ARF indicates acute renal failure; PCF, pharyngocutaneous fistula; SSI, surgical site infection; TEP, tracheoesophageal puncture.
Graboyes EM, Yang Z, Kallogjeri D, Diaz JA, Nussenbaum B. Patients Undergoing Total LaryngectomyAn At-Risk Population for 30-Day Unplanned Readmission. JAMA Otolaryngol Head Neck Surg. 2014;140(12):1157-1165. doi:10.1001/jamaoto.2014.1705
Patients undergoing total laryngectomy are at high risk for hospital readmission. Hospital readmissions are increasingly scrutinized because they are used as a metric of quality care and are subject to financial penalties.
To determine the rate of, reasons for, and risk factors that predict 30-day unplanned readmission for patients undergoing total laryngectomy.
Design, Setting, and Patients
Retrospective cohort study at a single academic tertiary referral medical center. The study population comprised 155 patients who underwent total laryngectomy with or without flap closure between January 2007 and December 2012 as either a primary treatment or salvage treatment for prior nonsurgical management.
Main Outcomes and Measures
Rate of 30-day unplanned readmission, readmission diagnoses, and risk factors for unplanned readmission. Univariable and multivariable logistic regression were performed to identify risk factors for unplanned readmission within 30 days of discharge.
The 30-day unplanned readmission rate for patients following discharge after total laryngectomy was 26.5% (41 of 155). The most common readmission diagnoses were pharyngocutaneous fistula (27% of readmissions; n = 11) and stomal cellulitis (16% of readmissions; n = 7). The median time to unplanned readmission was 7 days. Thirty-four percent of readmissions (14 of 41) occurred within 3 days of discharge. Significant predictors of 30-day unplanned readmission on multivariable analysis were postoperative complication after discharge (odds ratio [OR], 11.50; 95% CI, 4.10-32.28), visit to the emergency department within 30 days after discharge (OR, 5.25; 95% CI, 1.84-14.99), salvage total laryngectomy (OR, 3.52; 95% CI, 1.56-13.12), and chyle fistula during the index hospitalization (OR, 5.25; 95% CI, 0.86-29.92). The discriminative ability of the model to predict unplanned readmission, as measured by the C statistic, was 0.88.
Conclusions and Relevance
Patients undergoing total laryngectomy are an at-risk patient population with a high rate of unplanned readmission within 30 days of discharge. By identifying the risk factors that predict 30-day unplanned readmission, these data can be used to design and implement quality-improvement interventions to decrease readmissions.
Hospital readmissions are increasingly scrutinized because they are used as a metric of quality care and are subject to financial penalties. The “Hospital Readmission Reduction Program” of the Affordable Care Act (ACA), which penalizes hospitals with excess readmission rates for patients with certain conditions, may expand to cover surgical patients in 2015.1 As a result, there has been a burgeoning research effort to identify risk factors for hospital readmissions for various populations of surgical patients.2- 5 Previous work examining risk factors for readmission of otolaryngology patients demonstrated that patients undergoing total laryngectomy (TL) are a high-risk population.6 For the 1 year of otolaryngology discharges analyzed, the rate of 30-day unplanned readmission independent of complications following TL was 40%, more than a 5-fold increased risk of the 7.3% overall otolaryngology readmission rate. The details of the readmissions after TL were not elucidated in this prior study.
The present study analyzing exclusively unplanned readmissions in patients undergoing TL was performed to characterize this high-risk population. The objectives of this study are 3-fold: (1) determine the rate of 30-day unplanned readmission in patients undergoing TL; (2) describe the characteristics of these patients; and (3) enumerate the risk factors that predict 30-day unplanned readmission in this patient population.
