The relationship of postoperative complications with in-hospital outcomes and costs after renal surgery for kidney cancer

Authors


Correspondence: R. Houston Thompson, Department of Urology, Mayo Clinic, 200 First Street Southwest, Rochester, MN 55905, USA.

e-mail: Thompson.Robert@mayo.edu

Abstract

What's known on the subject? and What does the study add?

  • Postoperative complications for open radical nephrectomy (ORN), laparoscopic radical nephrectomy (LRN), and open partial nephrectomy (OPN) and its relationship with hospitalisation costs and mortality remain poorly described.
  • The present population-based study suggests modest differences in postoperative complications estimated at 27%, 23%, and 24% among patients with kidney cancer undergoing ORN, LRN, and OPN, respectively. Moreover, postoperative complications were associated with higher mortality, length of stay and total costs of hospitalisation.

Objectives

  • The association of complications after renal surgery for renal cell carcinoma (RCC) with in-hospital mortality and costs remains to be defined.
  • To describe the incidence of complications after open radical nephrectomy (ORN), laparoscopic RN (LRN), and open partial nephrectomy (OPN); and to evaluate its relationship with in-hospital mortality and total costs.

Patients and Methods

  • We identified 49 983 individuals who underwent ORN (35 712), LRN (5327), or OPN (8944) for RCC at 2037 hospitals from the Nationwide Inpatient Sample 2001–2008.
  • The outcomes assessed were in-hospital mortality and total hospitalisation costs.
  • Multivariable logistic regression and generalised estimating equations were used to test the associations between complications and in-hospital mortality and total costs.

Results

  • With 26.0% of patients experiencing postoperative complications, there were modest differences in the proportion of patients with complications after ORN, LRN, and OPN at 27.0%, 22.6%, and 24.0%, respectively (P < 0.001).
  • After adjusting for patient and hospital variables, postoperative complications resulted in higher odds of in-hospital death for ORN (odds ratio [OR] 7.20; P < 0.001), LRN (OR 12.04; P < 0.001), and OPN (OR 7.82; P < 0.001).
  • Adjusted total costs also rose significantly with the presence of any postoperative complications compared with those without any complications for ORN ($21 242 vs $13 183; P < 0.001), LRN ($19 548 vs $12 555; P < 0.001), and OPN ($18 883 vs $12 098; P < 0.001).

Conclusions

  • With about a quarter of patients experiencing postoperative complications, adverse events for ORN, LRN, and OPN carry a significant risk of in-hospital death and higher total costs.
  • Efforts to reduce postoperative complications may correlate with substantial reductions in hospital mortality and total costs.
Abbreviations
(O)(L)PN

(open) (Laparoscopic)partial nephrectomy

(O)RN

(open) radical nephrectomy

NIS

Nationwide Inpatient Sample

ICD-9

International Classification of Disease Modification, 9th edition

DVT

deep vein thrombosis

PE

pulmonary embolus

GEE

generalized estimating equation

LOS

length of stay

Introduction

Postoperative complications have become an increasingly recognised healthcare dilemma due to its role in increasing the risks of adverse outcomes and higher costs during hospitalisation for patients undergoing surgery [1, 2]. With the national annual costs attributable to medical harm from hospitalisation estimated at $17 billion [3], postoperative complications have been identified as one of the leading contributors to higher healthcare costs [4, 5]. As a result, increased attention has been placed on promoting surgical quality by identifying key processes of care and reducing postoperative complications in efforts to improve patient outcomes and lower costs.

Against this backdrop, RCC remains the third most commonly diagnosed genitourinary malignancy with an estimated 58 240 incident cases and 8210 cancer-related deaths in 2010 [6]. With the annual incidence of small renal masses rising [7, 8], there has been a corresponding increase in use of both partial nephrectomy (PN) and radical nephrectomy (RN) [9, 10]. With the introduction of PN and laparoscopy, safety and effectiveness were initially reported from retrospective single-institutional studies showing that open RN (ORN), laparoscopic RN (LRN), and open PN (OPN) had equivalent risks of mortality, yet modest differences in surgical morbidity [11-13]. OPN and ORN both appear to have comparable complication rates, while ORN and LRN are markedly different in the risks of morbidity and re-admission [14-16]. Yet, the risks of surgical complications and their relationship with in-hospital outcomes and total costs for ORN, LRN, and OPN remain to be defined. In this context, we sought to describe the characteristics of postoperative complications for ORN, LRN, and OPN; and evaluate the relationship of in-hospital mortality, total costs, and prolonged length of stay (LOS) attributable to each type of complication by different types of renal surgery.

