Table 1 Cox proportional hazard models.

Models also include generalized estimating equation cluster terms for individual and village. For each model, the number of individuals (n), number of medical observations (obs), and number of observed pregnancies (preg) are given. Dashes indicate variables not applicable for a given model or excluded by AIC. Details and additional excluded variables are given in tables S2 and S3.

Age of first pregnancy
(n = 425, obs = 639, preg = 87)
Time to next pregnancy
(n = 561, obs = 1623, preg = 405)
VariableExp(β)95% CIPExp(β)95% CIP
Age (decades)*1.00(0.80–1.25)0.992
Age4 (decades)*0.95(0.93–0.96)<0.001
Hookworm0.34(0.20–0.58)<0.0010.74(0.60–0.91)0.004
A. lumbricoides3.06(1.91–4.91)<0.0011.64(1.16–2.33)0.005
A. lumbricoides × age*0.68(0.51–0.89)0.006
Treatment with antihelminthic0.43(0.19-0.97)0.0420.75(0.58–0.97)0.027
Education (years)0.92(0.86–0.99)0.017
Speaks Spanish0.74(0.57–0.95)0.018
Distance to town (10 km)0.96(0.91–1.00)0.075
Season (P-spline)<0.001<0.001

*Age is centered at 20 years. Age was continuous to the nearest tenth of a year but is shown in decades to make the parameters more easily interpretable. Because age-related changes in fecundity are nonlinear, transformations ranging from age2 to age5 were compared by AIC to select the age transformation (age4) that best fit the data (fig. S3). †For the time to next pregnancy model, the roundworm parameter represents the hazard ratio at age 20.