Appendix II–Definitions and Methods ŸFollowing is an alphabetical listing of the definitions ofkey terms and methods used in this chartbook.Accident—The term accident is used in thischartbook to mean an unintentional injury death oran unintentional nonfatal event. The word is usedsparingly because of its history of being considered bysome to be inappropriate to describe events that arepreventable. See the introduction for the chartbook.Activity—An activity is used to describe what theinjured person was doing when the injury occurred.The data source is the National Health InterviewSurvey. Categories include driving, working at paidjob, working around house or yard, attending school,sports, leisure activities (excluding sports), and other.More than one activity can be checked for the sameepisode (Figure 24).Age—A person’s age is reported as age at lastbirthday, that is, age in completed years. Presentingthe data by single year of age rather thanpredetermined age groups provides a level of detailthat traditional 5- or 10-year age groupings of fataland nonfatal injuries can obscure. For example, thecommonly used age group, 15–19 years, is a poorchoice for motor vehicle traffic death rates and othercauses of injury; the rate at 19 years of age is threetimes the rate at 15 years of age. Combining the agesmakes this higher rate less obvious.Age-adjusted rates (R’) are calculated by the directmethod by applying unrounded age-specific rates (Ri)to the U.S. standard population (wi).The application of unrounded age-specific rates tothe standard population differs from the currentmethod used to calculate death rates in other reportspublished by the National Center for Health Statistics.In very few instances do the final age-adjusted ratesdiffer, but when they do, they differ by no more than0.1 per 100,000. For example, in comparing homiciderates from 1984 through 2004 shown in Figure 4,only data year 2003 differs from what is published inNational Vital Statistics Reports (6.1 compared with6.0 per 100,000 population).MortalityBeginning with 2003 data, the traditional standardmillion population, along with correspondingstandard weights to six decimal places based onthe projected year 2000 population, were replacedby the unrounded projected year 2000 populationage distribution (see Table VII). The effect of thechange is negligible and does not significantly affectcomparability with age-adjusted rates calculated usingthe previous method (Figures 4, 6, and 11).Age adjustment—Age adjustment is used to comparerisks of two or more populations at one point in timeor one population at two or more points in time. Ageadjusted rates are computed by the direct methodof applying age-specific rates in a population ofinterest to a standardized age distribution to eliminatedifferences in observed rates that result from agedifferences in population composition. Age-adjustedrates should be viewed as relative indexes rather thanactual measures of risk. Age-adjusted rates for twodifferent outcome measures at the same point in timeshould not be compared (Figures 4, 6, and 11).Injury in the United States: 2007 Chartbook163

Appendix II–Definitions and Methods ŸTable VII. United States standard population and agegroups used to age adjust mortality dataAge groupTotalUnder 1 yearPopulation274,633,6423,794,9011–4 years15,191,6195–14 years39,976,61915–24 years38,076,74325–34 years37,233,43735–44 years44,659,18545–54 years37,030,15255–64 years23,961,50665–74 years18,135,51475–84 years12,314,79385 years and over4,259,173SOURCE: National Institutes of Health, National Cancer Institute.Surveillance, Epidemiology, and End Results (SEER), Standardpopulation in single year of age. Available from: eages.html.National Health Interview Survey—Estimates basedon the National Health Interview Survey are ageadjusted to the same 2000 U.S. standard population.Adjustment is based on 4 age groups as shown belowwith their corresponding standard population (seedata tables for Figures 22 and 23). (See Table VIII.)Table VIII. United States standard population and agegroups used to age adjust National Health InterviewSurvey dataAge groupTotalPopulation274,633,642Under 15 years58,963,13915–24 years38,076,74325–64 years142,884,28065 years and over34,709,480SOURCE: National Institutes of Health, National Cancer Institute.Surveillance, Epidemiology, and End Results (SEER), Standardpopulation in single year of age. Available from: eages.html.Alcohol-impaired driving—Alcohol-impaired drivingis defined as operating a motor vehicle when legally164intoxicated. Legally intoxicated drivers have ameasurable or estimated blood alcohol concentration(BAC) of 0.08 grams per deciliter (g/dl) or above.Trends in alcohol impaired driving are tracked usingthe proportion of fatally injured drivers who werelegally intoxicated among all fatally injured drivers.The Insurance Institute for Highway Safety calculatesthe proportion using the Fatality Analysis ReportingSystem (FARS), which captures the number of fatallyinjured drivers on public roads from all 50 states andthe District of Columbia. Multiple imputation is usedfor estimating the BACs for those with missing valuesusing the U.S. Department of Transportation’s multipleimputation model.References:Insurance Institute for Highway Safety. IIHS fatality facts 2005,alcohol. Available from: facts/alcohol.html). 