Technical Appendix for 1991 QUALITY OF DATA Completeness of registration All States have adopted laws that require the registration of births and deaths and the reporting of fetal deaths. It is believed that more than 99 percent of the births and deaths occurring in this country are registered. Reporting requirements for fetal deaths vary somewhat from State to State (see "Comparability and completeness of data"). Overall reporting is not as complete for fetal deaths as for births and deaths, but it is believed to be relatively complete for fetal deaths at a gestation of 28 weeks or more. National statistical data on fetal deaths include only fetal deaths occurring at a stated or presumed gestation of 20 weeks or more. Massachusetts data The 1964 statistics for deaths exclude approximately 6,000 events registered in Massachusetts, primarily to residents of that State. Microfilm copies of these records were not received by NCHS. Figures for the United States and the New England Division are affected also. Alabama data The 1988 statistics for deaths show no deaths assigned to the City of Prattville in Autauga County. The death records that should have been assigned to this area were instead assigned to the Balance of County due to a processing error. Amended records for Alabama, Alaska, Hawaii, and New Jersey Numbers of deaths occurring in Alabama, Alaska, Hawaii, and New Jersey for 1991 are in error for all causes of death combined and for selected causes because NCHS did not receive changes resulting from amended records. An estimate of the effect of these omissions can be derived by comparing NCHS counts of records processed through the VSCP with counts prepared by the respective States as shown in table E. Differences are concentrated among selected causes of death, principally Symptoms, signs, and ill-defined conditions (ICD-9 Nos. 780-799) and external causes. - 1 - Table E. Numbers of deaths and ratios of deaths for selected causes as tabulated by State of occurrence and NCHS, 1991 [Data by place of occurrence include deaths of nonresidents. Numbers after causes of death are category numbers of the Ninth Revision, International Classification of Diseases, 1975] _____________________________________________________________________________ Ratio Causes Alabama NCHS AL/NCHS _____________________________________________________________________________ All causes........................... 39,753 39,753 1.00 Symptoms, signs, and ill-defined conditions...............780-799 879 1,034 0.85 Accidents and adverse effects ..........................E800-E949 2,210 2,215 1.00 Motor vehicle accidents ..........................E810-E825 1,243 1,216 1.02 All other accidents and adverse effects.........E800-E807,E826-E949 967 999 0.97 Suicide....................E950-E959 571 534 1.07 Homicide and legal intervention ..........................E960-E978 603 580 1.04 All other external causes..E980-E999 41 44 0.93 _____________________________________________________________________________ Ratio Causes Alaska NCHS AK/NCHS _____________________________________________________________________________ All causes........................... 2,212 2,201 1.00 Symptoms, signs, and ill-defined conditions...............780-799 45 62 0.73 Accidents and adverse effects ..........................E800-E949 380 387 0.98 Motor vehicle accidents ..........................E810-E825 128 105 1.22 All other accidents and adverse effects.........E800-E807,E826-E949 252 282 0.89 Suicide....................E950-E959 108 74 1.46 Homicide and legal intervention ..........................E960-E978 53 45 1.18 All other external causes..E980-E999 5 5 1.00 - 2 - Table E. Numbers of deaths and ratios of deaths for selected causes as tabulated by State of occurrence and NCHS, 1991 [Data by place of occurrence include deaths of nonresidents. Numbers after causes of death are category numbers of the Ninth Revision, International Classification of Diseases, 1975] _____________________________________________________________________________ Ratio Causes Hawaii NCHS HI/NCHS _____________________________________________________________________________ All causes........................... 7,042 7,035 1.00 Symptoms, signs, and ill-defined conditions...............780-799 45 158 0.28 Accidents and adverse effects ..........................E800-E949 296 298 0.99 Motor vehicle accidents ..........................E810-E825 148 147 1.01 All other accidents and adverse effects.........E800-E807,E826-E949 148 151 0.98 Suicide....................E950-E959 129 110 1.17 Homicide and legal intervention ..........................E960-E978 50 48 1.04 All other external causes..E980-E999 35 15 2.33 _____________________________________________________________________________ Ratio Causes New Jersey NCHS NJ/NCHS _____________________________________________________________________________ All causes........................... 