Patients and Methods

Data for all patients who underwent ORN, LRN, or OPN for RCC were abstracted from the Nationwide Inpatient Sample (NIS) from 2001 to 2008. The NIS, which is maintained by the Healthcare Cost Utilization Project (HCUP), contains about one fifth of all admissions from a stratified sample in the USA [17]. To identify our analytic cohort, we adopted a methodology described previously with the NIS for renal surgery, which has been further validated as a claims-based algorithm using International Classification of Disease Modification, 9th edition (ICD-9) codes (Appendix I) [18, 19]. All adult patients (aged ≥18 years) were then categorised into one of three types of nephrectomy: ORN (35 712), LRN (5327), or OPN (8944). From this cohort, cost data from the NIS were available for 44 777 (89.6%) patients.

Clinical information included as covariates were patient age, gender, race, primary health insurance (private health insurance, Medicare, Medicaid, and other), type of admission (elective vs emergent), median zip code income ($1–$38 999; $39 000–$47 999; $48 000–$62 999; and ≥$63 000), and time interval of surgery (2001–2002, 2003–2004, 2005–2006, and 2007–2008). With about a quarter of individuals missing a race designation, we created an indicator variable for this group in order to include these patients in our analysis. Secondary diagnostic codes were used to define the Elixhauser comorbidity index [20].

From the NIS, hospital features included as covariates were teaching status (non-teaching vs teaching), location (urban vs rural), and annual case volume. To adjust for the relationship of nephrectomy volume with complications and outcomes, the unique hospital identifier was used to divide hospital volume into the following quartiles of average nephrectomy volume per year: very low (<3 cases/year), low (3–6 cases/year), medium (7–14 cases/year), and high (≥15 cases/year) volume.

To define in-hospital complications, we selected relevant complications for renal surgery from a validated methodology (Appendix II) [21, 22]. We then categorised the complications into the following groups: cardiac, respiratory, bowel, genitourinary, deep vein thrombosis (DVT)/pulmonary embolus (PE), infectious, perforation, haemorrhage, and wound complications. In-hospital complications were defined as the presence of one or more complications.

The primary outcomes were in-hospital mortality, defined as a postoperative death during hospitalisation, and total costs. The secondary outcome evaluated in our study was prolonged LOS. In-hospital total costs were calculated from the cost to charge ratio as provided by the NIS [23]. All total hospitalisation costs were adjusted to U.S. dollars using the Consumer Price Index [24]. We further defined a procedure-specific, prolonged LOS as the top 10th percentile for ORN (LOS >10 days), LRN (LOS >7 days), and OPN (LOS >8 days).

Bivariate associations for patient and hospital covariates, complications, and in-hospital mortality were tested by the Pearson chi-square. We fitted multi-level/mixed effects multivariable logistic regression models to test the associations of patient and hospital variables and postoperative complications with the primary outcome of in-hospital mortality by each subgroup of surgical treatment, while adjusting for clustering of patient covariates to the hospital level.

To determine the predicted probabilities of in-hospital mortality and prolonged LOS and the adjusted costs for each type of complication by surgical treatment subgroup, we used generalized estimating equations (GEE) to adjust for patient and hospital covariates, the type of complication, and clustering of patients to the hospital level. In-hospital death and prolonged LOS were defined as binary variables in their respective models. In the GEE model for total in-hospital costs, we specified a Gamma distribution and log link to account for over-dispersion in costs [25]. We then calculated the adjusted probabilities and costs based on the estimates produced by each model. A two-sided P ≤ 0.05 was considered to indicate statistical significance.

Results

Table 1 presents the patient and hospital characteristics of the 49 983 patients who underwent nephrectomy for RCC at 2037 hospitals. Overall, the proportions of patients who underwent ORN, LRN, and OPN for RCC during the 8-year study interval were 71.4% (35 712), 10.7% (5327), and 17.9% (8944), respectively. On bivariate analysis, there were statistically significant differences in patient and hospital variables by the type of nephrectomy.

