Variation in length of hospital stay after lung cancer surgery in the Netherlands† (original) (raw)

Abstract

OBJECTIVES

Length of stay (LOS) in the hospital after lung cancer surgery is influenced by patient characteristics, tumour characteristics, surgical technique and perioperative care. Our objective was to determine whether there were variation in LOS between hospitals that could not be accounted for by these known parameters. Residual variation in LOS would suggest important differences in perioperative care protocols and discharge criteria.

METHODS

This study analysed data from the Netherlands National Cancer Registry (NNCR) on 10 195 anatomical lung resections for primary lung cancer from 2010 to 2015. Multivariable analysis was performed for multiple factors, using hierarchical linear regression analysis of the mean LOS. Information on comorbidity and socio-economic status was not available. Association between LOS and postoperative mortality was evaluated in multivariable logistic regression.

RESULTS

The median LOS was 7 days (interquartile range 5–10 days), and the mean LOS was 8.3 days. LOS was negatively affected by larger resections, open surgery and advancing age. Histology and tumour stage had little influence. Overall, 30-day and 90-day mortality were 2.1% and 3.8%, respectively; 1.7% and 3.3% (not significant) in the group of hospitals with shorter LOS. After case-mix correction, residual between-hospital variation in the mean LOS was observed, ranging from 1.5 days shorter to almost 2.5 days longer.

CONCLUSIONS

A clinically relevant between-hospital variation in LOS after lung cancer surgery is observed in the Netherlands. Although residual confounding by comorbidity or socio-economic status cannot be excluded, this variation is deemed to be largely due to differences in perioperative care protocols. Evaluation of best practices can help to improve perioperative care for lung surgery patients and optimize LOS.

INTRODUCTION

Perioperative care is aimed at limiting the impact of surgery and promoting optimal recovery after surgery, giving patients the opportunity to return home and resume their daily activities as safely and quickly as possible [1]. Length of stay (LOS) is frequently used as a measure of postoperative recovery, with a short LOS indicating a rapid postoperative recovery [1, 2]. LOS is not only influenced by many patient characteristics and factors related to the operation but is also determined by perioperative treatment protocols and discharge criteria [2–4]. Postoperative treatment protocols focusing on enhanced recovery after thoracic surgery (ERATS) can potentially reduce LOS, complications, readmissions and cost [1, 4–6]. Such an all-encompassing protocol does not exist for thoracic surgery yet.

LOS after anatomical resection for non-small-cell lung cancer (NSCLC) is reported to be influenced by surgical technique, hospital volume, postoperative complications, gender, age, extent of disease, insurance coverage and availability of healthcare services [2, 7]. Netherlands National Cancer Registry (NNCR) records all lung cancer cases in the Netherlands, along with several patient and treatment characteristics mentioned above, enabling analysis of LOS variation between Dutch healthcare providers and comparison to international data. Netherlands is a small, densely populated country, with easily accessible community healthcare facilities, universal healthcare insurance coverage and an even spread of hospitals providing thoracic surgery service. Therefore, geographical and economic influences will be of lesser influence on LOS in the Dutch situation.

We hypothesize that if differences in LOS after anatomical resection for NSCLC are present between hospitals in the Netherlands, which cannot be explained by patient characteristics, tumour characteristics and/or surgical technique, these would reflect differences in perioperative care and discharge criteria [2–4]. A significant variation in LOS would prompt further research into the reasons for this variation, identifying best practices and aiding the development of a standardized ERATS protocol.

MATERIALS AND METHODS

Source

Information regarding patients who underwent lung cancer surgery from 2010 to 2015 was retrieved from the NNCR, after approval by the Privacy Review Board. In accordance with the regulations of the Central Committee on Research involving Human Subjects (CCMO), this type of study does not require approval from an ethics committee in the Netherlands. The NNCR collects data on all cancer patients diagnosed in the Netherlands, based on notification of newly diagnosed malignancies by the national automated pathological archive and of hospital discharge diagnoses. Information on demographics, diagnosis, staging and treatment is extracted routinely from the medical records by specially trained NNCR personnel. Dates of surgery and hospital discharge are recorded; the date of hospital admission prior to surgery is not available. Coding for video-assisted thoracic surgery also includes robotic surgery. Tumour stage is recorded according to the 7th edition of the tumour, node and metastasis (TNM) Classification of Malignant Tumours from the International Union Against Cancer (UICC). Information on survival status is updated annually using a computerized link with the national civil registry. Information on socio-economic status, comorbidity, complications or readmission was not available.