The study was approved by the Washington University School of Medicine institutional review board, which waived written informed consent for this retrospective analysis. Both departmental billing records and an institutional database (the Clinical Investigation Data Exploration Repository) were searched for patients undergoing TL, with or without flap reconstruction, at our academic medical center from 2007 to 2012. The search identified 158 patients. Three patients were excluded: 2 patients who received known follow-up care at a different institution and 1 patient who died during the index hospitalization. The study cohort thus comprised 155 patients who underwent a TL and were discharged from the otolaryngology service with follow-up at our academic medical center. Twenty-five of the patients in this study were included in a previous study examining risk factors for readmission for all otolaryngology patients discharged over a 1-year period.6
The primary outcome was the 30-day unplanned readmission rate for patients undergoing TL. The study defined a 30-day unplanned readmission as an admission to any service in our hospital within 30 days of discharge from the otolaryngology service that was not expected.6 A complication-associated readmission was defined as a readmission for a complication, diagnosed prior to or after discharge, that resulted from an otolaryngology intervention during the index admission. Patients with planned readmission that occurred within 30 days of discharge (n = 5: biopsy of lung nodule [n = 2], completion thyroidectomy [n = 1], re-resection of positive margins [n = 1], and carotid artery coiling [n = 1]) were excluded from data analysis.
By direct review of the medical record, data were collected on 55 variables relating to sociodemographics and oncologic data, comorbidities, surgical details, hospital course, and postdischarge follow-up. Race was self-identified by the patient at hospital registration and was recorded from the medical record. The Charlson comorbidity index was calculated for each patient.7 Variables analyzed for complications included nonsurgical site infection, surgical site infection or wound dehiscence (including pharyngocutaneous fistula), postoperative bleeding requiring a trip to the operating room or blood transfusion, hypocalcemia requiring consultation from endocrinology, chyle fistula, new venous thromboembolism, and other (eg, myocardial infarction, atrial fibrillation, acute kidney injury).
Standard descriptive statistical operations were performed to describe the distribution of clinical characteristics, hospital course and follow-up, and readmissions. The independent samples t test, or its nonparametric equivalent Mann-Whitney U test, was used to compare continuous variables between the groups with and without unplanned readmission. The Pearson χ2 test or Fisher exact test was used to compare categorical variables between the 2 groups. Univariable and multivariable logistic regression were performed to identify predictors of 30-day unplanned readmission. Variables significant at the 0.10 α level were considered for inclusion into the multivariable logistic regression model. A backward stepwise elimination approach was used to determine the final multivariable model. When colinearity was identified, clinical relevance was used to choose between the collinear variables. Performance of the multivariable model was assessed by Nagelkerke R2 and the C statistic. All statistical tests were 2 sided. Statistical significance was set at P < .05. Data analysis was performed using SPSS software, version 20.0 (IBM SPSS Inc).
The 30-day unplanned readmission rate for patients following TL was 26.5% (41 of 155). Patient demographics and the univariable analysis of the risk of unplanned readmission are summarized in Table 1. No sociodemographic variables predicted unplanned readmission on univariable analysis. Overall, the patients most commonly had squamous cell carcinoma (SCC) of the supraglottis (37%), glottis (35%), and hypopharynx (17%). The other 11% had posttreatment (chemo)radiation laryngeal dysfunction (n = 7), non-SCC disease (thyroid carcinoma, n = 3; parathyroid carcinoma, n = 1; chondrosarcoma, n = 1), SCC of the subglottis (n = 4), and SCC of the oropharynx (n = 1). Of the patients with SCC, most had pathologic tumor staging of pT4 (59%) or pT3 (26%) and neck staging of pN0 (52%) or pN2c (18%). Tumor subsite, T stage, and N stage did not differ between the patients with and without unplanned readmission.
Of the 12 comorbidity variables examined, 2 were found to be significant on univariable analysis (Table 2). Patients undergoing salvage TL for recurrent or persistent disease following nonsurgical management (approximately 25% of patients) had a 30-day unplanned readmission rate of 42% (15 of 36), which was associated with a 2.5-fold increased risk of readmission on univariable analysis (odds ratio [OR], 2.56; 95% CI, 1.16-5.64). Long-term steroid use also predicted 30-day readmission on univariable analysis (OR, 7.78; 95% CI, 1.50-41.80). A weighted measure of comorbidity, the Charlson comorbidity index, did not predict unplanned readmission.
Variables related to the surgical technique and hospital course are detailed in Table 3. A neck dissection was performed in 85% of patients (bilateral in 66%, unilateral in 19%). A primary tracheoesophageal puncture (TEP) was performed in nearly 50% of patients. Primary TEP was associated with a 2-fold decreased risk of readmission on univariable analysis (OR, 0.45; 95% CI, 0.21-0.95).
