Ann Thorac Surg 2009;88:1749-1756. doi:10.1016/j.athoracsur.2009.08.006
© 2009 The Society of Thoracic Surgeons
Original Articles: General Thoracic
Health Care Utilization Among Surgically Treated Medicare Beneficiaries With Lung Cancer
Farhood Farjah, MD, MPH,
Douglas E. Wood, MD,
Thomas K. Varghese, MD,
Nader N. Massarweh, MD,
Rebecca Gaston Symons, MPH,
David R. Flum, MD, MPH*
Department of Surgery, University of Washington, Seattle, Washington
Accepted for publication August 6, 2009.
* Address correspondence to Dr Flum, University of Washington, Department of Surgery, Box 356410, 1959 NE Pacific St, Seattle, WA 98195-6410 (Email: daveflum{at}u.washington.edu).
 |
Abstract
|
|---|
Background: Markers of increased health care utilization are surrogates for adverse events, and one such metric—prolonged length of stay greater than 14 days (PLOS)—was recently endorsed as a provider-level performance measure.
Methods: This is a cohort study (1992 through 2002) aimed to describe increased health care utilization among 21,067 operated lung cancer patients using the Surveillance, Epidemiology, and End-Results-Medicare database. Increased utilization was defined by PLOS, discharge to an institutional care facility (ICF), or readmission within 30 days.
Results: Twelve percent of patients had a PLOS, 13% were discharged to an ICF, and 15% were readmitted. In multivariate analyses, factors associated with a higher odds ratio of PLOS, discharge to ICF, or readmission included age older than 80 years, increasing comorbidity index, not being married, and pneumonectomy (all p < 0.05). Relative to patients living in the West, those in the Midwest or South had a higher odds ratio of PLOS and readmission but a lower odds ratio of discharge to an ICF (all p < 0.05). Adjusted rates of PLOS decreased significantly with time, whereas adjusted ICF and readmission rates increased (all p < 0.01). Patients who required increased utilization had higher adjusted 2.5-year mortality rates compared with those who did not (PLOS, 42% versus 20%; ICF, 32% versus 20%; readmission, 33% versus 19%; all p < 0.001).
Conclusions: Baseline health status and nonclinical factors were associated with increased utilization, nonuniform trends in utilization were observed with time, and increased utilization was associated with worse long-term outcomes. These findings have implications for quality-improvement initiatives that measure increased health care utilization as a surrogate for provider performance.
|
Introduction
|
|---|
There is growing interest in measuring health care utilization as a surrogate for the quality of surgical care. For instance, the National Quality Forum (NQF) recently endorsed hospital stays greater than 14 days after elective lobectomy for lung cancer as a measure of clinician-level performance [1]. Little is known about increased health care utilization after pulmonary resection for lung cancer in the United States [2–4]. A better understanding of health care utilization might improve the ability to measure and use this information to improve quality.
This investigation examined three markers of increased health care utilization—prolonged length of stay greater than 14 days (PLOS), discharge to an institutional care facility (ICF), and readmission within 30 days of discharge. We describe factors associated for each of these three indicators of increased utilization, trends with time, the relationship between different markers of utilization, and associated long-term outcomes.
|
Material and Methods
|
|---|
Data Source
A cohort study was conducted using the Surveillance, Epidemiology, and End-Results (SEER)-Medicare database to examine lung cancer patients diagnosed between 1992 and 2002. An overview of the SEER-Medicare database is provided elsewhere [5]. The University of Washington Institutional Review Board approved this study, and waived consent because we used existing, de-identified data.
Patient Population
Of the 221,208 patients within the database, sequential exclusions were made as follows: diagnosis at death or autopsy (n = 5,109), nonoperated-on patients (n = 188,469), age younger than 66 years (n = 2,376), identification of another malignancy between 3 months before and 6 months after diagnosis (n = 1,445), patients without both Part A and B Medicare or with concurrent health maintenance organization enrollment between 1 year before and 9 months after diagnosis (n = 2,208), histologic diagnosis incompatible with lung cancer (n = 45), and patients without identifiable inpatient claims data (n = 499). Reasons for these exclusions are detailed elsewhere [6].