Patient and hospital selection

Patient and hospital selection criteria were applied to improve comparability and relevance of the findings. Patients who underwent anatomical lung resections for invasive NSCLC were included in our study, excluding those with metachronous tumours. Wedge resections were excluded, as were patients operated abroad and patients with missing information on LOS (n = 910). Information on LOS was more frequently missing in the 1st year of the study period. Several hospitals discontinued performing lung resections during the study period. To limit the influence of low volume facilities, hospitals performing less than 20 operations per year after 2012 were excluded from the analyses. A total of 10 195 patients in 46 hospitals were analysed.

Analysis

Statistical analyses were performed using Stata/SE 14.1. LOS was calculated from the day of surgery to day of discharge and is presented as 25, 50, 75 and 90 percentiles. Patients with a LOS more than 30 days were maximized at 30 days to avoid a confounding impact of individual outliers. The median LOS (P50) is thought to reflect the LOS for a typical patient receiving standard clinical care. The interquartile range reflects the extent of the variation in LOS. The P90 reflects LOS for patients with extensive complications or other abnormal circumstances. Patients who died in-hospital during the primary admission (n = 207) were excluded from LOS analyses. Postoperative 30-day and 90-day mortality were tabulated including patients who died in-hospital.

Multivariable analysis of the mean LOS was performed using hierarchical linear regression with the Stata ‘mixed’ command. The analysis included age, gender, year of surgery, extent of surgery, surgical approach, TNM stage and histology as fixed effects, while hospital was included as a random effect. The statistical significance of the fixed parameters was evaluated in backward analysis with the likelihood ratio test. Parameters with a _P_-value >0.05 were eliminated from the model. Significant predictive parameters are represented by the mean difference in LOS and a corresponding 95% confidence interval. Residuals were assessed by quantile plots against a normal distribution. Due to the major sample size, even small differences can be asserted statistically significant.

The posterior estimates of the random effects for individual hospitals, controlling for case mix, were obtained by post-estimation commands. These estimates reflect the mean difference in LOS from the overall mean and were plotted against the total number of operations during the study period to display the amount of residual variation. The significance of the random component was evaluated by a likelihood ratio test comparing the hierarchical model with standard linear regression. Using the posterior estimates of the multivariable analysis, hospitals were categorized into 3 summary groups: medium LOS (n = 20), longer LOS (mean LOS greater than +0.5 days, n = 13) and shorter LOS (mean LOS less than −0.5 days, n = 13).

The association between LOS summary hospital group and 30-day and 90-day postoperative mortality was evaluated using multivariable logistic regression analysis. Age (categorical), gender, extent of surgery, surgical approach, TNM stage and histology were included as case-mix parameters. The prognostic impact of the LOS summary hospital group is represented by odds ratios and 95% confidence intervals, adjusted for the other parameters. The medium LOS group was defined as the reference category. Model fit was assessed by the area under the receiver operating characteristic curve, which was 0.76 and 0.74 for 30-day and 90-day postoperative mortality, respectively.

RESULTS

The median LOS after anatomical lung resection for NSCLC in the Netherlands (2010–2015) was 7 days (interquartile range 5–10 days) (Table 1). The mean LOS was 8.3 days in 2010 and decreased by more than a day from 2010 to 2015, controlling for other predictive factors. Multivariable analysis was necessary to control for concurrent changes. The use of video-assisted thoracic surgery, for example, increased from 18% in 2010 to 59% in 2015. The mean LOS increased with advancing age and was longer for patients with squamous carcinoma. The mean LOS was a day longer after bilobectomy, when compared to LOS after lobectomy or pneumonectomy. After open resection and converted operation, mean LOS increased by 1.3 and 1.8 days, respectively, compared to minimally invasive approaches. TNM stage had a minor influence on LOS, and the results did not differ by gender.

Table 1:

Univariable (25, 50, 75 and 90 percentile) and multivariable analysis of the LOS in days after surgery for lung cancer (n = 9988a)

a

Excluding patients who died in-hospital during the primary admission.

b

Determined by multivariable analysis.

CI: confidence interval; LOS: length of stay; ns: not significant; TNM: tumour, node and metastasis; VATS: video-assisted thoracic surgery.