Table 1. The patients' characteristics by the type of nephrectomy for kidney cancer (n = 49 983)
VariableORNLRNOPNP
Number of patients35 71253278944 
%:    
Age, years:   <0.001
<6040.340.447.7
60–6927.326.028.5
70–7923.424.019.5
≥809.09.64.3
Race:   <0.001
White58.162.059.4
Black6.86.17.7
Hispanic6.05.75.5
Other3.33.33.9
Missing25.822.923.5
Gender:   <0.001
Female38.841.737.9
Male61.258.362.1
Primary insurance:   <0.001
Private44.745.553.8
Medicare45.745.636.5
Medicaid4.54.25.0
Other5.14.74.7
Annual household income, $:   <0.001
<38, 99918.618.517.7
39 000–47 99924.023.722.3
48 000–62 99926.126.025.7
≥63 00031.331.834.4
Elixhauser comorbidity index:   <0.001
0–150.156.157.9
1–240.436.235.8
≥39.57.76.3
Hospital teaching status:   <0.001
Teaching54.665.570.2
Non-teaching45.435.529.8
Hospital location:   <0.001
Rural7.84.24.4
Urban92.295.895.6
Annual nephrectomy volume:   <0.001
Very low, <3 cases/year25.716.514.9
Low, 3–6 cases/year26.025.221.3
Medium, 7–14 cases/year25.828.726.7
High, ≥15 cases/year22.529.637.1

Table 2 details the types of complications, as well as in-hospital mortality and total costs from the hospitalisation by the type of nephrectomy. While the overall proportion of patients with complications during hospitalisation for nephrectomy was 26.0%, the proportion of patients with postoperative complications was modestly lower in LRN (22.6%) and OPN (24.0%) than in ORN (27.0%; P < 0.001). Although about a quarter of patients had complications, the overall in-hospital mortality remained low at 0.9%. However, in-hospital mortality differed by the type of surgery with a higher mortality rate in ORN (1.1%) than in LRN (0.6%) and OPN (0.4%; P < 0.001).

Table 2. In-hospitals complications, mortality, and LOS by the type of nephrectomy (n = 49 983)
VariableORNLRNOPNP
  1. IQR, interquartile range.
Number of patients35 7125 3278 944 
%:    
Type of complication:    
Cardiac2.32.01.80.006
Respiratory5.13.64.6<0.001
Bowel10.711.79.5<0.001
Genitourinary1.61.72.10.02
DVT/PE1.20.80.6<0.001
Infection1.71.51.30.01
Perforation/puncture4.12.63.1<0.001
Haemorrhage8.55.37.0<0.001
Wound0.40.40.20.01
In-hospital mortality1.10.60.4<0.001
Median (IQR) in-hospital cost, $11 558 (8 647–16 660)11 676 (9 212–15 613)11 589 (8 822–15 553) 
Median (IQR) LOS, days5 (3–10)3 (2–5)4 (3–6) 

Figure 1A–C presents the predicted probabilities of in-hospital mortality attributable to each type of complication for ORN, LRN, and OPN, after adjusting for patient and hospital features. In ORN, while all complications were associated with a higher predicted probability of inpatient death, DVT/PE posed the highest predicted mortality (7.42%), followed by cardiac (6.11%) and infectious (5.25%) complications. While cardiac events also resulted in a similarly higher predicted mortality (6.85%) for LRN, infectious complications (6.87%) represented the leading adverse event associated with predicted mortality. For OPN, our analysis also showed the highest predicted mortality with infectious (4.37%) and cardiac (2.35%) complications, both of which were moderately lower than in the other types of nephrectomy. Interestingly, bowel, DVT/PE, and genitourinary complications did not confer a higher predicted mortality for LRN, while complications related to the bowel, DVT/PE, genitourinary, and accidental perforation were similar in failing to predict mortality for OPN.

Figure 1.

Predicted probability of in-hospital mortality by type of complication for: A, ORN; B, LRN; C, OPN. Predicted probability of prolonged LOS by type of complication for: D, ORN; E, LRN; and F, OPN.

