How to Read an Output Table from WISH

Outline

I. Understand rows, columns, and labels
II. Read the numbers
III. Interpret results and the confidence interval

I. Understand rows, columns, and labels

A table is composed of rows and columns. A row is a horizontal line of cells in a table, and a column is a vertical line of cells.

In the example table, the rows are labeled according to region of residence: "All," "Southern," "Southeastern," etc. The columns are labeled "Number of Low Birthweight Births," "Total Number of Births," "Percent Low Birthweight," and "95% Confidence Interval."

As indicated by the column headings, the example table provides four kinds of information related to low birthweight in Wisconsin.

Column 1. "Number of Low Birthweight Births." This reports the number (frequency) of the event that is the subject of the query (births characterized by low birthweight).

Column 2. "Total Number of Births." This reports the total number of live births.

Column 3. "Percent Low Birthweight." This provides the proportion of all births that were low birthweight. It was calculated by dividing Column 1 by Column 2, then multiplying by 100:

Column 3 = (Column 1/Column 2) * 100 = (4,262 / 66,490) * 100 = 6.41%

Column 4. "95% Confidence Interval." A confidence interval is provided for all rates and percentages shown in WISH. See Part III (below) for more information about the need for and interpretation of a confidence interval.

II. Read the numbers

If you are interested in Wisconsin as a whole, you can read the row labeled "All." Or you can look at a row for a particular region of the state. For example, looking at the row labeled "Southern," we can see that in 1997 among residents of the Southern region of Wisconsin:

There were 685 babies born at low birthweight.

A total of 11,385 babies were born.

Low birthweight babies made up 6.02 percent of total births for the region.

We can be quite confident that the " true or underlying " low birthweight percentage for the Southern region lies in the range of 5.59% to 6.47%. (See Part III for more explanation.)

III. Interpret results and the confidence interval

Columns 1, 2 and 3 of the example table provide basic descriptive information, which is more straightforward than the information in Column 4 (confidence intervals). The following is a brief explanation of a confidence interval, its basic statistical meaning, and how to use it. You may also want to read Paul Buescher’s article, "Problems With Rates Based On Small Numbers."

What are confidence intervals?

Output tables provide confidence intervals for all rates and percentages. A confidence interval is a range around a rate; this range has a 95% probability of containing the "true" or underlying value.

Why present confidence intervals?

Confidence intervals are most often presented with rates based on sample data, to estimate the possible difference between the sample rate and the true rate. However, confidence intervals can also be used when all events in an entire population are known. This is because "a rate observed in a single year can be considered as a sample or estimate of the true or underlying rate" (Buescher, cited above).

"This idea of an ‘underlying’ rate is an abstract concept, since the rate observed in one year did actually occur, but it is this underlying rate that health policies should seek to address rather than annual rates which may fluctuate dramatically. The larger the numerator [number of events] of the observed rate, the better the observed rate will estimate the underlying rate" (Buescher, cited above).

How should confidence intervals be used?

Confidence intervals can assist in making comparisons between geographic areas and between years. For example, look at the low birthweight percentage for 1997 in the Southern region of Wisconsin (6.02%), along with the confidence interval around that percentage (5.59%-6.47%). This means we can be 95% confident that the 1997 low birthweight percentage for the Southern region, for purposes of comparison with other years for the Southern region and with other regions for 1997, lies in the range of 5.59% to 6.47%.

Example. Using the example table, we can assess whether the low birthweight rate in the Southern region of Wisconsin (6.02%) is significantly higher than that for the Northeastern region (5.71%). Looking at the confidence intervals for each of those percentages (5.59%-6.47%, and 5.34%-6.12%), we see that they overlap. As a result, we can say we are 95% confident that the two rates do not differ from one another.

Cautions

The size of a confidence interval will be affected by the number of events (sample size) and the p value (probability). Please consult the statistical literature or experts on interpreting confidence intervals, especially when you find small numbers of events in an output table.

Note about cell suppression (X):

If you are requesting data for a geographic area (such as a single county) where the annual number of births is small, it is often useful to combine years of data. WISH suppresses small numbers (when cell size is less than 5) to comply with Wisconsin vital records data privacy guidelines. In the Infant Mortality and Percent of Births to Teens modules, the rates, percents, and confidence intervals are suppressed when the denominator value (total births) is less than 20.

Reference:

This material on confidence intervals was adapted from the following publication: Buescher, Paul A. "Problems with rates based on small numbers," Statistical Primer 12:1-6 (published by the North Carolina Department of Environment, Health and Natural Resources; included in MCH Model Indicators, distributed by HRSA/MCHB, 1998). In 2008 the paper was revised as Buescher, Paul A. "Problems with Rates Based on Small Numbers," Statistical Primer 12 (published by the North Carolina Public Health, State Center for Health Statistics, 2008) (PDF).