Health Care Utilization
Health care utilization was ascertained using the Medicare Provider Analysis and Review file (inpatient claims). Prolonged length of stay was defined by a hospitalization greater than 14 days. Patients who died within 14 days of resection as an inpatient (n = 532, 2.5%) did not have an opportunity to be at risk for PLOS, and thus they were excluded from analyses involving PLOS. Discharge to a facility rather than a home was considered discharge to an ICF. Readmission was defined by rehospitalization for any reason within 30-days of discharge from the hospitalization for resection. Patients who died during their initial hospitalization (n = 947, 4.5%) could not have been discharged to an ICF or readmitted, and they were excluded from analyses involving these two variables. Additionally, those who died within 30 days of discharge without having been readmitted (n = 198, 0.98%) did not have an opportunity to be at risk for readmission and were excluded from analyses involving readmission.
Outcomes
Death data were available in the Medicare Enrollment Database, which was updated nightly by the Social Security Administration through 2005 [7]. Although all patients had at least 3 years of complete follow-up from the time of diagnosis, they only had 2.5 years of complete follow-up from the last potential exposure that could have occurred after diagnosis—readmission within 30 days of hospital discharge. Long-term mortality was defined by vital status 2.5 years after assessing readmission status (30 days after hospitalization).
Covariate Definitions
Patient demographics, socioeconomic and marital status, and cancer stage and histology were available through the SEER database [8]. A Klabunde-modified Charlson comorbidity index was calculated based on claims within the Carrier and Outpatient Claims files in the year before diagnosis [8]. The type of resection and perioperative radiation or chemotherapy were defined by their respective billing codes within Carrier and Outpatient Claims files.
Analysis
All statistical analyses were performed using STATA (Special Edition 9.2; Statacorp, College Station, TX). Logistic regression was used to conduct unadjusted and adjusted analyses of factors associated with increased health care utilization, trends in utilization with time, the relationship between length of stay (LOS) and discharge to an ICF and readmission, and the relationship between increased utilization and long-term outcomes. All models adjusted for clustering at the hospital level.
Variables used for adjustment included age category, sex, race, income, education, marital status, geographic location, area of residence, history of prior malignancy, comorbidity index, histologic type, cancer stage, neoadjuvant therapy, and extent of resection. Conditional standardization of the regression results based on modal values of the covariates was used to obtain an adjusted estimate for the proportion of interest.
Twenty-one percent of patients had at least one missing covariate value. The proportion of patients with missing data did not vary significantly between those with or without PLOS (22% versus 20%; p = 0.07) or those discharged home versus to an ICF (21% versus 21%; p = 0.81), although it did between those who were and were not readmitted (23% versus 20%; p < 0.001). Long-term mortality varied significantly between patients with and without missing data (47% versus 43%; p < 0.001). Variation in missing covariate data was deemed clinically insignificant and assumed to occur completely at random. Multivariate imputation by chained equations was performed [9], and all multivariate analyses were repeated using five sets of imputed covariate data. The results did not vary substantially between analyses using case-complete or imputed covariate data, and thus only the results of case-complete analyses were reported.
|
Results
|
|---|
Twelve percent of patients had a PLOS, 13% were discharged to an ICF, and 15% were readmitted. The median LOS was 7 days (range, 0 to 233 days). Among patients who were readmitted, the median time to readmission was 7 days.
Table 1
shows variation in health care utilization events across subgroups of patients. There was at least twofold variation in PLOS by histologic subtype and extent of resection; threefold variation in discharge to an ICF by age and twofold variation by marital status, geography, and comorbidity index; and nearly twofold variation in readmission by histologic subtype and comorbidity index.
Table 2
shows the results of multivariate analyses evaluating factors associated with each indicator of increased utilization. Age older than 80 years, not being married, increasing comorbidity index, and pneumonectomy were associated with a higher odds ratio of PLOS, discharge to ICF, and readmission. Black race, income, and education were not associated with any marker of increased utilization. Patients living in the Midwest or South had a significantly higher odds ratio of PLOS and readmission but a lower odds ratio of discharge to an ICF compared with patients living in the West. The risk of increased health care utilization did not vary significantly between patients living in the West or East. Adjustment for calendar year did not have an impact on these findings.
Figure 1
shows changes in health care utilization with time. The unadjusted proportion of patients who required discharge to an ICF or readmission increased significantly with time (both p < 0.001), whereas the proportion of patients who required PLOS decreased (p = 0.010). Coincident with these trends were changes in cohort characteristics during the study period (data available on request). After adjustment for these changes, temporal trends in utilization remained statistically significant for PLOS (p = 0.010), discharge to ICF (p < 0.001), and readmission (p = 0.001).
