Age-adjusted death rate - Deaths per 100,000 population, adjusted to a standard population (such as US 1940 or US 2000), by the direct method. Age-adjusted rates are often better for making comparisons than unadjusted rates, because they adjust for differences in age distribution between populations. An age-adjusted rate is a summary measure. Besides calculating overall age-adjusted rates, it is also recommended to compare age-specific rates.
Age-adjustment standard - A standard population for calculating an age-adjusted death rate. The 1940 and 2000 US Census population are the most common standards.
Apgar score - A summary measure of the condition of the infant based on heart rate, respiratory effort, muscle tone, reflex irritability, and color. Each factor is given a score of 0, 1, or 2; the sum of these five values is the Apgar score, ranging from 0 to 10.
Area set - One or more areas combined.
ASCII file - A text file, with only alphabetical, numerical, and punctuation characters, like you would see in normal text. Vitalnet can produce output in ASCII format.
Tabular chart - A section of a Vitalnet table. Gives an scaleable graphical representation of the data. May be omitted from the output table.
Birth rate - Births per 1,000 female or total population.
Birth weight - The weight of an infant at delivery, expressed in grams.
CDC - Centers for Disease Control and Prevention. US federal health agency.
Cell - A space for a single numerical result in a table, at a row-column intersection.
Cell suppression - An asterisk "*" is placed in cells with fewer deaths than a limit set by the user. Row / column totals with exactly one suppressed cell in the row / column are also suppressed. If more than one cell in the row / column is suppressed, the row / column total is displayed.
Cesarean rate - Number of cesareans / number of births.
Cesarean rate - primary - FC / (FC + FV). Numerator is number of first-time cesarean deliveries (FC). Denominator is number of first-time cesarean deliveries (FC) plus number of first-time vaginal deliveries (FV).
Cesarean rate - repeat - RC / (RC + VA). Numerator is number of repeat cesarean deliveries (RC). Denominator is number of repeat cesarean deliveries (RC) plus number of vaginal deliveries after previous cesarean (VA).
CNM - Certified Nurse Midwife (birth attendant category)
Columns - Vertical groupings of data in a Vitalnet table, such as a column for each race group.
Confidence interval (confidence limits) - A range of values within which the true value of a variable is thought to lie, with a specified level of confidence. For a result of 23.5, a confidence interval might be (23.1-23.9). The smaller the interval, the more reliable the result. If the 95% confidence intervals do not overlap, there is probably a statistically significant difference. Vitalnet uses several methods to calculate confidence intervals. The output table documents which method was used.
Confidence level - The likelihood that the true value of a variable is within a confidence interval. For example, for confidence intervals at the 95% level, we are statistically 95% certain that the actual value of the variable is within the interval.
CSV format - Comma-separated-value format. CSV files are readily imported into spreadsheet software. Each output item is separated by a comma from surrounding items, and each output text item is surrounded by "double quotes". A comma-separated-value file has "csv" extension. Similar to TSV format.
Cursor key - Arrow key, PgDn, or PgUp key. Used to navigate a web page or computer program.
Data mining - Finding unexpected relationships in a data set. Similar to exploratory data analysis. Vitalnet is excellent at data mining. Of course, keep in mind that the more you look, the more unusual events you will find, just by chance.
Data warehouse - Software system, such as Vitalnet, making large complex databases readily available for querying and analysis.
dBASE III format - A widely used file format derived from the database software of the same name. Files in this format may be readily imported into almost any data analysis, graphing, mapping, or other presentation software. Has dbf extension. Suppressed cells are represented as the number "-1".
DIF format - Data interchange format. DIF files are readily imported into spreadsheet software. The DIF format is too complex to explain in this glossary. Has "dif" extension.
DO - Doctor of Osteopathy (birth attendant category)
Denominator - The number on the bottom of a fraction. Population data are often referred to as "denominator data", as they are used as denominators to calculate population-based rates.
ENTER key - A key on your keyboard. Sometimes called RETURN key. Often used to select an item or complete an operation.