Twenty-eight percent of patients (44 of 155) received a vascularized flap to facilitate closure of the pharyngeal defect. This was a pectoralis flap in 80% and a free-tissue transfer in 20%. Neither type of closure nor type of flap differed between patients with and without unplanned readmission. In the subset of patients undergoing salvage TL after previous nonsurgical management (n = 36), 19 had a vascularized flap closure for reconstruction, and 17 did not. Three (18%) of the 17 patients who had a closure without a flap developed a pharyngocutaneous fistula (PCF), whereas 6 (32%) of the 19 patients who had a vascularized flap developed a PCF.
The median duration of hospitalization following TL was 10 days (range, 5-55 days). The presence of a complication during the index hospitalization did not predict 30-day unplanned readmission (OR, 1.01; 95% CI, 0.50-2.07). The number of complications during the index hospital stay was also not predictive (OR, 0.98; 95% CI, 0.68-1.39). When analyzed by specific postoperative complications during the index hospital course, no individual postoperative complication was statistically significant, although chyle fistula was associated with a trend toward increased unplanned readmission (OR, 3.82; 95% CI, 0.97-15.00). At the time of discharge, the chyle fistula was resolved in 4 patients and improving in the remaining 5 patients.
The median time to first contact with a physician following discharge was 5.5 days (range, 0-30 days). The service of first postdischarge contact was the head and neck cancer team (otolaryngology, radiation oncology, or medical oncology) for most patients (85%), and the emergency department (ED) for the rest (15%). First postdischarge contact with the ED rather than a member of the head and neck cancer team increased the risk of unplanned readmission 3-fold (OR, 3.10; 95% CI, 1.22-7.87). Scheduled follow-up with an otolaryngologist was not protective against unplanned readmission. Of the 41 patients with unplanned readmissions, almost 50% (19 of 41) attended a planned clinic visit with their otolaryngologist prior to the unplanned readmission.
In addition to association with frequent unplanned readmissions, ED utilization within 30 days of discharge was also common following TL. Twenty-one percent of patients (33 of 155) visited the ED at least once within 30 days, of whom 61% (20 of 33) were directly admitted, and 39% (13 of 33) were discharged home. The most common ED diagnoses (49%; 16 of 33) were related to poor care for the stoma, stomal appliances, or TEP feeding tube (eg, mucus plugging, inability to replace laryngectomy or tracheostomy tube, and feeding tube in TEP site displaced or clogged). Other diagnoses were much less common (dyspnea of uncertain cause, n = 2; pneumonia, n = 2; PCF, n = 2). Most patients with mucus plugging in the ED (4 of 6) were readmitted. All patients with other stomal care or TEP care issues were discharged home from the ED (14 of 14). A visit to the ED within 30 days of discharge was associated with a 9-fold increased risk of 30-day unplanned readmission on univariable analysis (OR, 8.92; 95% CI, 3.79-21.01).
Forty percent of patients (62 of 155) had at least 1 new complication diagnosed following discharge. The most common were stomal and/or TEP complications (13% of patients; 20 of 155), stomal cellulitis (12%; 19 of 155), and PCF (7%; 11 of 155). Four of the 20 patients who developed stomal and/or TEP complications after discharge were readmitted, all with the diagnosis of mucus plugging. Of the 19 patients who developed postdischarge stomal cellulitis, 12 were treated as outpatients, and 7 required inpatient admission. All of the patients diagnosed as having a PCF after discharge were readmitted. A new complication following discharge resulted in a nearly 19-fold increased risk of unplanned readmissions on univariable analysis (OR, 18.80; 95% CI, 7.14-49.47). Eighty percent of unplanned readmissions (33 of 41) were complication-associated readmissions.
