190 likes | 198 Views
Accuracy of Cancer Mortality Measured by Death Certificates Project. Two Presentations: Accuracy of Cancer Mortality measured by Death Certificates: Variability in Accuracy by Characteristics of Death Certifiers – Chris Johnson
E N D
Accuracy of Cancer Mortality Measured by Death Certificates Project • Two Presentations: • Accuracy of Cancer Mortality measured by Death Certificates: Variability in Accuracy by Characteristics of Death Certifiers – Chris Johnson • Using the Accuracy of Cancer Mortality (ACM) Study to Evaluate Survival Outcomes for Colon and Rectal Cancers in California – Daixin Yin • Plus Two Posters: • Linking Cancer Registry and Death Certificate Data: Concordance between Cancer-Specific Cause of Death and Primary Cancer Site at Diagnosis – Aliza Fink • Accuracy of Cancer Mortality: Investigation of Patterns of Cancer Site Misclassification – Ronaldo Iachan
Agencies and Persons Involved CDC’s Division of Cancer Prevention and Control Robert R. German, DrPH, MPH Sherri L. Stewart, PhD Stephanie Foster, MPH, MA (now with ATSDR) ICF Macro Aliza Fink, D.Sc. Ronaldo Iachan, PhD CDC’s National Center for Health Statistics Melonie Heron, PhD Robert N. Anderson, PhD The findings and conclusions in this presentation are those of the authors and do not necessarily represent the views of CDC.
Agencies and Persons Involved California Cancer Registry Janet Bates, MD, MPH Mark Allen, MS Daixin Yin, MS Gretchen Agha Colorado Central Cancer Registry Randi Rycroft, BA, BS, CTR Jack Finch, MS Cancer Data Registry of Idaho Chris Johnson, MPH Stacey Carson, RHIT, CTR Pam Harder, BS
Background • The death certificate is the source for state and national cancer mortality statistics. • Used to set public health goals and to measure health status. • Cause of death is collected by physician or medical examiner/coroner. • Incidence data come from central cancer registries. • (No kidding!) • The concordance of primary site in these two data sources has not been evaluated in the past 20 years, during which ICD-10 was introduced.
Methods • Registry records from California, Colorado, and Idaho Central Cancer Registries were linked with state vital statistics data. • Two study designs: • Prospective • Retrospective • In both study designs the primary site of cancer listed in the cancer registry is compared with the primary site of cancer listed on the death certificate.
Prospective Study Design • Compared diagnoses reported in the cancer registry from 1993-1995 to death certificate data from 1993-2004 • Concordance measured via detection rate (DR)
Retrospective Study Design • Compared death certificate data from 2002-2004 to cancer registry diagnoses from 1993-2004. • Concordance measured via confirmation rate (CR)
Results • Overall, there was good agreement between the cancer site listed in the cancer registry and the cancer site listed as the underlying cause of death. • The overall confirmation rate (retrospective study design) was 82.8%. • The overall detection rate (prospective study design) was 84.7%.
VARIABILITY IN DEATH CERTIFICATE ACCURACY BY CHARACTERISTICS OF DEATH CERITIFIERS Chris Johnson, MPH, Cancer Data Registry of Idaho Christine Hahn, MD, Idaho Division of Health
Background • Errors in death certificates are fairly common. • In the United States, deaths may be certified by physicians, other medical professionals, such as physicians’ assistants, coroners, or medical examiners. • Coroners are usually elected county-level officers who often lack medical qualification. • Less than one third of states with coroners require training. • Approximately 31% of counties in the United States are served by medical examiners at present.
Methods • Idaho data used in the ACM were linked to a dataset containing non-identifying death certifier characteristics provided by the Idaho Division of Health. • Type of certifier (coroner or physician) was available from the death certificate. • For physician certifiers, Dr. Hahn determined their specialty area(s) of practice using rosters from the Idaho State Board of Medicine and the Idaho Medical Association.
Methods • This study was restricted to individuals with a single primary. • The study was restricted to malignant tumors except: 1) malignant non-melanoma skin cancers were excluded and 2) in situ urinary bladder tumors were included. • Cancer cases that were found via autopsy or death certificate only were excluded from the study.
Methods • Certifier type and physician specialty were combined and grouped into five categories: • coroner and physician, • coroner who is not a physician (“coroner”), • physician with a cancer-related specialty • general practice physician, • physician with any other specialty. • Death certification experience (3 sets: 1984-2004, prospective, retrospective) • Years of experience certifying deaths in Idaho • Total number of deaths certified • Rate of death certification (deaths/year)
Results – Prospective Study Arm • Overall primary site concordance for 3,550 Idaho resident cancer cases 1993-1995 linked to Idaho resident cancer deaths 1993-2004 was 92.9%. • Detection rate varied significantly by: • primary site, age at death, and race, • certifier type/physician specialty • but not by sex, ethnicity, rurality/urbanicity, SEER summary stage, year of death or ICD version.
Results – Retrospective Study Arm • Overall primary site concordance for 3,848 Idaho resident cancer deaths 2002-2004 linked to Idaho resident cancer cases 1993-2004 was 93.7%. • Confirmation rates varied significantly by: • primary site and SEER summary stage, • certifier type/physician specialty • not by age at death, sex, race, ethnicity, rurality/urbanicity, year of death.
Odds of Concordance by Study Arm and Certifier Type/Physician Specialty
Conclusions • Based on USCS site categories, concordance between cancer registry and death certificate primary site is quite high (~93%). • This study showed significant differences by certifier type/physician specialty in the accuracy of cancer mortality measured by death certificates. • Coroners who are not also physicians have lower accuracy rates than physicians.