Model birth spacing and fetal health

Short interpregnancy intervals are associated with elevated risk of low birth weight and preterm birth (Conde-Agudelo et al. 2006, JAMA; WHO recommends ≥24 mo between pregnancies). FPsim tracks interpregnancy interval; Starsim’s FetalHealth module tracks fetal growth and delivery timing per pregnancy. Wired together, they answer the question: how many low-birth-weight and preterm babies would a postpartum contraception intervention avert?

You will need FPsim and Starsim installed (FetalHealth lives in starsim.library.mnch).

Sim skeleton

FetalHealth goes into the sim’s custom slot; the interval-to-fetal-damage link goes into connectors; the postpartum intervention is the same one as in Integrated postpartum care, with offer_prep=False:

import starsim as ss
import starsim.library as ssl
import fpsim as fp

def make_sim(scenario, seed):
    intvs = []
    if scenario == 'pp_fp':
        intvs.append(PostnatalPackage(offer_fp=True, offer_prep=False))
    return fp.Sim(
        location='kenya', n_agents=20_000, start_year=2000, end_year=2035, rand_seed=seed,
        custom=ssl.mnch.FetalHealth(),
        connectors=[SpacingFetalPenalty()],
        interventions=intvs,
    )

The spacing connector

FetalHealth exposes two levers on any pregnancy: apply_growth_restriction (reduces birth weight) and apply_timing_shift (brings delivery forward, raising preterm risk). The spacing connector reads months since the mother’s most recent live birth at conception and applies a penalty when the interval is short:

import starsim as ss

class SpacingFetalPenalty(ss.Connector):
    def __init__(self, pars=None, name='spacing_fetal', **kwargs):
        super().__init__(name=name)
        self.define_pars(
            short_cutoff_mo=18.0,
            medium_cutoff_mo=24.0,
            growth_penalty_short=0.10,
            growth_penalty_medium=0.05,
            timing_shift_short=ss.lognorm_ex(mean=ss.weeks(2.0), std=ss.weeks(0.5)),
            timing_shift_medium=ss.lognorm_ex(mean=ss.weeks(1.0), std=ss.weeks(0.5)),
        )
        self.update_pars(pars, **kwargs)

    def init_pre(self, sim):
        super().init_pre(sim)
        self.fp = sim.get_module(ss.Pregnancy)
        self.fh = sim.custom['fetal_health']
        self.fh.add_conception_callback(self._on_conception)

    def _on_conception(self, uids):
        parous = uids[self.fp.ti_live_birth.notnan[uids]]
        if not len(parous):
            return
        months = self.sim.ti - self.fp.ti_live_birth[parous]
        short = parous[months < self.pars.short_cutoff_mo]
        medium = parous[(months >= self.pars.short_cutoff_mo) & (months < self.pars.medium_cutoff_mo)]

        if len(short):
            self.fh.apply_growth_restriction(short, self.pars.growth_penalty_short)
            self.fh.apply_timing_shift(short, self.pars.timing_shift_short.rvs(short))
        if len(medium):
            self.fh.apply_growth_restriction(medium, self.pars.growth_penalty_medium)
            self.fh.apply_timing_shift(medium, self.pars.timing_shift_medium.rvs(medium))

    def step(self):
        pass

The three-bin penalty structure (short / medium / long) is a stylised version of the spacing-outcomes literature. Effect sizes are chosen so short-interval pregnancies carry roughly the Conde-Agudelo 2006 odds ratios for LBW and preterm; they are illustrative, not tuned to any specific setting.

What to look at

• sim.results.fp.short_intervals – births at <24 mo interval.
• sim.results.fp.n_preterm – preterm births (from the base ss.Pregnancy).
• sim.results.fetal_health.n_lbw – low birth weight births.
• sim.results.fetal_health.n_svn – small vulnerable newborns (preterm | lbw | sga).

Compare the intervention arm against baseline scenario by scenario, seed by seed.

Interpreting the numbers

The LBW and preterm counts produced by this composition are the spacing-attributable subset – other drivers (maternal nutrition, infection, adolescent pregnancy) are not modelled here. A real total burden requires either additional connectors for those drivers or a calibration of the FetalHealth baselines against setting-specific data. Averted counts also include two effects: fewer conceptions overall, and better outcomes among the pregnancies that still occur. A rate-based decomposition (LBW-per-birth by arm) will separate them.

Full reference project

A Kenya version of this analysis lives at hiv_fp_kenya as demo_spacing.py (sim and grid) and spacing_fetal.py (connector). The full run, two-panel figure, and narrative writeup are in that repo.