Abstract
PURPOSE:
Did the creation of an urgent care clinic specifically for patients with cancer affect emergency department visits among adults newly diagnosed with cancer?
PATIENTS AND METHODS:
We applied an interrupted time series analysis to adjusted monthly emergency department visits made by adults age 18 years or older who were diagnosed with cancer between 2009 and 2016 at a comprehensive cancer center. Cancer registry patients were linked to a longitudinal regional database of emergency department and hospital visits. Because the urgent care clinic was closed on weekends, we took advantage of the natural experiment by comparing weekend visits as a control group. Our primary outcome was emergency department visits within 180 days after a cancer diagnosis, compiled as adjusted monthly rates of emergency department visits per 1,000 patient-months. We analyzed subsequent hospitalizations as a secondary outcome.
RESULTS:
The rate of weekday emergency department visits was increasing at a rate of 0.43 visits (95% CI, 0.29 to 0.57 visits) per month before May 2012, then fell in half to a rate of 0.19 visits (95% CI, 0.11 to 0.28 visits) per month (P = .007) after the urgent care clinic was established. In contrast, the weekend visit rate was growing at a rate of 0.08 visits (95% CI, −0.03 to 0.19 visits) per month before May 2012 and 0.05 (95% CI, −0.02 to 0.13 visits; P = .533) afterward. By the end of 2016, there were 15.3 fewer monthly weekday emergency department visits than expected (P = .005). Trends in weekday hospitalizations were not significantly changed.
CONCLUSION:
Although only one in eight emergency department–visiting patients also used the urgent care clinic, the growth rate of emergency department visits fell by half after the urgent care clinic was established.
INTRODUCTION
In the first 6 months after their cancer diagnosis, adult patients experience a high volume of unplanned acute care visits via the emergency department (ED). These visits are predominantly for commonly anticipated complaints: nausea, vomiting, dehydration, pain, fever, cough, and shortness of breath.1,2 After an ED visit, patients with cancer are hospitalized roughly 60% of the time,3 exposing them to the harms of hospitalization, disrupting treatment schedules, generating a majority of early cancer care expenditures, and contributing to patient financial burden.4,5
Several cancer centers have created an urgent care clinic (UCC) for their patients with cancer aimed at reducing ED visits.6-10 However, it is not clear whether UCCs actually reduce the volume of ED visits or whether they serve as a conduit to ED referral.
Our comprehensive cancer center established a UCC for its oncology practice in May 2012. When patients call the clinic with a health concern, emergent conditions are triaged and directed to the ED, and patients suitable for an urgent care visit are scheduled that day or the next day. The UCC is open only during business hours (Monday to Friday from 8am to 4pm) but can administer intravenous fluids and medications, obtain basic laboratory tests with rapid results, and obtain common imaging tests such as chest x-ray and extremity duplex ultrasound.
We conducted a quasiexperimental interrupted time series analysis11 using a longitudinal database of ED visits of all nonfederal hospitals in the region to determine the impact of the cancer UCC on ED visits made by adults newly diagnosed with cancer. Because the cancer UCC was closed on weekends, we took advantage of this natural experiment to compare weekday ED visits with an internal control group of weekend ED visits from the same patient population. We also analyzed the impact of the UCC on ED hospitalization rates as a secondary outcome.
PATIENTS AND METHODS
Population and Covariates
We identified adults age 18 years or older from the hospital cancer registry who had an incident cancer diagnosis from July 2008 through December 2016, excluding patients with leukemia and nonmelanoma skin cancer. Patients with leukemia were excluded primarily because of prolonged inpatient treatments after diagnosis as well as disease-specific heterogeneity in treatment.
