What is the data?
The data presented in the EHDI Data Hub are a combination of reported metrics in CDC’s EHDI Data and calculated metrics from these. Metrics are related as shown in the flow diagram below or as noted in the formulas in tables below.
Figure: Flow diagram showing connections between select metrics of the EHDI process.
We have maintained metric names close to how they are recorded in CDC’s data. Three things to keep in mind:
- One word worth noting is referred as in “Referred Inpatient” or “Referred Outpatient.” In these cases, it should be understood as an alternate term for not passed or failed.
- From 2021, the CDC began presenting “Prevalence per 1,000 births” – we present per 1,000 screened, as you will also find in CDC’s 2020 and earlier data.
- From 2022, CDC changed some terms within columns of the PDFs (e.g., “Permanent Hearing Loss” in 2021 and “Deaf or Hard of Hearing (D/HH)” in 2022) – we present terms as you will also find in CDC’s 2021 and earlier data.
When we present metrics, we list calculations first (prevalence, rate, percentage, total) followed by reported metrics. We also aim to group metrics within these as they are related to screening, diagnostics, and early intervention, to present a flow in viewing the data that is similar to that in the diagram above.
Decision maker takeaways note how we believe a metric could be used to support EHDI work.
How was the data prepared?
Annual state average, median, and standard deviation (SD) were calculated across the available data for states and District of Columbia within a year – no territories were included.
Annual US values were across the available data for states, District of Columbia, and territories.
Calculated metrics
We have calculated 35 metrics as described in the table below. We calculated a prevalence metric, rates per 1000 screened, percentages, and total counts.
To maintain consistency in the dataset across the years, we utilized 1) a single formula for each metric as documented in the table below, and 2) the data as reported to the CDC in our calculations (such that, if a zero was in the CDC report, we use it as 0 unless we otherwise make note for metrics in formulas).
There are ten compound metrics that combine two or more variables through addition or subtraction. If a variable used in a calculation is missing, that value is ignored (i.e., treated as zero) in the calculation. If all variables for the calculation are missing, the metric is also missing. This is consistent with how the data come from the CDC. For example, when summing the total enrolled in an intervention, if the total enrolled in Part C EI was known but the total enrolled in Non-Part C EI was missing, the total enrolled in Part C & Non-Part C EI matched the total enrolled in Part C EI.
In efforts to present as many metrics as possible, the following two rules were applied for compound metrics:
- Tscr1, OPS, EI, EI6. These compound metrics are each the sum of two variables. If at least one of the constituent variables is available, the metric is calculated and available.
- CLEAR, NPIS, NPOS, PIS, POS, ReferToOPS. These compound metrics are each calculated with three or more variables If at least two of the constituent variables are available, the metric is calculated and available.
PerTNscr, PerPIS, Per NPIS, PerReferToOPS, PerReDirDiag, PerOPS, PerNPOS, PIS, NPIS, ReferToOPS, OPS, POS, NPOS. These metrics were calculated from detailed screening metrics provided to the CDC only from 2015 onward due to changes in reporting structure (noted below). They therefore appear as missing from 2007 through 2014.
Utilizing the data as reported led to some calculations that may show values or trends that may not be expected as you review the data. For example, percentages that are greater than 100% or totals of 0.
As depicted in the flow chart above, we use the Total Not Pass Last Screening metric in our understanding of the EHDI program. At this time, the CLEAR metric was calculated using Total Not Pass Last Screening and may lead to some unexpected results. For example, with South Dakota in 2021. We are reviewing this and a second metric on not passing – Total Not Pass – that includes babies who were not screened. We will make adjustments as decisions are made.
Rates
We calculated rate metrics by dividing by the “Total Screened” for each state. “Hearing Loss Prevalence,” a metric that we calculated and is also reported by the CDC, is an example of this approach. This allows many other metrics to be more easily compared by and between states, to state averages, and the US as a whole. The CDC has traditionally used percentages for many metrics. An example is “Loss to Follow-Up” divided by “Not Passed” expressed as a percentage. “Not Passed” – unique to each state’s specific inpatient and outpatient screening protocols – is unique to each state and not comparable. The resulting quotient is thus not a comparable quantity state to state. Using “Total Screened” as a denominator to create what we have called a “rate” avoids this problem and facilitates state comparisons. These rate metrics would typically be read as the metric per thousand babies screened.
Reported metrics
Calculated metrics were prepared from 27 as-reported metrics, gathered from EHDI data as reported to the CDC by states. All data were extracted from PDFs reported on CDC website. We provide links to all the PDF sources in the FAQ.
We reviewed the data from each year to identify metrics consistently provided in order to maintain consistency with the data over the years. For example, with the early intervention (EI) metrics, Part C EI and Non-Part C EI constituents were available in more years than the Part C & Non-Part C EI sums. We utilized the constituent metrics and calculated the sum EI metric for our dataset rather than a case-by-case basis of metric availability.
TNscr, NoInPaOut, NoInReOut, PaInPaOut, PaInReOut, PaInNoOut, ReInPaOut, ReInReOut, ReInNoOut, ReInStDia. These detailed screening metrics were provided to the CDC from 2015 onward due to changes in reporting structure. These metrics – and any metrics calculated from them (noted above) – therefore appear as missing from 2007 through 2014.
Utilizing the data as reported may show values or trends that may not be expected as you review the data. For example, values reported as 0.
As depicted in the flow chart above, we use the Total Not Pass Last Screening metric in our understanding of the EHDI program. At this time, the CLEAR metric was calculated using Total Not Pass Last Screening and may lead to some unexpected results. For example, with South Dakota in 2021. We are reviewing this and a second metric on not passing – Total Not Pass – that includes babies who were not screened. We will make adjustments as decisions are made.
Data availability
The Data Availability Table below summarizes all the years in which a metric (column) was not available in each state (row). A blank cell indicates the state provided a value for all years for that variable, even a value of 0. For example, data were unavailable for Alabama in 2014 on the number of babies who failed inpatient screening and were referred straight to diagnostics (ReInStDia).
More information
If you have any questions on how the data was prepared or any other information, please see our FAQ or reach out to us.