SEARCH

SEARCH BY CITATION

Keywords:

  • length of stay (LOS);
  • intensive care;
  • ovarian carcinoma;
  • resources

Abstract

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

BACKGROUND

The objective of the current study was to determine the association of age, comorbid illness, and length of stay (LOS) in the intensive care unit (ICU) in women who underwent oophorectomy for ovarian carcinoma.

METHODS

The authors conducted a population-based analysis of all women with a primary or secondary diagnosis of ovarian carcinoma who underwent oophorectomy between 1994–1999. Chi-square tests and Student t tests were used to determined differences in means or proportions. Multivariate regression methods were used to build predictive models.

RESULTS

Of 8109 women who were admitted, 1412 women underwent oophorectomy, 1045 of 1412 women (74%) underwent hysterectomy, 325 of 1412 women (23%) underwent intestinal surgery, and 296 of 1412 women (21%) were admitted to the ICU. Overall (± standard deviation) LOS was 8.3 days ± 6.90 days, and the total charges were $16,675 ± $15,590 (1999 dollars). Patients who underwent intestinal surgery were older (62.5 years vs. 57.1 years; P = 0.01), had a longer LOS (11.62 days vs. 7.33 days; P = 0.01), had a longer ICU stay (1.15 days vs. 0.58 days; P = 0.01), and had a higher mean Charlson Comorbidity Index (CCI) (16.01 vs. 8.73; P = 0.01) compared with patients who did not undergo intestinal surgery. Multivariate regression analysis revealed that age, intestinal surgery, CCI, ICU stay, and African-American race were associated with LOS and contributed indirectly to total charges, whereas age and ICU say were the two most important direct determinants of total charges.

CONCLUSIONS

Advancing age, ICU stay, intestinal surgery, African-American race, and comorbid illness were the most prominent predictors of LOS, whereas age and ICU stay were the most important factors predicting total charges in women who underwent oophorectomy for ovarian carcinoma. Cancer 2002;95:1457–62. © 2002 American Cancer Society.

DOI 10.1002/cncr.10872

Ovarian carcinoma is a major cause of cancer death in women. Most women present with advanced disease and undergo complex oncologic procedures for diagnostic and therapeutic indications. Although advances in surgical technique, anesthesia, and critical care allow elderly women to undergo surgery successfully to remove the tumor bulk, population-based data evaluating hospital-based outcomes of these women are not available. Measures of comorbid illness and their relation to treatment of patients for carcinoma are relevant, in that nearly 70% of older patients with carcinoma have chronic disease or comorbid conditions.1 Emerging data indicate that the effect of comorbidity should be separated from the effects of age in evaluating therapy for patients with carcinoma.2–4 Given the correlation of comorbid illness with treatment for patients with carcinoma, the objective of this study was to determine whether age and comorbidity were associated independently with intensive care unit (ICU) stay, total charges, length of stay (LOS), and death during a single admission for oophorectomy in a population-based study of women with ovarian carcinoma.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

This study was a population-based analysis of patients who were admitted with a diagnosis of ovarian carcinoma (International Classification of Disease [ICD-9] 183.0) who underwent oophorectomy during the period 1994–1999. The Maryland Health Services Cost Review Commission statewide discharge data base records all admissions that occur throughout the state's 53 acute care hospitals. The data base captures single episode inpatient hospital stay data by Current Procedural Terminology coding and ICD-9 coding. Thus, information regarding histology, presurgical therapy, International Federation of Gynecology and Obstetrics staging, and postsurgical therapy was not available. The data base would not permit evaluation of peritoneal cytology or of preoperative or postoperative treatment.

The unit of analysis is the hospitalization and not the individual; thus, unique individuals and physicians cannot be identified. However, due to the nature of our inquiry, (oophorectomy), in fact, each admission represents one patient. The period covered by this analysis was 1994–1999 (currency-related variables are converted to 1999 U.S. dollars by using the Medical Consumer Price Index). The numbers of patients undergoing oophorectomy are compatible with the number of incident cases in the state for 1995–1999.

