The evaluation covered primary health care delivery in three districts of Sri Lanka. The system had been heavily restructured in the years after the civil conflict ended. New facilities were built, staffing rose and supply chains improved. Donors had invested a great deal, and the Ministry of Health had spent considerably on process monitoring systems. When we began, the facility data looked reasonable.
In health system evaluation, facility data usually covers inputs and processes: staffing levels, availability of essential medicines, whether equipment works, waiting times as recorded by facility staff, patient volumes, and compliance with clinical protocols. All of these can be measured, verified and audited. Together they show only part of what a health system is doing.
The patient experience data told a different story. Exit interviews in all three districts showed the same pattern. Patients reported longer waits than facility records indicated, poor communication from providers about diagnosis and treatment, and low trust in referrals to secondary facilities. Patients referred upward often did not go. When we asked why, they talked mostly about the referral process itself, and distance and cost came up less often. They described being handed referral slips with no explanation of what the referral was for, what to expect at the facility they were sent to, or how to use its appointment system. Some had gone, waited, and come home without the service they had been referred for.
Health workers, interviewed separately, identified the gap themselves. They knew which metrics were reported upward (patient volumes, medicine availability, protocol adherence), and they knew these were not what decided whether patients found the health system trustworthy or effective. The system had been designed to meet its funders' reporting requirements. It answered to patients much less systematically.
Sri Lanka did not create this problem in health system evaluation and will not solve it alone. Most health system monitoring was built around supply-side indicators because they are easier to measure. Medicine stockouts and staffed facilities can be counted, and an inspector can check that a clinical protocol is written on the wall. Patient trust, the quality of a provider's explanation of a diagnosis, and whether a referral pathway works for the patient who has to follow it are much harder to count.
For evaluation design, this means an assessment from a single perspective will miss what determines patient experience, however carefully it is done. An evaluation of what a health system is doing needs, at a minimum, supply-side facility data and demand-side patient experience data together.
Patient experience data costs more to collect and is harder to standardise. Exit interviews need trained interviewers who can build rapport quickly. Sampling needs access to patients as they leave the facility, which some administrators resist. Patient-reported measures need careful validation, particularly after a conflict, when trust in institutions varies sharply from one community to the next.
In this evaluation, a sub-sample of health workers described one more problem. In some cases staff were working to the reported metric at the expense of the goal it was meant to stand for. Waiting time was recorded from registration to first contact with a health worker. Time spent waiting after that contact, for test results, for medicines or for a follow-up consultation, did not appear in the waiting time data. The recorded figures were correct for the interval they measured and said little about how long patients actually waited.
The WHO Service Availability and Readiness Assessment (SARA) methodology covered the supply side. Sri Lanka Ministry of Health annual health statistics and WHO Health Observatory data gave district-level comparisons. Patient experience measurement drew on the People-Centred Health Care framework literature.