68,615 68,795 1.00 Symptoms, signs, and ill-defined conditions...............780-799 276 958 0.29 Accidents and adverse effects ..........................E800-E949 1,957 1,921 1.02 Motor vehicle accidents ..........................E810-E825 830 808 1.03 All other accidents and adverse effects.........E800-E807,E826-E949 1,114 1,113 1.00 Suicide....................E950-E959 572 498 1.15 Homicide and legal intervention ..........................E960-E978 431 404 1.07 All other external causes..E980-E999 144 31 4.65 _____________________________________________________________________________ Quality control procedures Demographic items on the death certificate--As previously indicated, for 1991 the mortality data for these items were obtained from two sources--photocopies of the original certificates furnished by the Virgin Islands and Guam and records on data tape furnished by the 50 - 3 - States, the District of Columbia, New York City, and Puerto Rico. For the Virgin Islands and Guam, which sent only copies of the original certificates, the demographic items were coded for 100 percent of the death certificates. The demographic coding for 100 percent of the certificates was independently verified. As part of the quality control procedures for mortality data, each registration area goes through a calibration period, during which it must achieve the specified error tolerance level of 2 percent per item for 3 consecutive months, based on independent verification by NCHS of a 50-percent sample of that area's records. When the area has achieved the required error tolerance level, a sample of 70-80 records per month is used to monitor quality of coding. All areas providing data on computer tapes before 1991 have achieved the specified error tolerance; accordingly, the demographic items on about 70-80 records per area per month were independently verified by NCHS. The estimated average error rate for all demographic items in 1991 was 0.25 percent. These verification procedures involve controlling for two types of error (coding and entering into the data record tape) at the same time, and the error rates are a combined measure of both types. It may be assumed that the entering errors are randomly distributed across all items on the record, but this assumption cannot be made as readily for coding errors. Although systematic errors in coding infrequent events may escape detection during sample verification, it is probable that some of these errors were detected during the initial period when 50 percent of the file was being verified, thus providing an opportunity to retrain the coders. Medical items on the death certificate--As is true for demographic data, mortality medical data are also subject to quality control procedures to control for errors of both coding and data entry. Each of the 31 registration areas that furnished NCHS with coded medical information in 1991 according to NCHS specifications had to qualify for sample verification first. During an initial calibration period, the area had to demonstrate that its staff could achieve a specified error tolerance level of less than 5 percent for coding all medical items. After the area had achieved the required error tolerance level, a sample of 70-80 records per month was used to monitor quality of medical coding. For the 31 reporting States, the average coding error rate in 1991 was estimated at just over 4 percent. For the remaining 19 States, the District of Columbia, New York City, Puerto Rico, the Virgin Islands, and Guam, NCHS coded the medical items for 100 percent of the death records. A 1-percent sample of the records was coded independently for quality control purposes. The estimated average error rate for these areas was about 3 percent. The ACME system for selecting the underlying cause of death through computer application contributes to the quality control of medical items on the death certificate. (See "Automated selection of underlying cause of death.") The MICAR system automates the coding of multiple causes of death. The quality of the data produced by MICAR is better than the quality of the data produced using manual multiple cause-of-death coding. The version of MICAR used to process 1990 and 1991 records processed about 85 percent of the mortality records with an average error rate of 0.42 percent on an underlying-cause basis and a rate of 0.74 percent on a multiple-cause basis. - 4 - Demographic items on the report of fetal death--For 1991, all data on fetal deaths, except for New York State (excluding New York City), were coded under contract by the U.S. Bureau of the Census. Coding and entering of information on data tapes were verified on a 100-percent basis because of the relatively small number of records involved. Other control procedures--After coding and entering on data tape are completed, record counts are balanced against control totals for each shipment of records from a registration area. Editing procedures ensure that records with inconsistent or impossible codes are modified. Inconsistent codes are those, for example, indicating a contradiction between cause of death and age or sex of the decedent. Records so identified during the computer editing process are either corrected by reference to the source record or adjusted by arbitrary code assignment (37). Further, conditions specified on a list of infrequent or rare causes of death are confirmed by the certifier or a State Health Officer. All subsequent operations in tabulating and in preparing tables are verified during the computer processing or by statistical clerks. Estimates of errors arising from 50-percent sample for 1972 Death statistics for 1972 in this report (excluding fetal-death statistics) are based on a 50-percent sample of all deaths occurring in the 50 States and the District of Columbia. A description of the sample design and a table of the percent errors of the estimated numbers of deaths by size of estimate and total deaths in the area are shown in the Technical Appendix from Vital Statistics of the United States, 1972. - 5 -