Clinical Indications for Necessary and Discretionary Hospital Readmissions after Radical Cystectomy

Abstract

Background:

To assess predictors, indicators and medical necessity of readmissions after neoadjuvant chemotherapy (NAC) and radical cystectomy in order to identify opportunities for reducing readmission rates.

Methods:

Records for patients treated with cisplatin-based NAC followed by radical cystectomy between 2007 and 2017 were reviewed for 90-day complications and readmission. Readmissions were classified as necessary vs. discretionary based on independent clinician review. The association between postoperative complications and necessary or discretionary readmission were examined with adjusted regression models.

Results:

Among a total of 250 patients, 76 patients (30.4%) were readmitted within 90 days of surgery (19 discretionary and 57 necessary). Age, insurance coverage, and comorbidity were similar between readmitted and non-readmitted patients. Readmission was more likely after neobladder than ileal conduit (39% vs. 23%, p=0.02). Major (grade ≥ 3) complications within 90-day of surgery including index admission and post-discharge period were significantly more common among re-admitted patients compared to patients who were not readmitted (40% in necessary, 21% in discretionary, 3% in none, p<0.001). Median length of stay on readmission was twice as long in necessary cases compared to discretionary cases (5 vs. 2.5 days, p<0.001). Gastrointestinal and infectious complications were associated with discretionary readmission in adjusted analyses, while infectious, renal/genitourinary and thromboembolic complications were associated with necessary readmission.

Conclusions:

Twenty-five percent of readmissions were categorized as discretionary and were driven primarily by low-grade gastrointestinal complications, marginal oral intake and failure to thrive, suggesting that better coordinated post-discharge supportive care could help avoid a substantial proportion of readmissions.

Introduction

While surgical quality and patient-centered outcomes have come into focus in the last decade, radical cystectomy remains a complex surgery that is often performed in older patients who have competing health issues that contribute to surgical frailty.¹ For example, 90-day complication rates exceed 50% and perioperative mortality range between 2% to 6%.^2,3 Furthermore, approximately 30% of cystectomy patients are readmitted within 90 days of surgery, some for serious complications that require intensive medical care.^4,5,6 These figures make cystectomy the most complication-prone cancer surgery when compared to other complex procedures such as esophagectomy, pancreatectomy, thoracic resection.⁷

Radical cystectomy has been subject to extensive examination in efforts to improve recovery after surgery.⁸ In addition to previous research examining risk factors for postoperative complications, interventions such as enhanced recovery after surgery (ERAS), nutritional optimization, and preoperative exercise programs have been implemented to mitigate adverse recovery courses.^9–11 While such interventions have improved perioperative care and decreased index length of stay (LOS), they have not significantly impacted hospital readmissions.¹² Minimally invasive surgery has also failed to decrease rates of complications or readmission among cystectomy patients.^13,14 Even quality of care measures that are associated with reduced perioperative morbidity and mortality such as high case volume have not been clearly linked to decreased readmission rates.^15–17

The degree to which readmission risk is modifiable is not well defined, but some readmissions are likely avoidable. One study, for example, estimated that 26% of 30-day readmissions after cystectomy may have been avoided.¹⁸ However, indications, decision-making, and thresholds to readmit patients can be nuanced and vary among physicians.¹⁹ Though prior research has defined the risk and common reasons associated with readmission after cystectomy, less work has focused on assessing the clinical condition, symptom severity and medical necessity prompting readmission. The objective of this study was to assess reasons for and necessity of readmission after radical cystectomy through a quality assessment framework and using implicit physician case review. The results of such efforts may help identify opportunities to improve in this phase of post-cystectomy care.

Materials and Methods

Patients treated with neoadjuvant cisplatin-based chemotherapy followed by radical cystectomy between 2007 and 2017 at a high-volume tertiary referral center were included in the study. Patients were excluded if they did not have muscle-invasive bladder cancer or did receive other neoadjuvant chemotherapy regimens (Supplementary figure 1). Relevant demographic and clinical characteristics, treatment information, perioperative complications and readmissions were identified by querying two prospectively maintained institutional data registries, including a departmental cystectomy database and a center-wide health and research informatics resource, followed by review of clinical records using a standardized chart review process. Study compliance and regulation were overseen by the Moffitt Cancer Center Scientific Review Committee and Institutional Review Board (IRB# MCC16525).

