Definition and Analysis of Textbook Outcome: A Novel Quality Measure in Kidney Transplantation

Abstract

BACKGROUND:

“Textbook outcome” (TO) is a novel composite quality measure that encompasses multiple postoperative endpoints, representing the ideal “textbook” hospitalization for complex surgical procedures. We defined TO for kidney transplantation using a cohort from a high-volume institution.

METHODS:

Adult patients who underwent isolated kidney transplantation at our institution between 2016 and 2019 were included. TO was defined by clinician consensus at our institution to include freedom from intraoperative complication, postoperative reintervention, 30-day intensive care unit or hospital readmission, length of stay >75^th percentile of kidney transplant patients, 90-day mortality, 30-day acute rejection, delayed graft function, and discharge with a Foley catheter. Recipient, operative, financial characteristics, and post-transplant patient, graft, and rejection-free survival were compared between patients who achieved and failed to achieve TO.

RESULTS:

A total of 557 kidney transplant patients were included. Of those, 245 (44%) achieved TO. The most common reasons for TO failure were delayed graft function (N=157, 50%) and hospital readmission within 30 days (N=155, 50%); the least common was mortality within 90 days (N=6, 2%). Patient, graft, and rejection-free survival were significantly improved among patients who achieved TO. On average, patients who achieved TO incurred approximately $50,000 less in total inpatient charges compared to those who failed TO.

CONCLUSIONS:

TO in kidney transplantation was associated with favorable post-transplant outcomes and significant cost-savings. TO may offer transplant centers a detailed performance breakdown to identify aspects of perioperative care in need of process improvement.

INTRODUCTION

Kidney transplantation (KTx) is the preferred treatment for patients with end-stage renal disease and the most commonly performed solid organ transplant in the US [1]. It has become a routine procedure with improvements in patient and graft survival over time and across institutions [1–3]. While patient and graft survival are monitored and reported, there is no national system that oversees perioperative quality [4]. The Transplant National Surgical Quality Improvement Project (Transplant-NSQIP) database may eventually fill this gap [5]; however, appropriate metrics to capture perioperative quality of care remain uncertain [6].

Presently, transplant center performance is evaluated primarily using one-year patient and graft survival [7,8]. Composite measures may be more useful as they capture multiple domains of overall surgical and hospital performance [9–11]. “Textbook” outcome (TO) is a novel way to define composite measurements that reflect these domains [12–15], as it includes multiple postoperative endpoints that may represent the ideal “textbook” hospitalization. This definition generally includes important markers of perioperative quality such as perioperative morbidity, mortality, early readmissions, and procedure-specific variables, such as margin status and lymph node retrieval for cancer operations.

TO may improve understanding of perioperative quality of care for patients and transplant programs and serve as a standardized metric to aid comparison and guide quality improvement across centers. TOs have been developed for several complex procedures, especially in surgical oncology [13,16–20]. Recently, we defined TO for liver transplantation [21]. We herein defined TO for KTx using a cohort from a high-volume institution and evaluated its ability to predict clinically and financially-relevant outcomes.

METHODS

Data sources and study population

We conducted a retrospective cohort analysis using institutional and United Network for Organ Sharing data. Adult (age≥18) patients who underwent isolated KTx at Duke University Hospital between 2016 and 2019 were included. This study was approved by our Institutional Review Board (Pro00103325).

Definition and impact of textbook outcome

TO was defined by clinician consensus at our institution to include freedom from intraoperative complication, 30-day reintervention (surgical, endoscopic, radiologic), 30-day intensive care unit or hospital readmission, length of stay (LOS) >75^th percentile of KTx patients, 90-day mortality, 30-day biopsy-proven acute rejection, delayed graft function (DGF), and discharge with a Foley catheter. Freedom from all listed complications constituted a TO.

Patient, graft, and rejection-free survival were compared between recipients who achieved and failed TO. As a sensitivity analysis, patient, graft, and rejection-free survival were compared among recipients with and without DGF as DGF has been associated with reduced patient and rejection-free survival [22].

Financial impact of textbook outcome

Financial data from the Duke Transplant Center was obtained for patients in the study cohort. Patient-level charge data was determined as the sum of charges billed from the date of transplant to 30 days post-discharge. Transplant-related charges were those billed to the transplant center during the same period.

