Using the Selection Screen

Table Variables
As indicated on the selection screen, tables can be formed by making a selecting from the row variables and then the column variables. If desired, the table may also be restricted by selecting values from the variables in step three of the selection screen (as in restricting a table to head injuries, males only, etc.)
If causes and/or counties are displayed in the table, they can be sorted (so that the most frequent causes or counties are listed first in the table) by checking the Sort option in step one.
Single or multiple years of data can be included by checking one or more years from step four. Multiple counties or health districts can be included by holding down the Ctrl key on your computer as you select each county or health district. (Do not select both counties and health districts for the same table, because the results are likely to be incorrect.) If you create more than one table, be sure to unselect any options you used in the prior table and do not need in your next table-the earlier selections are maintained after a table is created unless you unselect them. A value can be unselected by selecting it while holding down the Ctrl key.

One- Vs Two-Variable Tables
Generally, a two-variable table is desired, as in a table comparing the rate of head injuries by both age and race. If you want to create a one-variable table, as in looking just at the head injury rate by age, the easiest way is to select Age as the row or column variable and then Year as the other variable, and then select the latest year available in the Year selection row. If this is not possible, as in wanting to look at the head injury rate by age for all five years combined, the best strategy is to select a simple variable like Sex as your second variable.

Comparing Vs Combining Variable Values
Values for the Year, County, District and Cause variables can be either combined or compared. For example, if you select Year as a row or column variable and Cause as the other variable, and then select more than one year in the Year row, the table will show each year, allowing you to look at the trend of causes over the year. If you select Age instead of Year as your first variable and again select more than one year, the years you select will be combined rather than compared, and you will see the average percent or rate of each cause for the combined years. (This is often done when examining rates for one or more counties, since a head or spinal cord injury rate averaged over a number of years of data is likely to be more reliable than a rate based on a single year, especially if a county has a small population.) The same principle holds for the County and Health District variables. If they are not selected as one of the row or column variables but more than one county or health district is selected, their values will be combined rather than compared.

Multiple Level Variables-Drilling Down
A number of variables have more than one level of specificity that can be displayed in a table. These include the Race, Age, Cause and Health District variables. For example, if you select Motor Vehicle as the Cause variable, the Motor Vehicle heading in the resulting table will be underlined and highlighted, indicating that you can select it to develop the table for a more specific selection within Motor Vehicle. For the Motor Vehicle variable, the more specific selections relate to the type of person injured in the crash (motor vehicle driver or passenger, motorcyclist, etc.). For Race, the 'Total' heading in the table can be selected to show the options included in the 'Other' race category (Asian, Native American, etc.). Drilling down on one of the health districts will result in a table that shows the counties that make up that health district.

Rates by Race
If Race is selected as a row or column variable and 'Frequencies and Rates' is selected as the measure of interest, rates will be displayed only for those areas that have a significant number of minority populations. These include the County- menu selections of Missouri, St Louis, St Louis City, Boone, Greene, Jackson, and Pemiscott. Race-specific rates will not be displayed for any other selections from the County menu or for any of the health districts.

Confidence Intervals (CI)
Confidence intervals can be used to assess the level of precision of a reported rate for a population. If the rate of head injuries among whites is 111 and the CI is 108 - 114, it is likely that the CI includes the true population rate. It is also true that the true rate may not be exactly 111 (see extended discussion below for the reasoning behind this). If the rate among Black/African Americans is 120 and the confidence interval is 111-130, we say that the estimate for Whites is more precise than the estimate for Blacks because the CI for Whites encompasses a smaller range (108-114) than the CI for Blacks (111-130). Because rates based on a small number of patients or events are likely to have large confidence intervals, indicating that our estimates of them are very imprecise, rates based on 20 events or less are not shown in the tables.
Another use of CI's is to see whether two rates are significantly different from one another. In the example above, we would say that the rates in the Black and White populations are probably not different, because their confidence intervals overlap (the upper limit for Whites, 114, is above the lower limit for Blacks, 111). Generally, researchers use the 95% confidence interval in estimating rates and comparing them.