Clinical staging was determined by transurethral resection pathology, exam under anesthesia and preoperative radiological imaging. Cystectomy, including pelvic lymph node dissection, was performed using standard approaches and surgical principles. The extent of pelvic lymph node dissection was left to the surgeon’s discretion. Health conditions contributing to patient comorbidity were captured from chart review of documented health problems and administrative diagnosis codes and codified according the a modified Charlson comorbidity index defined by Klabunde et al.²⁰ All postoperative complications that occurred during index admission and post-discharge period within 90 days of surgery were included in data collection. Postoperative complications within 90 days after surgery were graded according to the Clavien-Dindo classification system,²¹ and classified by organ system as previously described.² Gastrointestinal complications consisted of ileus, Clostridium difficile colitis, gastrointestinal bleed, and anastomotic bowel leak. Infectious complications consisted of fever, urinary tract infection, and pelvic effusion/abscess. Wound complications consisted of wound infection and wound dehiscence. Renal and genitourinary complications consisted of acute kidney injury, ureteral obstruction, and ureteral leak. Cardiopulmonary complications consisted of arrhythmia and myocardial infarction. Thromboembolic complications consisted of deep vein thrombosis and pulmonary embolism. Hemorrhagic complications consisted of anemia requiring blood transfusion. Neurologic complications consisted of delirium, agitation, cerebrovascular accident and/or transient ischemic attack. Grade ≥ 3 were categorized as major complications.

Readmissions were categorized as necessary or discretionary using structured implicit chart review.²² This process directs physicians to review and analyze specific parts of the medical record in order to assess and rate patient presentation, processes of care and outcomes using their own criterion based on clinical experience, knowledge and judgement. Any clinical indications that led to readmissions, including postoperative complications, were reviewed independently by two clinical urologic oncology fellows (AMA, SKC) and readmissions were categorized based on judgement of the indication and clinical need for inpatient hospitalization and acute management regardless of the root cause, complication or patient symptom. Disagreements in categorizing readmissions were resolved by a consensus review with a senior urologic oncologist (SMG). Necessary readmissions were generally defined by symptoms or conditions that would always necessitate inpatient management based on the severity of illness, associated comorbidities, and needed medical intervention. In contrast, discretionary readmissions were defined by conditions for which alternative ambulatory services are available and could reasonably be employed in order to avoid acute inpatient rehospitalization.

The objective of the analysis was to examine the association between type and grade of postoperative complications (that occurred during index stay and/or outpatient post-discharge period within 90 days of surgery) with the probability of discretionary or necessary readmission, accounting for other risks for readmission. Associations between demographic and clinical characteristics, and outcome measures were tested using Chi-square, Fisher’s exact and Mann-Whitney U tests. Separate logistic regression models were used to estimate odds ratios and 95% CIs for readmission type (dependent variable) according to postoperative complications (independent variable). Adjusted models were constructed to account for age, comorbidity, insurance, pathological stage and urinary diversion (UD) type. P-values were 2-sided and p<0.05 was considered statistically significant, and statistical analyses were performed with SAS statistical software (version 9.4; SAS Institute Inc).

Results

A total of 250 patients were included for the analysis. Median age was 66 years (range 33–82). Most patients (89.6%) received ≥3 cycles of NAC, 203 (81.2%) received GC (gemcitabine, and cisplatin) and 47 (18.8%) received ddMVAC (dose-dense methotrexate, vinblastine, doxorubicin, and cisplatin). One hundred sixty-four (66%) patients received an ileal conduit, while 64 (26%) received a neobladder and 22 (9%) received a catheterizable colon pouch or alternative UD. Twenty one (8%) and 133 (53%) patients experienced major or any grade complications, respectively, and 54 (21.6%) patients were readmitted within the first 30 days after cystectomy.

Within the first 90-days after cystectomy, a total of 76 (30.4%) patients were readmitted, of which 57 (75%) were classified as necessary and 19 (25%) that were classified as discretionary (Table 1). The most common indication for discretionary readmissions were gastrointestinal problems followed by dehydration and fatigue/failure to thrive (Table 2). Readmissions were significantly more likely following neobladder than after ileal conduit (39% vs. 23.7%, p=0.02). Median LOS on readmission was twice as long in necessary cases compared to discretionary cases (5 vs. 2.5 days, p<0.001), although median LOS of the index hospitalization were similar (7 days for each group, p=0.09).

