A nomogram predicting severe adverse events after ureteroscopic lithotripsy: 12 372 patients in a Japanese national series

Authors


Correspondence: Toru Sugihara, Department of Urology, The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-8655, Japan.

e-mail: ezy04707@nifty.com

Abstract

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

  • Ureteroscopic lithotripsy sometimes causes severe complications, e.g. septic shock, and the relationship between long operative duration and complication rate has been empirically recognised. But due to the rarity, evidence is limited.
  • We analysed 12372 cases and showed that the complication rate increased according to operative duration, especially for operations taking >90 min. Also, we found that high-volume centres had lower complication rates.

Objective

  • To develop a nomogram to predict severe adverse events (AEs) after ureteroscopic lithotripsy (URSL) including the effects of operative duration and hospital volume.

Patients and Methods

  • We identified patients undergoing URSL from the Japanese Diagnosis Procedure Combination database between 2007 and 2010, and defined severe adverse events as (i) in-hospital mortality; (ii) postoperative medication including catecholamine, γ globulin, protease inhibitors, medications for disseminated intravascular coagulation and transfusion; and (iii) postoperative interventions including percutaneous nephrostomy, central vein catheterisation, intensive care unit, dialysis, mechanical cardiopulmonary support.
  • Univariate and multivariate logistic regression models addressed the occurrence of severe AEs.

Results

  • Of 12 372 patients, 296 patients (2.39%) had severe AEs. Multivariate analysis showed a positive linear trend of operative duration and severe AEs (odds ratio [OR] 1.58 in 90–119 min to OR 4.28 in ≥210 min compared with ≤ 59 min; each P < 0.05) and an inverse relationship between hospital volume and severe AEs (OR 0.64 in ≥39 URSLs/year compared with ≤ 15 URSLs/year; P = 0.004) with adjustment for other significant factors including sex, age, Charlson comorbidity index, type of anaesthesia and type of admission.
  • A nomogram and a calibration plot based on these results were well-fitted to predict a probability between 0.01 and 0.10 (concordance index 0.677).

Conclusion

  • Severe AEs after URSL were associated with longer operative duration and lower hospital volume, and were accurately predicted using the present nomogram.
Abbreviations
AE

adverse event

CCI

Charlson Comorbidity Index

DPC

Diagnosis Procedure Combination (database)

EAU

European Association of Urology

HV

hospital volume

ICU

Intensive Care Unit

OR

odds ratio

URSL

ureteroscopic lithotripsy

Introduction

Treatment of patients with urolithiasis has been dramatically altered by the introduction of the ureteroscope during the past two decades. Now that ureteroscopes have become more miniaturised, flexible and less invasive, ureteroscopic lithotripsy (URSL) has replaced conventional open surgeries. The reported rates of overall complications and severe complications decreased from 20% and 6.6%around 1985 to 12% and 1.5% in 1992, respectively [1]; this improvement would be attributable to advances in ureteroscope design and accumulated surgical experience.

Despite the lessening of surgical trauma and the improvement of stone-free outcome rates, guidelines on urolithiasis by European Association of Urology (EAU) have mentioned that the overall rate of complications after URSL is 9–25%, including a sepsis rate of 1.1% [2, 3].

To avoid postoperative febrile events, some experts have recommended shortening the duration of water utilisation for ureteroscopy, if possible, to <1 h [4]. Theoretically, this is reasonable, because during the URSL procedure, it is common to use pressurised irrigation, which causes a high intrarenal pressure, resulting in pain, septic shock and traumatic damage to the mucosa of the renal pelvis and ureter owing to irrigation backflow [5].

To date, the relationship between URSL operative duration and overall perioperative complications, consisting mainly of non-febrile events, e.g. ureteric injury was described [6-8]. However, to our knowledge, there has been no quantitative clinical evidence to verify the relationship between URSL operative duration and postoperative severe adverse events (AEs). We infer that the rarity of postoperative septic events has made it difficult to assess this relationship. If there is a strong relationship between operative duration and severe AEs, an appropriate prediction tool for postoperative severe AEs would help physicians to avoid this complication by reducing operative duration.

