Abstract
Background
As value-based care models continue to gain emphasis, along with the need for improved profiling across the continuum of lung cancer care, a better understanding of geographic variation in utilization of services surrounding episodes of care is needed.
Methods
In this retrospective cohort study of patients undergoing lung cancer resection from 2017 to 2019, we examined geographic variation in utilization of services surrounding episodes of lung cancer resection. We utilized hierarchical logistic regression models to determine risk-adjusted utilization of services. This study utilized inpatient and ambulatory databases across 4 states: New Jersey, Pennsylvania, Florida, and Maryland. All patients undergoing lung cancer resection were included. The primary outcome was risk-adjusted utilization of services.
Results
Mean risk-adjusted utilization of ambulatory procedures across all hospital referral regions (HRRs) was 34.1% (95% CI 30.7%-37.6%), while the individual HRR utilization varied from 10.9% to 54.9% (P < .01). Mean risk-adjusted utilization of inpatient admissions in the 6 months prior to surgery was 15.3% (95% CI 13.9%-16.7%), ranging from 7.4% to 24.7% (P = .07) across HRRs. Finally, mean risk-adjusted utilization of inpatient hospitalizations in the 6 months following surgery was 19.4% (95% CI 17.7-21.0%), ranging from 10.0% to 33.6% (P = .19) across HRRs.
Conclusions
Overall, we observed that utilization of ambulatory services varied significantly across HRRs, while inpatient utilization did not demonstrate significant variation. Given these findings, there may be geographic drivers of variation in the utilization of services surrounding lung cancer resection.
In Short.
-
▪
This retrospective cohort study, including 13,548 patients who underwent lung cancer resection across 41 hospital referral regions in 4 states, reveals risk-adjusted geographic variation in the utilization of ambulatory resources surrounding episodes of lung cancer resection.
-
▪
This study demonstrates that geographic variation in lung cancer treatment and resource utilization exists, and highlights that understanding the key drivers of variation may identify opportunities to improve the value of care.
There is known variation in lung cancer care. For example, length of stay after a minimally invasive lung cancer resection has been shown to vary from a median of 3 to 8 days at the hospital level.1 However, hospitals and health systems have not been fully evaluated for the value of care provided across the disease process. For instance, previous research has continued to profile and assess variation in silos, such as cost, mortality, and readmission.2,3 As a result, there is limited understanding of the key drivers of variation across the full episode of lung cancer from diagnosis to postoperative care.
Understanding the key drivers of variation may identify opportunities to improve the quality of care. Subsequently, this could help create profiles of hospitals, health systems, or geographic areas that reflect the full continuum of care. The focus on improving the quality of lung cancer surgery care needs to translate across the entire disease process. Identifying key areas of variation in utilization may aid in discriminating performance.
In this context, we utilized state inpatient and ambulatory databases to identify geographic variation in the utilization of ambulatory procedures and inpatients services in the 6 months before and after lung cancer surgical resection. Utilization was calculated at the level of hospital referral regions (HRRs) and compared across geographic areas. We hypothesized that geographic variation exists in the utilization of ambulatory and inpatient services surrounding lung cancer resection.
Patients and Methods
Design and Study Population
This was a retrospective cohort study of Pennsylvania, Maryland, Florida, and New Jersey inpatient and ambulatory databases across a 2-year period, either 2017-2018 or 2018-2019. Maryland, Florida, and New Jersey data were provided by Healthcare Cost and Utilization Project, and Pennsylvania data by Pennsylvania Health Care Cost Containment Council. Geographic distribution was determined by Dartmouth Atlas Hospital Referral Regions.4 HRRs were chosen based on criteria including hospital service areas (HSAs), at least 1 hospital performing major cardiovascular and neurosurgical procedures, and a minimum population of 120,000.5 The inclusion criteria was all lung cancer patients with a surgical resection of their lung cancer between the age of 18 and 99 years. This study followed Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines.
Data Source and Covariates
Patients were grouped based on HRR and identified in the databases. Utilization was measured by determining the number of ambulatory pulmonary procedures done in the 6 months prior to the index surgery. Inpatient admissions in the 6 months prior to and after the index surgery were identified. Additional patient characteristics obtained from the databases were age, sex, race, ethnicity, method of surgical resection, and Charlson-Deyo Comorbidity Index Score. This information was retrieved utilizing International Classification of Diseases, 10th revision Clinical Modification and Procedure Coding System codes (Supplemental Material).
Study Exposure and Outcomes
The primary outcome was risk-adjusted utilization of services in the 6 months prior to and after the index surgery. Ambulatory utilization was subdivided into bronchoscopy, pulmonary biopsy, and miscellaneous pulmonary procedures. Inpatient utilization was identified by additional inpatient admissions prior to and after the index surgery. Risk-adjusted utilization was compared based on HRRs. In a subanalysis, to identify whether there is variation within HRRs, all the HSAs within a high utilizer HRR and low utilizer HRR were analyzed.