View larger version (12K):
[in this window]
[in a new window]
|
Fig 1. Unadjusted (A) and adjusted (B) utilization rates are shown by calendar year for prolonged length of stay (PLOS) discharge to an institutional care facility (ICF), and readmission to the hospital.
|
|
Figure 2
describes the relationship between groups defined by LOS and the risk of discharge to an ICF or readmission. The risk of discharge to ICF increased with increasing LOS. A U-shape relationship existed between readmission and LOS, such that the risks of readmission increased as LOS increased by more than 4 days or LOS decreased by less than 4 days. Wide 95% confidence intervals did not rule out the possibility that chance alone might explain this relationship. Despite this apparently nonlinear relationship, readmission rates were higher among patients who required PLOS compared with those who did not (25% versus 9%; p < 0.001).
View larger version (11K):
[in this window]
[in a new window]
|
Fig 2. Adjusted discharge to an institutional care facility (ICF) rate (A) and adjusted readmission rate (B) as a function of length of stay are presented. Error bars represent 95% confidence intervals.
|
|
Table 3
shows the relationship between long-term mortality and markers of increased utilization. Adjusted long-term mortality was higher among patients who required any type of increased utilization. In stratified, adjusted analyses, among those in the non-PLOS strata, 2.5-year mortality was still higher among patients who required discharge to an ICF or readmission compared with those who did not.
|
Comment
|
|---|
Surgical quality improvement depends, in part, on fair and acceptable measures of quality. Lung resection volume for any one hospital or surgeon is at least one order of magnitude lower than coronary artery bypass grafting volume [10, 11], and operative mortality rates for pulmonary resection are low (2% to 5%) [12–15]. This combination of low volume and low mortality presents a challenge when attempting to discriminate provider performance [16], and has led to interest in measuring PLOS—a more frequent event than early death—as a surrogate for surgical morbidity and mortality [17]. Increased health care utilization may be a useful inverse surrogate for surgical quality, but little is known about health care utilization after lung resection. The descriptive information from this study provides a framework for future investigations and discussions about how best to use measures of increased health care utilization for surgical quality improvement.
Previous studies have described a relationship between increased health care utilization and adverse events. Compared with those who did not require a PLOS, those who did experienced a higher frequency of complications such air leaks greater than 5 days, pneumonia, adult respiratory distress syndrome, and in-hospital death [17]. Forty percent of readmissions after lung resection were related to pulmonary complications or infection in a single-institution study [18]. Reasons for discharge to an ICF have not been described among patients undergoing lung resection, but in a study of coronary artery bypass grafting patients, at least 36% of those who were discharged to an ICF went there for medical reasons [19].
Not surprisingly, increased health care utilization is also a marker of baseline health status. Wright and associates [18] demonstrated that patients who required PLOS were older and had more comorbidities and severe underlying disease compared with those without a PLOS. Our study demonstrated that older age and a greater number of comorbid conditions were not only associated with PLOS but also discharge to an ICF and readmission. These collective findings support the notion of risk-adjusted measures of health care utilization.
Increased utilization also appears to be a marker of nonclinical determinants of care and outcomes. We show that marital status and geography were associated with increased health care utilization, although we were unable to explore the potential mechanisms underlying these associations. Marital status may be a surrogate for social support. Lack of social support accounted for 15% of discharges to an ICF after coronary artery bypass grafting [19]. Geographic variation in the delivery of care is well documented, although the reasons for this variability remain elusive [20]. Nonclinical reasons for increased utilization may account for some degree of utilization in the absence of an adverse event. From a measurement and quality-improvement perspective, the importance of these findings is that fair comparisons of provider performance based on health care utilization will likely require adjustment for nonclinical determinants of increased utilization.
Health care utilization has changed with time. Another nationally representative study demonstrated the mean LOS after lung resection decreased with time coincident with an increase in discharges to an ICF [4]. We corroborate these findings and report an increasing trend in readmissions. These unadjusted findings do not take into consideration other well-documented changes in the characteristics of operated-on lung cancer patients with time [4, 21]. After adjustment, trends in health care utilization remained statistically significant but may not have been important after 1996. Interestingly, several lines of evidence suggest that despite an increasingly higher-risk group of operated-on lung cancer patients with time, morbidity and mortality has decreased with time [12, 22]. Only trends in PLOS paralleled these declining trends in adverse events, which raises questions about the suitability of discharge to an ICF and readmission as surrogates for quality.