Table 1.

Demographic and clinical variables of 250 patients treated with cisplatin-based neoadjuvant chemotherapy and radical cystectomy by 90-day readmission status.

Variable	Readmission status			p value
	None (N=174)	Discretionary (N=19)	Necessary (N=57)
Median age, (range), years	66 (42–80)	64 (48–81)	63 (33–82)	0.19
Gender, n (%)				0.35
Female	44 (25.3)	5 (26.3)	20 (35.1)
Male	130 (74.7)	14 (73.7)	37 (64.9)
Race, n (%)				0.48
White	167 (95.9)	18 (94.7)	54 (94.7)
Black	5 (2.9)	1 (5.3)	1 (1.8)
Other	2 (1.2)	0 (0)	2 (3.5)
Ethnicity, n (%)				0.73
Not Hispanic or Latino	171 (98.3)	19 (100)	55 (96.5)
Hispanic or Latino	3 (1.7)	0 (0)	2 (3.5)
Marital status, n (%)				0.94
Single	22 (12.6)	3 (15.8)	5 (8.8)
Married/partnered	124 (71.3)	13 (68.4)	41 (71.9)
Divorced/separated	15 (8.6)	2 (10.5)	5 (8.8)
Widowed	13 (7.5)	1 (5.3)	6 (10.5)
Education				0.07
High school or less	40 (23.0)	2 (10.5)	14 (24.6)
College or some college	47 (27.0)	9 (47.4)	13 (22.8)
Graduate/professional degree	20 (11.5)	0 (0)	2 (3.5)
Unknown/missing	67 (38.5)	8 (42.1)	28 (49.1)
Insurance, n (%)				0.13
Medicare	101 (58.0)	9 (47.4)	21 (36.8)
Private	54 (31.0)	7 (36.8)	25 (43.9)
Medicaid	10 (5.8)	2 (10.5)	4 (7.0)
Self-paying/uninsured/other	9 (5.2)	1 (5.3)	7 (12.3)
Charlson comorbidity score, n (%)				0.62
0–2	69 (39.7)	6 (31.6)	22 (38.6)
3–5	53 (30.5)	8 (42.1)	14 (24.6)
>5	52 (29.8)	5 (26.3)	21 (36.8)
Clinical stage, n (%)				0.19
Stage II (T2N0M0)	130 (74.7)	17 (89.4)	37 (64.9)
Stage III (T3N0M0)	32 (18.4)	1 (5.3)	12 (21.1)
Stage IV (T4N0M0)	12 (6.9)	1 (5.3)	8 (14.0)
Pathological stage, n (%)				0.13
Stage 0/I (T0/Ta/Tis/T1N0M0)	68 (39.1)	14 (73.7)	21 (36.8)
Stage II (T2N0M0)	18 (10.4)	0 (0)	6 (10.5)
Stage III (T3N0M0)	38 (21.8)	2 (10.5)	11 (19.4)
Stage IV (T4N0–3M0–1)	50 (28.7)	3 (15.8)	19 (33.3)
Diversion type, n (%)				0.029
Ileal conduit	125 (71.7)	10 (52.6)	29 (50.9)
Neobladder	39 (22.4)	6 (31.6)	19 (33.3)
Pouch	3 (1.7)	2 (10.5)	3 (5.3)
Othera	7 (4.2)	1 (5.3)	6 (10.5)
Surgical approach, n (%)				0.24
Open	144 (82.8)	19 (100)	49 (86.0)
Laparoscopic/Robotic	30 (17.2)	0 (0)	8 (14.0)
Index length of stay
Median (Range)	7 (2–30)	7 (5–29)	7 (4–43)	0.09

^aCutaneous ureterostomy in 12 patients, sigmoid conduits in 2 patients.

Table 2.

Clinical indications for discretionary 90-day readmission in 19 patients.

