Birth and Delivery · Labor and Birth

Predictive value: symptoms of preterm labor

Strickland DM

Published February 1991 American Journal of Obstetrics and Gynecology, 164(2), 700-701
DOI 10.1016/s0002-9378(11)80052-9 PMID 1992726
Read our synopsis

In response to Early signs and symptoms of preterm labor (Katz M et al., 1990)

RRM Academy Synopsis

Preterm labor symptom predictive value depends on population risk

This 1991 letter reworked numbers from an earlier preterm labor study. The letter added no new patient data. A symptom's predictive value changes with how common preterm labor already is. At an assumed 7% rate, backache and contractions linked most closely to real preterm labor.

Key Findings

  • At a 7% assumed incidence of preterm labor, positive predictive values ranged from 30% for vaginal discharge amount to 65% for constant backache.
  • Negative predictive values ranged from 94% to 98% across all eight symptoms at the 7% assumed incidence.
  • The original study's predictive value table was built on a population the authors chose to have a 50% incidence of preterm labor.
  • National preterm birth incidence in 1980 was 3.55% overall and 6.77% among Black women, both far below the 50% used in the original table.
  • The original authors acknowledged their retrospective, patient-recall design as a limitation that could bias which symptoms patients remembered reporting.

Interpretation

This is a critique of another study. The critique adds nothing new. The original article, Katz et al. 1990, built its table on a group set at 50% incidence of preterm labor. The 50% figure is far higher than the 3.55% national rate or the 6.77% rate among Black women cited in this letter. Strickland's letter redoes the numbers at a 7% rate. Predictive values shift when the rate changes. Both versions use the same old data, gathered by asking patients to recall symptoms after the fact, which the original authors said could bias results. Neither table shows what a symptom predicts using the general population's own preterm birth rate.

RRM Context

This 1991 exchange did not study restorative reproductive medicine. The exchange's main point still applies here. A symptom's predictive value depends on how common the condition already is in the group being tested. The point matters for reading any symptom-based or cycle-charting sign. A sign that predicts well in a high-risk group may predict poorly in a group with a lower rate of the same condition.

Abstract not indexed.

Topics

By this author

Related research

Birth and Delivery › Labor and Birth › Physiologic Birth · Pregnancy › Preterm Birth › Prediction · Research Methods › Measurement and Statistics › Statistical Methods
PMID 1992726 1992726 DOI 10.1016/s0002-9378(11)80052-9 10.1016/s0002-9378(11)80052-9 Strickland et al. 1991, Strickland 1991

Cite this article

Strickland, D. M. (1991). Predictive value: symptoms of preterm labor. American Journal of Obstetrics and Gynecology, 164(2), 700-701. https://doi.org/10.1016/s0002-9378(11)80052-9