Statistical analysis

Recipient, operative, and financial characteristics were compared between TO and non-TO groups using Wilcoxon rank-sum tests for continuous variables and Chi-squared and Fisher exact tests for categorical variables. Patient, graft, and rejection-free survival were estimated using the Kaplan-Meier method and compared between groups using log-rank tests.

Operative characteristics and outcomes of TO and non-TO groups were compared in aggregate and separately for living donor (LDKT) and deceased donor KTx (DDKT) recipients. The association between TO and donor type was explored using analysis of variance (ANOVA) comparing rates of TO among living, donation after brain death (DBD), and donation after circulatory death (DCD) donor kidney recipients; post-hoc analyses were conducted using pairwise student t-tests for independent samples.

Logistic regression was used to investigate associations between recipient and operative characteristics and TO. Training and validation datasets were developed by randomly sampling 2/3 and 1/3 of the dataset without replacement (Table S1). Model selection was performed on the training dataset using backward stepwise regression with an AIC criterion. The final model was used to develop a nomogram to predict the probability of TO [23]. Model discrimination was assessed using the c-statistic with 95% confidence intervals (CI) determined based on 1000 bootstrapped resamples. A two-sided p-value less than 0.05 was considered statistically significant. All analyses were performed using R version 3.6.1 (Vienna, Austria).

RESULTS

Patient and operative characteristics

Of 557 KTx recipients, 245 (44%) achieved TO. Patients who achieved TO were more likely to be female and less likely to be Black. Additional patient characteristics were similar between groups (Table 1).

Table 1.

Recipient characteristics stratified by achievement of a textbook outcome (TO).

Characteristica	Achieved TO N = 245 (44%)	Failed TO N = 312 (56%)	P-value
Age (years)	52 (43–62)	54 (44–62)	0.4
Sex			<0.01
Female	123 (50%)	116 (37%)
Male	122 (50%)	196 (63%)
Race			0.013
White	124 (51%)	116 (37%)
Black	107 (44%)	173 (55%)
Asian	4 (2%)	4 (1%)
Other	10 (4%)	19 (6%)
Ethnicity (Hispanic)	6 (2%)	12 (4%)	0.4
Body mass index (kg/m2)	28.9 (24.8–33.0)	28.6 (24.6–33.2)	0.8
Panel reactive antibody at transplant (%)b
Class I	0 (0–13)	0 (0–12)	0.8
Missing	0 (0%)	0 (0%)
Class II	0 (0–0)	0 (0–0)	0.8
Missing	0 (0%)	1 (0.3%)
Etiology of kidney disease			0.6
Alport syndrome	4 (2%)	4 (1%)
Calcineurin inhibitor nephrotoxicity	11 (5%)	14 (5%)
Chronic glomerulonephritis	11 (5%)	11 (4%)
Chronic nephrosclerosis	1 (0.4%)	1 (0.3%)
Congenital obstructive uropathy	8 (3%)	6 (2%)
Focal glomerular sclerosis	28 (11%)	25 (8%)
HIV nephropathy	2 (0.8%)	2 (0.6%)
Hypertensive nephrosclerosis	45 (18%)	54 (17%)
IgA nephropathy	18 (7%)	15 (5%)
Lithium toxicity	3 (1%)	0 (0%)
Malignant hypertension	3 (1%)	5 (2%)
Nephrolithiasis	3 (1%)	0 (0%)
Polycystic kidney disease	21 (9%)	32 (10%)
Renal cell carcinoma	1 (0.4%)	1 (0.3%)
Sickle cell anemia	2 (0.8%)	2 (0.6%)
Systemic lupus erythematosus	9 (4%)	12 (4%)
Type 1 diabetes mellitus	7 (3%)	13 (4%)
Type 2 diabetes mellitus	48 (20%)	82 (26%)
Vasculitis	1 (0.4%)	3 (1%)
Other	19 (8%)	29 (9%)
History of prior transplant	37 (15%)	46 (15%)	0.9
Kidney	29 (12%)	38 (12%)	0.9
Liver	3 (1%)	4 (1%)	>0.9
Lung	2 (0.8%)	3 (1%)	>0.9
Heart	4 (2%)	3 (1%)	0.7
History of diabetes	80 (33%)	124 (40%)	0.08
Estimated Post-Transplant Survival Score (%)	36 (16–58)	46 (24–72)	<0.01
Pre-transplant dialysis	178 (73%)	273 (88%)	<0.01
Dialysis duration (years)	2.54 (1.27–4.20)	3.83 (1.99–5.75)	<0.01
Induction immunosuppression			<0.01
Basiliximab	38 (16%)	72 (23%)
Anti-thymocyte globulin	89 (36%)	114 (37%)
Campath	28 (11%)	17 (5%)
High-dose corticosteroids	67 (27%)	91 (29%)
Other	12 (5%)	13 (4%)
None	11 (5%)	4 (1%)
Missing	0 (0%)	1 (0.3%)
Initial maintenance immunosuppression
Tacrolimus	216 (88%)	284 (91%)	0.3
Cyclosporine	1 (0.4%)	3 (1%)	0.6
Mycophenolate mofetil	213 (87%)	286 (92%)	0.07
Azathioprine	2 (0.8%)	1 (0.3%)	0.6
Sirolimus	27 (11%)	20 (6%)	0.052
Belatacept	30 (12%)	22 (7%)	0.04
Corticosteroids	217 (89%)	289 (93%)	0.1
Other	1 (0.4%)	2 (0.6%)	>0.9

