Multi-State Results of Linking Death Certificates to Indian Health Service Patient Records 2007 NAPHSIS Annual Meeting June 6, 2007 Salt Lake City

Background Previous reports have shown that American Indian race is often not accurately reported on death certificates. For example, a 1999 study by Rosenberg, et al. 1 indicated that death rates of American Indians are understated in official national publications by approximately 21 percent.

Background The denominators of death rates, usually based on Census population data, tend to undercount minorities, which will inflate the death rates. Death certificates, used in the numerators, tend to undercount minorities, which will deflate the death rates. In the Rosenberg study, the net effect of these two opposing influences was that death rates for American Indians were understated by 21 percent.

Background Rosenberg, et al. also used data from the National Longitudinal Mortality Study (NLMS) to compare race reported on death certificates with matched data from the Current Population Survey (CPS). The race information in the CPS is reported for households in which the decedent was alive at the time of interview.

Background In contrast, race information on the death certificate is that reported by the funeral director based on responses from an informant, often a surviving family member, or based on observation by the funeral director. They found that 37 percent more persons were identified as American Indians in the CPS than on the corresponding death certificates.

Background There has been a large increase in self- reporting of American Indians in recent U.S. Censuses. About 33 percent of the growth in the American Indian population in the United States between 1960 and 1990 has been attributed to increased self-reporting as American Indian. We do not know the effect of this trend on misclassification of American Indian race on death certificates.

Background The total American Indian/Alaska Native (AIAN) population of the United States according to the 2000 Census was 2,476,000, counting those who reported American Indian/Alaska Native alone as their race. Counting also those people who reported American Indian/Alaska Native in combination with one or more other races, the AIAN population for the United States was 4,119,000.

Background To address the misclassification of American Indian race on death certificates, the Indian Health Service (IHS) National Epidemiology Program has undertaken a comprehensive linkage of state-provided death certificates to IHS patient records. Goals of this project were to check the quality of the race data on the death certificates and provide information for targeting state-level training for collection of better race data.

Description of IHS Linkage Project Fourteen states submitted their death certificate records to the IHS, generally for the years , and these records were linked to the IHS patient data base. These states are: Arizona, Arkansas, California, Idaho, Michigan, Minnesota, Montana, New Mexico, North Carolina, North Dakota, Oklahoma, Oregon, Utah, and Washington.

Description of IHS Linkage Project Linkages between death certificates and IHS patient records were conducted with probabilistic methods (LinkPlus software) using name, date of birth, gender, social security number, and other identifiers.

Description of IHS Linkage Project Not all American Indian tribes in the United States are served by the Indian Health Service, so not all American Indian death certificates would be expected to match to the IHS patient records. The IHS provides health care services to approximately 66% of the 2.5 million American Indians in the United States (40% if the larger AIAN census population figure of 4.1 million is used). This percentage varies substantially by state.

Description of IHS Linkage Project A total of 8,855,141 death certificate records were submitted by the 14 states for this project; these states include 27% of all deaths in the United States. These 14 states contain 62% of the total American Indian population of the United States (who reported AIAN race alone).

Description of IHS Linkage Project Summary linkage results and a linked data file were given back to the participating states by the IHS. Each of the 14 states gave permission for the IHS to provide the summary linkage results to Paul Buescher (in North Carolina) for the purpose of this presentation.

Overall Results The death certificate race codes indicated 103,185 American Indian deaths in the 14 states over the 14-year period. Matching to the IHS data base identified an additional 16,986 American Indian deaths, for a total of 120,171. The ratio of this “enhanced” number to the number identified from death certificates only is Therefore, 16% more American Indian deaths were identified in these 14 states through the linkage process.

Overall Results One problem with this approach is that not all American Indian tribes in the United States receive health care services from the Indian Health Service; therefore not all American Indian deaths would be expected to match to the IHS data base. For example, in North Carolina only one tribe out of eight receives IHS services (the Eastern Band of Cherokee), representing about 10% of the total American Indian population in North Carolina.

Cherokee in North Carolina

Lumbee in North Carolina

Overall Results A better approach to calculating a “misclassification rate” is to take the total number of death records that matched to the IHS data base and see what percentage of those were not indicated as American Indian by the race code on death certificates. For the 14 states, 87,927 death records matched to the IHS data base. Of these, 16,986 did not have American Indian race recorded on the death certificate, for a misclassification (underreporting) rate of 19%.

