Cancer Mortality Data Trends for 2000-2010 DS 12-12000
Number of Events – The number of events provides a description of how a disease affects a population, but it is not useful for examining trends or comparison across groups because the number of events largely depends on population size.8
Crude Rates, Age-Specific Rates, and Age-Adjusted Rates – The crude death rate (number of deaths per population size) is a widely used mortality measure.8 This rate represents the average chance of dying during a specified period for persons in the entire population. However, crude death rates are influenced by the age distribution of the population. As such, crude death rate comparisons over time or between groups may be misleading if the populations being compared differ in age composition.
The age specific death rate is defined as the number of deaths occurring in a specified age group divided by the population for the specified age group, usually expressed per 100,000 population. Age-specific death rates allow one to compare mortality risks of a particular age group over time or between age groups at a particular point in time. Although effective in eliminating the effect of differences in age composition, age-specific comparisons can be cumbersome, because they require a relatively large number of comparisons, one for each age group.9
To control for the effect of age on death rates and provide a single measure, age-adjusted death rates are used.8 Age-adjusted rates are computed by separating deaths into their respective age groups based on the age of the decedent, and computing age-specific rates. These age-specific rates are then weighted according to the 2000 U.S. Standard Population, and are summed to produce the age-adjusted rate. Age-adjusted death rates are highly effective for making comparisons among population groups and among geographical areas because they remove the effects of dissimilar age distributions.
Three important caveats apply when using age-adjusted rates. First, the age-adjusted death rate does not reflect the mortality risk of a “real” population. The actual risk of mortality is represented by the crude death rate. The numerical value of an age-adjusted death rate depends on the standard used and, as a result, is not meaningful by itself. Age-adjusted death rates are appropriate only when comparing groups or examining trends across multiple time periods. A comparison of age-adjusted death rates among groups or periods over time will reflect differences in the average risk of mortality.
Second, age adjusting may mask important information if the age-specific rates between comparison groups do not have a consistent relationship. As an example, Anderson and Rosenberg (1998)8 demonstrate that the trend in the age-adjusted death rate for cancer does not reflect the complexities in the underlying age-specific rates. As averages, age-adjusted rates, like other averages, may be misleading, especially when age-specific rates reflect divergent trends over time. However, usually age-specific rates move roughly in parallel. Thus, age-adjusted death rates are a widely accepted and useful convention for analyzing trends.
Finally, because age-adjusted death rates are averages, they represent merely the beginning of an analytical strategy that should proceed to age-specific analyses, and then to an examination of additional sociodemographic, temporal, and geographic variables.
Data Sources – Numerator data are taken from California Department of Public Health death records, and denominator population data are obtained from the Department of Finance 2012 “Race/Hispanics Population with Age and Gender Detail, 2000–2010. Sacramento, California, September 2012”. Due to the revised population estimates, rates may differ from previously published reports. The 2000 U.S. Standard Population was used for calculating age-adjustments in accordance with statistical policy implemented by NCHS.8 Age-adjusted death rates are not comparable when rates are calculated with different population standards, e.g., the 1940 U.S. Standard Population.
Variability of Rates – Rates are sensitive to size variations in both the numerator (the number of vital events that occurred) and the denominator (the estimated population at risk). For example, in small counties a numerator variation of only a few cases might cause a relatively large shift in a rate, while in a large county could cause no difference in the rate. Likewise, a minor revision in a small county population estimate may cause a relatively major change in a county’s vital event rate. Therefore, caution needs to be exercised when analyzing small numbers, including the rates derived from them.
Rates that are calculated from fewer than 20 deaths are considered unreliable (Tables 2a-2c). These rates are not shown, and are indicated with an asterisk (*). Unreliable age-adjusted rates by race/ethnicity and sex (Table 3) and county of residence (Table 5), are displayed with an asterisk (*) and are provided only as a point of information for further investigation. Rates based on no events are denoted with a dash (-).
Sampling Error and Vital Statistics – Vital events are essentially a complete count, because more than 99 percent of all vital events are registered. Although these numbers are not subject to sampling error, they may be affected by nonsampling errors in the registration process.
The number of vital events is subject to random variation and a probable range of values can be estimated from the actual figures, according to certain statistical assumptions. This is because the number of vital events that actually occurred can be thought of as one outcome in a large series of possible results that could have occurred under the same (or similar) circumstances.
A 95 percent confidence interval is the range of values for a measurement that would be expected in 95 out of 100 cases. The confidence intervals are the highest and lowest values of the range. Confidence intervals tell you how much a measurement could vary under the same (or similar) circumstances.
Confidence Intervals Based on 100 Deaths or More – When there were 100 deaths or more, a normal approximation was used to calculate confidence intervals.
Confidence Intervals Based on Fewer than 100 Deaths – When there were fewer than 100 deaths, a gamma distribution was used to calculate confidence intervals.