^aPresented as median (interquartile range) for continuous variables and frequency (proportion) for categorical variables.

^bMost recent pre-transplant panel reactive antibody levels.

Failure to achieve TO was associated with longer ischemic times, greater intraoperative blood loss, use of machine perfusion, receipt of a DCD donor kidney, higher Kidney Donor Profile Index (KDPI), increased human leukocyte antigen (HLA) mismatches, and intraoperative ureteral stent and drain placement. Patients who failed TO were less likely to have received living donor kidneys (Table 2).

Table 2.

Operative characteristics among all kidney transplant recipients, stratified by achievement of a textbook outcome (TO).

Characteristica	Achieved TO N = 245 (44%)	Failed TO N = 312 (56%)	P-value
Cold ischemic time (minutes)
All kidneys	868 (109–1320)	1032 (531–1460)	<0.01
Non-pumped kidneys only	107 (66–272)	133 (81–692)	0.03
Warm ischemic time (minutes)	28 (23–32)	29 (24–33)	0.03
Missing	0 (0%)	1 (0.3%)
Total ischemic time (minutes)	915 (136–1348)	1058 (540–1492)	<0.01
Missing	0 (0%)	1 (0.3%)
Estimated blood loss (mL)	100 (100–200)	150 (100–200)	<0.01
Missing	1 (0.4%)	3 (1.0%)
Transfusion requirement (units)
Packed red blood cells	0 (0–0)	0 (0–0)	0.2
Missing	0 (0%)	1 (0.3%)
Fresh frozen plasma	0 (0–0)	0 (0–0)	0.2
Missing	0 (0%)	0 (0%)
Machine perfusion used	127 (52%)	199 (64)	<0.01
Donor type
Living donor	89 (36%)	74 (24%)	<0.01
Donation after brain death donor	121 (50%)	166 (53%)	0.4
Donation after circulatory death donor	35 (14%)	72 (23%)	<0.01
US Public Health Service increased risk for disease transmission donorb	63 (26%)	70 (22%)	0.02
Extended criteria donorb	19 (8%)	36 (12%)	0.4
Donor serologies
Epstein-Barr virus positive	235 (96%)	297 (95%)	0.9
Missing	0 (0%)	2 (0.6%)
Cytomegalovirus positive	141 (58%)	185 (59%)	0.7
Kidney Donor Profile Index (%)b	48 (26–66)	56 (34–71)	0.048
Human leukocyte antigen mismatch	4 (3–5)	4 (3–5)	<0.01
Missing	1 (0.4%)	0 (0%)
Graft laterality			0.056
Right	83 (34%)	134 (43%)
Left	160 (65%)	173 (55%)
Dual	2 (0.8%)	5 (2%)
Ureteral stent used	147 (60%)	224 (72%)	<0.01
Drain placed in the operating room	46 (19%)	84 (27%)	0.02

^aPresented as median (interquartile range) for continuous variables and frequency (proportion) for categorical variables.