Table 3 presents the adjusted total hospitalisation costs attributable to the presence of any complication and each type of complication. For those cases without any complications, similar adjusted total costs were observed for ORN ($13 183), LRN ($12 555), and OPN ($12 098). The presence of complications overall and by each type of specific complication were associated with significantly higher adjusted costs across all surgical treatment groups (all P < 0.001). Wound and infectious complications were responsible for the highest adjusted costs from hospitalisation for all surgical groups, which were two to three times higher than in those cases without these specific complications. However, certain complications resulted in higher total costs by the type of nephrectomy, e.g. DVT/PE having higher adjusted costs in ORN and OPN, and perforation and cardiac complications responsible for higher adjusted costs in LRN.

Table 3. Total adjusted in-hospital cost by the type of nephrectomy and presence of complications (n = 44 777)*
VariableORN (n = 31 867)PLRN (n = 4 895)POPN (n = 8 015)P
Adjusted cost (95% CI)Adjusted cost (95% CI)Adjusted cost (95% CI)
  1. *GEE for total in-hospital cost adjusted for patient demographics, median income, Elixhauser comorbidity, primary insurance, time interval of surgery, hospital covariates (teaching status, location, and annual case volume), and clustering of patients to the hospital-level.
Complication      
Absent13 183 (12 851–13 516)<0.00112 555 (12 143–12 967)<0.00112 098 (11 664–12 531)<0.001
Present21 242 (20 652–21 833) 19 548 (18 699–20 372) 18 833 (18 022–19 643) 
Cardiac      
Absent15 183 (14 805–15 561)<0.00113 964 (13 491–14 437)<0.00113 575 (13 099–14 051)<0.001
Present27 429 (25 663–29 196) 26 063 (22 252–29 874) 27 082 (23 682–30 482) 
Respiratory      
Absent14 815 (14 441–15 189)<0.00113 748 (13 297–14 198)<0.00113 308 (12 837–13 778)<0.001
Present27 354 (26 142–28 567) 25 677 (23 020–28 334) 24 381 (22 441–26 321) 
Bowel      
Absent14 882 (14 439–15 205)<0.00113 396 (12 953–13 841)<0.00113 342 (12 860–13 824)<0.001
Present21 141 (20 373–21 909) 19 983 (18 824–21 139) 18 374 (17 222–19 526) 
Genitourinary      
Absent15 364 (14 975–15 753)<0.00114 104 (13 360–14 577)0.00913 651 (13 175–14 127)<0.001
Present23 346 (21 574–25 118) 19 224 (16 189–22 260) 23 158 (20 390–25 925) 
DVT/PE      
Absent15 263 (14 890–15 636)<0.00114 106 (13 631–14 582)<0.00113 768 (13 287–14 248)<0.001
Present33 500 (30 603–36 398) 24 125 (18 412–29 837) 28 720 (22 136–35 303) 
Infection      
Absent15 122 (14 745–15 499)<0.00113 780 (13 365–14 194)<0.00113 581 (13 124–14 038)<0.001
Present34 646 (32 130–37 162) 43 282 (37 760–48 582) 34 401 (29 528–39 274) 
Perforation      
Absent15 222 (14 841–15 603)<0.00113 880 (13 437–14 324)<0.00113 771 (13 284–14 259)0.003
Present22 078 (20 938–23 217) 26 994 (23 822–30 066) 16 913 (15 173–18 652) 
Haemorrhage      
Absent14 711 (14 337–15 085)<0.00113 898 (13 427–14 370)<0.00113 338 (12 855–13 820)<0.001
Present23 507 (22 606–24 407) 19 161 (17 384–20 939) 20 033 (18 594–21 471) 
Wound      
Absent15 420 (15 029–15 812)<0.00113 780 (13 365–14 194)<0.00113 813 (13 326–14 300)<0.001
Present38 232 (32 712–43 751) 43 171 (37 760–48 582) 42 877 (24 531–61 244) 

Postoperative complications also showed statistically significant trends for prolonged LOS for all surgical treatment groups (all P < 0.05). However, the various types of complications were associated with different predicted prolonged LOS for ORN, LRN, and OPN (Fig. 1D–F). In ORN, the predicted probabilities of prolonged LOS were >40% for wound, infectious, and DVT/PE complications (all P < 0.001). Conversely, patients who underwent LRN had a predicted probability of a prolonged LOS >40%, if they had only infectious or wound complications. Although all of the complication groups were associated with a prolonged LOS (all P < 0.05), DVT/PE and wound complications yielded the highest predicted prolonged LOS for OPN.