2007.Subramanian, R. Transitioning to multiple imputation—A newmethod to impute missing blood alcohol concentration (BAC) valuesin FARS. Technical report no. DOT HS-809-403. Washington, DC:U.S. Department of Transportation. 2002.Average percent change and test of trends—Joinpoint software, developed by the National CancerInstitute, was used to estimate the annual percentchange in death rates and in hospital discharge rates.The software uses trend data and fits the simplestjoinpoint model that the data allow. The user suppliesthe minimum and maximum number of joinpoints.The program starts with the minimum number ofjoinpoints (i.e., 0 joinpoints, which is a straight line)and tests whether more joinpoints are statisticallysignificant and must be added to the model (up to thespecified maximum number). This enables the user totest that an apparent change in trend is statisticallysignificant. The tests of significance use a Monte CarloPermutation method. The models used in these figuresincorporate estimates of both the data points as wellas the standard error of each point. Estimates wereincluded and calculated using only one decimal placeunless the standard error was greater than 0.0 but lessthan 0.5; in those cases, two decimal places were usedInjury in the United States: 2007 Chartbook

Appendix II–Definitions and Methods Ÿfor the standard errors. In addition, the models arelinear on the log of the response (i.e., for calculatingannual percentage rate change). Models used for thefigures were the ones determined by Joinpoint to bestfit the data (Figures 4, 6, 13, 15-1, and 15-2).For details see:, however, a blood sample is not necessary todetermine a person’s BAC. It can be measured moresimply by analyzing exhaled breath. All 50 states andthe District of Columbia (D.C.) have laws defining it asa crime to drive with a BAC at or above the proscribedlevel of 0.08%. The data in Figure 13 are based ondata from all 50 states and D.C. with imputations formissing BACs provided by the U.S. Department ofTransportation’s multiple imputation model.Reference:Kim HJ, Fay MP, Feuer EJ, Midthune DN. Permutation tests for joinpoint regression with applications to cancer rates [published correc tion appeared in Stat Med 20(4):655. 2001]. Stat Med 19(3):335–51.2000.Barell Injury Diagnosis Matrix (Barell Matrix)—Thematrix is a two-dimensional array of InternationalClassification of Diseases, Ninth Revision, ClinicalModification (ICD–9–CM) diagnosis codes for injury(updated as of 2002) grouped by body region of theinjury and the nature of the injury. This matrix providesa standard format for reporting injury data. This injurydiagnosis matrix is a product of the participants inthe International Collaborative Effort (ICE) on InjuryStatistics. For more information about the BarellMatrix, refer to atrix.htm. The matrix was adoptedfor use by the State and Territorial Injury PreventionDirectors Association and recommended as the basisfor defining injury hospitalizations. The matrix isincluded in Table I (see Figures 17-1, 17-2, 18, 21, and26).References:Barell V, Aharonson-Daniel L, Fingerhut LA, MacKenzie EJ, et al. Anintroduction to the Barell body region by nature of injury diagnosismatrix. Inj Prev 8:91–6. 2002.Injury Surveillance Workgroup. Consensus recommendations forusing hospital discharge data for injury surveillance. Marietta, GA:State and Territorial Injury Prevention Directors Association. 2003.Blood alcohol concentration (BAC)—A BACdescribes the amount of alcohol in a person’s blood,expressed as weight of alcohol per unit of volumeof blood. For example, 0.08% BAC indicates 80mg of alcohol per 100 ml of blood. For most legalInjury in the United States: 2007 ChartbookReference:Subramanian R. Transitioning to multiple imputation—A newmethod to impute missing blood alcohol concentration (BAC) valuesin FARS. Technical report no. DOT HS-809-403. Washington, DC:U.S. Department of Transportation. 