This data base includes a higher percentage of minorities than a national sample. Nonwhites comprise 24.9% of the population nationally, whereas, in Maryland, the proportion of nonwhites is 36%, 78% of whom are African Americans.5

Each discharge abstract supplies patient demographic characteristics (e.g., age, gender, and race), primary expected payment source, vital status at discharge (alive or dead), up to five diagnosis codes, procedure codes, and measures of resource use (LOS, daily charges, and total charges) associated with the each hospitalization. Total charges are comprised of all charges recorded by the hospital incurred during each hospitalization and include operating room fees, laboratory fees, bed charges, pharmacy charges, and radiology charges. Professional fees are not included in the charges. Hospital charge rates are standardized throughout the state and are 0.1% above the national average.6

The independent variables included age, race, and type of surgical procedure (i.e., oophorectomy with or without intestinal surgery). ICU stay and comorbid illness, as measured by the Charlson Comorbidity Index (CCI), were independent variables used as measures of severity of illness. The CCI uses ICD-9 codes obtained from chart abstraction and assigns points to each comorbid illness (i.e., diabetes, hypertension, arrhythmia, etc.), and the sum of these points equals the CCI. However, the age factor was included in set of variables separate from the CCI, because the effect of age was different for the total charge (nonlinear effect) and the LOS (linear effect).7 Age, CCI, and ICU stay were treated as continuous variables, and all other independent variables were categorical variables. Total charges and LOS were treated as continuous variables and were the resource-use measures that were analyzed as dependent variables in this study. We analyzed inpatient deaths during the exploratory analysis; however, given the small number of deaths (34 of 1412 patients) and the potential of this variable to bias the results of the other exploratory variables, it was not included in the multivariate regression analysis.

Statistical Analyses

We used STATA software (version 7.0; Stata, Inc., Tulsa, OK) for data management and analysis. Differences in population proportions were tested using Student t tests and chi-square analysis with a preset level of significance equal to P < 0.05. Linear and stepwise regression analyses were used to formulate the final three-stage regression analysis for the dependent variables LOS and total charges per patient per admission. We included variables that were significant at the P ≤ 0.05 level in the final model and reported 95% confidence interval limits (except for the nonlinear relation, for which the significance of the highest order term is only determined by this criterion). The regression coefficient (R2) assessed the predictive ability of each model. The (age)2 variable allowed analysis of the nonlinear relation of age and total charge.

RESULTS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

Of 8109 women who were admitted with a primary diagnosis of ovarian carcinoma, 1412 women underwent oophorectomy during admission. These 1412 women are the subjects of our analysis. Eighty-one percent of women were white, 14% were African American, 4% were of other races, and the mean (± standard deviation) age of the women who were admitted was 58.4 years ± 15.5 years. The average LOS in the hospital for women who underwent oophorectomy was 8.3 days ± 7.0 days, and the average total hospital charge in 1999 dollars was $16,675. The mean CCI score was 10 ± 8 (range, 2–26). Ninety-one percent of patients were discharged home after undergoing oophorectomy, 6% were transferred to other facilities, and 3% died during hospitalization.

Surgical Procedures During Admission

Fourteen hundred twelve patients underwent oophorectomy, and 325 of those 1412 patients (23%) also underwent intestinal surgery. Patients who underwent intestinal surgery were older and had higher levels of comorbidity compared with patients who did not undergo intestinal surgery. (Table 1). Other procedures, such as the destruction of peritoneal tissue and lymphadenectomy, did not influence LOS or total charges. One thousand forty-five patients (74%) underwent hysterectomy in addition to oophorectomy. There were no significant differences in the mean age, CCI, LOS, or total charges for women who underwent hysterectomy compared with women who did not undergo the procedure. Hysterectomy did not influence LOS or total charges. African-American patients were less likely to undergo hysterectomy at the time of admission compared with white patients (P = 0.04).

Table 1. Comparison of Mean Total Charges, Length of Stay, and Days in the Intensive Care Unit with or without Intestinal Surgery
VariableNo intestinal surgeryIntestinal surgeryP value (t test)
  • SD: standard deviation: ICU: Intensive Care Unit; CCI: Charlson Comorbidity Index.

  • a

    Neither normal nor log normal: The student t test could not be performed because of the nonnormality; however, these variables showed significant differences (P < 0.0001) in a Wilcoxon rank-sum test (also known as the Mann–Whitney U test).

  • b

    Chi-square test.

