Ninety-Day Mortality as a Reporting Parameter for High-Grade Serous Ovarian Cancer Cytoreduction Surgery

Abstract

OBJECTIVE: 

To evaluate the utility of using 90-day as an adjunct to 30-day mortality rates after surgical cytoreduction for serous ovarian cancer and to compare them across hospitals of differing case volumes over time.

METHODS: 

We performed a retrospective cohort study using the National Cancer Database of women undergoing cytoreductive surgery for high-grade serous carcinoma between 2004 and 2012. The primary outcome of the study was mortality rate by hospital volume. The secondary outcome was to evaluate the performance of hospital rankings based on 30- and 90-day mortality rates. Hospitals were categorized by cases per year as low volume (10 or fewer), intermediate (11–20), high (21–30), and ultra-high (31 or more).

RESULTS: 

A total of 24,827 women from 602 hospitals were included. Overall 30-day mortality was 2.1% (95% CI 1.95–2.3) compared with 90-day mortality of 5.1% (95% CI 4.8–5.4%, P<.001). For each hospital volume category, the 90-day mortality was approximately double that of the 30-day mortality. Substituting 90-day in place of 30-day mortality for hospital ranking, 57 hospitals (9.5%) changed ranks (26 worsened and 31 improved). Based on the logistic regression model (after controlling for age, race–ethnicity, income, Charlson comorbidity index, insurance status, hospital volume, distance from place of residence to the hospital, receipt of neoadjuvant chemotherapy, and year of diagnosis), care at the ultra–high-volume centers was an independent predictor of lower odds of death at 90 days [adjusted odds ratios (OR) 0.60, 95% CI 0.38–0.96, P=.034] but not at 30 days (adjusted OR 0.64, 95% CI 0.35–1.18).

CONCLUSION: 

Compared with low-volume centers, ultra–high-volume centers are associated with significantly lower 30- and 90-day risk-adjusted mortality. The 90-day mortality rate is double that of the 30-day rate and may be a better metric for assessing the initial quality of care for patients with ovarian cancer.

In Brief

Ninety-day mortality rates are more than twice the 30-day mortality rate after cytoreductive surgery, and both are consistently lower at hospitals performing a high volume of serous ovarian cancer surgeries.

Deaths among patients with serious treatable complications after surgery is now a key measure used by the Centers for Medicare & Medicaid Services in evaluating hospitals.¹ There is considerable debate regarding the time period after the index surgical procedure that should be included in this metric. Currently, 30 days after the surgical procedure is the most widely used metric and is included in major quality improvement initiatives such as the National Surgical Quality Improvement Program and the Michigan Surgical Quality Collaborative.^2,3

In ovarian cancer, and specifically in high-grade serous ovarian cancer, there are several reasons to believe that 30-day mortality might not be the most appropriate measure of surgical quality. Managing high-grade serous ovarian carcinoma requires a level of “appropriate aggressiveness” and improvement in survival depends on the efforts to achieve no residual disease and timely delivery of cytotoxic chemotherapy after surgery.^4,5 Most of these patients receive chemotherapy after surgery, and the 30-day metric does not consider morbidity and mortality during this phase of care. Also, with the advent of more intensive life-prolonging measures, 30-day mortality might not be a true metric of surgical outcomes. It is possible that improved medical care can prolong life even in patients with poor surgical outcomes. The result could be more patients alive within 30 days of surgery, but not necessarily representing a meaningful prolongation of life.⁶

One alternative to 30-day mortality is to increase the time period of mortality reporting. Studies in other cancer sites have proposed 90 days as an alternative time point.^7–9 We therefore sought to compare 30- and 90-day mortality in patients undergoing debulking surgery for high-grade serous ovarian carcinoma.

MATERIALS AND METHODS

We performed a retrospective cohort study using the National Cancer Database of women diagnosed with high-grade serous carcinoma between 2004 and 2012 who were undergoing cytoreductive surgery during the primary treatment course. The National Cancer Database, referred to here as “the database,” is a joint program of the American Cancer Society and the Commission on Cancer of the American College of Surgeons. This database uses a hospital-based registry populated by approved programs that are required by the Commission on Cancer to abstract and follow all malignant tumors either diagnosed or treated at the hospital.¹⁰ Approximately 70% of the ovarian cancer cases in the United States are reported to the database.¹¹ Data are deidentified to ensure confidentiality; therefore, this study is exempt from obtaining informed consent by the study participants and met the criteria for exempt status by the University of Michigan institutional review board.

Cases were selected using the International Classification of Diseases, 10th Revision diagnostic code C56.9 and were restricted to high-grade serous carcinoma with International Federation of Gynecologists and Obstetricians stage III or IV. We included patients undergoing debulking surgery as part of initial treatment of ovarian cancer. We included both primary debulking cases and those with interval debulking after neoadjuvant chemotherapy.