The 2 most common readmission diagnoses were PCF (27% of unplanned readmissions; n = 11) and stomal cellulitis (17%; n = 7) (Figure). Most patients were readmitted from the otolaryngology clinic (54%; n = 22), and 80% (33 of 41) were readmitted to the otolaryngology service. The median time to unplanned readmission was 7 days (range, 1-28 days). Thirty-four percent of readmissions (14 of 41) occurred within 3 days of discharge; 27% (n = 11) occurred between 4 and 10 days after discharge; and 39% (n = 16) occurred more than 10 days after discharge. No unifying diagnosis was seen in the patients with early readmissions. Those readmitted with PCF tended to be readmitted later than other patients with unplanned readmissions (median time to readmission, 12 days), and 73% (8 of 11) were admitted more than 1 week after discharge. The patients readmitted with mucus plugging (4 of 41) did not show a temporal pattern to the timing of readmission (1, 3, 8, and 10 days after discharge). The median duration of readmission was 4 days (range, 1-37 days).
Using a backward stepwise elimination approach to multivariable logistic regression analysis, we created a multivariable model of predictors of 30-day unplanned readmission (Table 4). The presence of a postoperative complication following discharge was the strongest predictor of unplanned readmission, with an 11.5-fold increased risk of unplanned readmission in the multivariable model (OR, 11.5; 95% CI, 4.10-32.28). Visiting the ED within 30 days of discharge increased the odds of unplanned readmission by 5-fold (OR, 5.25; 95% CI, 1.84-14.99). The model’s predictive discrimination, as measured by the C statistic, was 0.88.
Prior research analyzing unplanned readmission in a heterogeneous cohort of otolaryngology patients reported that following TL patients are at high risk for unplanned readmission.6 The present study confirms this finding in a much larger cohort comprising exclusively patients undergoing TL. These patients are an at-risk population with a 30-day unplanned readmission rate of 26.5% at our institution. In relation to other surgical procedures, this is twice the risk of readmission (planned and unplanned) found for a group of 6 major surgical procedures evaluated recently (coronary artery bypass grafting [CABG], pulmonary lobectomy, endovascular repair of abdominal aortic aneurysm [AAA], open repair of AAA, colectomy, and hip replacement).2 It is also twice the rate of readmission for CABG found in other studies, a procedure rumored to be on the short list for the expansion of the Readmission Program in 2015, and one that is already tracked as a quality metric.8 This is also higher than readmission rates for other published “high-risk” procedures, such as pancreaticoduodenectomy, which was reported to have a 26% risk of readmission (unplanned and planned) over the patient’s lifetime.9 Since approximately 3500 TLs are performed annually in the United States,10 this would potentially equate to over 900 unplanned readmissions within 30 days of discharge.
A postoperative complication after discharge was the most predictive risk factor for 30-day unplanned readmission, with an 11-fold increased risk of unplanned readmission on multivariable analysis. The most common postoperative complications following discharge were stomal and/or TEP complications, stomal cellulitis, and pharyngocutaneous fistula. Twenty percent of patients with stomal and/or TEP complications, 37% with stomal cellulitis, and all patients with PCF required readmission for management of the complication. The most common readmission diagnosis was PCF. No studies have been published about the rate of PCF following discharge from the hospital. The overall rate of PCF in our study (during the index hospitalization and after discharge) was 15% (23 of 155), which is consistent with published rates.11,12 For patients undergoing salvage TL, the PCF rate was 25% (9 of 36), which is again consistent what has been published in the recent otolaryngology literature13 and major clinical trials.14
One of the major findings of this study is that, following TL, postoperative complications arising after discharge are the major determinant of 30-day unplanned readmissions. Neither the presence of an in-hospital complication (yes/no) nor the number of in-hospital complications (1, 2, 3, 4, and so on) predicted unplanned readmission. Only the timing of the postoperative complication (during hospitalization vs after discharge) was predictive. This suggests that when patients have postoperative complications during the index hospitalization, as nearly half of them did, the complications are detected during the hospital stay and treated appropriately prior to discharge. Other studies of surgical patients have identified the timing of complications (ie, during hospitalization or after discharge) as being prognostic.4 This is an important finding with regard to implementing transition-of-care policy changes because caretakers and clinicians must remain vigilant for complications even after the patient is discharged from the hospital. It also has implications for public policy, since readmissions among patients undergoing TL were not necessarily associated with poor care but rather with the timing of the complications occurring.