Statistical Analyses
A logistic regression model was created utilizing method of surgical resection, age, sex, and Charlson-Deyo Comorbidity Index Score. These independent variables passed a collinearity test. The descriptive statistics were analyzed using analysis of variance and χ2 tests. The study was approved by the institutional review board of Thomas Jefferson University Hospital. Statistical analyses were done using Stata-SE v. 17.0 (StataCorp LLC).
Results
Patient Treatment and Characteristics
A total of 13,548 patients across 41 HRRs underwent lung cancer resection during our study period. Patients had a mean age of 68 (SD 9) years and 7424 (55%) were female. The HRRs were split into quartiles by ambulatory utilization. Race, ethnicity, and Charlson-Deyo Comorbidity Index Score were similar between quartiles. The predominant surgical approaches were video-assisted thoracoscopic surgery lobectomy (39%), video-assisted thoracoscopic surgery segmentectomy/wedge (26%), and open lobectomy (23%) (Table 1).
Table 1.
Patient Characteristics by Hospital Referral Region Risk-Adjusted Ambulatory Utilization Quartiles
|
Characteristic |
Hospital Referral Region Ambulatory Utilization Quartiles |
P Value | |||
|---|---|---|---|---|---|
|
1 (Lowest) |
2 |
3 |
4 (Highest) |
||
| (10.9%-23.5%) | (23.6%-36.9%) | (37.0%-41.1%) | (41.2%-55.0%) | ||
| Number of HRRs | 11 | 10 | 10 | 10 | |
| Number of hospitals | 125 | 133 | 117 | 129 | |
| Number of patients | 3840 | 4053 | 2711 | 2944 | |
| Age, y | 69.1 ± 9.2 | 68.4 ± 8.8 | 68.7 ± 8.9 | 69.1 ± 9.2 | <.01 |
| Race | <.01 | ||||
| White | 2747 (71.5) | 3360 (82.9) | 2282 (84.2) | 2569 (87.3) | |
| Black | 266 (6.9) | 457 (11.3) | 152 (5.6) | 206 (7.0) | |
| Asian | 74 (1.9) | 48 (1.2) | 34 (1.3) | 27 (0.9) | |
| Native American | 2 (0.1) | 6 (0.1) | 2 (0.1) | 4 (0.1) | |
| Other | 725 (18.9) | 168 (4.1) | 235 (8.7) | 125 (4.2) | |
| Ethnicity | <.01 | ||||
| Hispanic | 677 (17.6) | 76 (1.9) | 174 (6.4) | 100 (3.4) | |
| Not Hispanic | 3054 (79.5) | 3946 (97.4) | 2513 (92.7) | 2801 (95.1) | |
| Sex | .47 | ||||
| Male | 1724 (44.9) | 1813 (44.7) | 1262 (46.6) | 1325 (45.0) | |
| Female | 2116 (55.1) | 2240 (55.3) | 1449 (53.4) | 1619 (55.0) | |
| Comorbidity scorea | <.01 | ||||
| 0 | 1048 (27.3) | 988 (24.4) | 644 (23.8) | 751 (25.5) | |
| 1 | 1213 (31.6) | 1233 (30.4) | 879 (32.4) | 908 (30.8) | |
| 2 | 562 (14.6) | 601 (14.8) | 395 (14.6) | 442 (15.0) | |
| ≥3 | 1017 (26.5) | 1231 (30.4) | 793 (29.3) | 843 (28.6) | |
| Surgical approach | <.01 | ||||
| VATS seg/wedge | 1126 (29.3) | 1102 (27.2) | 579 (21.4) | 779 (26.5) | |
| Open seg/wedge | 339 (8.8) | 366 (9.0) | 358 (13.2) | 273 (9.3) | |
| VATS lobectomy | 1626 (42.3) | 1604 (39.6) | 899 (33.2) | 1181 (40.1) | |
| Open lobectomy | 708 (18.4) | 915 (22.6) | 824 (30.4) | 675 (22.9) | |
| Pneumonectomy | 41 (1.1) | 66 (1.6) | 51 (1.9) | 36 (1.2) | |
Values are presented as mean ± SD or n (%), unless otherwise noted.
HRR, hospital referral region; seg, segmentectomy; VATS, video-assisted thoracoscopic surgery.
Charlson-Deyo Comorbidity Score.