State-Specific Results: Ratio of Total A.I. Deaths after Linkage to A.I. Deaths As Indicated by Death Certificates Arizona1.04New Mexico1.02 Arkansas1.69N. Carolina1.02 California1.21N. Dakota1.02 Idaho1.10Oklahoma1.51 Michigan1.06Oregon1.15 Minnesota1.09Utah1.05 Montana1.08Washington1.15

State-Specific Results: Total A.I. Deaths after Linkage as a Percentage of All Deaths Occurring in the State Arizona3.7%New Mexico7.5% Arkansas0.5%N. Carolina0.8% California0.4%N. Dakota4.0% Idaho1.1%Oklahoma6.2% Michigan0.5%Oregon0.9% Minnesota1.0%Utah1.1% Montana4.9%Washington1.4%

State-Specific Results: Percentage of Linked Deaths Not Indicated as American Indian on the Death Certificates Arizona5%New Mexico2% Arkansas55%N. Carolina17% California35%N. Dakota2% Idaho12%Oklahoma38% Michigan18%Oregon23% Minnesota12%Utah6% Montana8%Washington18%

Results Data from both North Carolina and Oklahoma show that 97% of the American Indian deaths that were misclassified were counted as “white.” Data from Oklahoma indicate that the highest rate of misclassification was among decedents ages 5-24 years.

Results In North Carolina, for the matching American Indian deaths where the decedent was a resident of the two western counties containing the Cherokee Indian Reservation (Qualla Boundary) (710 records), the misclassification rate was 5%. In contrast, for the American Indian deaths where the decedent was not a resident of the two reservation counties (322 records), the misclassification rate was 43%.

Results One of the items of information captured in the IHS patient data base is “blood quantum.” The categories are: full Indian, ½ to full, ¼ to ½, and Indian but less than ¼. Data from North Carolina show that 21% of the death certificates that matched to an IHS patient record had a blood quantum of full, 33% were ½ to full, 14% were ¼ to ½, 31% were Indian but less than ¼, and 2% had unknown or unspecified blood quantum.

Percentage of Misclassification of American Indian Race on North Carolina Death Certificates by Blood Quantum 5% 9% 41%

Conclusions Of the 87,927 death records that matched to the Indian Health Service patient data base, 16,986 were not identified as American Indian on the death certificates, for an overall underreporting rate of 19%. The rate of underreporting among the 14 states ranged from 2% to 55%.

Conclusions In general, states with the lowest percentages of total deaths that were American Indian had the highest rates of underreporting of American Indian race on death certificates. The correlation coefficient between these two values across the 14 states was -.32, though the difference from 0.0 was not statistically significant (p =.27).

Conclusions New Mexico tied with North Dakota for the lowest underreporting rate of the 14 states. In New Mexico: the Vital Records agency has a good working relationship with the tribes; tribal officials often act as funeral directors and therefore are involved in recording race on the death certificates; ten percent of the population is American Indian and therefore many funeral directors do not make assumptions about American Indian race.

Conclusions Oklahoma has used the information on American Indians from the IHS linkage to enhance the race coding on their death certificate data files. Oklahoma has both the original race code from the death certificate and the code from the IHS linkage on their analysis files and on their public data query web site.

Conclusions North Carolina “closes” its death files each year and therefore has not gone back to revise the race data. The benefits of using the IHS linkage process to enhance death certificate race codes would be maximized if this linkage project was continued on a regular basis.

Conclusions This linkage project was carried out for death years , when a single race was captured on the death certificates. Many states are or will be modifying their death certificates to record multiple races for the decedent. It will be interesting to see how American Indian race misclassification rates (as determined by the IHS linkage process) change as multiple races are recorded on the death certificates.

Acknowledgments Staff of the 14 states who released their data for this presentation Melissa Jim and Marguerite Adams-Cameron of the IHS National Epidemiology Program for doing the linkages CDC Division of Cancer Prevention and Control for funding the linkages Matthew Avery and Sidney Evans of the State Center for Health Statistics in North Carolina for assistance in compiling data

Reference (1) Rosenberg HM, Maurer JD, Sorlie PD, Johnson NJ, et al. Quality of death rates by race and Hispanic origin: A summary of current research, National Center for Health Statistics. Vital and Health Statistics 2(128)

Contact Information Paul A. Buescher, Ph.D. Director State Center for Health Statistics 1908 Mail Service Center Raleigh, NC (919)