We noted cancer type and categorized stage at diagnosis into advanced (stage IIIb or higher for lung cancer, stage III or higher for pancreatic cancer, and stage IV for all others except brain cancer) versus nonadvanced disease.4 The cancer registry also provided the following information from the date of diagnosis: age, sex, race and ethnicity, zip code of residence, comorbid conditions, primary payer, initial cancer treatment modalities (chemotherapy, radiation therapy, surgery, and/or immunotherapy), and date of death. We organized comorbidities into a Charlson comorbidity index.12-14 Finally, by matching residential addresses to census tracts and 2009 American Community Survey poverty-level and educational attainment characteristics, we used validated measures15-17 to characterize neighborhoods as low education (≥ 25% of individuals did not graduate high school) or high poverty (≥ 10% of households below poverty level).
Integrated Data Set of Regional Hospital Visits
Patients may visit EDs at multiple different hospital systems, which may not be captured by any single local electronic health record (EHR). To overcome this, we identified incident cancer diagnoses in our local cancer registry and linked patients to a regional database containing longitudinal hospital use data. This regional resource, the Dallas-Fort Worth Hospital Council Education and Research Foundation, collects information for approximately 12 million unique patients and their 65 million hospital encounters and warehouses claims data from 98% of the nonfederal hospitals in the North Texas region. Visit-level data are organized into a master patient index that assigns a unique identifier to each patient, allowing linkage and longitudinal tracking of any patient over time, independent of insured status, insurance type, and health system visited.
Using a combination of first and last names, date of birth, zip code, and medical record number, we matched patients to their unique foundation identifier. The foundation database provided dates of ED arrival and discharge, name of hospital and health system, diagnosis codes associated with the visit, and discharge disposition (including died in ED, transferred to another hospital, observation stay, inpatient admission, and discharge home). Patients were considered hospitalized if they were transferred to another hospital or had an observation stay or inpatient admission.
We also linked our cancer registry cohort to our institutional comprehensive EHR (Epic Systems, Verona, WI) to identify UCC visits made during the study period. We collected dates and times the UCC visits were scheduled and completed, whether patients had an ED visit within 24 hours of the UCC visit, the reason for the ED visit, and whether this ED visit resulted in hospitalization.
Study Design
We generated descriptive statistics for the patient cohort, their ED visits to all hospitals, subsequent hospitalizations, and any local UCC visits. Our primary outcome of interest was the monthly ED visit rate during the first 180 days (6 months) after a cancer diagnosis. For the denominator, we used patient-months so that our rates were not affected by early patient deaths. Each partial month a patient remained alive during each 30-day period amounted to 1 patient-month contributed to the underlying denominator, for up to 6 months per patient. We reported the rate as monthly ED visits per 1,000 patient-months.
We applied an interrupted time series analysis to the trends in ED visit rates before and after the creation of the UCC in May 2012, with a 6-month phase-in period (through October 2012) that was excluded from analysis.11,18 We chose a 6-month phase-in period to allow the oncology practice to adapt to the establishment of this new service. Because the UCC was only open during weekdays, our intervention outcome was weekday ED visit rate. Although the UCC was not open during weekday nights, the foundation database did not report the time of ED arrival.
To account for underlying trends in ED use, we used a control group of ED visits not affected by the intervention11: the ED visit rate when the UCC was closed on weekends and weekday holidays. We compared weekday ED visit rates with the control group of weekend ED visit rates, both before and after creation of the cancer UCC. We conducted the same pre- versus postanalysis for our secondary outcome of weekday ED hospitalization rate.
Statistical Analysis
After compiling unadjusted monthly visit rates, we used generalized estimating equations with a Poisson distribution to adjust rates for changes in cohort characteristics over time (sex, age, race and ethnicity, insurance type, education, poverty level, advanced cancer stage at diagnosis, and initial treatment modalities), accounting for the correlation between visits generated by the same patient. We used marginal effects methods to calculate the adjusted monthly visit rates, seasonally adjusted these rates, and then applied interrupted time series segmented regression, adjusting the SEs for autocorrelation across time.19,20
We modeled changes in visit rates over time to generate baseline rate, baseline trend (slope), and postintervention change. We defined statistical significance at the level of α = 0.05 and reported 95% CIs of the coefficients. Finally, we calculated the difference in visit rates in December 2016 between the post-UCC trend and the predicted rate on the basis of pre-UCC trend.21 This gave us the absolute difference in visits associated with creation of the UCC. We conducted the same analyses for ED visit rates and ED hospitalization rates using the same 6-month phase-in period. For sensitivity analyses for ED visits, we reanalyzed without a phase-in period, as well as with phase-in periods of 5 and 7 months.