Table 1:

Univariable (25, 50, 75 and 90 percentile) and multivariable analysis of the LOS in days after surgery for lung cancer (n = 9988a)

a

Excluding patients who died in-hospital during the primary admission.

b

Determined by multivariable analysis.

CI: confidence interval; LOS: length of stay; ns: not significant; TNM: tumour, node and metastasis; VATS: video-assisted thoracic surgery.

Despite controlling for case mix, a clear residual variation in LOS was observed between hospitals (Fig. 1), which was confirmed by formal testing (P < 0.0001). In 3 hospitals, LOS was more than 1.5 days longer than the general mean. In 6 hospitals, the mean LOS was more than a day shorter than the general mean. There was no apparent association between LOS and hospital volume.

Figure 1:

Residual variation between hospitals in the mean LOS, after controlling for age, gender, tumour, node and metastasis stage, type of surgery, surgical approach and histology. LOS: length of stay.

Overall, 30-day and 90-day mortality were 2.1% and 3.8%, respectively. In the group of hospitals with shorter LOS, outcomes were better, though not statistically significant (Table 2). Mortality rates were slightly higher in the group of hospitals with longer LOS.

Table 2:

Association between hospital discharge practices and the 30-day and 90-day postoperative mortality

a

Multivariable analysis controlling for age, gender, TNM stage, type of surgery, surgical approach and histology.

CI: confidence interval; LOS: length of stay; TNM: tumour, node and metastasis.

Table 2:

Association between hospital discharge practices and the 30-day and 90-day postoperative mortality

a

Multivariable analysis controlling for age, gender, TNM stage, type of surgery, surgical approach and histology.

CI: confidence interval; LOS: length of stay; TNM: tumour, node and metastasis.

DISCUSSION

In this retrospective database analysis, we observed an unexplained difference in postoperative LOS after lung resection for NSCLC between hospitals, ranging from a mean LOS of 6.7 days to 10.5 days. After case-mix correction, residual between-hospital variation in the mean LOS was observed, ranging from 1.5 days shorter to almost 2.5 days longer. Because LOS was calculated after correcting for patient, treatment and tumour characteristics, we hypothesize that this variation is largely attributable to differences in postoperative care programmes [3].

Using LOS as a measure for quality of perioperative care has limitations. Use of the median LOS is an approximation of the LOS in a typical case and does not take all treatment and patient factors into account. Because LOS is dependent on many factors, care should be taken not to equate short LOS with good perioperative care, even after correction for known case-mix variables [2, 7].

Several of these variables were analysed in our series. In contrast to previous publications, hospital volume and patient gender did not seem to influence LOS [2]. Insurance and geographical influences were not analysed, considering the small size of the country, the distribution of hospitals providing lung cancer surgery and universal healthcare insurance coverage. Age, extent of resection and surgical approach were confirmed as important factors determining LOS. The increased use of video-assisted thoracic surgery in the Netherlands from 18% to 59% between 2010 and 2015 coincided with a decrease in median LOS from 8 to 7 days during the study period. Postoperative LOS after anatomical resection for NSCLC is comparable to data from the USA and compares favourably to the European Society of Thoracic Surgeons (ESTS) data [7, 8] (Table 3).

Table 3:

Outline of studies reporting postoperative LOS

ESTS: European Society of Thoracic Surgery; IQR: interquartile range; LOS: length of stay; NCDB: National Cancer Database; STS: Society of Thoracic Surgery; VATS: video-assisted thoracic surgery.

Table 3:

Outline of studies reporting postoperative LOS

ESTS: European Society of Thoracic Surgery; IQR: interquartile range; LOS: length of stay; NCDB: National Cancer Database; STS: Society of Thoracic Surgery; VATS: video-assisted thoracic surgery.

A short LOS, however, could increase the risk of patients being discharged before complications become apparent. A recent publication showed no increase in readmissions after a short LOS, when short LOS was the norm in the specific hospital. However, when patients were discharged early from a hospital where a longer LOS was common, the readmission rates increased [7]. This finding underscores the notion that a short LOS is a consequence of a good perioperative care programme and should not be considered as a goal in itself [1, 3, 7]. Apart from increased readmission rates, increased post-discharge mortality after a short LOS would signify detrimental effect of an early discharge protocol [7, 9]. In our analysis, shorter LOS was not associated with a higher 30-day or 90-day mortality.