2002.Body region—Body region refers to one of the twodimensions of the Barell Injury Diagnosis Matrix andthe Injury Mortality Diagnosis matrix. This dimensionclassifies the part of the body that was injured and isbased on ICD–9–CM codes in the Barell Matrix andICD–10 codes in the Injury Mortality Diagnosis Matrix.For a detailed listing of the body regions see Table I.Cause of death—For the purpose of nationalmortality statistics, every death is attributed to oneunderlying condition, based on information reportedon the death certificate and using the internationalrules for selecting the underlying cause of deathfrom the conditions stated on the death certificate.For injury deaths, the underlying cause is definedby the World Health Organization (WHO) as thecircumstances of the accident or violence thatproduced the fatal injury. Generally more medicalinformation is reported on death certificates than isdirectly reflected in the underlying cause of death. Theconditions that are not selected as underlying cause ofdeath constitute the nonunderlying cause(s) of death,also known as multiple causes of death.Cause of death is coded according to the appropriaterevision of the ICD. Effective with deaths occurringin 1999, the United States began using the TenthRevision of the ICD (ICD–10) to code cause of death.During 1979–1998, causes of death were coded and165

Appendix II–Definitions and Methods Ÿclassified according to the Ninth Revision (ICD–9).Each of these ICD revisions has produceddiscontinuities in cause-of-death trends. Thesediscontinuities are measured using comparabilityratios. These ratios are essential to the interpretationof mortality trends.For more information, see Comparability ratio.See also Multiple-cause-of death data and injurydiagnoses.Reference:Miniño AM, Anderson RN, Fingerhut LA, Boudreault MA, Warner M.Deaths: Injuries, 2002. National vital statistics reports; vol 54 no 10.Hyattsville, MD: National Center for Health Statistics. 2006.Cause-of-death ranking—Selected causes of deathof public health and medical importance are rankedaccording to the number of deaths assigned tothese causes. The top-ranking causes are the leadingcauses of death. For deaths from injuries in 2004, 3causes ranked in the top 15 rankable causes basedon ICD–10. They are accidents (unintentional injuries),intentional self-harm (suicide), and assault (homicide).Causes that are tied receive the same rank; the nextcause is assigned the rank it would have received hadthe lower-ranked causes not been tied, that is, it skipsa rank. See ICD.References:Heron MP, Smith BL. Deaths: Leading causes for 2003. National vitalstatistics reports; vol 55 no 10. Hyattsville, MD: National Center forHealth Statistics. 2007.NCHS. ICD–10 cause-of-death lists for tabulating mortality statistics(updated October 2002). NCHS instruction manual; part 9. Hyatts ville, MD: National Center for Health Statistics; 2002.Cause-of-death ranking for leading mechanisms ofinjury death—Leading mechanisms of injury death areranked according to the number of deaths assignedto rankable mechanisms in the external cause of injurymortality matrix; rankable mechanisms are indicatedby the number symbol (#) using a procedure consis tent with that used to rank leading causes of death.166Vaguely defined categories are summarily excludedfrom selection as rankable mechanisms. These includeall categories beginning with the words ‘‘other’’ or“unspecified.’’ Among the remaining mechanismcategories, decisions were made to select as rankablethe mechanism of injury considered most useful from apublic health perspective with the following condition:the rankable mechanisms must be mutually exclusive.If a category representing a subtotal (such as mo tor vehicle traffic or fire or hot object or substance)is selected as a rankable mechanism, its componentparts are not selected as rankable. The external causeof injury mortality matrices for ICD–9 and ICD–10 areshown in Tables II and III.Civilian population, civilian noninstitutionalizedpopulation—See Population.Comparability ratio—About every 10 to 20 years, theICD is revised to stay abreast of advances in medicalscience and changes in medical terminology. Eachof these revisions produces breaks in the continuityof cause-of-death statistics. Discontinuities acrossrevisions are due to changes in classification and rulesfor selecting underlying cause of death. Classificationand rule changes affect cause-of-death trend databy shifting deaths away from some cause-of-deathcategories and into others.Comparability ratios are based on a comparabilitystudy in which the same deaths were coded by boththe Ninth and Tenth Revisions. The comparabilityratio was calculated by dividing the number ofdeaths classified by ICD–10 by the number of deathsclassified by ICD–9. The resulting ratios represent thenet effect of the Tenth Revision on cause-of-deathstatistics and can be used to adjust mortality statisticsfor causes of death classified by the Ninth Revision in1998 to be comparable with cause-specific mortalitystatistics classified by the Tenth Revision in 1999.The application of comparability ratios to mortalitystatistics helps make the analysis of change between1998 and 1999 more accurate and complete. The 1998Injury in the United States: 2007 Chartbook