Also, little is known about the effect of hospital volume (HV) on severe AEs after URSL. HV is the annual operative caseload and is known to be associated with operative outcomes, including for several urological surgeries [9-11].We hypothesised that HV was inversely associated with the occurrence of severe AEs in URSL, as well as extracorporeal shockwave lithotripsy and percutaneous nephrolithotomy [12, 13].

In the present study, we collected a large number of contemporary cases throughout Japan using a Japanese nationwide database: (i) to reveal how operative duration and HV affect the occurrence of severe AEs after URSL and (ii) to develop a nomogram predicting severe AEs.

Patients and Methods

IN the present study, we used the Diagnosis Procedure Combination (DPC) database, a Japanese inpatient administrative claims database, which has been on-going since 2002 as a project for the development of a Japanese case-mix classification system [10, 11, 13, 14]. The database contains: (i) main diagnoses, pre-existing comorbidities at admission and complications after admission, which are coded with International Classification of Diseases and Related Health Problems, 10th Revision codes; (ii) surgical procedures coded with Japanese original surgical codes (K-codes), operative duration, and the date of performing each operation; (iii) discharge status (dead or alive); (iv) a list of drugs and blood products used and the dates of using each drug or product; and (v) use of other resources including Intensive Care Unit (ICU), mechanical ventilation and dialysis, which were coded with Japanese original procedural codes, and the dates of using each resource. The DPC database has collected this data from 1 July to 31 December every year (6 months each year). The numbers of cases in the database were 2.99, 2.86, 2.57 and 3.19 million in 2007, 2008 2009 and 2010, respectively, which represented ≈40% of all acute care inpatient hospitalisations in Japan. Study approval was obtained from the Institutional Review Boards and the Ethic Committee of The University of Tokyo. Given the anonymous nature of the data collection process, informed consent was not required.

In the present study, we consulted the DPC database for data on in-patients undergoing URSL (Japanese operation K-code: K781) between 2007 and 2010. To detect patients with severe AEs, we extracted information on postoperative medications and interventions that suggested postoperative urosepsis or shock. We defined a severe AE as a status involving at least one of the following conditions: (i) in-hospital morality; (ii) postoperative medication including catecholamines (strong vasopressors: dopamine, dobutamine, adrenaline and noradrenaline), γ-globulin products (used for severe septic shock), protease inhibitors (used for shock status: gabexate mesilate, nafamostat mesilate, and ulinastatine), medications for disseminated intravascular coagulation (dalteparin sodium, danaparoid sodium, human anti-thrombin III, thrombomodulin α) and blood transfusion; and (iii) postoperative interventions consisting of percutaneous nephrostomy (K775), central vein catheterisation (Japanese procedure code: G005.2), ICU admission (A301), dialysis support (J038.x–039, J041.x–042), mechanical ventilation (J044.1, J045.x) and cardiac support including chest compression (J046), use of defibrillator (J047), open cardiac massage (K545), intra-aortic balloon pumping (K600), and use of extracorporeal circulation devices such as percutaneous cardiopulmonary support (K602) and ventricular assisting devices (K603). We were able to differentiate between preoperative and perioperative treatments because the DPC administrative data included information on the dates when each drug, device or procedure was used on the individual patients.

Other extracted data were age, sex, comorbidities at admission, type of anaesthesia (epidural, spinal or general anaesthesia), operative duration, type of admission (elective, emergent without ambulance transfer, and emergent with ambulance transfer) and type of hospital (academic or non-academic).

Comorbidities at admission were converted into Charlson Comorbidity Index (CCI) based on the Quan's protocol, which is a well-known pre-existing risk adjuster widely used to weigh burden of comorbidities [15].

HV was determined using the unique hospital identifier and was categorised into three groups (i.e. low-, medium- and high-volume) according to tertile volumes.

Operative duration included time under epidural or spinal anaesthesia, management of general anaesthesia and preparation for monitoring and positioning.

We excluded patients who: (i) lacked data of operative duration, (ii) preoperatively received medications or interventions which met our severe AE criteria and/or (iii) underwent any operation other than URSL and placement/removal of a ureteric stent (K783.2, K783.3)

A univariate logistic regression analysis was performed with each covariate for the prediction of severe AE. We then constructed a multivariate logistic regression model excluding insignificant factors at a <5% significance level in the univariate model. Based on the results, we built a nomogram to predict the occurrence of severe AEs. Internal validation was performed via a bootstrap method with 1000 resamples and a calibration plot was derived to evaluate the relationship between predicted probabilities by the nomogram and the observed rates [16].

