The day I leave Ampara on Sri Lanka's east coast, a wild elephant kills a woman and severely injures two others on the road near my house. This is the second fatal attack in town this year and, as before, the animal is rounded up and bundled back to the jungle in a truck. The incident seems to encapsulate something important about the nature of Sri Lanka: dark forces coiled beneath an appearance of calm. In the past month, for example, three security guards have been gunned down at hospitals in Ampara, Batticaloa and Sammanthurai. Yet the world of crisp nursing bonnets and clinical order remains intact throughout. No one knows who the killers were or how they chose their victims, but in this smoke and mirror conflict, rumours are fuelled of a final push by one side or the other. Then nothing happens, just more of the same, daily isolated encounters, as if it were in no one's interest to go for all-out war. Meanwhile the world's attention moves on to Lebanon.

I sit in the departure lounge at Bandaranaike International Airport wondering whether to buy another T-shirt with an elephant on it. In fact it's hard to find elephant-free souvenirs. I try to work out what good I’ve been to Sri Lanka during my 3-month stint of training doctors on the east coast. Could I have even done harm?

The plan, in keeping with current foreign aid theory (Reference Degnbol-Martinussen and Engberg-PedersenDegnbol-Martinussen & Engberg-Pedersen, 2003), had been to work collaboratively with the existing health system to increase its capacity to detect and treat people with severe mental illness. At face value things had been fine. Doctors at half a dozen new regular clinics had been intensively supervised in both assessing and treating patients, and in training their community staff to identify and bring in new patients; 30 further doctors had been given clinical and teaching skills-based workshop training; around 400 community staff were given training so that they could recognise and support those with severe mental illness; and nearly 150 people were engaged in supervised psychiatric treatment they would otherwise not have received.

But, could the aid have caused unintended harm? Or damage from friendly fire as Amartya Sen puts it (Reference SenSen, 2005)? Was there anything that should have been done differently?

Although I avoided the hit and run model of teaching characterised by context-free training, minimal skills development and lack of follow through, I only had 3 months before handing over to another trainer. Three months of fortnightly clinics with busy doctors who had many other duties wasn't enough to really consolidate their psychiatric skills or bed down the new primary care mental health system. Effective supervision relies on fostering good relationships and providing continuity and containment over a significant period. Something that was tricky to deliver in 3 months across a network of distant clinics, particularly since it takes a few weeks to get orientated and a few more to disengage when you leave. Not to mention the dodgy roads, super-cool bus drivers with a death wish and menacing road blocks (where you go real slow, turn off the music, remove the shades, switch on the interior light if it's dark and make sure everyone's hands are visible). Moreover, a third of clinics were cancelled or unsupervised, mainly because of security concerns. So, a trainer staying for a full 6 months would have been better, ideally with top-up visits during a 1- or 2-year follow-up period to support the programme's sustainability, which is something non-governmental organisations (NGOs) have been accused of sidelining (Reference Degnbol-Martinussen and Engberg-PedersenDegnbol-Martinussen & Engberg-Pedersen, 2003).

The other possible source of damage I worried about was the way free aid might let local services off the hook by sorting out problems for them, a pattern that could be taken for granted. Relying on NGOs for building projects, vehicles and training acts as a disincentive for government to sort these things out. Yet the dilemma is that since government resources are so stretched, these things may only happen if NGOs get involved. However, the trick that local health leaders appear to have learnt was impressive. They were certainly dependent on aid for the building of new hospitals and clinics, and the provision of many training programmes (to the point that some health staff felt burnt out by so much foreign aid training), but in adapting to this ‘ window of opportunity’ aid-rich environment, they had become pretty effective at getting NGOs to do what they wanted. So, in practice, a balance had been struck between the inevitably Westernised agenda of donors and NGOs and the strategic concerns of local health service leaders. Maybe that's as good as it can get: a potentially creative, although sometimes tense interface between outside agencies and local leaders producing what happens on the ground. All lubricated by the donors’ need to shift money and the NGOs’ need to spend it.

The worst case it seemed to me would be a strong yet poorly coordinated NGO community operating in a place where there were weak local leaders who lacked any sort of strategic vision. Perhaps in an ideal world local leaders in disaster-hit areas should be given the money to commission services from NGOs (rather than Western donors, often governments, being the commissioners). This would certainly maximise local empowerment and responsibility, although such power is rarely transferred voluntarily.

In the end I go for the elephant bookends and a coffee that doesn't taste of earth, and I think of the Sri Lankan doctors who are now unlikely to receive psychiatric training in the UK because of the new discriminatory work permit requirements. However, every cloud has a silver lining and maybe it's more effective taking trainers to the trainees, rather than trainees to the trainers. Both models involve cultural adjustment, but the great thing about exporting the trainer is that the trainees are supervised in the sociocultural context they will be working in. They may also be less tempted to emigrate and more inclined to continue working locally where they are certainly needed.

I glance at the papers. They contain the by now familiar pages of deliberations on the undeclared war, as well as brief reports of yesterday's killings (900 in the 3 months I’ve been here, half of them civilians), and news that 50 000 people have now been internally displaced by the disturbances so far this year. There is also a little piece on Sri Lanka's chances in the Miss Universe competition in Los Angeles. Miss Sri Lanka is the TV business affairs presenter Jackie Fernandez. A columnist argues that a Miss Foreign Exchange would have been more appropriate, in recognition of the apparel workers, tea pluckers and housemaids on whom the country depends. I remember the half a dozen women I’ve seen whose illnesses had been triggered by bad experiences in the Middle East or difficulties adjusting to returning home. A source of stress that featured in clinics almost as often as the tsunami.

My flight is announced. Well, that's it, time has run out. I have an image of my last visit to the Sea Breeze rooftop café. Presents exchanged, photographs taken, a feeling of guilt on my part for leaving so soon. Along the beach much of the rubble has now been cleared and steel rods are sticking out of newly poured concrete, marking out the foundations of a new school.