What the Webuye 36 should teach us and what it reveals about our blind spots

@Rhoda Nakhosi This case study illustrates a critical breakdown between policy intent, data collection, and frontline service delivery, demonstrating that improved reporting alone does not guarantee better care or protection for adolescent girls. The reality that 36 pregnancies surfaced in a single school while Bungoma County recorded over 10,000 adolescent pregnancies in 2025 highlights how health and protection data typically capture isolated visibility events rather than the full, systemic scope of community harm. The near-total silence of parents, with only one official police report filed out of 36 cases, reveals how informal community negotiations, social stigma, and deep-seated fears bypass formal statutory child protection mechanisms. Furthermore, while Kenya’s 2026 Ministry of Education circulars guarantee a legal right to school re-entry, the reality that nearly 90 percent of teenage mothers never return proves that a policy exists in a vacuum without localized infrastructure, such as childcare, mental health support, and economic relief. Framing this crisis as a "school problem" requiring dormitories dangerously shifts the burden onto restricting girls' mobility rather than holding adult perpetrators accountable for statutory offenses. Ultimately, without redesigning monitoring instruments to directly trigger funded, multi-sectoral care, better data collection will only create a cleaner dashboard of systemic neglect while the girls themselves quietly disappear from the education system.

This case should be understood not simply as an episode of adolescent pregnancy within one school, but as a potential sentinel event revealing weaknesses across the adolescent health, education, safeguarding, and community protection systems.

From an epidemiological perspective, the first question is not only “How many girls became pregnant?” but “What population-level processes produced these pregnancies, who was exposed to risk, what protective factors failed, and what happened after identification?” A cluster of 36 pregnancies in one school warrants a structured, ethically conducted investigation rather than assumptions about causation. The analysis should establish the girls’ ages at conception and identification, school attendance patterns, distance and mode of travel to school, socioeconomic circumstances, access to adolescent-responsive health services, contraceptive and SRH information, relationships with partners, and any indicators of coercion, exploitation, abuse or other safeguarding concerns. These data should be collected with strict confidentiality and child-protection safeguards, particularly because some affected learners may be minors.

1. The denominator is as important as the headline number

The figure of 36 is highly visible, but epidemiologically it is only meaningful when placed against an appropriate denominator. We would want to know the number of female learners at risk, age distribution, person-time or school-year exposure where appropriate, and whether similar schools in the same sub-county demonstrate comparable patterns.

Likewise, the reported county figure should not automatically be interpreted as evidence that every school faces the same magnitude of risk. County-level aggregates can conceal substantial geographical, socioeconomic and age-specific heterogeneity. A useful surveillance system should therefore allow disaggregation by age, school, ward/sub-county, pregnancy outcome, school status, and referral/service outcome, while protecting individual identities.

This is where the event becomes particularly valuable for public-health learning: a highly visible school cluster may be a sentinel event rather than an isolated phenomenon.

2. The reporting gap is itself an epidemiological signal

The fact that only one parent reportedly made a formal police report should not automatically be interpreted as parental indifference or acceptance. Non-reporting can arise from multiple mechanisms: stigma, fear of retaliation, economic dependence, family pressure, lack of knowledge about reporting pathways, distrust of institutions, informal dispute resolution, or concerns about exposing the child to further trauma.

Consequently, surveillance data need to distinguish between absence of reporting and absence of harm. A case that does not enter the police or child-protection database may nevertheless appear in school, health-facility, community-health-worker or social-service records—or disappear from all formal systems.

This creates an important data linkage problem. If education, health, child protection and justice systems operate as separate silos, the same child can become simultaneously visible in one dataset and invisible in another.

3. Better reporting is necessary—but it is not the endpoint

I strongly agree with the concern that improved reporting does not automatically translate into improved outcomes.

A stronger monitoring framework should follow the entire continuum of care and protection, for example:

Identification → safeguarding assessment → clinical care → psychosocial support → protection/legal response where indicated → school re-entry → retention → completion.