In univariate comparisons, categorical variables were compared by chi-square tests and continuous variables were compared by Mann–Whitney U-tests. The threshold for significance was a value of P < 0.05. Statistical analyses were conducted using IBM SPSS version 19.0 (IBM SPSS, Armonk, NY, USA) and a nomogram was built by R version 2.11.1 (R Foundation for Statistical Computing, Vienna, Austria) with the rms library [17, 18].

Results

Among 11.6 million inpatients in the DPC database, in the period 2007–2010, we identified 14 270 patients who underwent URSL. After the exclusion process, 12 372 eligible patients from 625 hospitals remained. Of these, 296 patients (2.39%) met the severe AE criteria including eight fatalities (0.06%). Postoperative medication with catecholamines accounted for 81% (240/296) of all severe AEs. HV levels of low-, medium- and high-volume were defined as ≤15 (n = 3840), 16–38 (n = 4210) and ≥39 URSLs/year (n = 4322), respectively. Table 1 shows the patient characteristics and details of severe AEs.

Table 1. Characteristics of patients and distribution of severe AEs after URSL.
Variablen(%)
  1. CV, central vein catheter; DIC, disseminated intravascular coagulation; ICU, intensive care unit; PCN, percutaneous nephrostomy.
Total12 372(100.00)
Sex:  
Male7 918(64.00)
Female4 454(36.00)
Age, years:  
≤595 902(47.70)
60–693 330(26.92)
70–792 342(18.93)
≥80798(6.45)
CCI:  
09 484(76.66)
11 837(14.85)
2767(6.20)
≥3284(2.30)
Type of anaesthesia:  
epidural or spinal7 219(58.35)
General5 153(41.65)
Operative duration, min:  
≤593 044(24.60)
60–893 422(27.66)
90–1192 446(19.77)
120–1491 614(13.05)
150–179861(6.96)
180–209448(3.62)
≥210537(4.34)
Hospital volume, URSLs/year:  
Low: ≤ 153 840(31.04)
Medium: 16–384 210(34.03)
High: ≥394 322(34.93)
Type of admission:  
Elective11 254(90.96)
Emergency without ambulance transfer955(7.72)
Emergency with ambulance transfer163(1.32)
Season:  
July–September6 699(54.15)
October–December5 673(45.85)
Type of hospital:  
Non-academic10 545(85.23)
Academic1 827(14.77)
Details of severe AEs
Overall296(2.39)
Mortality8(0.06)
Postoperative medications:  
Transfusion45(0.36)
Catecholamine240(1.94)
γ globulin46(0.37)
Protease inhibitor45(0.36)
Anti-DIC medication15(0.12)
Postoperative interventions:  
PCN11(0.09)
CV insertion27(0.22)
ICU78(0.63)
Dialysis18(0.15)
Mechanical ventilation17(0.14)
Cardiac support5(0.04)

Figure 1 shows the relationship between operative duration and each severe AE. Generally, incidences gradually increased with increasing operative duration.

Figure 1.

Relationship between severe AEs (sAEs) and URSL operative duration. Frequencies of severe AEs gradually increased according to operative duration. CV, central vein catheter; DIC, disseminated intravascular coagulation; ICU, intensive care unit; PCN, percutaneous nephrostomy.

Table 2 presents univariate and multivariate logistic regression analyses on severe AEs. The univariate analysis showed that longer operative duration and lower HV were significantly associated with severe AEs, as are sex, age, CCI, type of anaesthesia and type of admission. Season and type of hospital did not reach significance. A multivariate analysis showed a linear trend positive relationship between operative duration and severe AEs (odds ratio, OR 1.58 in 90–119 min to 4.28 in ≥210 min compared with ≤ 59 min; each P < 0.05). Inverse HV-outcome relationship was also clarified. Other significant risk factors were female sex, older age, higher CCI, general anaesthesia and emergent admission.