Race and ethnicity were used to construct single mutually exclusive race–ethnicity variables (non-Hispanic white, non-Hispanic black, Hispanic, Asian, or other or unknown). Insurance status was categorized as uninsured, Medicaid, private, Medicare, or unknown. Age was categorized in deciles (40 or younger, 41–50, 51–60, 61–70, 71–80, and older than 80 years). The database provides four category quartiles of the median household income (derived from zip code of residence based on the 2012 American Community Survey data), which were used as a proxy for socioeconomic status. To adjust for comorbidities in the cohort, we used the Charlson scores of 0, 1, or 2 or greater, as provided by the database.¹¹ Mean hospital volume was calculated by dividing the total number of cases reported by the hospital by the time period during which the hospital reported the cases to the database. Mean hospital volume was then divided into four categories of annualized hospital volume (low: 10 or fewer cases per year, intermediate: 11–20 cases per year, high: 21–30 cases per year, and ultra-high: 31 or more cases per year). Distance traveled to the hospital (as reported in the database) was categorized into four categories by dividing the entire range of values into quartiles. The resulting categories were first quartile (5 miles or less), second quartile (5.1–15 miles), third quartile (15.1–40 miles), and fourth quartile (40 miles or greater).

Patients receiving chemotherapy before surgery were categorized as having undergone neoadjuvant chemotherapy. Patients undergoing major cytoreductive surgery were identified using the Facility Oncology Registry Data Standards as per the database guidelines.¹⁰ These patients included 1) partial or total unilateral or bilateral salpingo-oophorectomy with omentectomy, with or without hysterectomy; 2) debulking or cytoreductive surgery with the colon (including appendix), small intestine resection, or partial resection of the urinary tract (not incidental); or 3) anterior, posterior, total, or extended pelvic exenteration.

We compared the covariate distributions across outcome groups using χ² tests and independent two-sample t tests. Linear changes in study variables over the timeframe of the study were assessed using a nonparametric trend test (Stata, nptrend).¹² Factors associated with postoperative mortality based on prior knowledge or clinical plausibility were included in multivariable analyses. To account for clustering of the cases at the hospital level, a random-effects multivariable logistic regression model, with hospital as a random intercept, was used to assess factors associated with mortality at 30 and 90 days. Factors considered in the model included patient medical and sociodemographic factors (age, race–ethnicity, income, Charlson comorbidity index, and insurance status) and treatment factors (hospital volume, distance from place of residence to the hospital, receipt of neoadjuvant chemotherapy, and year of diagnosis). Using the odds ratios derived from logistic regression analysis, we calculated the risk-adjusted probability of 30- and 90-day mortality. This was obtained by postestimation computation of predicted or expected values from the fitted logistic regression model using methodology previously described.^13,14 These risk-adjusted mortality rates were then compared across the four categories of hospital volume.

The primary outcome of the study was mortality rate (at 30 and 90 days) by hospital volume. The secondary outcome was hospital rankings based on mortality rate. To evaluate the performance of 30- compared with 90-day mortality, we used the previously calculated risk-adjusted mortality to create three ranking categories for the hospitals. Hospitals with 30-day mortality rates lower than the 10th percentile were categorized as “outperforming at 30 days,” those higher than the 90th percentile were categorized as “underperforming at 30 days,” and the rest were categorized as “baseline performers at 30 days.” A second set of ranking variables with similar categories was created using the 90-day risk-adjusted mortality. The agreement between hospital rankings based on 30- and 90-day mortality was assessed by weighted κ as previously described in similar studies.¹⁵

Last, we performed sensitivity analysis by excluding patients receiving neoadjuvant chemotherapy.

RESULTS

We identified 24,827 patients with high-grade serous ovarian carcinoma undergoing debulking surgery in 602 hospitals from the database. A total of 493 low-volume hospitals performed 13,139 (53%) of the surgeries, 75 intermediate-volume hospitals performed 8,010 (32%) of the surgeries, 11 high-volume hospitals performed 1,963 (8%) of the surgeries, and five ultra–high-volume hospitals performed 1,693 (7%) of the surgeries. The sociodemographic information about the overall cohort and by hospital case volume is presented in Table 1. Patients treated in high– and ultra–high-volume hospitals were more likely to be non-Hispanic white, carry private insurance, have higher income, travel a greater distance, and have stage IV disease (Table 1). No significant differences were noted in the age and comorbidity distributions across the hospital case volumes.