Age Adjustment, Standard Population
Rates of a health condition may differ in two groups simply because the groups differ in age. For example, a county that has a higher percentage of elderly people may have a higher rate of head injuries than a county with a more middle-aged population merely because older people are more likely to fall. To avoid this confounding factor in comparing rates, the rates are 'adjusted' for age. This is done by multiplying the actual populations by the proportions that each age group accounted for in the U.S. population in either 1940, 1970, or 2000, and thereby creating new, artificial populations that are now similar in the percent of people they have in each age group. This negates the distorting effect of age and allows us to see the effect of other variables that we are more likely to be able to influence with a health program. (Select Age-Adjustment if you want to be linked to a more complete explanation of age adjusting.) Unless you have some specific reason to do otherwise, we recommend that you use the 2000 standard population.

Definition of Variables
Year - Year is the year in which the person is discharged from the hospital; if a person dies, it is the year of death; if they are treated in the emergency room and transferred to another hospital, it is the year of the date they were transferred.
Race - Races are grouped into the major categories of 'White', 'Black/African American' and 'Other'. If the 'All Races' column is selected in any table containing race, the categories that make up the 'Other' category are displayed (Hispanic, Native American, Asian/Pacific Islander and Other).
Sex - The gender of the patient is either Male, Female or Unknown.
Age - Age is shown in years and is calculated by subtracting the date of birth from the date of arrival at the hospital and converting the resulting number of days to years of age.
Pay Source - Pay source is the source of payment that the patient expects to use at the time they arrive at the hospital.
County/District - These are the county or health district of the patient's residence at the time of arrival at the hospital.
Alcohol - Alcohol is defined by the BAC (Blood Alcohol Content) noted on the hospital record. BAC's of at least 1.0 milligrams alcohol per deciliter of blood is the legal requirement for a finding of driving while under the influence of alcohol and is used to define the 'Inebriated' category.BAC's above 0.0 and below 1.0 defines 'Some Alcohol', and a reading of 0.0 defines the 'No Alcohol' category.
Injury Type - Injury Type refers to whether the person received a head injury or a spinal cord injury. Some patients may have incurred both, so if this option is selected for a table (the 'Head or Cord' option), the sum of head injuries plus cord injuries will be greater than the total shown in the table (because the table shows the total number of patients and not the total number of injuries). The ICD-9-CM codes that are used to define head and spinal cord injuries are as follows: Head Injury: ICD-9CM Codes 800-801, 803-804, 850-854, 95901; 837 can be used if the patient also dies. Spinal Cord Injury: ICD-9CM Codes 806, 952.
Place of Injury - This variable refers to the setting of the injury, such as 'home', 'street', etc.
Safety Device - Safety Device refers primarily to the safety equipment used while driving a vehicle of some kind. This category includes such devices as safety belts, child safety seats, helmets, etc; therefore, if 'Safety Device' is one of the variables selected as a row or column variable, you should usually restrict the table to 'Motor Vehicle' or 'Other Road Vehicle' as a cause of injury. Because helmets are used in sports, and because there is an 'Other' safety device option (which would include walkers, etc.), some tables which are not restricted to motor vehicles may also indicate the use of a small number of safety devices. These tables will also show that most records fall in the 'Unknown' or 'Missing'categories.