Detailed procedures and examples for each type of calculation are given in Technical Notes of Deaths: Final Data for 2009; National Vital Statistics Reports; National Center for Health Statistics, 2011.10
Cause of Death – One of the most important uses for vital statistics data is the study of trends by cause of death. Vital statistics trend research yields valuable information about population health status, emerging public health problems, and at-risk populations, and can be used to develop strategies and allocate resources to improve public health.
Cause-of-death statistics are derived from the medical information reported on the death certificate by the certifying physician or coroner. The medical portion of the death certificate has fields for up to four causes of death (immediate, two intervening, and underlying) plus additional fields for recording contributing causes of death. Up to 20 causes can be entered onto a single death certificate. The cause-of-death field selected for coding and tabulation in this report is the "underlying cause of death." This is generally defined as the disease, injury, or complication that initiated the morbid events sequence leading directly to death.
Leading Causes of Death – This report presents death by leading cause of death in California by sex and race/ethnicity. The leading cause-of-death are ranked according to the frequency of deaths within groups. Because the rank order of any particular cause of death will depend on the list of causes from which selection is made and on the rules applied in making the selection, a clearly defined, uniform method for ranking causes of death is vital to maintain consistency in the reporting of leading causes by federal and state agencies.11 To maintain consistency and to determine rankable causes of death, the ‘‘List of 113 Selected Causes of Death and Enterocolitis due to Clostridium difficile’’ was used to select 50 rankable causes (denoted by #) from which the leading causes. The rankable causes must be mutually exclusive.
Cause-of-death ranking is a popular method of presenting mortality statistics and is a useful tool for illustrating the relative burden of cause-specific mortality, but it must be used cautiously with a clear understanding of the limitations underlying the method. Rankings do not illustrate cause-specific mortality risk or absolute burden as depicted by mortality rates. When comparing rankings across groups or over time, careful note should be made of the age distribution of the populations being compared. It is also important to note that rankings do not necessarily denote the causes of death of greatest public health importance.11
Detailed procedures and further information regarding the 113 Selected Causes-of-Death are provided in NCHS Instruction Manual, Part 9: ICD-10 Underlying Cause-of-Death Lists Tabulating Mortality Statistics. Updated March 2011; National Center of Health Statistics.12
Deaths by Place of Residence – Mortality data analysis in this report are based on records for all California resident deaths occurring in the fifty states, the District of Columbia, US territories, and Canada; all other worldwide resident deaths are excluded. Deaths to non-California residents were excluded from analysis.
Age Groups – The following age groups were used to compute age-specific and age-adjusted rates: under 1 year, 1-4 years, 5-14 years, 15-24 years, 25-34 years, 35-44 years, 45-54 years, 55-64 years, 65-74 years, 75-84 years, and 85 and older.
International Classification of Diseases, Tenth Revision (ICD-10) – Beginning in 1999, cause of death has been coded using ICD-10.13 For more information, see the National Center for Health Statistics ICD-10 page.
Race/Ethnicity – Beginning in 2000, the federal race/ethnicity reporting guidelines changed to allow more than one race to be recorded on death certificates. California initiated use of the new guidelines on January 1, 2000, and collects up to three races per certificate. To be consistent with population data, current reports tabulate race of decedent using all races identified on the certificate.
To meet the U.S. Office of Management and Budget minimum standards for race and ethnicity data collection and reporting, and to be consistent with the population data obtained from the Department of Finance, this report presents Hispanic and the following non-Hispanic race/ethnic groups: American Indian, Asian, Black, Pacific Islander, White, and Two or More Races. Hispanic origin of decedents is determined first and includes decedents of any race group or groups. Non-Hispanic decedents who were reported with two or more races are subsequently placed in the Two or More Races group. Single non-Hispanic race groups are defined as follows: the “American Indian” race group includes Aleut, American Indian, and Eskimo; the “Asian” race group includes Asian Indian, Asian (specified/unspecified), Cambodian, Chinese, Filipino, Hmong, Japanese, Korean, Laotian, Thai, and Vietnamese; the “Pacific Islander” race group includes Guamanian, Hawaiian, Samoan, and Other Pacific Islander; the “White” race group includes White, Other (specified), Not Stated, and Unknown.
Caution should be exercised in the interpretation of mortality data by race/ethnicity. Misclassification of race/ethnicity on death certificates may contribute to underreporting of deaths in American Indians, Asians, Hispanics, and Pacific Islanders.14 This could contribute to artificially low rates for these groups and the Two or More Races group. Race groups’ data that are not individually displayed on the tables or figures due to unreliable rates are collectively included the state data totals.
Trend Analysis – In this report, linear regression was performed to establish the presence of statistically significant trends over the period examined. The trends identified in the report as statistically significant are those for which an F test yielded a p-value less than or equal to 0.05 and had R-square values greater than 0.50 unless otherwise specified. Trend analyses were not performed in cases where rates for one or more years examined were unreliable.