Limitations

Our study has several limitations due to its retrospective design and the limited data that were available within the NNCR. Detailed information on comorbidity is not available in the NNCR, therefore limiting the extent of case-mix correction. The lack of information on socio-economic status may also introduce confounding, despite the compulsory healthcare insurance and highly accessible healthcare system in the Netherlands. Lack of readmission data and complication data limits our ability to fully appreciate the relationship between LOS and quality of postoperative recovery. Careful evaluation of morbidity, mortality and readmission rates in relation to case-mix-adjusted LOS should be applied when trying to identify best practices in perioperative care pathways and discharge criteria.

CONCLUSION

Limiting surgical impact and promoting rapid and uneventful recovery after surgery are key points in surgical care. As ERATS benefits patients by enabling them to return to their homes, families and daily activities as safely and quickly as possible, it may also reduce complication rates and hospital LOS and hence potentially reduce pressure on limited healthcare budgets [1, 3, 5]. Our analysis of the NNCR data shows a clinically significant difference in case-mix-adjusted LOS after anatomical lung resection for NSCLC, without signs of increased mortality related to early discharge. These findings justify further research into the differences in postoperative treatment protocols and discharge criteria leading to these results, facilitating the development of an optimal ERATS protocol.

Conflict of interest: none declared.

REFERENCES

1

Ljungqvist

O

,

Scott

M

,

Fearon

KC.

Enhanced recovery after surgery

.

JAMA Surg

2017

;

152

:

292

–

8

.

2

Giambrone

GP

,

Smith

MC

,

Wu

X

,

Gaber-Baylis

LK

,

Bhat

AU

,

Zabih

R

et al.

Variability in length of stay after uncomplicated pulmonary lobectomy: is length of stay a quality metric or a patient metric?

Eur J Cardiothorac Surg

2016

;

49

:

e65

–

70

.

3

Krell

RW

,

Girotti

ME

,

Dimick

JB

,

Mw H

JM

,

Rs

M.

Extended length of stay after surgery

.

JAMA Surg

2014

;

149

:

815

–

20

.

4

Senturk

JC

,

Kristo

G

,

Gold

J

,

Bleday

R

,

Whang

E.

The development of ERAS across surgical specialties

.

J Laparoendosc Adv Surg Tech A

2017

;

27

:

863

–

70

.

5

Paci

P

,

Madani

A

,

Lee

L

,

Mata

J

,

Mulder

DS

,

Spicer

J

et al.

Economic impact of an enhanced recovery pathway for lung resection

.

Ann Thorac Surg

2017

;

104

:

950

–

7

.

6

Fiore

JF

,

Bejjani

J

,

Conrad

K

,

Niculiseanu

P

,

Landry

T

,

Lee

L

et al.

Systematic review of the influence of enhanced recovery pathways in elective lung resection

.

J Thorac Cardiovasc Surg

2016

;

151

:

708

–

15.e6

.

7

Rosen

JE

,

Salazar

MC

,

Dharmarajan

K

,

Kim

AW

,

Detterbeck

FC

,

Boffa

DJ.

Length of stay from the hospital perspective. Practice of early discharge is not associated with increased readmission risk after lung cancer surgery

.

Ann Surg

2017

;

266

:

383

–

8

.

8

Seder

CW

,

Salati

M

,

Kozower

BD

,

Wright

CD

,

Falcoz

PE

,

Brunelli

A

et al.

Variation in pulmonary resection practices between the Society of Thoracic Surgeons and the European Society of Thoracic Surgeons General Thoracic Surgery Databases

.

Ann Thorac Surg

2016

;

101

:

2077

–

84

.

9

Salati

M

,

Brunelli

A

,

Xiumè

F

,

Refai

M

,

Pompili

C

,

Sabbatini

A.

Does fast-tracking increase the readmission rate after pulmonary resection? A case-matched study

.

Eur J Cardiothorac Surg

2012

;

41

:

1083

–

7

.

Author notes

Presented at the 31st Annual Meeting of the European Association for Cardio-Thoracic Surgery, Vienna, Austria, 7–10 October 2017.

© The Author(s) 2018. Published by Oxford University Press on behalf of the European Association for Cardio-Thoracic Surgery. All rights reserved.