SAS software (version 9.4; SAS Institute, Cary, NC) and STATA/MP software (version 15.1; STATA, College Station, TX) were used for statistical analyses. The University of Texas Southwestern Medical Center institutional review board approved study protocol (STU 112017-042).
RESULTS
Patient Cohort and Visits
Our entire study cohort contained 33,316 adults age 18 years or older who contributed patient-months between January 1, 2009, and December 31, 2016. The median age was 60 years (interquartile range, 51-69 years); 47.4% were female, 65.2% were non-Hispanic white, 19.1% had advanced stage cancer at the time of diagnosis, and 93.9% had health insurance at the time of diagnosis (34.2% in Medicare). We list patient characteristics in Table 1, divided into populations before and after the UCC was established.
TABLE 1.
Characteristics of Patients With Newly Diagnosed Cancer, 2009 to 2016
This patient cohort generated 17,835 total ED visits within 180 days after diagnosis to 70 different adult acute care hospitals within 17 hospital systems. During the ED visit, 4.2% patients died or left against medical advice, and 44.4% of the remaining ED visits resulted in hospitalization (observation or inpatient). Only 23.6% of all ED visits took place at the primary hospital where the patient was receiving cancer treatment, although this proportion had increased from 20.0% in 2009 to 30.1% by 2016. This rise was in the setting of the opening of our new, much larger university hospital ED at the end of 2014. These ED visits led to 8,107 hospitalizations, for a total hospitalization rate of 45.4%, with a rate of 47.7% in 2009, decreasing slightly to 44.7% by 2016.
During the postintervention period, 4,846 patients had at least one ED visit within 180 days after their cancer diagnosis; only 589 patients (12.2%) completed 861 UCC visits during the same period. The UCC visit was completed the same day the appointment was made for 74.0% of visits. After 80 of UCC visits (9.3%), there was an ED visit within 24 hours, and after 65 of UCC visits (7.5%), an ED visit was followed by a hospitalization.
ED Visit Rate
Modeled trend analysis results are listed in Table 2 Appendix Table A1 (online only), and Figure 1. In January 2009, the adjusted weekday ED visit rate was 54.73 visits (95% CI, 51.83 to 57.63 visits) per 1,000 patient-months, and the weekend ED visit rate was 23.56 visits (95% CI, 21.63 to 25.49 visits) per 1,000 patient-months.
TABLE 2.
Trends in Monthly ED visits and Hospitalizations (per 1,000 patient-months)
Fig 1.
Monthly emergency department (ED) visit rates before and after creation of an urgent care clinic (UCC) for patients with cancer. ED visits within 180 days after cancer diagnosis. Monthly number of ED visits to all regional hospitals per 1,000 patient-months. Rates adjusted for varying population characteristics over time: sex, age, race/ethnicity, education, poverty, insurance type, cancer type, stage of cancer at diagnosis, initial treatment modalities (chemotherapy, surgery, radiation therapy, and immunotherapy); we also accounted for correlation of multiple visits made by the same patient; rates were seasonally adjusted. For weekends (blue) and weekdays (red), dots indicate observed ED visit rate, dotted lines indicate predicted baseline trend before UCC, and solid lines indicate modeled trend before and after UCC. (*) P = .007 compared with pretrend slope, indicating statistical difference in ED visits post-UCC.
Before the UCC, monthly weekday ED visit rates were increasing at a slope of 0.43 visits (95% CI, 0.29 to 0.57 visits) per 1,000 patient-months and slowed after the UCC was established to a slope of 0.19 visits (95% CI, 0.11 to 0.28 visits) per 1,000 patient-months, a significant difference (P = .007). Monthly weekend ED visit rates were not significantly increasing, with a slope of 0.08 visits (95% CI, −0.03 to 0.19 visits) per 1,000 patient-months, and remained unchanged after the UCC was established, at 0.05 visits (95% CI, −0.02 to 0.13 visits; P = .533) per 1,000 patient-months.