Ambulatory Utilization
Mean risk-adjusted utilization of ambulatory procedures across HRRs was 34.1% (95% CI 30.7%-37.6%), while individual HRR utilization varied from 10.9% to 54.9% (P < .01) (Figure 1). When split into quartiles by ambulatory utilization (Table 2), the ambulatory utilization varied from lowest to highest at 20.2% (95% CI 17.7%-22.8%, 31.7% (95% CI 28.3%-35.0%), 38.8% (95% CI 38.0%-39.5%), and 47.2% (95% CI 43.3%-51.1%), with P < .01 (Figure 2). Mean bronchoscopy utilization varied across quartiles from 11.2% to 21.1% (P < .01), and pulmonary biopsy utilization varied across quartiles from 10.5% to 31.6% (P < .01).
Figure 1.
Risk-adjusted ambulatory utilization rate split into quartiles by hospital referral regions (HRR) in Florida, Maryland, Pennsylvania, and New Jersey.
Table 2.
Risk-Adjusted 6-Month Inpatient and Ambulatory Utilization Rates by Hospital Referral Region Quartiles
|
Characteristic |
Hospital Referral Region Ambulatory Utilization Quartiles |
P Value | ||||
|---|---|---|---|---|---|---|
| Overall | 1 (Lowest) | 2 | 3 | 4 (Highest) | ||
| Number of HRRs | 41 | 11 | 10 | 10 | 10 | |
| Ambulatory utilization, % | ||||||
| All procedures | 34.1 (30.7-37.6) | 20.2 (17.7-22.8) | 31.7 (28.3-35.0) | 38.8 (38.0-39.5) | 47.2 (43.3-51.1) | <.01 |
| Bronchoscopy | 16.5 (14.4-18.6) | 11.2 (8.9-13.5) | 15.9 (10.9-20.9) | 18.3 (13.4-23.3) | 21.1 (17.3-24.9) | <.01 |
| Pulmonary biopsy | 20.5 (17.3-23.8) | 10.5 (8.5-12.6) | 18.0 (13.5-22.4) | 23.0 (17.0-29.0) | 31.6 (25.2-38.0) | <.01 |
| Miscellaneous | 1.6 (1.1-2.1) | 0.7 (0.3-1.0) | 1.2 (0.7-1.8) | 2.1 (0.9-3.2) | 2.6 (0.8-4.3) | .04 |
| Inpatient utilization,a% | ||||||
| Admissions prior | 15.3 (13.9-16.7) | 12.6 (9.6-15.5) | 17.0 (13.4-20.5) | 17.1 (14.3-19.8) | 14.9 (12.6-17.2) | .07 |
| Admissions after | 19.4 (17.7-21.0) | 17.2 (12.4-22.0) | 19.6 (15.6-23.6) | 21.6 (19.2-24.0) | 19.3 (16.5-22.1) | .19 |
Values are presented as mean (95% CI).
HRR, hospital referral region.
Admissions in the 6-month surrounding the admission for the index surgery.
Figure 2.
Risk-adjusted ambulatory utilization ranked in ascending order of utilization by hospital referral regions (HRRs). Horizontal line represents average ambulatory utilization across all HRRs. Each HRR’s vertical bar represents 95% CIs surrounding their average utilization rate.
Inpatient Utilization
Mean risk-adjusted utilization of inpatient admissions in the 6 months prior to surgery was 15.3% (95% CI 13.9%-16.7%), ranging from 7.4% to 24.7% (P = .07) across HRRs. The highest ambulatory utilization quartile (quartile 4) had the second lowest inpatient utilization at 14.9% (95% CI 12.6%-17.2%). Finally, mean risk-adjusted utilization of inpatient hospitalization in the 6 months following surgery was 19.4% (95% CI 17.7%-21.0%), ranging from 10.0% to 33.6% (P = .19) across HRRs. Quartile 4 had the second lowest inpatient utilization after the index surgery at 19.3% (95% CI 16.5%-22.1%), while having the most ambulatory utilization.
Analysis of HSAs
In a subanalysis of the data, the HSAs from 1 high and low utilizer HRR were evaluated for ambulatory utilization. Mean risk-adjusted utilization of ambulatory procedures was 30.2% (95% CI 25.0%-35.4%), while it ranged from 0% to 49% in the low utilizer HRR and 0% to 61% in the high utilizer HRR. Though not significant, there are trends towards variation in the ambulatory utilization within HRRs at the level of HSAs (Supplemental Figure 1).
Comment
Overall, there was significant geographic variation in the utilization of ambulatory resources for patients undergoing lung cancer resection across HRRs. Furthermore, there were trends towards variation in ambulatory utilization across HSAs, highlighting variation within HRR. However, there was no significant variation observed in inpatient utilization. Our findings highlight that variation in the utilization of resources surrounding episodes of surgical care for lung cancer may be nested within geographic areas.