ESCAPE key - A key on your keyboard, often on the upper left. Tells Vitalnet to return to a higher-level (previous) menu. Pressing 'Z' usually does the same thing.
Export - Produce output that can be read into other computer programs. Vitalnet produces ASCII text (txt), comma-separated-value (csv), HTML (htm), and dBASE III (dbf) files for export.
Fetal death - A fetus showing no evidence of life after a complete birth.
Fetus - Term applied to the unborn offspring from the date of conception until the completion of pregnancy.
Filter variable - A variable solely used to filter which records are included in the output. For example, for a single table with race rows and sex columns, age is a filter variable.
Footer - Last part of a Vitalnet table. Lists required but less important details of the analysis, such as the date and time produced, and data sources. The footer contains a unique ID to assist in keeping track of analyses.
General fertility rate - Total live births (to all women) per 1,000 women age 15-44 in a given year.
Gestational age - Number of completed weeks elapsed between the first day of the last normal menstrual period and the date of delivery. Gestational age is expressed in completed weeks.
Gestational age - Number of completed weeks elapsed between the first day of the last normal menstrual period and the pregnancy outcome. Gestational age is expressed in completed weeks.
General fertility rate - Live births per 1,000 women age 15-44 in a given year.
Header - First part of a Vitalnet table. Lists key analysis parameters, such as years analyzed.
Hepatitis case rate - Number of cases per 100,000 people at risk. For example, if Smith County has 40 cases, and 200,000 people, the hepatitis case rate is 20 per 100,000.
Hepatitis death rate - Number of deaths per 100,000 people at risk. For example, if Smith County has 4 deaths, and 200,000 people, the hepatitis death rate is 2 per 100,000.
High birth weight - A birth weight 4,000 grams or greater.
High birth weight percent - Percent of births with birth weight 4,000 grams or greater.
Import - Read information into a computer program. ASCII text, CSV, TSV, DIF, HTML, and dBASE III files from Vitalnet are easily imported into word processing, spreadsheet, data analysis, mapping, graphing, and other presentation software programs.
Infant - An individual less than one year of age.
Infant death - Death of a liveborn infant from the moment of birth to the end of the first year of life.
Intrauterine growth retardation - Birth weight in the lowest decile of birth weight for gestational age. An alternative definition is birth weight less than 2,500 grams (low birth weight) in a full-term infant (born at or after 37 weeks gestation).
Jackknife Method - A method for calculating variances and confidence intervals. Jackknife is accurate for complex survey designs, such as BRFSS. Jackknife will work with any statistic, such as percent, mean, or median. The jackknife method repeatedly calculate a replicate statistic. For each replicate, it leaves out one observation (or group of observations), and reweights the observations left in. The variance of the replicates is the same as that of the original data. It is called "jackknife" because it is so generally useful.
Kessner adequate percent - Numerator is number of births with adequate Kessner index. Denominator is number of births with known Kessner index (unknown Kessner scores are not included in denominator). This is a measure of the adequacy of prenatal care. A higher Kessner adequate percent for a population indicates better prenatal care.
Kessner index - Method of categorizing adequacy of prenatal care, based on month of pregnancy care started, number of visits, and length of gestation. This takes both amount and start of prenatal care into account, and adjusts for the fact that women with short gestations have less time in which to make prenatal care visits.
Least-squares - A standard method for fitting the best straight line to a set of points. Produces a Y-intercept and a slope defining the least-squares line.
Light bar - A highlighted area on the computer screen that you can move by pressing an arrow key or other cursor keys. The light bar highlights an item that you may select, add or delete.
Live birth order - Live birth order is the number of children born alive to a mother, including the current baby. If the mother has three previous live births, the live birth order for the next birth is four. Another example: If the mother has one previous live birth, and has twins this time, the live birth order for the first twin is two, and is three for the second twin.
Local area network (LAN) - A computer networking product, such as Novell Netware or Windows NT. Vitalnet may be accessed from a local area network.