Appendix Table A1 lists the difference between pre-UCC predicted ED visit rates and modeled post-UCC ED visit rates. The predicted weekday ED visit rate for December 2016, on the basis of the pre-UCC trend, was 93.09 visits (95% CI, 82.92 to 103.26 visits) per 1,000 patient-months, whereas the modeled ED visit rate for December 2016 was 77.75 visits (95% CI, 75.09 to 80.40 visits) per 1,000 patient-months. Therefore, by the end of the study period, the ED visit rate was 15.34 visits (95% CI, 4.78 to 25.91 visits) per 1,000 patient-months lower after the UCC than predicted by the pre-UCC trend. This totaled 488 fewer ED visits during the post-UCC period (95% CI, −838 to −138 visits), excluding the 6-month phase-in period.
The predicted weekend ED visit rate for December 2016, on the basis of the pre-UCC trend, was 30.64 visits (95% CI, 22.26 to 39.02 visits) per 1,000 patient-months, and the adjusted rate was 29.49 visits (95% CI, 27.58 to 31.39 visits) per 1,000 patient-months, for a nonsignificant difference of 1.16 visits (95% CI, −7.36 to 9.67 visits) per 1,000 patient-months.
Sensitivity analyses varying the length of the intervention phase-in period yielded no substantive differences from these results. Details are listed in Appendix Table A1.
Hospitalization Outcome
Modeled trend analysis results are listed in Table 2 and Appendix Table A1. In January 2009, the adjusted weekday hospitalization rate was 29.30 ED hospitalizations (95% CI, 25.44 to 33.15 visits) per 1,000 patient-months, and the weekend ED hospitalization rate was 13.17 visits (95% CI, 11.85 to 14.48 visits) per 1,000 patient-months.
Before the UCC, adjusted monthly weekday ED hospitalization rates were not significantly increasing, at a slope of 0.11 visits (95% CI, −0.04 to 0.25 visits) per month; after the UCC was established, the slope was 0.07 visits (95% CI, −0.01 to 0.16) per month, for a nonsignificant difference in slope of −0.03 visits (95% CI, −0.20 to 0.14 visits; P = .695) per month. Monthly weekend ED hospitalization rates were nonsignificantly declining, at −0.05 visits (95% CI, −0.11 to 0.02 visits) per month before the UCC and 0.00 visits (95% CI, −0.05 to 0.05 visits) per month after the UCC, for a nonsignificant difference of −0.05 visits (95% CI, −0.13 to 0.03 visits; P = .247) per month.
Similarly, the absolute differences between the predicted pre-UCC trend and the modeled post-UCC values were nonsignificant for both weekday and weekend ED hospitalizations. The difference for weekday ED hospitalizations was 0.48 visits (95% CI, −9.63 to 10.59 visits) per 1,000 patient-months, and the difference for weekend ED hospitalizations was −3.65 visits (95% CI, −8.39 to 1.09 visits) per 1,000 patient-months.
DISCUSSION
More than 4 years after implementation, the creation of a cancer UCC was associated with a significant reduction in ED visits made within the 180 days after a new cancer diagnosis. In contrast, the rate of ED visits made on weekends, when the UCC was closed, remained unchanged over the same period of time. This increases our confidence that the UCC was associated with a real reduction in ED visits when the UCC was open. However, our secondary outcome of hospitalizations after ED visits showed no significant changes after the UCC was established. Although the number of UCC visits completed was almost twice the estimated number of ED visits presumably avoided, this may still represent a favorable tradeoff for overall expenditures and other avoided harms. We also undertook a thorough review of potential cointerventions surrounding the introduction of the UCC and found none.
