Ethical Data Practices for Sensitive Programs (Mental Health & GBV) in Uganda

Imagine a GBV survivor sitting across from an enumerator, answering questions about what happened to her not because the program needed that level of detail, but because no one thought carefully enough about what the form was actually asking. The data got collected. She left the interview visibly distressed. And the data never got used in any report.

This happens more than we like to admit. In mental health, GBV, child protection, and trauma care programs, the stakes of poor data practices are not abstract. They show up in real people's lives in re-traumatization, in breached confidentiality, and in eroded trust between communities and the organizations meant to serve them.

Data ethics is a wide field, and no single blog post covers all of it. But drawing from my work at Mouthpiece Initiative for Mental Health and Addiction (MIMHA) and previous child protection roles, here are some practices I've found essential when collecting sensitive data in Ugandan contexts.

Start by asking what you actually need.

When designing data collection tools, it is tempting to include questions that "might be useful at some point" or "would be interesting to know." In sensitive programs, that temptation needs to be resisted more firmly than anywhere else.

For every question on your form, ask, “How does this directly answer my research or monitoring question?” If the honest answer is "it doesn't, but it would give us useful context," that question should go. Burdening respondents with unnecessary questions is an ethical issue on its own, and in mental health or GBV contexts, every extra question about sensitive experiences carries a real cost.

Don't treat informed consent as a formality.

Informed consent is often reduced to a signature on a form that most respondents haven't fully understood. In Northern Uganda, many program participants have low literacy, have had negative experiences with institutions, and may agree to things out of politeness rather than genuine understanding.

A signature doesn't mean someone understood what they consented to. In contexts where respondents distrust data collectors, a complicated consent form can make things worse, not better. At MIMHA, we read consent aloud in Luo or Acholi for clients who prefer it, explain specifically who will see the data and in what form, and make it clear that refusing won't affect their access to services. We also revisit consent when we introduce new tools not as a bureaucratic requirement but because trust is built over time, not in a single session.

For programs collecting data from children, parental or guardian consent is non-negotiable, and the child's own assent still matters even when a parent has agreed.

Be thoughtful about how sensitive questions are asked and by whom.

The way a question is asked matters as much as what it asks. In GBV programs, asking a survivor to recount the details of what happened to her is clinical work done by trained caseworkers in a controlled setting. A MEAL officer running a monitoring survey is not in that setting. Monitoring can capture whether someone accessed services, whether they felt safe throughout the process, and whether a referral happened without requiring anyone to relive trauma.

Enumerators working in sensitive programs need training not just on how to ask questions, but on how to respond if a respondent becomes distressed. In one household survey I reviewed, there were no protocols at all for what an enumerator should do if a respondent began to cry or disclose something serious. That's a gap that can cause real harm.

Relatedly, consider who is doing the data collection. A respondent may feel more comfortable speaking to someone of the same gender or someone who shares their language and cultural background. Local enumerators who understand community dynamics often build rapport faster and get more honest responses. People can tell when someone understands their context and when they don't.

Think about power dynamics in the room.

Power dynamics shape data quality in ways that are easy to miss. If you are collecting data on service quality on behalf of the organization providing that service, respondents may hesitate to give negative feedback out of concern that it could affect their access to services even if you tell them it won't. In mental health programs, clients may feel that honest responses about their well-being could influence their treatment.

Thinking through these dynamics before data collection and designing tools and processes that reduce them leads to more honest, reliable data. Anonymous feedback mechanisms, peer data collectors, and community scorecards are all approaches worth considering depending on the context.

Prioritizing ethics in data collection is not separate from doing good MEAL work. It is the same thing. When people feel safe, they give you better information. When communities trust your processes, they will engage more openly. The shortcuts don't save time but move the cost somewhere harder to see.