Log file - A computer file for saving Vitalnet tables, and other Vitalnet output.
Low birth weight - A birth weight less than 2,500 grams (5 pounds, 9 ounces).
Low birth weight percent - Percent of births with birth weight less than 2,500 grams (5 pounds, 9 ounces).
Main Menu - The top level Vitalnet menu. All parameters are summarized on the Main Menu, and you will always return to the Main Menu before producing a table.
Marriage rate - Number of marriages per 1,000 population. For example, if Smith County has 4,000 marriages, and 200,000 people, the marriage rate is 4 per 1,000. Due to data limitations, does not take into account the current marital status of the population.
MD - Doctor of Medicine (Physician) (birth attendant category)
Multiple age groups - One age group for each table row (or column). Example: 0-19, 20-59, 60-99+.
Multi-tables - Vitalnet option to automatically produce a series of tables. For example, there may be one table for each selected race.
NCHS - National Center for Health Statistics. US health statistics agency. Part of the CDC.
Neonatal death - Death of a liveborn infant within the first 27 days, 23 hours, and 59 minutes of life.
Neonatal death rate - Numerator = neonatal deaths x 1000. Denominator = number of liveborn infants.
Neonatal period - Period from birth through the first 27 days, 23 hours, and 59 minutes of life.
Neonate - A newborn infant during the first 27 days, 23 hours, and 59 minutes of life.
Perinatal period - Period from the 20th completed week of gestation (140 days) through the first 27 days, 23 hours, and 59 minutes of life.
Place of occurrence - The geographic location where a birth occurred.
Place of occurrence birth data - Data compiled by the location where the birth occurred, without regard to the place of residence of the mother.
Place of residence - The geographic location where a birth occurred.
Plurality - Number in a birth, such as twins or triplets.
Population - The number of people living in an area.
Postneonatal - Period between 28 days and one year of age.
Post term birth - Birth after the 41st week of gestation. Birth during or after the 42nd week of gestation.
Pregnancy outcome - A birth, fetal death, or abortion.
Pregnancy rate - The number of pregnancies per 1,000 total or female population.
Preterm birth - Birth occurring before the 38th week of gestation. Birth occurring during or before the 37th week of gestation.
Rows - Horizontal lines in a Vitalnet table, such as a row for each race group.
Row sort settings - Vitalnet rows may be sorted in ascending or descending order.
Set - A combination of one or more things. For example, several areas may be combined into an area set.
Single age group - Only one age group (30-49, for example) is selected. A single age group is used for tables that do not have age columns or age rows.
Stand-alone PC - A desktop or laptop computer running off its own local hard disk. Vitalnet may be run from a stand-alone PC.
Statistic (Main Statistic) - The basic type of numerical result displayed in a table, chart, or map. For example, birth rate. divorce rate. death rate. hepatitis case rate. hepatitis death rate. marriage rate. pregnancy rate.
Submenu - A menu accessed from a higher level menu. A submenu helps select or modify a parameter listed on the higher level menu.
Table - A set of results produced by Vitalnet. A table has several parts:
Term birth - Birth during the four week period after the 37th week of gestation and before the 42nd week of gestation. Birth during the 38th through 41st week of gestation.
TSV format - Tab-separated-value format. TSV files are readily imported into spreadsheet software. Each output item is separated by a tab from surrounding items, and each output text item is surrounded by "double quotes". A tsv-separated-value file has "tsv" extension. Similar to CSV format.
Unix - A popular set of computer operating systems. Vitalnet can run on Unix.
Unknown Values - Unknowns are automatically inserted into a Vitalnet table. For example, a separate row (or column) for unknown race. The rate is assigned as zero for an unknown category, since there is no population denominator to use. Some fields, such as sex for certain data sets, are never unknown, so unknowns are left off the table. When, such as for age-adjusted rates, the unknown variable (age) is different from the rows or columns, the number of unknowns for age is shown below the table.
