Intersecting birth weight-specific mortality curves: solving the riddle

Area of Science:

• Epidemiology
• Biostatistics
• Perinatal Health

Background:

• Small babies from high infant mortality populations often show better survival than those from low-risk populations.
• This counterintuitive survival pattern is frequently observed in perinatal and infant health studies.
• Existing explanations often attribute this to unmeasured confounding factors.

Purpose of the Study:

• To demonstrate how unmeasured confounding can explain the observed reversal of mortality risk among small infants.
• To model the impact of unmeasured confounders on birth weight-specific mortality curves.
• To re-evaluate the strength of the true mortality gradient by birth weight.

Main Methods:

• Utilized a previously developed model for birth weight-specific mortality.
• Introduced a simulated unmeasured confounder that decreases birth weight and increases mortality.
• Analyzed the resulting changes in mortality curves stratified by known risk factors.

Main Results:

• The model demonstrated that strong unmeasured confounders can cause mortality curves to intersect.
• The addition of a confounder produced a reversal of risk among small babies.
• Unmeasured confounders explain how high-risk babies within a birth weight stratum can have lower mortality.

Conclusions:

• Unmeasured confounding factors can fully explain the phenomenon of better survival in small babies from high-risk populations.
• These confounders can create the appearance of a reversed mortality gradient by birth weight.
• The true gradient of infant mortality with respect to birth weight may be weaker than currently observed.