Severity - The Severity variable on the selection table is formed by using a number of variables so that every injured person could be assigned a severity level. This was necessary because none of the variables that indicate the severity of an injury are completely reported or could be applied to both head-injured and spinal-cord injured patients. The philosophy behind the assignment of severity scores was that at least some attempt should be made to place people into categories of severity based on the available data, and that if a small number of broad categories were used, the ratings would tend to be more reliable and accurate. Thus, people were assigned either to a Minor, Moderate-Severe, or Death category based on the data items described below. A very few patients did not receive even a Minor severity score, but these were assigned to the Minor category based on the assumption that they would not be in the hospital if the did not have at least a Minor level of injury.
The severity score was assigned as follows: Thurman-scale scores of 1-3 were used to assign patients a score of Minor (Dr Thurman, of the Centers for Disease Control, developed a scale that uses ICD-9-CM codes relating to head injuries and has had some validation work done on it; it does not apply to spinal-cord injuries.). Patients were also assigned a score of Minor if they had an Injury Severity Score (ISS) score less than 16. (The ISS is an optional item assigned by personnel in some hospitals. If the hospital does not assign a score, a computer program assigns an ISS by using the ICD-9CM diagnosis codes: a diagnosis code representing injury to a given body region (head, chest, etc.) is assigned a score from 1 - 6 depending on the degree of severity the diagnosis represents (ranging from 1 - 2 for minor to 6 for not survivable). The three highest scores for different body regions are squared and summed to produce the ISS. Thus if scores for the head, chest and abdominal areas were the highest, say 3, 5, and 2, the ISS would be 32 + 52 + 22, or 38. Scores greater than 15 are considered severe.) Patients were assigned to the Moderate-Severe category if they met any of the following conditions: had an ISS score greater than 15, had been discharged to inpatient rehabilitation, had spent over 10 days in ICU, had a length of stay over 10 days, had a level-of dependence-score that ranged from 'dependent-total help needed' to 'dependent with a device' in the areas of eating, expressing oneself or locomotion, had any degree of paraplegia or quadriplegia (as defined by ICD-9-CM codes), or had scores of 4 or 5 on Dr. Thurman's scale. Finally, those who died received a severity score of Death.
Cause - This variable relates to the cause of the head or spinal cord injury. The ICD-9-CM codes which are used to define the causes are as follows:
Motor Vehicle: E810 - E825
--Driver: E810 - E825 + 5th digit of 0
--Passenger: E810 - E825 + 5th digit of 1
--Motorcycle Driver: E810 - E825 + 5th digit of 2
--Motorcycle Passenger E810 - E825 + 5th digit of 3
--Pedal Cyclist: E810 - E825 + 5th digit of 6
--Pedestrian: E810 - E825 + 5th digit of 7
--Others: All remaining E810 - E825
Other road vehicle: E826 - E829
--Pedal Cyclist: E826 - E829 + 5th digit of 1
--Other Vehicle: All remaining E826 - E829
Falls: E880 - E888
--Stairs: E8809
--Ladder: E8810
--Building: E882
--Same Level: E885
--Other Level: E8849
--Chair/Bed: E8842, E8844
--Other: All remaining E880 - E888
Assault: E960 - E969
--Firearms: E9650 - E9654
--Stabbing: E966
--Blunt Object: E9682
--Other: All remaining E960 - E969
Self-Inflicted: E950 - E959
--Firearms: E9550 - E9554
--Other: All remaining E950 - E959
All other: All remaining E800 - E999