Emergency department utilization following TL was common, and a visit to the ED within 30 days of discharge was associated with a 5-fold increased risk of readmission on multivariable analysis. Most patients visited the ED for complications related to the laryngectomy stoma or TEP. Although many of the patients who went to the ED for stomal or TEP complications were discharged home, merely visiting the ED increased the patient’s risk of 30-day unplanned readmission. Therefore, even though the ED represents a distal step in the chain to readmission relative to the event that prompts the ED visit, ED utilization remained predictive of unplanned readmission in the multivariable model, independent of a postdischarge complication.
We hypothesize that the prognostic significance of visiting the ED arises from a number of intangible factors that cannot be detected in a retrospective study: admission inertia (the patient is already in the hospital), time of day (not wanting to discharge the patient from the ED late at night), ED physician lack of familiarity with laryngectomy stoma (inappropriate concern that the laryngectomy stoma will close if the laryngectomy and/or tracheostomy tube is not replaced immediately), and caregiver preference for admission (the patient and/or the family expressing lack of comfort with the stoma and requesting more intensive laryngectomy stomal education). Educating patients and caregivers in excellent stomal care could decrease the number of stomal complications (and subsequent ED visits for these problems) and potentially decrease the risk of readmission.
Salvage TL carried a 42% risk of 30-day unplanned readmission, which was associated with a 3.5-fold increased risk of unplanned readmission on multivariable analysis. Almost 50% of the readmission diagnoses (7 of 15) in this group were for PCF. The PCF in the salvage TL group occurred late (median time to diagnosis, 8 days after discharge), and half of the patients had a scheduled postoperative visit with their otolaryngologist prior to the readmission. The high rate of PCF following salvage TL, occurring 2 to 4 weeks postoperatively, is consistent with what has been published,13,14 but it may make preventing readmissions in this subgroup of patients challenging.
Approximately 50% of the patients undergoing salvage TL received a vascularized flap in our cohort, and vascularized flap was not found to be protective against PCF or 30-day unplanned readmission in this study. However, since the use of vascularized tissue for reconstructing the pharyngeal defect in patients undergoing salvage TL has been reported to decrease the rate of PCF,13,15 a more standardized use of flaps might decrease the 30-day unplanned readmission rate in the salvage TL population. The small number of patients undergoing salvage TL in our study limits our ability to make any conclusions about this hypothesis. In addition, the 30-day unplanned readmissions for post–salvage TL pharyngocutaneous fistula represent only a small percentage of unplanned readmissions in our patient cohort (17%; 7 of 41).
Chyle fistula during the index hospitalization, although not a significant predictor of readmission in the final multivariable model, improved the model’s overall statistical performance for identifying high-risk patients for 30-day unplanned readmission. Nine patients in the study had a chyle fistula diagnosed, all prior to discharge. The status of the chyle fistula at discharge (resolved vs improving) did not differ between patients who were readmitted and those who were not.
Although associated with a decreased risk of readmission on univariable analysis, TEP was not significant in the final model owing to colinearity with healthier patients and primary, nonsalvage TL. Feeding tube complications related to TEP did prompt visits to the ED, but none of the patients with TEP-related complications had an unplanned readmission for the TEP. The TEP nevertheless represents another aspect of postoperative care, and thus complication, that may prompt visits to the ED and place patients at risk for readmission.
Designing interventions to reduce the rate of unplanned readmissions in patients following TL may prove difficult for a number of reasons. The 2 major predictors of unplanned readmission found in this study (postoperative complication following discharge and ED utilization <30 days after discharge) do not identify high-risk patients prior to discharge from the hospital. Since the highest-risk patients are not known prior to discharge from the hospital, clinicians and other members of the discharge planning team cannot concentrate time, energy, or resources on them as they transition to the postdischarge setting. In addition, prompt follow-up with an otolaryngologist did not appear to prevent readmissions: nearly 50% of readmitted patients attended a scheduled follow-up appointment with their otolaryngologist, at which time no imminent problems were identified.