^bRecorded for deceased donors only.

Operative characteristics stratified by TO among LDKT and DDKT recipients are shown in Tables S2 and S3, respectively. Rates of TO were different among living, DBD, and DCD donor kidney recipients (ANOVA p<0.01). Recipients of living donor kidneys were more likely to achieve TO than recipients of DBD and DCD donor kidneys. Rates of TO were similar among DBD and DCD donor kidney recipients (Table S4).

Prevalence of events determining failure to achieve textbook outcome

Among patients who failed TO (N=312), the most common reasons for TO failure were DGF (50%) and hospital readmission (50%); the least common was mortality (2%) (Table 3; Figure 1).

Table 3.

Prevalence of events leading to textbook outcome (TO) failure among patients who failed TO^a

Characteristic (N = 312)	N (%)
Intraoperative Complicationb	10 (3%)
Reintervention	64 (21%)
Reoperation	45 (14%)
Radiologic Intervention	23 (7%)
Intensive Care Unit Readmission Within 30 Days	32 (10%)
Hospital Readmission Within 30 Days	155 (50%)
Post-Transplant Length of Stay >75th Percentile	136 (44%)
Mortality Within 90 Days	6 (2%)
Delayed Graft Function	157 (50%)
Discharged With Foley Catheter	25 (8%)
Acute Rejection Within 30 Days	30 (10%)

^aFailure to achieve TO was defined by the occurrence of any of the above events, however each patient who failed TO could have had multiple events therefore may be counted more than once in this table.

^bIntraoperative complications included renal artery thrombosis, renal artery spasm with resultant kidney allograft ischemia, severe hyperkalemia (K>8), and cardiac arrhythmia (supraventricular tachycardia: one case; asystole: one case).

Figure 1. Individual feature prevalence and cumulative achievement of textbook outcome (TO).

The bars represent individual features of TO ranked by severity from left to right. The bar heights correspond to freedom from those complications (i.e., taller bars indicate less complication). The black dotted line represents the cumulative achievement of TO as each feature is added. It decreases with each feature added as more patients fail to achieve TO with more specifications that must be met.

Financial implications of textbook outcome

Patients who achieved TO incurred approximately $50,000 less in total inpatient charges. Transplant-related charges accounted for most inpatient charges but were still significantly lower among patients who achieved TO (Table S5; Figure 2).

Figure 2. Financial implications of textbook outcome (TO).

(a) Total inpatient charges during the index hospitalization. (b) Transplant-related inpatient charges during the index hospitalization.

Patient, graft, and rejection-free survival

Patient, graft, and rejection-free survival were decreased among patients who failed TO (Figure 3). In our sensitivity analysis, patient survival was similar between DGF and non-DGF groups; graft and rejection-free survival were decreased among patients with DGF compared to those without (Figure S1).

Figure 3. Kaplan-Meier survival analysis of the entire patient cohort stratified by achievement of textbook outcome (TO).

(a) Patient survival. (b) Graft survival. (c) Rejection-free survival.

Among LDKT recipients, patient and graft survival were similar between TO and non-TO groups; rejection-free survival was decreased among patients who failed TO (Figure S2). Among DDKT recipients, patient survival was similar between TO and non-TO groups; graft and rejection-free survival were decreased among those who failed TO (Figure S3).

Prediction of the probability of achieving a textbook outcome

Recipient sex, donor type, and number of HLA mismatches were independently associated with achievement of TO. Male sex was associated with 46% decreased odds of achieving TO (odds ratio [OR] 0.54, 95% CI 0.38–0.77, p<0.01); each HLA mismatch was associated with 16% decreased odds of achieving TO (OR 0.84, 95% CI 0.75–0.94, p<0.01). Receiving a DCD donor kidney decreased odds of achieving TO by 40% compared to receiving a DBD donor kidney (OR 0.60, 95% CI 0.37–0.97, p=0.042).

