Where the Other Half Dies: Analysis of Mortalities Occurring More Than 30 Days After Complex Cancer Surgery.

BACKGROUND: Nearly half of operative mortalities occur outside the traditionally studied 30-day period after surgery. To identify additional opportunities to improve surgical safety, the circumstances of deaths occurring 31-90 days after complex cancer surgery are analyzed. PATIENTS AND METHODS: Patients aged ≥ 65 years who died within 90 days of complex cancer surgery for nonmetastatic cancer were analyzed in the Surveillance, Epidemiology, and End Results (SEER)-Medicare and the Connecticut Tumor Registry (CTR) databases. RESULTS: Of the 36,114 patients undergoing complex cancer surgery from 2004 to 2013 in SEER-Medicare, 1367 (3.8%) died within 31-90 days ("late mortalities"). Seventy-eight percent of late mortalities were readmitted prior to death. The highest proportion of late mortalities occurred during a readmission (49%), and 11% were never discharged from their index admission. Cause of death (COD) was largely attributed to the malignancy itself (56%), which is unlikely to be the underlying cause. Of the noncancer COD, cardiac causes were most frequent (34%), followed by pulmonary causes (18%). Death was rarely attributed to thromboembolic disease (< 1%). The CTR provided location of death, which was most commonly in a hospital (65%) or nursing facility (20%); death at home was rare (6%). CONCLUSIONS: The vast majority of patients dying between 31 and 90 days of surgery were admitted to a hospital or nursing facility at the time of their death after initially being discharged, and few patients died at home. Greater clarity in death documentation is needed to identify specific opportunities to rescue patients from fatal complications arising in the later postoperative period.

Surveillance of the frequency and results of testing of incident oropharyngeal cancers for human papillomavirus: the potential role of population-based cancer registries.

Temporal increases in incidence rates for certain cancers of the oropharynx (OP), especially the base of tongue and tonsil (BTT), have been interpreted in relation to the epidemic of human papillomavirus (HPV) infection, but data on the actual presence of HPV in these tumors are limited. Data on the frequency and results of testing for HPV in OP cancers in defined populations also can be useful to clinicians. This study used the American Joint Committee on Cancer Collaborative Staging System's Site-Specific Factor 10 (SSF 10) for HPV status of OP tumors, collected by some registries for diagnoses since 2010. The study included 483 incident invasive BTT cancers diagnosed in 2010-2012 and reported to the Connecticut Tumor Registry of the Surveillance, Epidemiology and End Results (SEER) Program. Of the 483 cancers, 45.8 percent were reportedly tested for HPV in tumor tissue; the proportion coded as unknown declined from 54.6 percent for 2010 to 34.3 percent for 2012. The 153 cases reported as HPV-positive comprised 69.2 percent of the 221 cases with a known HPV test result, which is consistent with the proportions reported in the literature. Trends (2000-2010) in BTT cancer incidence rates in Connecticut were representative of trends in all 18 SEER registries combined. Similar studies are needed from other US central cancer registries that are collecting or want to start collecting HPV status of OP tumors, along with data on the specific types of HPV testing, for surveillance of the frequency and results of HPV testing of OP cancers.

Cancers coded as tongue not otherwise specified: relevance to surveillance of human papillomavirus-related cancers.

Data from US population-based cancer registries have shown increasing incidence rates for cancer of the base of the tongue, interpreted as related to the epidemic of human papillomavirus (HPV) infection, but rates could be underestimated due to miscoding of some base of tongue cancers to tongue "not otherwise specified" (NOS). Tongue NOS was the most commonly coded subsite among incident (2000-2011) invasive cancers of the oral tongue (tongue excluding base of tongue and lingual tonsil which together comprise the posterior one-third of the tongue) in the 18 Surveillance, Epidemiology and End Results (SEER) Program registries combined and in the Connecticut SEER registry. All 173 cases of tongue NOS cancer in the Connecticut SEER registry diagnosed in selected years were reviewed. Only 5% were recoded to base of tongue, decreasing from over time from 8% to 2%, resulting in minimal impact on the incidence rate for base of tongue cancer in Connecticut. Most (76%) of the 173 tongue NOS cases were recoded to anterior two-thirds of tongue NOS, ruling out base of tongue as the actual site but resulting in underestimation of incidence rates for anterior two-thirds NOS in Connecticut. Similar studies are needed on tongue NOS cancers in other US cancer registries, along with studies on the HPV status of tumors at specific subsites of the oral tongue, to enhance surveillance and interpretation of trends in cancer incidence in relation to the HPV epidemic.