Indication	No. of patients
Abdominal pain after stent removal	1
Constipation	2
Nausea ± vomiting	3
Dehydration ± mild electrolyte/glucose imbalance*	3
Presumed urinary tract infection, undocumented	3
Fatigue ± failure to thrive	3
Abdominal or pelvic fluid collection	2
Education ± care for maintenance of continent urinary diversion	2

^*One patient with clinical signs/symptoms of dehydration only, one patient with mild hypoglycemia, one patient with mild hyponatremia, mild hyperglycemia without increase in serum creatinine levels.

Within the first 90-days after cystectomy, 33 (13%) and 144 (58%) patients had major and any grade postoperative complications, respectively (Table 3). Gastrointestinal, infectious, renal/genitourinary and thromboembolic complications were significantly more common among re-admitted patients compared to patients who were not readmitted. The frequency of postoperative major complications was higher in necessary cases (40.4%) compared to discretionary cases (21.1%), albeit not statistically significant.

Table 3.

90-day postoperative complications in 250 patients stratified by readmission status.

		Readmission Status
Complications	All (N=250)	None (N=174)	Discretionary (N=19)	Necessary (N=57)	P value
Any complications	144 (58%)	77 (44%)	14 (74%)	53 (93%)	<0.001
Major (≥ grade 3)	33 (13%)	6 (3%)	4 (21%)	23 (40%)	<0.001
Gastrointestinal	42 (17%)	24 (14%)	7 (37%)	11 (19%)	0.03
Major (≥ grade 3)	5 (2%)	1 (1%)	0	4 (7%)	0.02
Ileus	31 (12%)	18 (10%)	7 (37%)	6 (11%)
C difficile colitis	14 (6%)	9 (5%)	1 (5%)	4 (7%)
Gastrointestinal bleed	1 (<1%)	0	0	1 (2%)
Anastomotic bowel leak	2 (1%)	0	0	2 (4%)
Infectious	55 (22%)	17 (10%)	6 (32%)	32 (56%)	<0.001
Major (≥ grade 3)	17 (7%)	3 (1%)	2 (11%)	12 (21%)	<0.001
Fever	25 (10%)	7 (4%)	1 (5%)	17 (30%)
Urinary tract infection	22 (9%)	10 (6%)	3 (16%)	9 (16%)
Pelvic effusion/abscess	10 (4%)	0	2 (11%)	8 (14%)
Wound	30 (12%)	18 (10%)	3 (16%)	9 (16%)	0.42
Major (≥ grade 3)	2 (1%)	0	0	2 (4%)	0.09
Wound infection	24 (10%)	15 (9%)	1 (6%)	8 (14%)
Wound dehiscence	7 (3%)	3 (2%)	2 (11%)	2 (4%)
Renal/Genitourinary	27 (11%)	9 (5%)	2 (11%)	16 (28%)	<0.001
Major (≥ grade 3)	11 (4%)	1 (1%)	1 (5%)	9 (16%)	<0.001
Acute kidney injury	9 (4%)	4 (2%)	0	5 (9%)
Ureteral obstruction	4 (2%)	1 (1%)	0	3 (5%)
Ureteral leak	14 (6%)	4 (2%)	2 (11%)	8 (14%)
Cardiopulmonary	15 (6%)	9 (5%)	2 (11%)	4 (7%)	0.45
Major (≥ grade 3)	2 (1%)	1 (1%)	1 (5%)	0	0.20
Arrhythmia	14 (6%)	8 (5%)	2 (11%)	4 (7%)
Myocardial infection	2 (1%)	1 (1%)	0	1 (2%)
Thromboembolic	11 (4%)	2 (1%)	1 (5%)	8 (14%)	<0.001
Major (≥ grade 3)	3 (1%)	0	1 (5%)	2 (4%)	0.004
Deep vein thrombosis	8 (3%)	1 (1%)	1 (5%)	6 (11%)
Pulmonary embolism	3 (1%)	1 (1%)	0	2 (4%)
Hemorrhagic	46 (18%)	28 (16%)	4 (21%)	14 (25%)	0.34
Major (≥ grade 3)	0
Neurological	5 (2%)	2 (1%)	1 (5%)	2 (4%)	0.19
Major (≥ grade 3)	0
Delirium/Agitation/CVA/TIA	5 (2%)	2 (1%)	1 (5%)	2 (4%)
Psychological illness	2 (1%)	1 (1%)	0	1 (2%)	0.52
Major (≥ grade 3)	0