Discussion

The present study has several important findings about the occurrence of complications after ORN, LRN, and OPN, and their relationship with in-hospital mortality, total costs, and prolonged LOS. First, we found that postoperative complications, which occur in about a quarter of all patients undergoing nephrectomy, are significant risk factors for in-hospital mortality, prolonged LOS, and higher hospitalisation costs. Overall, ORN, LRN, and OPN remain relatively safe surgical procedures in which the in-hospital mortality ranged from 0.4% to 1.1%. However, the presence of one or more postoperative complications led to a significantly higher predicted mortality for ORN, LRN, and OPN (1.1–2.6%), while the absence of any complications during hospitalisation resulted in a low predicted mortality (0.2–0.4%). The present findings are similar to previous studies in reporting low surgical mortality rates [12, 13, 15, 26, 27]. However, a higher proportion of patients had postoperative complications in the present study. Although Corman et al. [14] reported a similar 30-day morbidity for RN and PN (15.0% vs 16.2%) from a Veterans Affairs population, only one study has examined the relationship of complications with in-hospital outcomes and costs from a population-based cohort to our knowledge [15]. Using the 2000–2003 NIS, Joudi et al. [15] reported similar complication rates for RN and PN (17.2% and 18.2%). However, the present study differs in several important ways. Here, we elected to adopt a validated methodology in identifying postoperative complications, which probably explains our findings of a higher proportion of patients with in-hospital complications [21, 22]. Hence, the present findings more likely to reflect the actual complication rate for nephrectomy, as a recent SEER-Medicare study also reported postoperative complication rates for ORN and LRN at 38.8% and 31.7%, respectively [16]. Moreover, with the increasing use of laparoscopy, the present study advances the understanding of the differences in outcomes for ORN, LRN, and OPN in comparison to earlier population-based studies that categorised surgical treatment by PN vs RN only without adjusting for laparoscopy [14, 15].

Second, the present study is the first to document the additional costs attributable to different types of postoperative complications for ORN, LRN, and OPN to the best of the authors' knowledge. While previous studies have shown that postoperative complications generate higher costs [2, 15, 28], the results here provide important information for patients, urologists, and policymakers about the economic implications of complications on higher hospitalisation costs for ORN, LRN, and OPN. Moreover, in an era of rising annual healthcare expenditure, where in-hospital postoperative and infectious complications are responsible for an annual cost of $17 billion and $5 billion, respectively [3-5], urologists may face greater scrutiny with emphasis placed on reducing these adverse events to improve patient safety and lower costs [29, 30]. Indeed, if postoperative complications were reduced by 10% for each surgical group in the present analytic cohort, this would represent a significant overall cost savings for ORN ($75 859 430) LRN ($3 725 171), and OPN ($2 324 009) during the 8-year study period (as determined from the total adjusted costs model).

Third, the present study details the effects of the specific complications on mortality for each type of nephrectomy. While wound and infectious complications increased therisk of poor outcomes for all surgical groups, there were also notable differences as well. In the previous NIS study, postoperative infections had the highest hazards ratio for in-hospital death in both PN and RN, but the study did not include laparoscopy for either type of nephrectomy [15]. While DVT/PE and cardiac complications were associated with the highest predicted mortality for ORN in the present population-based cohort, cardiac and infectious complications resulted in the highest predicted mortality for LRN and OPN.