Recipient and operative characteristics independently associated with achievement of TO were used to construct a nomogram to predict the probability of TO (Figure 4). The logistic regression model was used to assign each factor in the nomogram a weighted point value. A patient’s probability of achieving TO is estimated by calculating the sum of points and determining its probability correlate on the nomogram. For example, a male patient (0 points) receiving a DBD donor kidney (47.5 points) with four HLA mismatches (32.5 points) would score 80 points, correlating to a TO probability of approximately 40%. A female patient (57.5 points) receiving a living donor kidney (85 points) with one HLA mismatch (82.5 points) would score 225 points corresponding to a TO probability of approximately 70%. The nomogram c-index was 0.687 (95% CI 0.611–0.729) and 0.620 (95% CI 0.525–0.661) for the training and validation sets, respectively.

Figure 4.

Nomogram for the prediction of the probability of achieving a textbook outcome (TO) after kidney transplantation (KTx).

DISCUSSION

In this study, we introduced TO as a novel composite quality index for KTx and demonstrated a strong association with post-transplant outcomes and cost. To define TO, we emulated existing models which include outcomes such as morbidity, mortality, LOS, and readmission [18,20,25]. We expanded the definition with KTx-specific metrics, defined by consensus among transplant clinicians at our institution.

44% of KTx patients achieved TO, similar to rates of TO for other complex surgical procedures [13,16,20,26]. Factors associated with TO failure included longer ischemic times, ureteral stent placement, and use of grafts with high KDPI. Unsurprisingly, TO failure was associated with significantly increased inpatient charges.

Transplant centers are primarily evaluated based on one-year patient and graft survival. The Centers for Medicare and Medicaid Services and some private insurance companies use publicly-reported data to determine in-network eligibility, grant transplant center certification, and flag programs with poorer patient or graft survival [27,28]. However, these metrics may no longer be best-suited to advance KTx. Currently, one-year patient and graft survival exceed 95% [7,32] with declining center-level variability amidst rising KTx volume and increasing experience nationwide [33]. In this context, there is growing recognition that longer-term outcomes may more robustly differentiate center quality, and inform ongoing management of this complex patient population that increasingly survives beyond five or ten years post-transplant [32].

In addition to highlighting long-term outcomes [7,33], a multi-faceted short-term metric such as TO may provide critical insight into care processes that can further improve perioperative care. Specific short-term outcomes are particularly important to patients and their view of the healthcare experience. Patients may be more likely to view perioperative outcomes holistically, basing transplant center quality assessments on occurrence of any complication rather than weighing performance in one domain versus another [18,20]. Unlike long-term prognosis, immediate outcomes including need for additional operations, or return to dialysis and prolonged LOS due to DGF may more directly influence patients’ decision-making and quality assessments of transplant centers. Rather than offering a new means by which to assess patient and graft survival, TO may be best understood as a patient-centered means of facilitating understanding of transplant center quality, and secondarily informing center-level initiatives to align quality improvement with care processes of greatest relevance to patients to optimize patient-centered care in KTx [34].

In contrast to prior studies identifying prolonged LOS as the primary obstacle to achievement of TO [18,20], early readmission was a leading reason for TO failure in KTx. High rates of acute care utilization and hospital readmission early post-KTx have previously been reported, portending poor post-transplant outcomes and higher healthcare costs [35–38]. While it remains difficult to identify preventable readmissions, patient factors including older age, Black race, elevated body mass index, and presence of medical comorbidities are associated with increased rates of early readmission [35,38,39], suggesting that pre-operative identification of high-risk patients may facilitate pro-active prevention of readmission through improved patient education and outreach in the perioperative period [37]. Concurrently, care processes such as intraoperative ureteral stent and drain placement, and induction and maintenance immunosuppression management may represent modifiable areas through which centers can mitigate complications necessitating readmission. At our institution, ureteral stent and drain placement vary across surgeons, for some representing routine practice, and for others potentially reflecting operative complexity. Regardless, any intervention may entail complications, potentially accounting for a higher rate of TO failure among these patients. While our findings suggest that early readmission may be the primary reflection of complications related to intraoperative or perioperative management decisions at our center, multi-institutional studies are needed to further elucidate the impact of readmissions on rates of TO nationally, and associations with early care processes, as post-transplant emergency department utilization and hospital readmission vary across centers [35,37,38].