Our findings add to the literature by rigorously evaluating the impact of a cancer UCC on ED use within 180 days after a cancer diagnosis. Prior published abstracts described unadjusted pre- and post-UCC ED visit volumes in a single health system and thus limited ability to distinguish the impact of changes in patient volume and characteristics over time.8-10 Our study used a regional longitudinal database of hospital use, allowing us to link important granular cancer and clinical details22 in the EHR to comprehensive use data, without having to subsegment our cohort to match any specific insurance claims database. Furthermore, we used robust statistical analyses to account for underlying trends in nontargeted weekend ED visits.
Given the outsized role of unplanned hospital care in the early expenditures of cancer treatment, this delivery innovation was highly effective and may be a promising strategy for organizations participating in the Oncology Care Model demonstration project and other risk-bearing contracts, such as accountable care organizations.23-25
In our study, although the trend of ED hospitalizations seemed to decline after the UCC, this was not a significant trend change. Despite hospitalization rates of nearly 50%, our study may have been underpowered to detect a change in this outcome. Relatedly, we note the limited uptake of the UCC; whereas 4,846 patients had at least one ED visit after creation of the UCC, only one in eight patients actually visited the UCC. Alternatively, it is possible that because lower-acuity ED visits were diverted to the UCC, higher-acuity visits may have continued to result in hospitalizations, leaving the hospitalization rate relatively unaffected.
It is intriguing that the weekend ED visit growth rate remained low throughout the study period, especially because nonemergent visits are often thought to result from a lack of access to other sites of care. Our study did not allow for robust explanations of the growth rates on weekdays and weekends, but it seems to suggest at least that patients’ perceived need for ED care on the weekends was unchanged; the decision making behind these trends should be studied further.
Finally, we note that by the end of 2016, more than two thirds of ED visits were still made to hospitals that were not within the primary cancer–treating health system. These EDs were likely unable to quickly access detailed records of the patients’ treatment and clinical histories. It is unclear whether visits to such unfamiliar hospitals were for higher-acuity conditions or whether visiting a hospital unfamiliar with the clinical history of a patient with cancer affects the decision by an ED to hospitalize.26 More work should be done to understand how to increase awareness and use of the cancer UCC, disseminate this delivery innovation to other practice settings, and understand its impact on unplanned hospitalizations.
Our findings should be interpreted in the context of certain limitations. Although interrupted time series is a robust quasiexperimental design, we cannot be certain that the associated decline in ED visits was solely a result of the creation of the UCC. However, we adjusted outcome rates for changes over time in patient age, sex, race, insurance type, education and poverty levels, cancer type and stage, initial treatment modalities, and correlation in visits originating from the same patient; we also found a nonsignificant change in the unaffected group of weekend visits when the UCC was closed.
Although our study population spanned a wide range of cancer types and patients over a 7-year period, it is the experience of an academic cancer center with a large regional referral base. It is not clear how generalizable these results may be to nonacademic centers or smaller oncology practice populations.
Finally, because the Dallas-Fort Worth Hospital Council Education and Research Foundation database does not record time of ED arrival, ED visits occurring on weekday nights were included in the intervention group and not the control group, even though the UCC was closed. We suspect this conservative approach might have led to an underestimation of the impact of the UCC.
In conclusion, establishing a UCC specifically for patients with cancer was associated with a significant decrease in ED use by adults newly diagnosed with cancer. Even so, the proportion of patients who actually used the UCC was small. Our findings suggest that this alternate site of nonemergent care holds promise in optimizing the delivery of acute care for patients undergoing cancer treatment. These results should spur additional work to manage more patients in this alternate acute care setting and understanding ED use in this high-use population.
ACKNOWLEDGMENT
Supported by the Texas Health Resources Clinical Scholars Program, by the National Center for Advancing Translational Sciences of the National Institutes of Health under UL1TR001105, by R24HS022418 from the Agency for Healthcare Research and Quality, and by National Cancer Institute Cancer Center Support Grant 1P30CA142543-03. The content is solely the responsibility of the authors and does not necessarily represent the official views of Texas Health Resources, University of Texas Southwestern Medical Center, the National Institutes of Health, or the Agency for Healthcare Research and Quality. The funders had no role in the design or conduct of the study; collection, management, analysis, or interpretation of the data; preparation, review, or approval of the manuscript; or decision to submit the manuscript for publication. We the Dallas-Fort Worth Hospital Council Education and Research Foundation, which contributed a portion of the data, although it did not have a role in the design or conduct of the study. We also acknowledge Hannah Fullington, MPH, Department of Clinical Sciences, University of Texas Southwestern Medical Center, who provided assistance with data gathering.