Opposing trends in PLOS versus discharge to an ICF and readmission have been observed among cardiac surgical patients and have raised the hypothesis that one form of utilization might simply be exchanged for another [19, 23]. If PLOS, discharge to an ICF, and readmission are surrogates for adverse events, then one might reasonably expect the risks of discharge to an ICF or readmission to increase with LOS. We observed this relationship between LOS and discharge to an ICF but not LOS and readmission, suggesting that some patients leave the hospital too early. Reasons for early discharge may have to do with patients (insistence to leave early), surgeons (failure to recognize an adverse event or high-risk situation), or hospitals (changing paradigms of resource utilization or economic pressure). Early discharge raises two issues from a measurement and quality-improvement perspective. If patient- and system-level reasons for early discharge vary across providers, then measuring only one type of health care utilization might misclassify provider performance. If providers are the key drivers of utilization and their performance is only measured by one type of utilization, then there is the possibility that some might attempt to improve their performance with earlier patient discharges, even though rates of adverse events may not change or even increase. A composite measure for utilization may overcome potential limitations of measuring only one marker of utilization.
The association between increased health care utilization and long-term outcomes is likely explained by multiple factors. One explanation is a higher prevalence of unfavorable, unmeasured baseline health characteristics among patients who required increased utilization. Another explanation is the higher incidence of adverse events resulting in increased utilization. Describing a link between increased health care utilization and long-term outcomes in the absence of confounding factors is important because it would help establish that increased utilization is a patient-centered surrogate measure of quality above and beyond being a convenient measure.
This study has a number of limitations not previously discussed. Our findings may not be generalizable to nonelderly patients or those with insurance other than Medicare, although the results are likely applicable to a majority of lung cancer patients in the United States. The mean age of operated-on lung cancer patients is 67 years, and Medicare covers 97% of the care received by elderly patients [24]. Another limitation has to do with the timeliness of the data. Purely administrative data or The Society of Thoracic Surgeons General Thoracic Database might offer a more contemporaneous examination of health care utilization. However, administrative data do not robustly identify cancer patients, describe cancer stage, or provide information about nonclinical variables such as marital status. The Society of Thoracic Surgeons General Thoracic Database only recently (January 2009) began collecting information on readmission or discharge status. Also, The Society of Thoracic Surgeons General Thoracic Database may be more representative of care provided by surgical specialists rather than care provided by all types of surgeons, and it has not yet been formally audited [13].
In summary, although others have demonstrated that increased health care utilization is a surrogate for surgical morbidity and mortality, we show that increased utilization is also a marker of baseline health status and nonclinical factors. Increased health care utilization is changing as a function of time, and one form of utilization might be replacing another. Patients who required increased utilization had worse long-term outcomes. These findings have implications for quality-improvement initiatives that use increased utilization as a surrogate for provider performance. Risk-adjustment may be necessary if risk factors for utilization vary across providers, and, importantly, risk-adjustment models may also have to account for nonclinical determinants of increased utilization. To the extent that discharge to an ICF and readmission are reasonable surrogates for surgical morbidity, a composite measure of utilization might offer advantages versus PLOS alone.
|
Acknowledgments
|
|---|
This study used the linked SEER-Medicare database. The authors acknowledge the efforts of the Applied Research Program, National Cancer Institute (NCI); the Office of Research, Development and Information, Centers for Medicare & Medicaid Services (CMS); Information Management Services, Inc; and the SEER Program tumor registries in the creation of the SEER-Medicare database. F.F. was supported by a Cancer Epidemiology and Biostatistics Training Grant (T32 CA09168-30) and National Research Service Award (F32 CA130434-01) from the NCI. The interpretation and reporting of these data are the sole responsibility of the authors. The views expressed in this article do not necessarily represent the official views of the NCI, National Institutes of Health, CMS, or the University of Washington (UW). We are grateful for statistical advice provided by Professor Patrick J. Heagerty, PhD, of the UW Department of Biostatistics.
|
References
|
|---|
- National quality forum endorses national consensus standards promoting accountability and public reportinghttp://www.qualityforum.org/news/releases/080508-endorsed-measures.aspAccessed October 16, 2008.
- Finlayson E, Fan Z, Birkmeyer JD. Outcomes in octogenarians undergoing high-risk cancer operation: a national study J Am Coll Surg 2007;205:729-734.[Medline]
- Goodney PP, Stukel TA, Lucas FL, Finlayson EV, Birkmeyer JD. Hospital volume, length of stay, and readmission rates in high-risk surgery Ann Surg 2003;238:161-167.[Medline]
- Memtsoudis SG, Besculides MC, Zellos L, Patil N, Rogers SO. Trends in lung surgery: United States 1988 to 2002 Chest 2006;130:1462-1470.[Abstract/Free Full Text]
- Warren JL, Klabunde CN, Schrag D, Bach PB, Riley GF. Overview of the SEER-Medicare data: content, research applications, and generalizability to the United States elderly population Med Care 2002;40(8 Suppl):IV-3-18.