Logistic regression analyses adjusted for age, comorbidity, insurance coverage, pathological stage and UD type showed that gastrointestinal complications were independently associated with discretionary readmission (OR 4.72, 95% CI 1.42–15.73, p=0.01) whereas renal/genitourinary and thromboembolic complications were independently associated with necessary readmission (OR 8.58, 95% CI 3.26–22.56, p<0.001 and OR 15.91, 95% CI 2.95–85.85, p=0.001, respectively) (Table 4). Infectious complications were independently associated with readmission regardless of readmission type (OR 7.82, 95% CI 2.13–28.70, p=0.002 for discretionary and OR 13.67, 95% CI 6.09–30.69, p<0.001 for necessary readmission).

Table 4.

Adjusted regression analysis for prediction of necessary and discretionary 90-day readmissions by postoperative complications.

	Discretionary readmission*			Necessary readmission*
Variable**	OR***	95% CI	P value	OR***	95% CI	P value
Any type of complication	4.37	1.32–14.49	0.02	24.70	7.46–81.85	<0.001
Any type of major complication	12.96	2.60–64.70	0.002	40.26	12.34–131.42	<0.001
Any GI complication	4.72	1.42–15.73	0.01	1.76	0.75–4.13	0.19
Any infectious complication	7.82	2.13–28.70	0.002	13.67	6.09–30.69	<0.001
Any wound complication	1.83	0.35–9.45	0.47	1.60	0.61–4.17	0.34
Any renal/GU complication	1.88	0.28–12.76	0.52	8.58	3.26–22.56	<0.001
Any cardiopulmonary complication	3.55	0.57–21.96	0.17	1.43	0.38–5.42	0.61
Any thromboembolic complication	-	-		15.91	2.95–85.85	0.001
Any hemorrhagic complication	1.39	0.39–4.93	0.61	2.05	0.93–4.50	0.07
Any neurological complication	6.98	0.48–100.87	0.15	2.92	0.35–24.51	0.32

^*Discretionary readmission vs. No readmission; Necessary readmission vs. No readmission. Adjusted for age (continuous), Charlson comorbidity score (1–2, 3–5, >5 comorbidities), insurance (Medicaid, Medicare, private, self-pay/uninsured), pathological stage (0/1, 2, 3, 4) and urinary diversion type (ileal conduit, neobladder, pouch, and other).

^**Each variable is adjusted for a total of 6 variables (age, Charlson comorbidity score, insurance, pathological stage and urinary diversion), and each row represents a separate adjusted regression model.

^***Referent is patients without complications.

Discussion

Complications account for up to 37% of the total costs committed to bladder cancer management,²³ and total 90-day costs of care are significantly higher among Medicare beneficiaries who are readmitted after cystectomy (median $45,000) compared to those who are not ($26,000).⁷ Nonetheless, applicable and actionable information regarding reasons for and patterns of readmission following cystectomy is scant. We explored factors leading to necessary and discretionary readmission and found that a quarter of readmissions after radical cystectomy could be classified as discretionary. Discretionary readmissions represent unplanned health encounters that could be managed in outpatient settings, potentially avoiding readmission and inpatient healthcare services and costs.

Reducing unnecessary readmissions after surgery through better care coordination and outpatient support has become a major healthcare priority in recent years. For example, the 2013 Hospital Readmission Reduction Program of the Affordable Care Act established financial penalties for hospitals with higher-than-expected 30-day readmission rates for specific conditions such as acute myocardial infarction, heart failure, and pneumonia.²⁴ Though cancer surgeries are currently not included in readmission reduction programs, policy efforts that target quality of care standards and healthcare spending among cancer patients treated with systemic therapy are in place.²⁵ Radical cystectomy might come under scrutiny under expanded policies in the future. Despite ground-level efforts to improve care and decrease complications after surgery, readmissions remain problematic and stubbornly common. Skolarus et al. recently suggested implementing regular check-ins within the first 2 weeks of surgery to help identify and proactively manage patients.²⁶ Daneshmand et al. reported on intravenous hydration at home and strict follow-up telehealth assessments from nursing teams to support at-home recovery after cystectomy.¹⁰ Likewise, we call for further research to optimize postoperative follow-up care based on our study findings, since there appears to be considerable room for improvement.

