Using Birth Defects Surveillance to Count Neonatal Abstinence Syndrome in Three States

This page summarizes a report in CDC's MMWR by CDC, the March of Dimes and the health departments of Illinois, New Mexico and Vermont, on infants born in 2015 and, in Illinois, 2016.

Neonatal abstinence syndrome (NAS) is a drug withdrawal syndrome that can follow exposure to opioids before birth. In the United States it is tracked mostly through diagnosis codes in hospital discharge data, which are neither validated nor checked against the medical record.

Counted that way, reported NAS incidence rose from 1.5 to 8.0 per 1,000 U.S. hospital births during 2004–2014, using ICD-9-CM codes. Little was known about how well those codes find true NAS, or how the switch to ICD-10-CM codes on October 1, 2015 changed the estimates. There is no standard way to run state NAS surveillance and no standard national system.

The pilot

Through a competitive process, the March of Dimes, a nonprofit working to improve the health of mothers and babies, worked with CDC to award grants to the CDC-funded birth defects programs of Illinois, New Mexico and Vermont. Each adapted birth defects surveillance methods to find NAS, either through active population-based surveillance or through passive case-finding with confirmation. CDC has supported population-based birth defects surveillance since 1967.

Finding possible cases. All three states searched hospital discharge data for these infant codes, plus other infant and maternal codes of their choosing:

Two states added sources:

• Illinois: its birth defects registry's Adverse Pregnancy Outcomes Reporting System, which records infants of Illinois residents with documented prenatal opioid exposure or withdrawal symptoms during the newborn stay; and reports from selected hospitals of infants with NAS scores above 8, on a scale that typically runs 0–37.
• Vermont: Medicaid claims, and commercial claims from the state's all-payer claims database, the Green Mountain Care Board's Vermont Health Care Uniform Reporting and Evaluation System, as part of a Birth Information Network set up for birth defects and other congenital conditions.

Confirming them. After removing duplicates, the states abstracted every available infant and maternal medical record for each possible case and applied one shared clinical case definition, which expands on an earlier published one. Records could not be obtained for under 2% of possible cases, for example when the infant was born or treated out of state.

Scoring the codes. For each code, the positive predictive value (PPV) was the number of confirmed cases divided by all possible cases the code flagged, times 100.

Findings

Confirmed NAS in 2015 births, per 1,000 births:

Vermont's Medicaid and commercial figures use each payer's births as the denominator.

Best codes. In all three states, the infant drug withdrawal codes had the highest PPVs: 779.5 (ICD-9-CM, 58.6%–80.2% across states) and P96.1 (ICD-10-CM, 58.5%–80.2%). In two of the three states, the PPV did not change when 779.5 gave way to P96.1.

What it means

• The method works. Existing birth defects programs, drawing on several data sources, can produce population-based NAS estimates.
• States differ widely, in line with maternal opioid use. Opioid use disorder documented at delivery in 2014 was 14.8 per 1,000 delivery hospitalizations in New Mexico and 48.6 in Vermont; Illinois had no data. Vermont has spent 15 years of perinatal quality improvement training birthing hospital staff to diagnose and treat NAS and building opioid agonist treatment capacity. States that look harder for mothers with opioid use disorder and infants with NAS may find more cases.
• Illinois's passive system may undercount. Its reporting system alone gave 2.2 per 1,000. Illinois's figure was lower than a 2012 discharge-data estimate for Illinois, Indiana, Michigan, Ohio and Wisconsin together (6.9) and similar to a 2013 Iowa estimate (2.2).
• Codes alone may slightly overcount. Earlier unconfirmed discharge-data estimates for New Mexico and Vermont were 8.5 and 33.3 per 1,000, a little above the confirmed figures here.
• Check codes before relying on them. A Tennessee Medicaid study reported higher PPVs, 91% for 779.5 among 950 possible cases in 2009–2011 and 98.2% for P96.1 among 217 in 2016, but it used a lower NAS score threshold (above 4, against above 8 here). Coding and case definitions vary across states, hospitals and providers, which may explain the spread.

Limitations

• Three states may not represent the rest of the country.
• Sensitivity could not be measured, because the true number of NAS cases in each state is unknown.
• The case definition required a newborn hospital stay of more than 2 days, so infants sent home sooner were missed. That is likely a small share: average stays for infants with NAS have been found to run 14.9 to 16.6 days.

Sources

• Lind JN, Ailes EC, Alter CC, et al. "Leveraging Existing Birth Defects Surveillance Infrastructure to Build Neonatal Abstinence Syndrome Surveillance Systems — Illinois, New Mexico, and Vermont, 2015–2016." MMWR 68(7). https://www.cdc.gov/mmwr/volumes/68/wr/mm6807a3.htm
• The report's figure is not reproduced, because the report was prepared jointly with the March of Dimes and state health departments.