A large dataset can estimate an association precisely while leaving important exposure details unknown. In a 2024 California pregnancy study, the key measurement was a delivery-record diagnosis code. It was not a standardized methamphetamine dose or a quantitative toxicology result.
What population was examined?
The retrospective cohort linked California vital statistics and hospital discharge records from 2008–2019. It included 4,775,463 singleton pregnancies at 23–42 weeks, of which 18,473 had the selected psychostimulant-use codes. The comparison group contained 4,756,990 pregnancies.
The authors used the label methamphetamine use for ICD-9 codes 304.4 and 305.7 and ICD-10 F15 diagnoses recorded at delivery. Those administrative categories do not provide quantity, route, exact timing or a distinction between episodic use and dependence. Recorded concurrent opioid, cocaine, alcohol or nicotine use was excluded, along with other eligibility restrictions.
What did severe maternal morbidity mean?
It was a composite defined by at least one of a set of serious conditions or procedures. Indicators included acute renal failure, eclampsia, sepsis, ventilation, hysterectomy and blood-product transfusion, among others. It was not a synonym for maternal death, and it did not require every listed complication to occur.
The composite occurred in 858/18,473 coded-exposure pregnancies, or 4.6%, versus 56,172/4,756,990 comparison pregnancies, or 1.2%. The adjusted risk ratio was 3.53, with a 95% confidence interval of 3.29–3.77.
How should the ratio and percentages be read together?
The percentages describe observed frequencies in these selected record groups. The adjusted ratio comes from a regression model accounting for maternal age, prepregnancy body mass index, education, insurance, parity, chronic hypertension and preexisting diabetes.
A ratio of 3.53 is not a 353% chance of morbidity. Nor can the adjusted ratio simply be multiplied by an unrelated patient’s baseline risk to produce a reliable personal forecast. The observed comparison frequency belongs to this cohort, and the adjusted estimate belongs to its model.
Why does the composite require care?
Different components can carry different clinical meanings and mechanisms. An association with the combined endpoint does not establish an identical increase for every component. It also does not identify which event a particular patient will experience.
This distinction matters when explaining research to a client. “More severe maternal morbidity” can be a faithful summary only if the reader understands that the study grouped several diagnoses and procedures together. Replacing that endpoint with death, permanent disability or one named complication would change the claim.
Does the large sample establish causation?
No. Coding may miss use, misclassify exposure or reflect differences in screening and clinical recognition. Excluding recorded polysubstance use does not prove that every remaining exposure was chemically isolated. The authors also identified unmeasured confounding and an inability to reconstruct dose or temporal patterns.
Large numbers reduce some sampling uncertainty; they do not recover information that the records never contained. The result supports a historical population association within the study’s eligibility rules, rather than an experimental demonstration that methamphetamine caused a particular complication.
Source
Prenatal methamphetamine use increases risk of adverse maternal and neonatal outcomes. American Journal of Obstetrics and Gynecology. 2024;231(3):356.e1–356.e15. DOI: 10.1016/j.ajog.2024.05.033. Full text; verified bibliographic record.