Another potential solution, given the high rate of readmissions we found within 3 days of discharge, might be to keep patients in the hospital 1 or 2 extra days. This strategy would not appear to mitigate readmissions in our group, however, because the reasons for early readmission were highly variable, evenly distributed among 9 different causes, and unrelated to complications in almost 50% of cases. Since delayed presentation of stomal cellulitis was a common readmission diagnosis, one might consider treatment with a longer course of postoperative antibiotics. But this would not likely decrease the rate of delayed postoperative cellulitis, since our patients routinely receive postoperative antibiotic prophylaxis following TL.
Despite these potential difficulties, the data still point to a few options for potentially decreasing the rate of 30-day unplanned readmission following TL. First, recognizing that postoperative complications will happen, one might consider a more concerted effort to educate patients and caretakers about early recognition of complications. Early recognition may facilitate prompt follow-up with the otolaryngologist with subsequent earlier interventions and outpatient management.
Second, an increased emphasis on stomal and TEP care, perhaps with formalized predischarge assessments, may help decrease readmissions by directly decreasing revisits for stomal and TEP complications and by decreasing the rate of delayed cellulitis. These 2 complications were the most frequent following discharge, and correction of stomal and TEP complications seems to be the one most easily achievable. Whether improved patient and family education will have an impact on decreasing 30-day unplanned readmissions is a focus of future research at our institution.
Finally, skilled nursing facilities may play an important role as a discharge destination, especially when caregivers feel overwhelmed by the new stomal care responsibilities. Perhaps better triaging of patients to a post–acute care setting with more nursing support will decrease postdischarge complications and 30-day unplanned readmissions following TL.
Despite its strengths, this study possesses limitations. First, the study has the limitations inherent to all retrospective studies such as dependency on the accuracy and completeness of the medical record. Second, the number of patients in the study is relatively small, which makes estimations of risk less precise and confidence intervals wide. Third, the study did not capture readmissions outside of the single institution. To minimize this limitation, patients known to have followed up elsewhere were excluded. In addition, in our tertiary care academic referral center, patients who experience a need for postoperative readmission following TL are transferred back to our institution for their care. Fourth, the study was conducted at a single academic medical center and therefore might not reflect all otolaryngology practices equally well with respect to this patient population.
This study confirms that patients undergoing TL are an at-risk patient population with a high rate of 30-day unplanned readmission following their initial procedure. The multivariable model identifying risk factors associated with 30-day unplanned readmission in patients undergoing TL included postoperative complication after discharge (OR, 11.50), visit to the ED within 30 days after discharge (OR, 5.25), salvage TL (OR, 3.52), and chyle fistula during the index hospitalization (OR, 5.25). These data can be used to start designing and implementing quality-improvement interventions to decrease readmissions.
Submitted for Publication: February 3, 2014; final revision received April 18, 2014; accepted April 24, 2014.
Corresponding Author: Brian Nussenbaum, MD, Department of Otolaryngology–Head and Neck Surgery, Washington University School of Medicine, 660 S Euclid Ave, Campus Box 8115, St Louis, MO 63110 (firstname.lastname@example.org).
Published Online: August 21, 2014. doi:10.1001/jamaoto.2014.1705.
Author Contributions: Drs Graboyes and Nussenbaum had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Graboyes, Diaz, Nussenbaum.
Acquisition, analysis, or interpretation of data: Graboyes, Yang, Kallogjeri, Diaz, Nussenbaum.
Drafting of the manuscript: Graboyes, Kallogjeri, Nussenbaum.
Critical revision of the manuscript for important intellectual content: Graboyes, Yang, Diaz, Nussenbaum.
Statistical analysis: Graboyes, Kallogjeri.
Administrative, technical, or material support: Diaz, Nussenbaum.
Study supervision: Diaz, Nussenbaum.
Conflict of Interest Disclosures: None reported.
Funding/Support: This research was supported in part by the P30 Research Center for Auditory and Vestibular Studies and by grant P30DC04665 from the National Institutes of Health, National Institute on Deafness and Other Communication Disorders, which funded the biostatistics work.
Role of the Funder/Sponsor: The funding institutions had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Previous Presentation: This study was presented at the Fifth World Congress of the International Federation of Head and Neck Oncologic Societies and the Annual Meeting of the American Head & Neck Society; July 27, 2014; New York, New York.