APPENDIX
TABLE A1.
Sensitivity Analyses of Trends in Monthly ED Visits (per 1,000 patient-months), With Varying Phase-In Periods for Intervention
AUTHOR CONTRIBUTIONS
Conception and design: Arthur S. Hong, Stephanie Clayton Hobbs, Simon J. Craddock Lee, Ethan A. Halm
Financial support: Ethan A. Halm
Administrative support: Ethan A. Halm
Data analysis and interpretation: Arthur S. Hong, Thomas Froehlich, Simon J. Craddock Lee, Ethan A. Halm
Manuscript writing: All authors
Final approval of manuscript: All authors
Accountable for all aspects of the work: All authors
AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST
Impact of a Cancer Urgent Care Clinic on Regional Emergency Department Visits
The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to www.asco.org/rwc or ascopubs.org/jop/site/ifc/journal-policies.html.
Arthur S. Hong
Consulting or Advisory Role: Janssen (I), AbbVie (I), Celgene (I)
Travel, Accommodations, Expenses: Janssen (I), Celgene (I), AbbVie (I)
Thomas Froehlich
Honoraria: National Comprehensive Cancer Network
Speakers’ Bureau: Genentech
Travel, Accommodations, Expenses: National Comprehensive Cancer Network, Genentech
No other potential conflicts of interest were reported.
REFERENCES
- 1.Mayer DK, Travers D, Wyss A, et al. Why do patients with cancer visit emergency departments? Results of a 2008 population study in North Carolina. J Clin Oncol. 2011;29:2683–2688. doi: 10.1200/JCO.2010.34.2816. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Manzano J-GM, Luo R, Elting LS, et al. Patterns and predictors of unplanned hospitalization in a population-based cohort of elderly patients with GI cancer. J Clin Oncol. 2014;32:3527–3533. doi: 10.1200/JCO.2014.55.3131. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Rivera DR, Gallicchio L, Brown J, et al. Trends in adult cancer-related emergency department utilization: An analysis of data from the nationwide emergency department sample. JAMA Oncol. 2017;3:e172450. doi: 10.1001/jamaoncol.2017.2450. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Brooks GA, Li L, Uno H, et al. Acute hospital care is the chief driver of regional spending variation in Medicare patients with advanced cancer. Health Aff (Millwood) 2014;33:1793–1800. doi: 10.1377/hlthaff.2014.0280. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Narang AK, Nicholas LH. Out-of-pocket spending and financial burden among Medicare beneficiaries with cancer. JAMA Oncol. 2017;3:757–765. doi: 10.1001/jamaoncol.2016.4865. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Coyle YM, Miller AM, Paulson RS. Model for the cost-efficient delivery of continuous quality cancer care: A hospital and private-practice collaboration. Proc (Bayl Univ Med Cent) 2013;26:95–99. doi: 10.1080/08998280.2013.11928928. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Klotz A, Martin S, Atoria CL, et al: The effect of observation units in cancer care on hospital admissions. J Clin Oncol 32, 2014 (suppl; abstr e17633)
- 8. Adelson KB, Dest V, Velji S, et al: Emergency department (ED) utilization and hospital admission rates among oncology patients at a large academic center and the need for improved urgent care access. J Clin Oncol 32, 2014 (suppl; abstr 19)
- 9. Sivendran S, Holliday R, De la Torre K, et al: Impact of a nurse practitioner-staffed, symptom-management clinic on emergency department utilization in a large community oncology cancer institute. J Clin Oncol 32, 2014 (suppl; abstr 58)
- 10. Scarborough B, Chai E, Holcombe RF, et al: Results from year 1 of an embedded supportive oncology practice. J Clin Oncol 32, 2014 (suppl; abstr 40)