No. of patients1093319
Resource use (SD)   
 Total charges (1999 $)$14,608.41 ($14,827.77)$23,754.29 ($16,078.30)a
 Length of stay (days)7.33 (6.24)11.62 (7.96)a
 ICU days0.58 (2.47)1.15 (2.59)a
 Room and board charge$5280.82 ($5944.10)$8616.05 ($6779.81)a
 Operating room charges$2778.55 ($1329.92)$3761.17 ($1976.12)a
Age (yrs)57.1462.54< 0.01
CCI8.7316.01< 0.01b
Inpatient death rate (%)1.94.10.03

Chemotherapy During Admission

Twenty-five patients who underwent chemotherapy during the same hospitalization that they underwent oophorectomy had higher total charges during admission ($25,292.00 vs. $15,214.51) and had a longer LOS (14.4 days vs. 8.2 days) compared with the 1387 patients who underwent oophorectomy without chemotherapy.

Admission to the ICU

Two hundred ninety-two patients (21%) were admitted to the ICU with a mean LOS of 0.71 days ± 2.5 days. Patients who were admitted to the ICU were older; had more comorbid illnesses, as determined by the CCI; and were more likely to die in the hospital compared with patients who were not admitted to the ICU (Table 2).

Table 2. Comparison of Age, Surgical Procedures, and Resource Utilization in Patients with and without Intensive Care Unit Admissions
VariableICUNo ICUP value (t test)
  • ICU: Intensive Care Unit; SD: standard deviation; CCI: Charlson Comorbidity Index.

  • a

    Neither normal nor log normal: A Student t test could not be performed because of the nonnormality; however, these variables showed significant differences (P < 0.0001) in a Wilcoxon rank-sum test (also known as the Mann–Whitney U test).

No. of patients2921120
Mean (SD) age (yrs)64 (13.9)57 (15.5)< 0.01
Mean (SD) CCI15 (7.7)9 (7.7)< 0.01
Mean (SD) total charge (1999 $)$28,148 ($24,535)$13,683 ($10,327)a
Intestinal surgery (%)108 of 292 (37)213 of 1120 (19)< 0.01
Inpatient death rate (%)19 of 292 (6.5)15 of 1120 (1.3)< 0.01

Mortality

Thirty-four patients died during hospitalization. There were no racial differences in the rates of death. Patients who died were older (67.5 years ± 13.5 years vs. 58.1 years ± 15.5 years; P < 0.01) and had a higher CCI (16.0 ± 8.4 vs. 10.3 ± 8.0; P < 0.01) compared with patients who were discharged alive. Total charges for patients who died were 3.5 times higher compared with the total charges for patients who were discharged alive ($54,562 ± $40,798 vs. $15,740 ± $13,147). Patients who were admitted to the ICU and died subsequently stayed in the hospital longer prior to death compared with patients who were not admitted to the ICU (19.0 days ± 12.8 days vs. 8.0 days ± 6.5 days). In addition, the mean LOS in the ICU for women who died was 6.24 days ± 8.1 days compared with a mean LOS in the ICU of 0.57 days ± 2.0 days for women who were discharged successfully.

Multivariate Analysis

We used multistage regression analysis to analyze health care resource utilization—total charges and LOS. Figure 1 shows the advantage of multistage regression analysis over a single-equation regression analysis. Single-equation regression models capture only the direct effect of independent variables, whereas multistage regression modeling can separate direct effects (Fig. 1, solid arrows) from indirect effects (Fig. 1, dotted arrows) through a change in an intermediate variable like LOS. The regression coefficient (R2) of the model was 0.91 for total charge and 0.40 for LOS, indicating that the model explained 91% and 40%, respectively, of the total charge and the LOS (Table 3). We determined that, although age, LOS in the ICU, intestinal surgery, African-American race, and CCI had an indirect effect on total charge through increased LOS, only the nonlinear effect of age and ICU stay had a strong direct effect on total charge. The nonlinear effect of age is described by an increase in total charges with advancing age with a plateau effect at the upper extremes of age, reflecting lower total charges for the very elderly (Table 3).

thumbnail image

Figure 1. The single equation regression model versus the multistage regression model for women with ovarian carcinoma. ICU: intensive care unit; LOS: length of stay.

