Tobacco Use During Pregnancy

Limitaciones

OASIS's percent of births to mothers who reported tobacco use in pregnancy, out of births of known status (unpublished), not all births. The certificate changed in mid-2007: compare from 2008 (payer rows start then), race groups from 2010 (Georgia) or 2013 (counties). Percent and count are NULL where over 20% of a row's births have unknown status, as in Catoosa, Dade and Walker before 2008 and in 2012; Hart and Charlton before 2008; McDuffie, Richmond and Columbia in 2019 (up to 61%; our mask). Live births to mothers who live in Georgia, placed by the mother's county of residence (not where the birth happened); OASIS counts births, not pregnancies, so twins are two births. Each percent divides by the group's own births of known tobacco status: a mother's age group, race group or payer by its births, not by all births in the place. demographic, mother_age and payer are never crossed: a row breaks down by one of them, the others at 'all' (OASIS's requests never cross them). OASIS publishes every count, however small, and withholds the percent on 1-4 births with the measure (NULL here); the platform adds no small-cell suppression of its own (its masks, below, are quality masks) and never recomputes a withheld percent. 31% of rows with births have no percent (withheld on 1-4 births with reported use, or NULLed by our masks), mostly small counties and groups. Where a group had no births in a place (num_births 0) the percent is NULL: OASIS shows 0, but a percent of no births is undefined (our mask; the counts stay 0). That covers 16,948 rows, mostly small race and age groups and payers. OASIS leaves births of unknown tobacco status out of the percent's denominator in every year but 2009 (proved on every row: in 2009 the percent equals num_births_tobacco_use / num_births exactly) and publishes no count of them. By our calculation (births minus count / percent) Georgia left out 1,115-2,559 births a year in 2000-2007 (0.79%-1.92% of its births) and 160-1,665 from 2008 (0.12%-1.31%), so Georgia's percent runs at most 0.16 percentage points above num_births_tobacco_use / num_births. In counties the gap was wide: in 40 county-years with 100+ births (8 counties) 20%-61% of births have unknown status (Catoosa, Dade and Walker in 2000-2007 and 2012; Hart in 2000-2003; Charlton in 2000-2003 and 2006-2007; McDuffie, Richmond and Columbia in 2019), and their percents and counts are NULL here (our 20% mask, below). In Catoosa in 2004, 496 of 808 births had unknown status, and OASIS's 26.3% compares with 82 / 808 = 10.1%; OASIS's percent there ran up to 22.6 points above the counts' ratio (Dade 2007). There OASIS's percent ran high, not just loose: on the same certificate, births of unknown status jump in 2012 (Catoosa, Dade, Walker) and 2019 (McDuffie, Richmond, Columbia) while the births with reported use hold, so OASIS's percent is 1.2-1.7 times the mean of the years either side while count / births stays close to theirs (Catoosa: 18.8%, 31.7% and 18.3% in 2011-2013, with 123, 138 and 135 births with reported use). So births of unknown status there are mostly births with no reported use (our reading), which is why those rows are masked rather than served. Our reading: records of births in out-of-state hospitals (the Tennessee, South Carolina and Florida borders) before 2008 and, on the Tennessee border, again in 2012, and a reporting gap in the Augusta area in 2019. Where more than 20% of a row's births have an unknown value for the measure (OASIS leaves them out of its denominator), the percent and its count are NULL (our mask: OASIS's own 20% threshold, which it applies only to whole years); num_births is kept. That covers 236 rows (42 county totals); 25 of them are 0% or withheld rows (6 at 0% and 19 withheld), whose percent implies no denominator but whose births of known status OASIS's other rows bound: at most the place's known-status births (count / percent of its total row) less the breakdown's other groups'. A 0% or a withheld percent implies no denominator, so it can't be tested unless OASIS's other rows bound it; unbounded, it stays as published. OASIS shows 0 where it has no births to divide by, so a group whose births all had unknown status probably shows 0% too (our reading): such a 0 cannot be told apart from no reported use unless OASIS's other rows bound the group's births of known status. Where they prove over 20% of its births unknown, the percent and count are NULL (the 20% mask above: 6 0% and 19 withheld rows, Georgia's unknown payer of 2016 among them); no 0% is provably 0 of 0. The other 0s are served as published, most likely in small groups of the counties above. The certificate break: Georgia adopted the revised (2003) U.S. standard birth certificate in mid-2007 (2008 is its first full year, as for the payer). The 1989 certificate asked whether the mother used tobacco during pregnancy; the 2003 one asks the cigarettes smoked in the three