLAi KT and advanced age increase the risk of EHR after living donor KT. EHR was also associated with inferior renal function, late hospital readmission, and death-censored graft loss. Thus, KT recipients who are at risk for EHR should be managed with great care, including monitoring of DSA, renal function, and infectious complications. Definitions. EHR was defined as at least one readmission to the transplant hospital within 30 days after discharge from the index hospitalization 7 . The primary reason for EHR was ascertained by chart review. Late hospital readmission was defined as that occurring during the year after the EHR window 9 .

Study population.
Desensitization and immunosuppression. All ABOi or HLAi KTs were performed according to the desensitization protocol as described previously 30 . Briefly, rituximab was administered within 7 days before kidney transplantation in cases of ABOi or HLAi KT. Plasmapheresis was performed until the target antibody titer (IgG titer ≤ 1:16 in ABOi KT, negative CDC T-cell crossmatch in HLAi KT) was achieved.
The maintenance immunosuppressive regimen was tacrolimus, prednisolone, and mycophenolate mofetil (MMF). MMF was administered 1 week before transplantation in ABOi KT or HLAi KT cases. Some patients who experienced recurrent adverse events associated with MMF (gastrointestinal trouble or leukopenia) discontinued MMF at the clinician's discretion. We administered tacrolimus 1 day before or on the day of transplantation in all patients, regardless of immunologic risk. The doses of maintenance immunosuppression were administered per institutional protocols.
Postoperative care and renal biopsy. Before index discharge, all patients received comprehensive education and clearance from transplant team members. After discharge, patients visited the transplant clinic every week for laboratory assessments during the first post-transplant month.
Renal biopsies were performed in cases of acute allograft dysfunction (increased serum creatinine >30% from baseline or proteinuria >1 g/day). We did not perform routine protocol biopsies during the study period. All acute rejections were biopsy proven and classified into antibody-mediated rejection (AMR) or T-cell mediated rejection (TCMR) using the 2015 Banff criteria 31 . All biopsy specimens were stained for C4d.
Prophylaxis protocol and infection monitoring. All patients received trimethoprim-sulfamethoxazole as Pneumocystis jirovecii pneumonia prophylaxis (3-6 months). Fungal prophylaxis consisted of 4 mL of oral nystatin four times daily for 3-6 months. No prophylaxis against cytomegalovirus infection was administered to any patient. Instead, a preemptive approach was used in high-risk patients (donor-seropositive and recipient-seronegative).
Study endpoints. The primary study endpoint was the incidence of EHR. The secondary endpoints included causes and risk factors of EHR, late hospital readmission, death-censored graft survival, patient survival, and graft function. GFR was estimated using CKD-EPI (Chronic Kidney Disease Epidemiology Collaboration) equation 32 .
Statistical analysis. Data were expressed as frequencies (percentages), means and standard deviations, or medians and interquartile ranges, as appropriate. Chi-square or Fisher's exact tests were used to compare categorical variables. Continuous variables were compared using one-way analysis of variance (parametric data) with post hoc Bonferroni analysis or Kruskal-Wallis (nonparametric data) with post hoc analysis with Dunn's