Appendix II–Definitions and Methods Ÿcomparability-modified death rate is calculated bymultiplying the comparability ratio by the 1998 deathrate. Comparability-modified rates should be used toestimate mortality change between 1998 and 1999.Comparability ratios measure the effect of changesin classification and coding rules. For all externalcauses of injury based on the external cause of injurymortality matrix, comparability ratios are shown inTable VI.For selected causes of death, the ICD–9 codes usedto calculate death rates for 1985 through 1998 differfrom the ICD–9 codes most nearly comparable withthe corresponding ICD–10 cause-of-death category,which also affects the ability to compare deathrates across ICD revisions. In this chartbook, ratesfor unintentional injuries and homicide in Figure 4and motor vehicle traffic fatalities and drowningin Figure 6 for 1985–1998 were recalculated usingICD–9 codes that are more comparable with codes forcorresponding ICD–10 categories. The codes used canbe found in Table VI.Final and preliminary comparability ratios for 113selected causes of death are available from: Statistics/NCHS/Datasets/Comparability/icd9 icd10/.the ICD–9–CM. For a given medical care encounter,the first-listed diagnosis can be used to categorizethe hospital discharge, or if more than one diagnosisis recorded on the survey abstraction form, thedischarge can be categorized based on all diagnosesrecorded. The first-listed diagnosis is often, but notalways, considered the most important or dominantcondition among all comorbid conditions. Forexample, a hospital discharge would be considered afirst-listed injury discharge if an ICD–9–CM diagnosiscode for injury was recorded in the first diagnosisfield on the NHDS abstract form. Other dischargesmay have an injury diagnosis in one or more of theremaining second through seventh diagnosis fields onthis abstract form.See related External cause of injury, Initial injuryemergency department visit, and Multiple cause-of death data and injury diagnoses.Education—In survey data, educational categories arebased on information about educational credentials,such as diplomas and degrees. In vital statistics,educational attainment is based on years of schoolcompleted. This chartbook does not include deathrates by educational attainment because of a changein the way the data are collected in states that haveimplemented the 2003 revised death certificate. See certs rev.htm.Reference:Anderson RN, Miniño AM, Hoyert DL, Rosenberg HM.Comparability of cause of death between ICD–9 and ICD–10:Preliminary estimates. National vital statistics reports; vol 49 no2. Hyattsville, MD: National Center for Health Statistics. Availablefrom: 02.pdf.2001.Death rate: See Rate: Death and related rates.Diagnosis—Diagnosis is the act or process ofidentifying or determining the nature and cause of adisease or injury through evaluation of patient history,examination, and review of laboratory data. Diagnosesin the National Hospital Discharge Survey, and theNational Hospital Ambulatory Medical Care Surveyare abstracted from medical records and coded toInjury in the United States: 2007 ChartbookNational Health Interview Survey (NHIS)—Beginningin 1997, the NHIS questionnaire was changedto ask “What is the highest level of schoolhas completed or the highest degree received?”Responses were used to categorize adults accordingto educational credentials (i.e., no high school diplomaor general educational development [GED] highschool equivalency diploma; high school diplomaor GED; some college, no bachelor’s degree; orbachelor’s degree or higher).Emergency department—According to the NationalHospital Ambulatory Medical Care Survey, anemergency department is a hospital facility that is167

Appendix II–Definitions and Methods Ÿstaffed 24 hours a day and provides unscheduledoutpatient services to patients whose conditionsrequire immediate care. Offsite emergencydepartments that are open less than 24 hours areincluded if staffed by the hospital’s emergencydepartment.Emergency department visit—In the NationalHospital Ambulatory Medical Care Survey, anemergency department visit is a direct personalexchange between a patient and a physician or otherhealth care providers working under the physician’ssupervision, for the purpose of seeking care andreceiving personal health services. See related Initialinjury emergency department visit.Episode of injury—See Injury or poisoning episode.Ethnicity—See Race and Hispanic origin.External cause of injury—The external cause ofinjury is used for classifying the circumstances in whichinjuries occur. The external cause is comprised of twoaxes, the mechanism or cause (e.g. firearm or motorvehicle) and the manner or intent (e.g. homicide orsuicide).External cause of injury matrix—The matrix

Age—A person's age is reported as age at last birthday, that is, age in completed years. Presenting the data by single year of age rather than predetermined age groups provides a level of detail that traditional 5- or 10-year age groupings of fatal and nonfatal injuries can obscure. For example, the commonly used age group, 15-19 years, is a poor