months before pregnancy and in each trimester, and NCHS treats the two as not comparable. Georgia's percent steps from 7.40% in 2006 and 7.05% in 2007 to 6.32% in 2008 (6.51% in 2009), and births of unknown status fall from 1.50% to 0.21% of Georgia's births. OASIS does not document how it maps the new item to tobacco use (our reading: smoking in any trimester, as NCHS counts it). The step is small beside the decline from 8.35% in 2000 to 1.30% in 2025, so every year is served; compare levels from 2008. The race groups' series break in 2007-2008, which looks like a change in how the mother's race was recorded (our reading; OASIS documents none): multiracial births rose from 648 in 2006 to 3,538 in 2007 and 6,961 in 2008, then held between 2,487 and 3,499 a year from 2010, so multiracial figures of 2007-2009 are not comparable with other years; white births fell from 66,940 to 57,568 and native_american births from 215 to 91 in 2008. race_unknown (unknown ethnicity, and in 2008-2016 also births recorded with no race) peaks at 5.1% of Georgia's births in 2009 (0.6% in 2025), so the race groups run that much low. Counties break harder and for longer: in 84 county-years of 2000-2012 (none since; 17 counties) more than 25% of births have no recorded race or ethnicity, up to 78%, most often in Catoosa, Dade and Walker (the Tennessee border) and in Tift and nearby counties from 2007, so OASIS's race groups for those county-years run far low (Tift: 414 of its births in 2011 had no recorded race or ethnicity, and OASIS counts its black births as 16 in 2011 and 199 in 2015). Hispanic births spiked in Sumter, Walton and Macon in 2008-2009 (Sumter: 92, 336, 290 and 56 in 2007-2010). Compare a county's race groups over time from 2013 only (its race_unknown share has been at most 17% since). Those county-years, and the others with more than 20% of births of unknown race, have their race groups NULL here (our mask), Tift's 2011 black births among them. Where the mother's race or ethnicity is unknown for more than 20% of a place-year's births (race_unknown / all num_births, the births topic's rule on the same births), every race group's percent, count and num_births for that place and year are NULL (our mask, OASIS's own 20% threshold for missing items): each group's figures would rest only on the births whose race was recorded. The all, mother's age, payer and race_unknown rows are kept, so the unknown share stays visible. That covers 97 county-years (2000-2012, 19 counties; never Georgia), 679 race-group rows. Race groups are the mother's race and ethnicity in the platform convention: hispanic is Hispanic of any race; white, black, asian, native_american (American Indian or Alaska Native), pacific_islander (Native Hawaiian or Other Pacific Islander) and multiracial are non-Hispanic. race_unknown is our calculation, the total minus those seven (unknown ethnicity, or non-Hispanic of unknown race), so the race groups' counts add up to all where every one is published (our masks NULL some, below); it has no percent: OASIS publishes one only for its unknown-ethnicity cell, which equals race_unknown except in 2008-2016 (when race_unknown also holds non-Hispanic births of unknown race), so none is served, keeping one definition. mother_age is the mother's age at the birth in OASIS's groups (10_14, 15_17, 18_19, then five-year groups to 40_44, and 45_55); the groups' counts add up to all where every one is published (our 20%-missing mask NULLs some, below). Payer rows start in 2008, as in births_by_payer: before the revised birth certificate (mid-2007, per OASIS) the payer was recorded only as medicaid or unknown (2000-2004) or not at all (2005-2006), and 2007 is the transition. medicare rows exist only in 2013 and 2014 and peachcare rows only in 2013, 2014 and 2016, the years either has births (the revised certificate has no box for either, our reading); in other years no birth could carry them, so those rows are not served. Small groups give extreme percents, preserved as published: the highest, 56% (Jasper 2003, mother_age 15_17), is 5 of the 9 births of known status among 9 births. In 2025, counties with 1,000+ births range from 0.15% (Cobb) to 4.26% (Bibb). 2025, the newest year, was added to OASIS on 2026-04-30; OASIS marks no year provisional. Counties add up to Georgia, and the race groups (with race_unknown), mother's ages and payers to all, in both counts, wherever every count is published. Our masks NULL some (the 20%-missing mask 236 counts, 42 of them county totals, and 43 race_unknown counts; the race mask 679 race-group counts and their num_births in 97 county-years), so a sum across counties or groups falls short where they apply; never rebuild a masked count by subtraction. Not served: OASIS's SES-vulnerability rows (their percents are valid, each dividing by the SES group's own births of known status, but the platform has no SES group yet) and the mother's education. tract_geoid is always NULL.