Download figure to PowerPoint

Table 3. Three-Stage, Least-Squares Regression Analysis Showing Variables that Affected Total Charge and Total Length of Stay for Patients Who Underwent Oophorectomy for Ovarian Carcinoma
VariableCoefficientSEP value95%CI
  • SE: standard error: 95% CI: 95% confidence interval; ICU: Intensive Care Unit; Age2: variable for the nonlinear relation of age and total change; CCI: Charlson Comorbidity Index.

  • a

    Correlation coefficient (R2) (total charge) = 0.91; R2 (length of stay) = 0.40.

  • b

    In addition to the reference group estimate of 1.60 days.

  • c

    Reference group (non-African American).

Total chargea    
 Length of stay2054.38117.733< 0.011823.628–2285.134
 ICU days1280.73189.852< 0.01908.630–1652.837
 Age39.7124.1770.100−7.675 to 87.097
 Age2−0.990.238< 0.01−1.454 to −0.521
Length of staya    
 Age0.070.009< 0.010.052–0.089
 ICU days1.420.058< 0.011.308–1.534
 African Americanb1.200.378< 0.010.460–1.942
 Intestinal surgery2.540.352< 0.011.850–3.230
 CCI0.080.018< 0.010.049–0.120
Constantc1.600.578< 0.010.470–2.734

DISCUSSION

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

In the current study, we were able to correlate variables that were associated with resource use and death in women undergoing surgery for ovarian carcinoma. We determined that advancing age, African-American race, intestinal surgery, and comorbid illness (CCI) were associated with LOS; however, advancing age and ICU stay were the only two variables that were associated independently and directly with total charges incurred during a hospital stay. Because total charges and LOS are correlated highly, we used linear, stepwise, and three-stage regression modeling techniques to determine the extent to which LOS influenced total charges. The three-stage regression model solves the error introduced into single regression equations when the relations between equations are not independent by plugging the fitted value of the LOS variable (free from error terms) into the total charge equation. These findings assist us in assessing the relative impact of clinical and demographic variables on resource use for the surgical treatment of patients with ovarian carcinoma.

We also were able to correlate on a population basis the association of ICU stay with inpatient deaths. The 3% death rate recorded by this report is five times the rate reported by our group for a population of women who underwent hysterectomy for endometrial carcinoma.2 In a post-hoc analysis, we determined that ICU stays ≥ 6 days were associated with a significantly higher rate of death compared with shorter ICU stays. Although the numbers of deaths were small in this study, our findings are compatible with those of Abbas et al., who reported a significant association between inpatient death and ICU stay beyond 5 days (i.e., stays ≥ 6 days).8 These findings should be confirmed with prospective studies to allow health care providers and payers to understand the variables associated with resource use and postsurgical death in women undergoing surgery for ovarian carcinoma.

Because administrative data bases do not provide information regarding surgical stage, we investigated a number of coded surgical procedures performed coincident with oophorectomy, such as hysterectomy, destruction of peritoneal tissue, lymphadenectomy, and intestinal surgery, to analyze the influence of the performance of these procedures on resource use. Of the procedures evaluated, only intestinal surgery was associated indirectly with total charges by its effect on LOS. The less frequent performance of hysterectomy in African Americans likely was due to a higher rate of prior hysterectomy in that group.

This report focuses attention on the performance of complex surgical procedures in the elderly in an effort to characterize and quantify variables that are associated with higher resource use. In a report evaluating a combined index of age and comorbidity, Charlson et al. determined that the estimated relative risk of death by an increment of 1 in the comorbidity score was equivalent to adding an additional decade of age. That report and the current study demonstrate the importance of incorporating both components when evaluating resource utilization or health outcomes.1, 9, 10 It is interesting to note that patients who underwent intestinal surgery, on average, were 5 years older and had a CCI that was nearly double (16.0 vs. 8.7) that of women who did not undergo such surgery, indicating the presence of more biologically aggressive or longer standing disease. Analysis of age as a squared factor, as we reported previously, revealed that, although expenditures increased with age, there was a flattening of this relation in the older age groups, indicating lower total charges for the very elderly. These findings indicate that elderly patients remain in the hospital for longer periods without receiving ongoing expensive medical procedures. Similarly, the longer LOS for African Americans may have been due to factors that cannot be measured with an administrative data base.1