- 11. Shadish WR, Cook TD, Campbell DT: Experimental and Quasi-Experimental Designs for Generalized Causal Inference. Boston, MA, Houghton Mifflin Company, 2001. [Google Scholar]
- 12.Charlson ME, Pompei P, Ales KL, et al. A new method of classifying prognostic comorbidity in longitudinal studies: Development and validation. J Chronic Dis. 1987;40:373–383. doi: 10.1016/0021-9681(87)90171-8. [DOI] [PubMed] [Google Scholar]
- 13.Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases. J Clin Epidemiol. 1992;45:613–619. doi: 10.1016/0895-4356(92)90133-8. [DOI] [PubMed] [Google Scholar]
- 14.Klabunde CN, Potosky AL, Legler JM, et al. Development of a comorbidity index using physician claims data. J Clin Epidemiol. 2000;53:1258–1267. doi: 10.1016/s0895-4356(00)00256-0. [DOI] [PubMed] [Google Scholar]
- 15.Krieger N. Overcoming the absence of socioeconomic data in medical records: Validation and application of a census-based methodology. Am J Public Health. 1992;82:703–710. doi: 10.2105/ajph.82.5.703. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Krieger N, Chen JT, Waterman PD, et al. Race/ethnicity, gender, and monitoring socioeconomic gradients in health: A comparison of area-based socioeconomic measures—The public health disparities geocoding project. Am J Public Health. 2003;93:1655–1671. doi: 10.2105/ajph.93.10.1655. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Trivedi AN, Zaslavsky AM, Schneider EC, et al. Relationship between quality of care and racial disparities in Medicare health plans. JAMA. 2006;296:1998–2004. doi: 10.1001/jama.296.16.1998. [DOI] [PubMed] [Google Scholar]
- 18.Wagner AK, Soumerai SB, Zhang F, et al. Segmented regression analysis of interrupted time series studies in medication use research. J Clin Pharm Ther. 2002;27:299–309. doi: 10.1046/j.1365-2710.2002.00430.x. [DOI] [PubMed] [Google Scholar]
- 19. Zhang F, Wagner AK, Soumerai SB, et al: Methods for estimating confidence intervals in interrupted time series analyses of health interventions. 62:143-148, 2009. [DOI] [PMC free article] [PubMed]
- 20.Linden A. Conducting interrupted time-series analysis for single- and multiple-group comparisons. Stata J. 2015;15:480–500. [Google Scholar]
- 21.Linden A. A comprehensive set of postestimation measures to enrich interrupted time-series analysis. Stata J. 2017;17:73–88. [Google Scholar]
- 22.Nipp RD, El-Jawahri A, Moran SM, et al. The relationship between physical and psychological symptoms and health care utilization in hospitalized patients with advanced cancer. Cancer. 2017;123:4720–4727. doi: 10.1002/cncr.30912. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Rocque GB, Williams CP, Kenzik KM, et al. Where are the opportunities for reducing health care spending within alternative payment models? J Oncol Pract. 2018;14:e375–e383. doi: 10.1200/JOP.2017.024935. [DOI] [PubMed] [Google Scholar]
- 24.Colligan EM, Ewald E, Ruiz S, et al. Innovative oncology care models improve end-of-life quality, reduce utilization and spending. Health Aff (Millwood) 2017;36:433–440. doi: 10.1377/hlthaff.2016.1303. [DOI] [PubMed] [Google Scholar]
- 25.Aviki EM, Schleicher SM, Mullangi S, et al. Alternative payment and care-delivery models in oncology: A systematic review. Cancer. 2018;124:3293–3306. doi: 10.1002/cncr.31367. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Wright B, Martin GP, Ahmed A, et al. How the availability of observation status affects emergency physician decisionmaking. Ann Emerg Med. 2018;72:401–409. doi: 10.1016/j.annemergmed.2018.04.023. [DOI] [PubMed] [Google Scholar]