Very low birth weight - A birth weight less than 1,500 grams (3 pounds, 5 ounces).
Very low birth weight percent - Percent of births with birth weight less than 1,500 grams (3 pounds, 5 ounces).
Windows - A widely used set of PC operating systems. Vitalnet runs under any version of Windows.
Wizard - An interactive utility that guides the user through a potentially complex task. Wizards are often implemented as a sequence of dialog boxes which the user can move forwards and backwards through, filling in the details required. The implication is that the expertise of a human wizard is encapsulated in the software wizard, allowing the average user to perform expertly.
World Wide Web (WWW) - A widely used part of the internet that may be easily accessed with a web browser. Vitalnet runs on the WWW.
The "Main Statistic" (numerical outcome) is the basic type of number in a Vitalnet table. It is best understood by looking at the examples below.
Valid Interviews - The number of interviews considered valid for inclusion in the BRFSS data set. This number is NOT adjusted for sample weights. It is NOT valid for epidemiological comparisons. Interviews with not sure, refused, don't know, unknown, and missing values are INCLUDED.
Valid Interviews (Weighted) - The number of interviews considered valid for inclusion in the BRFSS data set. This number IS adjusted for sample weights. It IS valid for epidemiological comparisons. This number approximates the population in a category. Interviews with not sure, refused, don't know, unknown, and missing values are INCLUDED.
'Yes' Responses to Question - The number of interviews with a Yes response for a particular variable. This number is NOT adjusted for sample weights. It is NOT valid for epidemiological comparisons.
'Yes' Responses (Weighted) - The number of interviews with a Yes response for a particular variable. This number IS adjusted for sample weights. It IS valid for epidemiological comparisons. This number approximates the population responding Yes for a variable.
'No' Responses to Question - The number of interviews with a No response for a particular variable. This number is NOT adjusted for sample weights. It is NOT valid for epidemiological comparisons.
'No' Responses (Weighted) - The number of interviews with a No response for a particular variable. This number IS adjusted for sample weights. It IS valid for epidemiological comparisons. This number approximates the population responding Yes for a variable.
% Responding 'Yes' (Weighted) - Percent of interviews responding Yes to a particular variable, with weighting factored in. It IS valid for epidemiological comparisons. The weighting means this is not a simple ratio of Yes / Total. Not sure, refused, don't know, unknown, and missing are EXCLUDED from the denominator.
% Responding 'No' (Weighted) - Percent of interviews responding No to a particular variable, with weighting factored in. It IS valid for epidemiological comparisons. The weighting means this is not a simple ratio of No / Total. Not sure, refused, don't know, unknown, and missing are EXCLUDED from the denominator.
Responses - The number of interviews with a valid response for a particular variable. Not sure, refused, don't know, unknown, and missing are EXCLUDED. This number is NOT adjusted for sample weights. It is NOT valid for epidemiological comparisons.
Definition - A "confidence interval" is a range of values within which the true value of a variable is thought to lie, at a certain "confidence level", such as 95%. A larger percentage (such as 99%) is more stringent than a smaller percentage (such as 80%). Use 95% if you are unsure.
Interpretation - The smaller the interval, the more reliable the result. Two results that overlap at the 95% level are less likely to be significantly different than results which don't overlap.
Methods - The method Vitalnet uses to calculate confidence intervals depends on the context. The method is listed in the footnotes to the table.
Z * Rate / Sqrt (Events) - This method is recommended by the NCHS. Technical Appendix of the Vital Statistics of the United States, Vol II, Mortality, Part A
Jackknife Replication - A method for calculating variances and confidence intervals. Jackknife is accurate for complex survey designs, such as BRFSS. Jackknife will work with any statistic, such as percent, mean, or median. The jackknife method repeatedly calculate a replicate statistic. For each replicate, it leaves out one observation (or group of observations), and reweights the observations left in. The variance of the replicates is the same as that of the original data. It is called "jackknife" because it is so generally useful.