Table 1.: Sociodemographic Variables of the Overall Cohort and by Hospital Case Volume

Overall 30-day mortality was 2.1% (95% CI 1.95–2.3%) compared with a 90-day mortality rate of 5.1% (95% CI 4.8–5.4%, P<.001). Based on the hospital volume, the unadjusted 30-day mortality rates for the 8-year period were 2.41% (95% CI 2.16–2.68%) for low-volume, 1.82% (95% CI 1.55–2.14%) for intermediate-volume, 2.09% (95% CI 1.54–2.82%) for high-volume, and 1.42% (95% CI 0.95–2.1%) for ultra–high-volume centers. Over the same timeframe, the overall unadjusted 90-day mortality rates were 5.66% (95% CI 5.28–6.07%) for low-volume, 4.6% (95% CI 4.16–5.07%) for intermediate-volume, 5.1% (95% CI 4.2–6.16%) for high-volume, and 3.37% (95% CI 2.6–4.34%) for ultra–high-volume centers.

Overall, the rate of both 30- and 90-day mortality declined over the study period (Fig. 1). The trend in decline in 30-day mortality was not significant (P=.089); however, the decline in 90-day mortality was noted to be significant (P<.001). Based on the hospital case volume, the most substantial decline in mortality was noted at the ultra–high-volume centers. The 30-day mortality rate for ultra–high-volume hospitals fell from 3.48% (95% CI 1.57–7.57%) to 0.5% (95% CI 0.07–3.5%) (P=.15) and the 90-day mortality rate fell from 5.81% (95% CI 3.1–10.5%) to 1.5% (95% CI 0.49–4.6%, P=.001). Mortality rates at 30 and 90 days also declined for the low-, intermediate-, and high-volume hospitals, although the rate of decline was lower than that of the ultra–high-volume centers (Fig. 1).

Fig. 1.: Unadjusted yearly 30-day and 90-day mortality rates in patients undergoing debulking surgery for ovarian carcinoma overall (A) and by yearly case volume of the treating facility (B–E). Ten or fewer cases per year (B), 11–20 cases per year (C), 21–30 cases per year (D), 31 or more cases per year (E).Spencer. Ninety-Day Mortality in Ovarian Cancer. Obstet Gynecol 2017.

Based on the regression models, after controlling for patient medical and sociodemographic factors (age, race–ethnicity, income, Charlson comorbidity index, and insurance status) and treatment factors (hospital volume, distance from place of residence to the hospital, receipt of neoadjuvant chemotherapy, and year of diagnosis), care at the ultra–high-volume centers was an independent predictor of lower odds of death at 90 days [adjusted odds ratios (OR) 0.604, 95% CI 0.38–0.96, P=.034], but not at 30 days (adjusted OR 0.64, 95% CI 0.35–1.18). Results of multivariable logistic regression models for factors associated with 30- and 90-day mortality are presented in Table 2. The rates of risk-adjusted 30- and 90-day mortality were significantly lower at the ultra–high-volume centers (Fig. 2).

Table 2.: Results of Logistic Regression Models for Factors Associated With 30- and 90-Day Mortality

Table 2-A.: Results of Logistic Regression Models for Factors Associated With 30- and 90-Day Mortality

Fig. 2.: Risk-adjusted 30-day (A) and 90-day (B) mortality rates based on the hospital case volume. Nonoverlapping CIs represent statistical significance.Spencer. Ninety-Day Mortality in Ovarian Cancer. Obstet Gynecol 2017.

Changes in the risk-adjusted ranking of the hospitals based on the 30- and 90-day mortality rates are presented in Figure 3. Overall, 57 hospitals (9.5%) changed ranks between the two mortality metrics. For example, 14 of the 19 hospitals ranked as “underperforming” by the 30-day mortality metric were “baseline performers” based on the 90-day mortality metric. Of the 35 hospitals ranked as “outperforming” by the 30-day metric, only 22 (63%) remained “outperforming” by the 90-day metric. There was moderate agreement between the 30-day and 90-day mortality-based rankings of the hospitals with a weighted κ of 0.491 (95% CI 0.479–0.498).

Fig. 3.: Changes in risk-adjusted ranking of hospitals based on 30-day and 90-day mortality rates.Spencer. Ninety-Day Mortality in Ovarian Cancer. Obstet Gynecol 2017.

We combined those who received upfront cytoreduction and those who received neoadjuvant chemotherapy with interval cytoreduction because the neoadjuvant chemotherapy group comprised only 15% of the total study population. We felt it was important to look at the mortality rates with this group included because this mimics the current state of care for patients with ovarian cancer. On sensitivity analysis, our results were unchanged. After excluding 3,643 (14.7%) of the patients receiving neoadjuvant chemotherapy, the rates of 30- and 90-day mortality over the study period showed an improvement in both 30- and 90-day mortality on the sensitivity analysis similar to our primary analysis (data not shown).

