Even with a causal effect, it’s important to consider the effectiveness as an intervention for a specific group (riders for this company).
There’s a reasonable causal story to tell wherein helmet-wearing leads people t take make riskier maneuvers on trips they’d be on, which leads to helmet wearing being associated with injury.
There’s another reasonable causal story wherein helmet-wearing leads people to chose to bike on trips they otherwise would take alternative transport, or skip entirely, which leads to helmet wearing being associated with injury.
In the first, a pedicab company might prohibit helmets to reduce risky behavior, but should really compare that intervention to just training and incentivizing lower-risk riding. In the second, the company is severing the connection between the intervention and the effect! The passengers tell riders where to go, so helmet wearing has no effect on risk-taking behavior _specifically for their riders_. Prohibiting helmets then dramatically increases risk/severity of injury.
I’d want to be very sure about the specifics before taking such a counterintuitive decision.