Jackknife Replication - The Jackknife procedure for calculating a BRFSS confidence interval has five parts (1-5), as follows: 1) SORTING: Sorting is only needed if more than one observation is removed for each replicate. In other words, sorting is not needed if "unlimited replicates". For the case where more than one observation is removed for a replicate, if the sorting is not done, the observations might be in a different order under different table scenarios, resulting in slightly different CI in different tables. Here is the series of sort operations (A-E): A) Sort cell observations by position within original BRFSS data file. B) Seed LINUX POSIX.1-2001 srand function with 1. C) Sequentially assign RandNdx to cell observations, using LINUX POSIX.1-2001 rand function. D) Sort cell observations by RandNdx. E) Sort all table observations by the cellNdx (table cell). 2) PARTITIONING: Partitioning is only needed if more than one observation is removed for each replicate. In other words, partitioning is not needed if "unlimited replicates". Partitioning divides observations in a cell into groups. Assume there are 1000 observations in a cell. For "replicates = 500", groups would be 1-2, 3-4, ..., 999-1000. For "replicates = 999", groups would be 1, 2, ..., 998, 999-1000. For "replicates = 998", groups would be 1, 2, ..., 996, 997-998, 999-1000. For "replicates = 333", groups would be 1-3, 4-6, ..., 994-996, 997-1000. For "replicates = 200", groups would be 1-5, 6-10, ..., 996-1000. The same logic can be used to partition groups for any number of replicates. Since the observations have been randomly sorted in the previous step, there is no risk of bias resulting from related records being adjacent. If "unlimited replicates", or "replicates = N" and number of observations in cell is less than N, each observation is a group, and partitioning is not needed. 3) REWEIGHTING: Removes one group of observations. Redistributes total weight among remaining observations in a cell. For example, if "unlimited replicates", remove observation #1. Or if 1000 cell observations and "replicates = 500", remove observations 1-2. The remaining observations are re-weighted, so that the total FINAL_WT for each State remains the same. Subtract FINAL_WT of the omitted observations from State total FINAL_WT. Multiply each remaining (non-omitted) observation by the following factor: (State total FINAL_WT) / (State FINAL_WT after omitting observations). This readjusts the FINAL_WT for remaining observations, so the State total FINAL_WT is the same as before. 4) REPLICATING: Produces one replicate estimate. After a group of observations is omitted, and the remaining observations reweighted, the outcome statistic (% yes, median, or other) is calculated as a "replicate", and stored for later use. 5) CI CALCULATION: Calculates the confidence interval, from the replicates. Each replicate estimate is subtracted from the overall outcome statistic (the value with all observations included), and the difference is squared. The variance is calculated as: sumOfSquares_Cell * (replicate_count - 1) / replicate_count. The replicate_count is the number of replicates done for the cell. The standard error is the square root of the variance. The upper or lower confidence limit is (Student's T Value) * (standard error). The Student's T Value is for a given confidence interval (such as 95%), with degrees of freedom one less than the number of cell observations. If only one observation, degrees of freedom is set to one. Confidence intervals with very few obervations are not reliable in any case. If the confidence interval goes below zero, it is cut off at zero. If the percent confidence interval goes over 100%, it is cut off at 100%.
Brief description of jackknife method, for calculating variance for one BRFSS cell - 1) Put interview records for cell into random order. 2) Remove one observation, or a group of observations. 3) Adjust weights for remaining observations. 4) Calculate replicate statistic (mean, median, percent) with remaining observations. 5) Repeat steps 2-4 for each observation or group of observations removed. 6) Calculate deviation of each replicate statistic from overall statistic. 7) Add the squared deviations to produce sum of squares (SS). 8) Variance = SS * (replicateCount - 1) / replicateCount. NOTE: More replicates provide slightly more accuracy, but take longer. 1000 replicates seems more than adequate.
Poisson distribution - This method is valid if events are relatively rare, which usually applies to health events. Scientific Tables, Diem and Lentner (ed), Giegy, 1970, page 189.