- Farjah F, Wood DE, Yanez 3rd D, Symons RG, Krishnadasan B, Flum DR. Temporal trends in the management of potentially resectable lung cancer Ann Thorac Surg 2008;85:1850-1856.[Abstract/Free Full Text]
- Bach PB, Guadagnoli E, Schrag D, Schussler N, Warren JL. Patient demographic and socioeconomic characteristics in the SEER-Medicare database applications and limitations Med Care 2002;40(8 Suppl):IV-19-IV-25.[Medline]
- Klabunde CN, Potosky AL, Legler JM, Warren JL. Development of a comorbidity index using physician claims data J Clin Epidemiol 2000;53:1258-1267.[Medline]
- Ambler G, Omar RZ, Royston P. A comparison of imputation techniques for handling missing predictor values in a risk model with a binary outcome Stat Methods Med Res 2007;16:277-298.[Abstract/Free Full Text]
- Birkmeyer JD, Siewers AE, Finlayson EV, et al. Hospital volume and surgical mortality in the United States N Engl J Med 2002;346:1128-1137.[Medline]
- Birkmeyer JD, Stukel TA, Siewers AE, Goodney PP, Wennberg DE, Lucas FL. Surgeon volume and operative mortality in the United States N Engl J Med 2003;349:2117-2127.[Medline]
- Allen MS, Darling GE, Pechet TT, et al. Morbidity and mortality of major pulmonary resections in patients with early-stage lung cancer: initial results of the randomized, prospective ACOSOG Z0030 trial Ann Thorac Surg 2006;81:1013-1020.[Abstract/Free Full Text]
- Boffa DJ, Allen MS, Grab JD, Gaissert HA, Harpole DH, Wright CD. Data from The Society of Thoracic Surgeons General Thoracic Surgery database: the surgical management of primary lung tumors J Thorac Cardiovasc Surg 2008;135:247-254.[Abstract/Free Full Text]
- Harpole Jr DH, DeCamp Jr MM, Daley J, et al. Prognostic models of thirty-day mortality and morbidity after major pulmonary resection J Thorac Cardiovasc Surg 1999;117:969-979.[Abstract/Free Full Text]
- Little AG, Rusch VW, Bonner JA, et al. Patterns of surgical care of lung cancer patients Ann Thorac Surg 2005;80:2051-2056.[Abstract/Free Full Text]
- Dimick JB, Welch HG, Birkmeyer JD. Surgical mortality as an indicator of hospital quality: the problem with small sample size JAMA 2004;292:847-851.[Abstract/Free Full Text]
- Wright CD, Gaissert HA, Grab JD, O'Brien SM, Peterson ED, Allen MS. Predictors of prolonged length of stay after lobectomy for lung cancer: a Society of Thoracic Surgeons General Thoracic Surgery Database risk-adjustment model Ann Thorac Surg 2008;85:1857-1865.[Abstract/Free Full Text]
- Handy Jr JR, Child AI, Grunkemeier GL, et al. Hospital readmission after pulmonary resection: prevalence, patterns, and predisposing characteristics Ann Thorac Surg 2001;72:1855-1860.[Abstract/Free Full Text]
- Lazar HL, Fitzgerald CA, Ahmad T, et al. Early discharge after coronary artery bypass graft surgery: are patients really going home earlier? J Thorac Cardiovasc Surg 2001;121:943-950.[Abstract/Free Full Text]
- The quality of medical care in the United States: a report on the Medicare program; the Dartmouth Atlas of Healthcare 1999In: Wennberg DE, Cooper MM, editors. Chicago: Health Forum Inc; 1999.
- de Perrot M, Licker M, Robert J, Spiliopoulos A. Time trend in the surgical management of patients with lung carcinoma Eur J Cardiothorac Surg 1999;15:433-437.[Abstract/Free Full Text]
- Goodney PP, Siewers AE, Stukel TA, Lucas FL, Wennberg DE, Birkmeyer JD. Is surgery getting safer?. National trends in operative mortality. J Am Coll Surg 2002;195:219-227.[Medline]
- Bohmer RM, Newell J, Torchiana DF. The effect of decreasing length of stay on discharge destination and readmission after coronary bypass operation Surgery 2002;132:10-15.[Medline]
- Little AG, Gay EG, Gaspar LE, Stewart AK. National survey of non-small cell lung cancer in the United States: epidemiology, pathology and patterns of care Lung Cancer 2007;57:253-260.[Medline]
Top
Abstract
Introduction