Coding of specific subgroups of myelodysplastic syndromes in a population-based cancer registry: prospects for improvement.

Myelodysplastic syndromes (MDS), reportable to US cancer registries for diagnoses since 2001, are a group of myeloid neoplasms heterogeneous in prognosis and treatment, and of growing importance in an aging population. In US registries that have reported incident MDS cases by subgroup, about 50%-67% of cases have been coded as MDS "not otherwise specified" (NOS) in the International Classification of Diseases for Oncology Version 3 (ICD-O-3). For this study, MDS cases diagnosed in 2001-2009 and reported to the population-based Connecticut Tumor Registry (CTR) were analyzed. MDS was coded as NOS for 573 (56.7%) of 1,011 cases, but the proportion varied among reporting facilities hospitals (ie, from 0 to 100%), with several statistical outliers. In pathology reports obtained for 130 CTR patients diagnosed with MDS in 2008-2009, 84% of the 62 patients coded as NOS had information on a key element (ie, % of blasts in bone marrow) in ICD-O-3 coding and other classifications of MDS subgroups. These findings suggest that central cancer registries may want to work with hospital tumor registrars in improving reporting of specific MDS subgroups using ICD-O-3. The addition of % blasts from pathology reports to the site-specific factors for MDS in the Collaborative Staging System could be proposed.

Leukemia as a cause of death among patients with myelodysplastic syndromes (MDS) in a population- based cancer registry: improving estimates of MDS-related mortality in the population.

Myelodysplastic syndromes (MDS), a heterogeneous group of myeloid neoplasms diagnosed mostly in elderly persons, are of increasing interest in an aging population and are associated with variable risk of progression to acute myeloid leukemia (AML). The numbers of deaths related to MDS in the population are underestimated in routine US cancer mortality statistics which are based on only the underlying cause (UC) rather than multiple causes (MCs) of death recorded on death certificates. Additional MDS-related deaths, however, may be missed if some MDS patients die with mention of leukemia but not MDS on their death certificate. This requires studies of MCs of death among all MDS patients in population-based tumor registries. This study examined MCs of death among patients diagnosed with MDS in 2001- 2009 and reported to the population-based Connecticut Tumor Registry. MDS was the UC for 199 deaths (25.7% of all 773) and was coded as other than UC for 160 (20.7%). Another 121 (15.7%) death records, however, had leukemia without mention of MDS; the majority were coded to AML and most of the others as unspecified type of acute leukemia. If these 121 deaths are added to the 359 with mention of MDS, the total of MDS-related deaths would be 480 (or 62.1% of all 773 deaths). A total of 178 deaths (23.0% of all 773) were coded to leukemia as the UC, and would be included with leukemia (not MDS) in routine cancer mortality statistics. Leukemia diagnosed since 2010 in MDS patients is reportable to registries as a new primary cancer. This new rule will help central cancer registries to confirm leukemia diagnoses coded on death records, as part of the process of improving surveillance of cancer mortality rates in the population.

Obtaining data on comorbid diabetes among patients in a U.S. population-based tumor registry.

Comorbid diabetes mellitus has been shown to be associated with outcomes among cancer patients, but population-based data have been limited to elderly patients through linkages between the US National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) Program database and Medicare databases. Reporting of comorbidity to the population-based Connecticut SEER registry is not required, but the extent of voluntary reporting of comorbid diabetes was assessed in this preliminary study. Of 15,145 Connecticut residents diagnosed at age 20+ years with invasive cancer in 2006, who were ascertained from 33 registry sources, 8688 (57.4%) from 21 sources were included in the analysis of comorbid diabetes. The prevalence of comorbid diabetes was 12.5%, and was lowest for patients with prostate cancer (8.5%) and highest for with liver-pancreas cancer (25.9%), consistent with the literature. Diabetes prevalence was substantial (9.5%) within the non-elderly subgroup aged 20-64 years at cancer diagnosis who comprised 45% of the 8688 patients. These results indicate an opportunity for future large-scale studies of the impact of diabetes on outcomes among all newly diagnosed cancer patients (both non-elderly and elderly) in the Connecticut SEER registry and other US central cancer registries.