Ranking/Grading

Ranking
Grading
Indicator Precision: Confidence Intervals and Small Differences  

Ranking
All of the indicators selected for the 2008 Colorado Health Report Card are reported in a consistent manner across all or most states,¹ allowing us to rank Colorado’s performance relative to other states. For each indicator, the “best” performing state is ranked first. Colorado ranks first on two indicators—a low incidence of working-age adults who are obese and adults with diabetes—to 45th for low rates of insurance coverage for children.

An “adjusted rank” was calculated for the 11 indicators for which fewer than 50 states reported data. The rationale for this is that being 10th out of 30 states, for example, is not as good as being 10th out of 50 states. To illustrate how the calculation was derived, for the “abstinence from smoking during the last three months of pregnancy” indicator, Colorado ranked fifth among the 19 states that report data for this indicator. To calculate an adjusted rank, we divided five by 19 and multiplied this number by 50. The resulting “score” yielded an adjusted rank of 13 (scores were rounded to the nearest whole number). The affected indicator fact sheets show actual rank and number of states reporting followed by the adjusted rank, for example, 5/19 = 13/50.

For several indicators more than one state may have the same value. For example, 9.2 percent of babies in both Colorado and North Carolina were born at a low birth weight; both states were assigned the rank of 41. The next state, Tennessee, with 9.5 percent of babies born at a low birth weight, was ranked 43rd.² 

Grading
A grade is assigned to each life cycle phase based on Colorado’s average rank for all indicators included in the phase. For example, the average rank for the six Healthy Children indicators is 28. The letter grade of C- was assigned according to the rank/letter grade equivalencies shown in the table to the right. Because most states are like Colorado in that they have inconsistent indicator rankings within a life cycle domain, it is unlikely that any state would have an average domain rank above 5 or below 45.³ 

Indicator Precision: Confidence Intervals and Small Differences
Most of the indicators are derived from responses to sample surveys and therefore are subject to sampling error, that is, the possibility that the sample differs from the population from which it is drawn in some significant way. For indicators based on a sample, it is appropriate to calculate a “confidence interval,” often referred to as a “margin of error.” The margin of error is important to consider when ranking states based on sample surveys because two or more states that appear to have different values may not be statistically different from each other. Similarly, changes in an indicator’s value from year to year or differences among subpopulations (age, gender, ethnic and income groups) may not reflect real differences in the population from which the sample was drawn.

Some indicators such as infant mortality and teen fertility come from the Vital Statistics System records, which are maintained on virtually all such vital events. For these indicators, differences are real and not subject to sampling error. Small differences in values from year to year, however, may represent short-term fluctuations that do not constitute a trend. Similarly, small differences between states or subpopulations with similar values may be the result of short-term fluctuations.

1. Values for all 50 states are available for 28 of the 39 indicators.
2. Most indicators are shown as percents to the first decimal place (e.g., 22.7 percent). When two or more states have the same value to the first decimal place they are given the same rank. (See Confidence Intervals and Small Differences.)
3. Average ranks and equivalent letter grades have not been calculated for the other 49 states.