Data Collection and Processing
Reporting of head, spinal cord and spinal column injuries to Missouri's Head and Spinal Cord Injury (HSCI) registry is mandated by law. All records for individuals injured seriously enough to die, be hospitalized or transferred from the E/D to another hospital are to be sent to the Bureau of Emergency Medical Systems in the Missouri Department of Health and Senior Services; however, due to a variety of circumstances there is a fair amount of underreporting. To counteract this, records from the Patient Abstract System(PAS) of hospital discharge data and from the mortality data are added in the following manner: All PAS records that note a head or spinal cord injury are matched against the mortality records that note a head or spinal cord injury. This results in three files: matched PAS/mortality, unmatched PAS and unmatched mortality, with all records noting either a head injury, a spinal cord injury or both. After standardizing the variables on these three files to the HSCI registry variables, the files are combined into one file and matched to the HSCI registry. All PAS mortality records that note a head or spinal cord injury and are not already in the HSCI are then added to it.
Some underreporting still remains, however. A small number of Missouri residents will be treated in hospitals outside Missouri, and they would not be in the registry. According to the mortality file, which records out-of-state deaths, approximately 3 percent of Missouri residents with head injuries had their deaths recorded outside Missouri.

Confidence Intervals, More Discussion
Confidence Intervals (CI's) can be understood in the context of the problems we have in measuring or assessing any injury or disease. Even if the real rate of a health indicator, such as the head injury rate, heart disease rate, etc., in a population does not change from year to year, there are chance factors that affect the observed rates (the rates you see in MICA tables, annual reports, etc). Thus, the rate of a health problem that we report is our best estimate of the true, underlying rate over a given time period. Large confidence intervals indicate that our estimate of the rate is not very precise-it fluctuates from year to year or time to time due to a strong influence of chance factors (This happens particularly when rates are based on a small number of patients or events-20 or less is the guideline used for this MICA; when the number of patients is less than 20, rates are usually not shown.)
One use of confidence intervals is to determine whether the values of two variables are different enough that their difference is unlikely to be due to chance factors and is more likely to be due to true differences in their underlying rates. For example, the head injury rate reported in1998 for males is 149 with a CI of 144 to 154, and the rate for Females is 82 with a CI of 79 to 86. Since the confidence intervals for Males and Females do not overlap (the higher CI limit for Females, 86, is less than the lower CI limit for Males, 144), it is likely that the rates for Males and Females is truly different rather than being due to the chance factors that affect our assessment of the rates. (Because of these chance factors, we cannot be sure the rates of 149 and 82 are the precise head injury rates for Males and Females-they are our best estimates-but we can be fairly certain that the true rates are substantially different from one another.).
Another use of confidence intervals is to indicate how precise our estimate of a rate is. Large confidence intervals indicate our estimates are not very precise. Beyond that, the interpretation of a confidence interval becomes somewhat technical. While a CI of 144 to 154 at the 95 % level is often interpreted to mean that there is a 95% chance that the rate is somewhere in the range of 144 to154, this is not strictly correct. Because the CI and rate are conceived of as being estimates from a sample (a sample that is the population, in this case) that could be taken multiple times, a more correct statement would be that in a large number of samples taken, about 95% of the samples would produce a CI that includes the rate; both the estimated rate and CI are likely to change from sample to sample, and it is unlikely that each sample would have exactly 144 and 154 as the CI's lower and upper limits. If a CI is very small, which often occurs when a sample is very large (as in a large county), one can assume the true rate is likely to be fairly close to the estimated rate, and that it is relatively unlikely to be much lower or higher than the CI limits.
One should also note that a CI is likely to include the rate only if there are not severe, non-chance factors affecting our estimate of the rate and confidence intervals. If, for example, one hospital consistently failed to identify patients with head injuries, the estimated rate of head injuries for Missouri would likely be too low, and we could not be sure that the confidence intervals tended to contain the true population rate.
The 95% confidence interval is typically used by researchers. If the 99% level is used, the confidence intervals are even larger (so that it is even more likely that the CI contains the true rate); thus, if the confidence intervals still don't overlap despite being very large, one can be even more sure that the rates for two diseases or injuries are really different and not due to chance factors. In other words, the 99% level of statistical significance is a stricter level for statistical significance than is typically used. (You may also notice that when the disease or injury rates are based on a large number of injured people, the 95% and 99% CI's in the MICA tables are the same.)
As noted above, rates based on small numbers are particularly subject to chance factors. A simple example of this is given by flipping a coin. If the coin is evenly balanced, flipping it a large number of times should result in about the same number of heads and tails; however, flipping it only 10 times is less likely to result in the same number of heads and tails; rather it is more likely to be something like 3 tails and 7 heads, 4 tails and 6 heads, etc. Thus, when you are examining a rate that is based on a small number of patients or events, as in the spinal cord injury rate in a county, it is very important to look at the confidence interval; it is likely to be very wide, indicating a large degree of uncertainty about what the real rate is Also, if you are comparing counties with a small number of events, the rates may look different, but if the confidence intervals overlap, it is likely that the underlying true rates in the populations are not really different; instead, the observed rates are being substantially influenced by chance factors, causing them to fluctuate widely from year to year.