We found that 90-day readmission rates were higher after neobladder than after ileal conduit UD (39% vs. 23.7%, respectively). Moschini and colleagues reported that shorter index hospitalization and older age were risk factors for 30-day readmissions in a European cystectomy cohort and found a protective effect of an index LOS in patients older than 70 years.²⁷ Though we did not observe these relationships, an analysis of Nationwide Readmissions Database (NRD) also identified continent diversion a risk factor for 90-day readmission, which is consistent with our findings.²⁸ Importantly, we demonstrated that any perioperative complications following radical cystectomy, and especially gastrointestinal complications, were predictive for discretionary readmission. These findings suggest that perioperative care and education should be further tailored according to known risk factors in order to prevent discretionary readmissions, such as for patients who undergo neobladder procedure or who develops postoperative GI complications during index hospital stay prior to discharge.

Several studies that previously explored causes and nature of readmissions utilized ICD-9 diagnosis and procedure codes in population-based and health insurance claims databases.^18,26 Instead, we used a structured protocol directed process consisting of two independent reviews and ratings for each patient chart. Implicit chart reviews, conducted at individual patient level by an expert reviewer, has been used to measure quality of medical care in various settings for decades.^29,30 Although implicit reviews can vary from reviewer to reviewer, they allow highly individualized assessments of each case.²⁹ Of note, inter-rater reliability between the two independent reviewers for this analysis were good (kappa=0.70) in our study. Moreover, the causes of readmissions after cystectomy, a complex and morbid surgery, are potentially multifactorial and differ in each case. Implicit reviews are advantageous in such scenarios compared to explicit chart review.²⁹

There are important limitations to our study mostly related to its retrospective design, which might be potentially effected by confounding and selection bias. Second, evaluation of the reason for readmission and its categorization was based on qualitative review of documented clinical information within the clinical record. However, subtleties underlying clinical decision-making supporting readmission may not have been documented or easily discovered during the review process in all cases. A third consideration has to do with evolving clinical practices in perioperative management during the study time interval. Components of ERAS protocols were implemented at different times during the study which could affect observed complication profiles and risk of readmission. For example, alvimopan was implemented within the context of a clinical trial between 2012 and 2014 and then adopted as a standard management practice at our center after May 2014. However, other components of surgical ERAS protocol such as omitting oral bowel preparation, early mobilization, avoidance of nasogastric decompression, and early feeding were routinely utilized for more than a decade. The frequency of postoperative GI complications, discretionary readmissions and necessary readmissions within 90 days of surgery were comparable between patients who received alvimopan as a component of surgical ERAS protocol and patients who received no alvimopan (data not shown). Additionally, a previous study from our institution also found no significant effect of ERAS protocols on readmission in patients treated with cystectomy, suggesting that the impact of temporal changes in ERAS practices were likely minimal in this study.³¹ Fourth, our results may not generalize to all bladder cancer patients such as those who are cisplatin-ineligible or unfit to receive chemotherapy. In the presented study, we included patients who received neoadjuvant ddMVAC or GC regimens only, which are currently preferred regimens for neoadjuvant treatment of muscle invasive bladder cancer owing to their high survival benefit, good tolerability and acceptable toxicity.³² Importantly, the incidence of postoperative complications and readmission were independent of type of the cisplatin-based NAC regimen, and comparable to those in other cystectomy cohorts which were heterogenous in terms of NAC use and type of preferred regimen.^2,6,18,33 Lastly, patient disposition at discharge and its effect on readmission were not assessed, and some readmissions to outside health care facilities might have not been captured. Despite these limitations, our findings identified a significant minority of readmissions that could potentially have been managed without inpatient resources and their association to postoperative complications.

Conclusions

The findings of the current study highlight that type and grade of complications predict both discretionary and necessary readmission. Notably, 25% of readmissions were categorized as discretionary and were driven primarily by low-grade gastrointestinal complications, marginal oral intake, and failure to thrive suggesting that better coordinated post-discharge supportive care could help avoid a substantial proportion of the readmissions.

Highlights:

• About 25% of readmissions after radical cystectomy could be classified as discretionary.

• Discretionary readmissions represent unplanned health encounters that could be managed in outpatient settings.

• There remains an opportunity to redesign post-operative follow-up care for some cystectomy patients to avoid discretionary readmissions, inpatient healthcare services and costs.