Although this report presents an innovative use of an administrative data base that affords access to data on a large number of patients, we acknowledge the limitations imposed by a retrospective analysis.11, 12 Ambiguities in coding, completeness of coding, and uncoded factors limit the ability of administrative data to adjust for risk.11, 13 However, we have classified our patients using diagnostic classifications, thus following the principles of prospective studies by categorizing patients into homogeneous groups with comparable key characteristics.10, 11 With respect to errors in coding, it has been determined that, although under-reporting may exist, when a diagnosis related to cancer is included on the abstract, it is likely to be reliable.11 Despite their limitations, administrative data bases provide population-based insight, allow assessment of resource use, and are used increasingly to study hospital-based outcomes for patients with cancer because of their unique ability to capture clinical, demographic, and resource use data. These studies also provide information that has not been published previously in single-institution or cooperative group studies.14

In summary, we determined that, although the demographic variables intestinal surgery and comorbid illness were the most significant predictors of LOS, total charges were predicted primarily by the variables age and ICU stay in women who underwent surgery for ovarian carcinoma. These results may form the basis of prospective studies to develop benchmarks for ICU use in patients with ovarian carcinoma. We also believe that these data add to the understanding of the relation of comorbidity to treatment of ovarian carcinoma and LOS in the hospital. This information also may be useful for managed care organizations when projecting expenditures for treatment of patients with carcinoma based on the underlying demographics of their insured.

Acknowledgements

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

The authors thank Ms. Mary McAllister for her assistance with the article.

REFERENCES

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES
  • 1
    Librero J, Peiro S, Ordinana R. Chronic comorbidity and outcomes of hospital care: length of stay, mortality, and readmission at 30 and 365 days. J Clin Epidemiol. 1999; 52: 171179.
  • 2
    Brooks SE, Ahn J, Mullins CD, Baquet CR, D'Andrea A. Health Care Cost and Utilization Project (HCUP) analysis of comorbid illness and complications for patients undergoing hysterectomy for endometrial cancer. Cancer. 2001; 92: 950958.
  • 3
    Chalfin DB, Carlon GC. Age and utilization of intensive care unit resources of critically ill cancer patients. Crit Care Med. 1990; 18: 694698.
  • 4
    Abbas FM, Sert MB, Rosenshein NB, Zahyrak ML, Currie JL. Prolonged stay of OB/GYN patients in the surgical intensive unit. J Reprod Med. 1997; 42: 179183.
  • 5
    U.S. Census Bureau. State and county quick facts [available from URL: http://quickfacts.census.gov/qfd/states/24000.html].Washington, DC: U.S. Census Bureau, 2000 [accessed August 8, 2002].
  • 6
    Maryland Health Care Commission. State health care expenditures experience from 1998. Bethesda: Maryland Health Care Commission, 2000.
  • 7
    Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987; 40: 373383.
  • 8
    Abbas FM, Sert MB, Rosenshein NB, Zahyrak ML, Currie JL. Gynecologic oncology patients in the surgical ICU. Impact on outcome. J Reprod Med. 1997; 42: 173178.
  • 9
    Charlson M, Szatrowski TP, Peterson J, Gold J. Validation of a combined comorbidity index. J Clin Epidemiol. 1994; 47: 12451251.
  • 10
    Ball JK, Elizhauser A. Treatment differences between African-Americans and whites with colorectal cancer. Med Care. 1996; 34: 970984.
  • 11
    Farley DE, Ball JK: The quality and availability of discharge abstract data: The HCUP-2 experience. Presented at the Annual Meeting of the National Association of Health Data Organizations, Washington, DC, November 9–10, 1989.
  • 12
    Berthelsen CL. Evaluation of coding data quality of the HCUP National Inpatient Sample. Top Health Inform Manage. 2000; 21: 1023.
  • 13
    Demlo LK, Campbell PM, Brown SS. Reliability of information abstracted from patient's medical records. Med Care. 1978; 16: 995905.
  • 14
    Best AEJ. Secondary databases and their use in outcomes research: a review of the area resource file and the Healthcare Cost and Utilization Project. Med Syst. 1999; 23: 175181.