DGF was also prevalent among patients who failed TO. It is well-established that DGF portends increased risk of graft failure and acute rejection [32,40,41]. Its occurrence may be related to use of potentially high-risk kidneys such as those from older or DCD donors [40,41], raising concern that its inclusion in TO may unintentionally deter use of these grafts. In this context, however, DGF is more appropriately viewed as a marker of center-level aggressiveness in which high-risk grafts facilitate transplantation of more waitlisted patients. Accordingly, centers that frequently manage recipients of high-risk kidneys may be equipped to offset negative impacts of DGF to achieve acceptable outcomes including short LOS, which likely indicate streamlined processes for transition between hospital-based and community dialysis centers, and mitigation of other perioperative complications to achieve preserved post-transplant survival in spite of DGF’s occurrence. While ability to investigate these associations in an institutional study is limited, future work should investigate implications of DGF within TO, and its differential impact across centers.

Several patient and procedural factors were associated with TO in KTx. Unlike donor type and number of HLA mismatches, recipient sex cannot be altered to influence odds of achieving TO. Ours is not the first study to identify a non-modifiable patient factor as a significant predictor of TO, with prior work identifying age and sex as predictors of TO in hepatopancreatic and esophagogastric surgery [16,18,20,42]. Along with our proposed nomogram, which may help pre-emptively estimate patients’ probabilities of achieving TO, understanding non-modifiable factors associated with reduced odds of achieving TO may guide resource allocation to high-risk patients, inform pre-transplant decision-making, and facilitate targeted donor-recipient matching to maximize odds of TO for all patients. While our nomogram demonstrated modest predictive ability, examination of TO in a national study may facilitate development of a more granular predictive model to improve pre-transplant prognostication and identify additional targetable areas to optimize perioperative outcomes in KTx.

Our study has several strengths and limitations. This study is the first to define TO in KTx, using data from a modern cohort at a high-volume institution. The granularity of available data allowed inclusion of particularly relevant parameters in TO, which are unavailable in national datasets. However, the single institution, retrospective nature of our analysis may decrease generalizability, particularly regarding institution-dependent practices including immunosuppression management and intraoperative ureteral stent and drain placement. Likewise, as our study reflects recipient and donor populations at a single institution, we did not adjust for case-mix heterogeneity in our assessment of TO. Future multi-institutional studies should make appropriate provisions for risk-adjustment to account for differences in recipient and donor characteristics that may influence center-level rates of TO. While we included DGF in our definition of TO due to its implications for patient experiences and outcomes, risk-adjusted multi-institutional studies are necessary to better elucidate appropriateness of its inclusion, weighing outcome optimization against inadvertently heightened center-level risk aversion. Amidst ongoing prospective data collection through the Transplant-NSQIP database, our study may inform important donor, recipient, and outcome parameters that should be included to maximize future ability to expand upon and validate novel metrics such as TO. Finally, TO was defined based on clinician consensus at a single institution. While this is consistent with processes reported in the literature [15,18,20,25,43,44], an ideal definition of TO would come from international consensus among experts in the field. Nevertheless, we demonstrate that TO in KTx as herein defined is a promising multidimensional indicator.

CONCLUSIONS

In this institutional analysis, TO in KTx was associated with favorable short-term post-transplant outcomes, and significant cost-savings. Although it remains difficult to predict TO before KTx, this composite metric may offer a powerful parameter to assess between-hospital variation, compare quality between institutions or audits, and identify actionable areas for quality improvement in KTx. Validation of this novel quality metric in a multi-institution study is warranted.

ABBREVIATIONS

ANOVA

analysis of variance

CI

confidence interval

DBD

donation after brain death

DCD

donation after circulatory death

DDKT

deceased donor kidney transplant

DGF

delayed graft function

HLA

human leukocyte antigen

KDPI

kidney donor profile index

KTx

kidney transplantation

LDKT

living donor kidney transplant

LOS

length of stay

NSQIP

National Surgical Quality Improvement Project

OR

odds ratio

TO

textbook outcome