There are a variety of factors influencing lung cancer surgery outcomes, such as preoperative rehabilitation5,6 and minimally invasive approaches,7 but these studies do not focus on the full continuum of care, which includes utilization of resources. Our study did not find significant variation in the utilization of inpatient resources, which is important as inpatient care is a key driver of healthcare costs. For example, pulmonary biopsies in the ambulatory setting have a median cost around $1000 compared to nearly $30,000 in the inpatient setting.8 However, understanding key drivers of geographic variation in the ambulatory setting may help identify additional factors to consider in assessing inpatient utilization.
Unlike previous literature that has studied larger geographic areas such as HRRs,9 this study demonstrates there are trends towards variation in the utilization of resources in more focused geographic areas at the level of HSAs. Variation within HSAs is important because differences may exist within health systems, and not just between different systems.
This study has limitations. We used large administrative databases, which may lack granular details in patient care and staging characteristics. Nevertheless, these datasets have a greater than 90% capture rate of hospital discharges and comprise a large sample size, increasing the statistical power of the study.10 Additionally, some HRRs that were on the border of 2 states may have missing data as only 4 states were analyzed. Details related to the distribution of the population within HRRs, or the rural–urban status of each HRR was outside the scope of this study. Lastly, the 6-month timeframe may have underestimated the true utilization rate in some cases as only one quarter before or after the index surgery was included in the analysis. Further study is warranted, as the generalizability of the study is limited due to the inclusion of the 4 selected states.
Overall, our study demonstrated that there is geographic variation in the utilization of resources surrounding the surgical management of lung cancer. We identified geographic variation in ambulatory utilization, but not inpatient utilization surrounding episodes of care. Further studies are necessary to elucidate the underlying drivers of the geographic variation and where that variation is nested across the entire continuum of the disease process, from diagnosis to postoperative care. Understanding the drivers of high utilization with future studies will allow for the implementation of mitigation strategies to reduce variation and thus improve the value of care surrounding lung cancer management.
Acknowledgments
The Supplemental Material can be viewed in the online version of this article [https://doi.org/10.1016/j.atssr.2024.02.007] on http://www.annalsthoracicsurgery.org.
Funding Sources
The authors have no funding sources to disclose.
Disclosures
The authors have no conflicts of interest to disclose.
Supplementary Data
Supplemental Figure 1.
References
- 1.von Meyenfeldt E.M., Hoeijmakers F, Marres GMH, et al. Variation in length of stay after minimally invasive lung resection: a reflection of perioperative care routines? Eur J Cardiothorac Surg. 2020;57:747–753. doi: 10.1093/ejcts/ezz303. [DOI] [PubMed] [Google Scholar]
- 2.Haneuse S., Dominici F, Normand SL, Schrag D. Assessment of between-hospital variation in readmission and mortality after cancer surgical procedures. JAMA Netw Open. 2018;1 doi: 10.1001/jamanetworkopen.2018.3038. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Medbery R.L., Fernandez F.G., Kosinski A.S., et al. Costs associated with lobectomy for lung cancer: an analysis merging STS and Medicare data. Ann Thorac Surg. 2021;111:1781–1790. doi: 10.1016/j.athoracsur.2020.08.073. [DOI] [PubMed] [Google Scholar]
- 4.Data Dartmouth Atlas of Health Care. https://data.dartmouthatlas.org/
- 5.FAQ Dartmouth Atlas of Health Care. https://www.dartmouthatlas.org/faq/
- 6.Templeton R., Greenhalgh D. Preoperative rehabilitation for thoracic surgery. Curr Opin Anaesthesiol. 2019;32:23–28. doi: 10.1097/ACO.0000000000000668. [DOI] [PubMed] [Google Scholar]
- 7.Sihoe A.D.L. Video-assisted thoracoscopic surgery as the gold standard for lung cancer surgery. Respirology. 2020;25(Suppl 2):49–60. doi: 10.1111/resp.13920. [DOI] [PubMed] [Google Scholar]
- 8.Zhu D., Shi X., Nicholas S., Ma Y., He P. Estimated annual prevalence, medical service utilization and direct costs of lung cancer in urban China. Cancer Med. 2021;10:2914–2923. doi: 10.1002/cam4.3845. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Cooper Z., Stiegman O., Ndumele C.D., Staiger B., Skinner J. Geographical variation in health spending across the US among privately insured individuals and enrollees in Medicaid and Medicare. JAMA Netw Open. 2022;5:e2222138. doi: 10.1001/jamanetworkopen.2022.22138. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Healthcare Cost and Utilization Project (HCUP) Agency for Healthcare Research and Quality; 2021. HCUP Databases. [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.