Table 2. Univariate and multivariate logistic regression models for severe AEs after URSL.
 Severe AEs, %Univariate modelMultivariate model
OR(95% CI)POR(95% CI)P
Sex (vs male)2.05 Reference    
Female3.011.48(1.18–1.87)0.0011.42(1.11–1.80)0.004
Age, years (vs ≤ 59)1.76 Reference  Reference 
60–692.551.46(1.09–1.95)0.0111.32(0.98–1.77)0.065
70–792.951.71(1.26–2.33)0.0011.35(0.98–1.86)0.064
≥804.772.82(1.93–4.11)<0.0012.10(1.42–3.12)<0.001
CCI (vs. 0)1.98 Reference  Reference 
13.271.69(1.26–2.27)0.0011.34(0.99–1.82)0.055
24.172.18(1.48–3.19)<0.0011.79(1.21–2.64)0.004
≥35.632.98(1.77–5.04)<0.0012.26(1.32–3.87)0.003
Type of anaesthesia (vs. epidural or spinal)1.76 Reference    
General anaesthesia3.281.87(1.48–2.36)<0.0011.38(1.08–1.78)0.010
Operative duration, min (vs ≤ 59)1.25 Reference  Reference 
60–891.991.60(1.08–2.39)0.0211.46(0.97–2.19)0.070
90–1192.251.82(1.20–2.76)0.0051.58(1.02–2.43)0.039
120–1493.222.58(1.69–3.95)<0.0012.24(1.43–3.49)<0.001
150–1793.482.76(1.69–4.50)<0.0012.40(1.44–4.00)<0.001
180–2094.463.70(2.13–6.41)<0.0012.90(1.63–5.16)<0.001
≥2106.155.18(3.22–8.34)<0.0014.28(2.57–7.12)<0.001
Hospital volume, URSLs/year (vs low: ≤15)3.23 Reference  Reference 
Medium: 16–382.450.75(0.57–0.98)0.0340.79(0.61–1.04)0.090
High: ≥391.600.49(0.36–0.66)<0.0010.64(0.47–0.87)0.004
Type of admission (vs elective):2.24 Reference  Reference 
Emergency without ambulance transfer3.461.58(1.09–2.28)0.0161.78(1.21–2.61)0.003
Emergency with ambulance transfer6.793.19(1.71–5.95)<0.0012.78(1.46–5.29)0.002
Season (vs July–September)2.52 Reference    
October–December2.240.87(0.69–1.10)0.246   
Type of hospital (vs non-academic)2.35 Reference    
Academic2.631.10(0.80–1.51)0.551   

Based on these results, we built a nomogram that graphically shows the multivariate impact of each variable (Fig. 2). The concordance index of this model was 0.677. Calibration plots are shown in Fig. 3. The differences between the observed and predicted probabilities were within 0.005 for 90% of patients.

Figure 2.

Nomogram to predict the probability of severe AE after URSL. In the nomogram, the patient's value of a given parameter was plotted on the appropriate scale and vertical lines were drawn up to the top line to obtain the associated scores. After repeating this process for each parameter, all scores were summed to obtain the total points. The total point score on the total points line (second from the bottom) were plotted and a vertical line was drawn down to the bottom line. The corresponding value represents a predicted probability of severe AEs after URSL for the given patient.

Figure 3.

Nomogram calibration. The ‘ideal’ line at 45 ° (dashed line) indicates the ideal nomogram reference line. The ‘apparent’ line (dotted line) was calculated directly from the dataset. The ‘bias-corrected’ line (continuous line) is an adjusted line by bootstrap with 1000 resamples.

Discussion

In the present study, we analysed 12 372 URSL patients in the DPC database between 2007 and 2010, and assessed how operative duration and HV affected incidence of severe AEs after URSL, with adjustments for several background information items. We also developed a nomogram predicting the incidence of severe AEs.

We focused on severe AEs by confining the definition of severe AEs to medications and interventions necessary for cardiopulmonary support or treatment of severe septic status. We think that postoperative demands for these medications or interventions reflected actual life-threatening conditions. The overall incidence of severe AE was 2.38%, including eight deaths (0.06%), which is consistent with the postoperative sepsis rate of 3% in the 2007 guidelines edited by the EAU and the AUA Panel and 1.1% in EAU guidelines 2010 [2, 19].

Presumably many urologists have experienced or witnessed postoperative shock-related events after URLS and may recognise that long operative duration increases such risks as an empirical lesson. However, to date, the quantitative evidence to show such a tendency has been lacking and current guidelines have failed to indicate an appropriate operative duration to avoid severe AEs [2, 19].