DISCUSSION

This report highlights a significant increase in mortality 90 days after surgery for ovarian cancer when compared with the standard 30-day mortality metric. By incorporating improvements in medical care, physicians and hospitals are becoming increasingly adept at keeping patients alive in the immediate postoperative period, so focusing only on the 30-day mortality rate may not capture the true landscape of postoperative morbidity and the complications experienced. Given that most women with ovarian cancer will be receiving chemotherapy after surgery, a longer time interval to assess the quality of care during the initial treatment of ovarian cancer may be more meaningful.

The increased usefulness of a 90-day mortality assessment is supported by findings in other cancer sites. Mortality rates after pancreatectomy for oncologic indications was also shown to double from 3.7% at 30 days to 7.4% at 90 days with mortality rates lower at high-volume hospitals. The authors hypothesized that the ability to care for critically ill postoperative patients might have played a role in that outcome.⁹ In et al¹⁵ reported a doubling of the mortality rate after esophagectomy for esophageal cancer, from 4.2% at 30 days to 8.9% at 90 days. They concluded that 90-day mortality represented a better outcome measure because it evaluated treatment decisions and factors beyond the perioperative phase of care. Specifically in ovarian carcinoma, analyses of the Surveillance, Epidemiology, and End Results Program–Medicare registry from 1995 to 2007 showed that 90-day mortality after the diagnosis of ovarian cancer was 26%. This 90-day mortality is much higher than our reported approximate 5% figure, which is likely the result of that database including women 65 years of age and older and included those not able to undergo surgery as a result of comorbid conditions.¹⁶

Our data also demonstrate a significant 40% reduction in 90-day mortality for patients treated at an ultra–high-volume center compared with low-volume centers. Possible explanations include a superior ability to rescue patients from complications,¹⁷ higher rates of guideline-based care,¹⁸ and more consistent processes for postoperative care and better surgical quality.¹⁹ The 30-day mortality differences were not significantly different among the hospital volume categories. This could be a result of the 30-day mortality rate measure not allowing for sufficient statistical power to differentiate between the quality of care at low- compared with ultra–high-volume centers.

This finding has important policy implications because Medicare's “Hospital Compare” and other hospital ranking websites evaluate only 30-day mortality.^20,21 These rankings may report spurious results, because only 22 (63%) of the 35 hospitals ranked as “outperforming” at 30 days in our study remained in this category at 90 days. Moreover, 14 of 19 (74%) hospitals that ranked as “underperforming” at 30 days improved to “baseline performers” at 90 days. These findings are not unique to ovarian cancer with similar findings in esophageal cancer and cardiovascular surgery.^15,22 Inclusion of mortality metrics beyond 30 days is urgently needed.

Numerous publications suggest that increased hospital volume results in increased rates of appropriate surgical staging and optimal surgical cytoreduction in patients with ovarian cancer.^23,24 Bristow et al have also shown that high-volume surgeons and high-volume surgical centers (21 cases per year or more) confer improved in-hospital mortality and overall survival.^24,25 Our findings add another dimension to these data by demonstrating a relationship between hospital volume and the trend in improvement of inpatient mortality at ultra–high-volume centers. These data favor centralization of ovarian cancer care,²⁶ because this could result in improved 30-day and 90-day mortality and improved rates of optimal cytoreduction²⁷ while reducing the cost of care.²⁸

Strengths of this study include the ability to analyze multiple hospitals with varying patient volumes while allowing for the evaluation of temporal trends in mortality stratified by volume. In addition, the National Cancer Database minimizes the risk of information bias by having data entered by certified tumor registrars trained in the acquisition of clinical information.

There are some important weaknesses that deserve discussion. There is a possibility of confounding given the inability to determine cytoreduction status, details of adjuvant therapy received, and specific cause of death. The possibility for selection bias also exists because different surgeons and centers likely vary regarding the aggressiveness of cytoreduction and the administration of neoadjuvant chemotherapy. Lastly, we focused only on stage III or higher high-grade serous carcinomas in this analysis because the role of aggressive surgical debulking and chemotherapy is most well established in this histology. This group represents roughly 70% of the entire ovarian carcinoma spectrum, and the mortality rates are the highest in this subgroup. Analysis of the entire ovarian cancer group is unlikely to affect the findings of this study substantially.

Finally, although these data support the use of 90-day mortality as a quality metric for ovarian cancer care, we are unable to determine whether timeframes longer than 90 days would be more beneficial as a result of the constraints of the data set.

These limitations notwithstanding, our findings support the use of 90-day mortality after surgery for serous ovarian cancer as a meaningful and robust quality metric pertaining to initial surgery and adjuvant therapy. Centers performing more than 30 surgeries per year show significant and striking advantages in both 30- and 90-day mortality. The factors that lead to these findings cannot be specifically determined by these data at this time. Future inquiries should investigate specific factors that could improve mortality outcomes in lower volume centers and summarize the data on hospital and surgeon-specific volume in ovarian cancer to establish quality-of-care benchmarks.