For each stage, we should be asking:

• Was the learner identified?

• Was a safeguarding assessment completed?

• Was appropriate clinical care provided?

• Was psychosocial support offered?

• Was referral completed rather than merely issued?

• Was the learner able to return to school?

• Did she remain in school after re-entry?

• Did she complete the academic year?

• Was childcare or social support available where necessary?

• Were protection concerns appropriately investigated?

• What happened to learners who did not return?

These are substantially more informative indicators than pregnancy counts alone.

4. School re-entry should be measured as retention, not merely admission

A learner technically being allowed to return to school is not equivalent to successful educational reintegration.

A meaningful programme indicator would therefore move beyond:

“Number of pregnant learners who returned to school.”

It should include:

return within a defined period → attendance after return → retention at 3/6/12 months → grade progression → completion.

The system should also capture why learners do not return. Possible barriers include childcare responsibilities, financial constraints, stigma, family circumstances, health complications, lack of psychosocial support, school-level discrimination, transportation difficulties and fear of returning to peers.

This distinction matters because a policy can be formally correct while implementation remains incomplete.

5. The proposed dormitory needs to be evaluated as an intervention, not assumed to be a solution

A boarding facility may address some structural barriers—for example, long commuting distances, unsafe routes or excessive travel time. However, it should be evaluated against the specific mechanism through which it is expected to reduce risk.

If the causal pathway is:

long distance → unsafe journey → increased exposure → pregnancy risk,

then reducing travel distance may plausibly address one component.

But if important contributors include sexual exploitation, coercion, poverty, transactional relationships, inadequate access to SRH services, gender-based violence, weak safeguarding or community-level vulnerabilities, a dormitory alone cannot address those pathways.

Therefore, the intervention should be accompanied by a theory of change and measurable indicators, rather than being adopted simply because it is a visible infrastructure response.

6. Accountability must be evidence-based and child-centred

The statement that “these girls did not impregnate themselves” appropriately draws attention to the need to examine the role of adult men and other potential perpetrators. However, the response should avoid assuming that every pregnancy resulted from the same mechanism.

The appropriate public-health approach is to establish, through confidential safeguarding and investigation processes, who was involved, whether coercion or exploitation occurred, whether there were age-related legal concerns, and what protection response was required in each case.

Where offences are identified, accountability should follow the applicable child-protection and criminal-justice mechanisms. At the same time, the health and education response should not be delayed while investigations proceed.

The affected girls are first and foremost children and learners requiring protection and support, rather than merely cases in a pregnancy statistic.

7. We also need to examine the denominator of the “missing girls”

One of the most important questions is what happens after the headline event disappears.

If 36 pregnancies are identified but only a proportion subsequently appear in school-re-entry records, health-service records or child-protection follow-up, then we have a potentially important loss-to-follow-up problem.

A programme dashboard could therefore include:

• pregnancies identified;

• learners receiving initial assessment;

• safeguarding referrals;

• completed referrals;

• clinical care received;

• psychosocial support received;

• learners returning to school;

• learners retained at 3, 6 and 12 months;

• learners completing the academic year;

• learners lost to follow-up;

• documented reasons for non-return.

That would transform the data from a descriptive statistic into an accountability and programme-improvement instrument.

8. This should trigger a multi-sectoral investigation and response

The case sits at the intersection of several systems:

Education: attendance, re-entry, retention, school environment and safeguarding.

Health: adolescent-responsive SRH services, antenatal/postnatal care, HIV/STI services, mental-health and psychosocial support.

Child protection: identification, referral, protection planning and follow-up.

Justice/security: investigation of suspected offences where applicable.

Community systems: parents, community health promoters, chiefs, religious/community leaders and local organisations.

Social protection: poverty-related barriers, childcare and household vulnerability.

The response therefore should not be owned exclusively by the school or education department.

9. There is also an important surveillance-system question

This event provides an opportunity to ask whether Kenya's existing information systems can actually connect the relevant pieces of the story.