In the present study, we succeeded in clarifying that severe AE risk increases with each additional 30-min increase in operative duration. Compared with the group with an operative duration of ≤ 59 min, every group with an operative duration of ≥90 min had a significantly higher risk (OR ranged from 1.54 to 4.07). The low severe AE incidence (2.39%) would make it difficult to show a significant association between operative duration and severe AEs with a standard case series. We overcame this difficulty by using a nationwide database and collecting a large number of contemporary cases from 625 hospitals.

As far as complications of ureteric injury are concerned, some reports have described a relationship between operative duration and complications [7, 8]. For paediatric cases, Dogan et al. [6] reported that operative duration was the sole significant factor affecting perioperative complications, including febrile events.

An inverse HV–outcome relationship is another finding. As far as we know, the impact of HV has not been assessed in URSL. The multivariate analysis showed that medium-volume hospitals (16–38 URSLs/year) had an ≈20% reduction in severe AEs compared with low-volume ones (≤15/year), and that high-volume (≥39/year) hospitals achieved an additional 20% reduction in severe AEs. URSL requires delicate technique and we agree with previous reports that indicate that safe and successful outcome of URSL depends on the experience of the attending surgeon [8, 20-22]. We could not assess effect of surgeon volume directly because the DPC database does not have a surgeon identifier, but HV, which we adopted in the present study, is also a widely-used and reliable quality indicator for surgery [9-11, 13]. Theoretically, HV involves not only surgeon skill but also the experience of all the staff, including anaesthesiologists and nurses, and comprehensive hospital resources [10].

In the multivariate analysis, several other independent factors for severe AEs were also confirmed. It is rational that elderly or more comorbid patients who need to be hospitalised urgently are more likely to have severe AEs, although previous studies had failed to clarify the significances of these factors in multivariate analysis [7, 8, 20-22]. Females could be at higher risk because women are more vulnerable to UTIs than men [23]. Interestingly, a previous report showed that complications relating to ureteric injury occur more frequently in males than in females, probably due to urethral anatomical differences [7].

The present nomogram enables easy calculation of individualised prediction of patient outcome. Physicians can recognise the risk for a severe AE before URSL, make a more informative explanation to their patients, and plan an appropriate operative duration, especially with training of residents. The calibration plot (Fig. 3) showed that our nomogram was well fitted to the observed data between the probabilities of 0.01–0.10. Out of that range, our nomogram overestimates the risk. We think that a 10% severe AE rate is a sufficiently high upper end of the range for clinical practice, even though users should take into account this potential overestimation.

Several limitations must be considered when interpreting the present results. First, the analysis of an administrative claims database could lead to an underestimate or overestimate of comorbidities. Second, we lacked information about potential risk factors for severe AEs including stricture or twist of the ureter, stone details (size, impaction, or location), type of ureteroscope (flexible, semi-rigid or rigid), use of ureteric access sheaths, method of lithotripsy (laser, shockwave, etc.), and preoperative urine contamination [4, 8, 21]. However, we consider that stone status or difficulty of access to a stone is reflected in operative duration to a certain degree. Third, we could not measure an accurate intra-ureteric time. Finally, because there is a bias toward large hospitals in the DPC database [14], the possibility of sampling bias cannot be completely excluded.

Despite these limitations, we think the present analysis and nomogram based on a large number of contemporary cases provides the best evidence on postoperative severe AEs after URSL available today.

In conclusion, longer operative durations, especially those of >90 min, and less-experienced hospitals were highly correlated with an increased incidence of severe AEs after URSL, as were female sex, older age, comorbidity score, need for general anaesthesia, and emergent admission. The present nomogram accurately predicted severe AEs after URSL.

Acknowledgments

This study was funded by a Grant-in-Aid for Research on Policy Planning and Evaluation from the Ministry of Health, Labour and Welfare, Japan (Grant number: H22-Policy-031), by a Grant-in-Aid for Scientific Research B (Grant number: 22390131) from the Ministry of Education and Science and by the Funding Programme for World-Leading Innovative R&D on Science and Technology (FIRST programme) from the Council for Science and Technology Policy, Japan (Grant number: 0301002001001).

Conflict of Interest

None declared.

Ancillary