The results and inferences from the present study have several limitations. In addition to the limitations of using administrative data in clinical research [31], the NIS does not provide clinical or pathological information, e.g. primary tumour size, pathological stage, or tumour histology. It is plausible that advanced tumour stage, e.g. renal tumours with inferior vena cava tumour thrombus, may explain the observed differences in complications, as many of these patients are more likely to undergo ORN and may be at higher risk for DVT/PE due to the extent of renal tumour involvement. Another limitation is that the possible effects of surgeon experience and training were not evaluated, as the NIS has limited information about provider characteristics. The present study is probably subject to some degree of selection bias, in that patients may be treated with a particular type of renal surgery due to the extent of disease, performance status, comorbidities, and surgeon or patient preference. Although the present study also used a validated methodology in identifying postoperative complications [21, 22], it may also be limited in that claims data may either identify the complications that are not clinically accurate or relevant to in-hospital outcomes. Hence, adopting this methodology may partially explain why the present study found higher postoperative complications compared with other institutional and population-based studies [12, 13, 15]. It is also important to recognise that different data sources may bias or limit the clinical findings, such that administrative data are unlikely to fully describe the surgical care provided and the causes of the complications that may be more accurately described from institutional data. However, each of the complications defined in the present study was associated with increased risks of in-hospital death or prolonged LOS as well as higher hospitalisation costs.

In summary, the present study has important implications for informing patients and urologists about the effect postoperative complications have on increasing the risk of poor outcomes and higher total costs for most types of nephrectomy. Moreover, postoperative complications were responsible for a substantial proportion of additional costs from hospitalisation for ORN, LRN, and OPN. In this context, the hospitalisation costs attributable to postoperative complications were about two to three times higher than the costs for those patients who did not have any complications. Efforts to reduce complications after ORN, LRN, and OPN may substantially decrease the risks of in-hospital mortality and prolonged LOS, along with improving patient safety and reducing healthcare costs.

Acknowledgments

Funding source: Healthcare Delivery Research Scholars Program, Mayo Clinic.

Conflict of Interest

None declared.

Appendix: Appendix I – Patient Selection from the NIS

To select our study population, data were abstracted from the NIS by using the International Classification of Disease Modification, 9th edition (ICD-9) for renal neoplasm (189 and 189.9) from primary or secondary diagnostic codes to define all cases of RCC during the study period. ICD-9 procedure codes identified whether hospitalised patients underwent either PN (55.4) or RN, as defined by complete nephrectomy (55.5), total nephrectomy (55.51), nephrectomy of the remaining kidney (55.52), or bilateral nephrectomy (55.54) from all procedure codes. To determine the use of laparoscopy all ICD-9 procedure codes were abstracted for laparoscopic exploration (54.21), laparoscopic lysis of adhesions (54.51), or laparoscopic cholecystectomy (51.23). Absence of these secondary ICD-9 codes for laparoscopy was designated as open renal surgery. We excluded paediatric cases (4471) and those individuals who underwent laparoscopic PN due to a relatively few cases identified within the NIS (838) from the 52 292 individuals initially identified.

Appendix: Appendix II – Codes Defining In-Hospital Complications

Complication categoryDefinitionICD-9 codes
CardiacAcute myocardial infarction410.xx
Cardiac complications997.1
RespiratoryBacterial pneumonia481, 482.xx, 483.0, 485, 486
Aspiration pneumonia507.00, 507.01
Respiratory complications997.3x
BowelGastrointestinal bleed530.82, 531.0x-531.6x, 532.0x-532.0x, 532.6x, 533.0x-533.2x, 533.4x, 533.6x, 534.0x-532.4x, 533.4x, 533.6x, 534.0x-534.2x, 534.4x, 534.6x, 535.0x-535.6x, 578.9
Acute vascular insufficiency of intestine557.0
Intestinal obstruction or ileus560.0, 560.1, 560.8x, 560.9
Acute pancreatitis577.0
Digestive complications997.4
GenitourinaryUrinary complications997.5
DVT/PEAcute PE415.1x
Extremity DVT451.2, 451.81, 453.8, 997.2
InfectionClostridium difficile/Pseudomembranous colitis8.45
MRSAV090
Empyema510.0, 510.9
Abscess511.1
Pyelonephritis590.1, 590.80
Postoperative infection998.5x
PerforationGastrointestinal perforation530.4, 569.83, 575.4, 576.3
Accidental puncture or laceration998.2
HemorrhageAcute post haemorrhagic anaemia285.1
Haemorrhage complicating procedure998.1, 998.11, 998.12
WoundSeroma998.13
Disruption of wound998.3x, 998.83

Ancillary