For example, can a programme identify that:

Learner A → pregnancy identified → health facility attended → safeguarding referral made → referral completed → school re-entry → retained after six months?

If the answer is no, then the problem is not simply inadequate data collection. It is a continuum-of-care information architecture problem.

However, any linkage system must use appropriate privacy protections. Personally identifiable information should not be casually merged across platforms merely for surveillance purposes. Unique identifiers, role-based access, minimum necessary data, secure referral mechanisms and clear governance arrangements would be essential.

10. Finally, we should avoid turning the girls into surveillance objects

Better data should ultimately improve the lives of the people represented by the data.

The most important question after collecting information is therefore:

“What action does this indicator trigger?”

If a pregnancy is recorded but no service follows, the indicator has limited practical value. If a safeguarding concern is identified but referral is not completed, the system has detected risk without providing protection. If school re-entry is recorded but the learner subsequently drops out, the programme may report success while the learner experiences failure.

For me, the central lesson from the Bakisa case is therefore this:

We should move from counting events to tracking pathways, from reporting outcomes to understanding causes, and from documenting referrals to measuring completed care and protection.

A genuinely responsive adolescent-health system should be capable of answering not only “How many pregnancies occurred?”, but also “Who was affected, what circumstances contributed, what protection was provided, what services were received, whether the learner returned and remained in school, and what happened next?”

That is where epidemiology, surveillance, implementation science and community protection intersect. The value of this case study will ultimately depend not on how prominently the 36 pregnancies are reported, but on whether the event produces better detection, stronger safeguarding, more responsive services, improved educational retention, and measurable prevention of recurrence—without placing additional blame or burden on the girls themselves.

Rhoda, thank you for framing this as a case study
rather than a headline — that distinction matters, and your four observations
(regional pattern, the single police report, the re-entry gap, the
dormitory-as-partial-answer) are exactly the right ones to sit with.

From peer-led prevention work, point 2 is the one
I'd push on hardest. The silence you're describing — one parent reporting out
of 36 — is not an absence of harm, it's evidence that communities have built
informal systems to absorb these cases before they ever reach a protection
system. In our own district-level work, we see the same pattern with addiction:
families and communities develop quiet coping mechanisms — shame management,
informal settlements, relocation — precisely because the formal system feels
distant, slow, or punitive. Those informal systems aren't failures of
awareness. They're rational adaptations to a protection system communities
don't trust to help them.

That reframes your question 3 for me: this isn't
only a "school problem" vs. "community protection failure"
binary. It's a trust problem. Until families believe reporting leads to protection
and accountability rather than exposure and stigma, the data will keep
disappearing into "other" — not because it isn't being generated, but
because no one is bringing it forward.

The dormitory addresses proximity. It won't touch
the trust deficit that's keeping 35 of 36 families silent.

Nonvicks, Rhoda, Dr. Magoba — this exchange on linkage vs. surveillance is
the most important turn in the thread. "Traceable protection, not
universal traceability" is a strong principle, but I want to offer a
different layer to the solution, alongside the data-architecture one.

The tension you've identified is that any system
traceable enough to coordinate care is also traceable enough to expose the
child. But the reason families avoid formal systems in the first place — what I
called the trust deficit earlier — is precisely that formal systems feel
state-facing, not community-facing. A protected unique identifier still sits
inside an institutional architecture families have learned to distrust.

In our peer-led recovery work, the referral
bridge that actually gets used is human, not technical: a trained community
peer who already has standing trust, who can walk a family toward formal
services without the family's first point of contact being an institution. That
person doesn't need — and shouldn't have — full case-file access. Their role is
narrower: recognize the need, accompany the family to the right door, and stay
present through the handoff.

Applied here, that suggests a hybrid worth testing at county level: pair Dr.
Magoba's minimum-data protected-identifier architecture with a trained
community-based referral layer — peer navigators who carry trust but not data —
so the "traceable protection" isn't only a technical safeguard but a
relational one the community already believes in before the system ever needs
to be legible to anyone.