Cost-effectiveness analyses of training: a manager's guide.

The evidence on the cost and cost-effectiveness of global training programs is sparse. This manager's guide to cost-effectiveness analysis (CEA) is for professionals who want to recognize and support high quality CEA. It focuses on CEA of training in the context of program implementation or rapid program expansion. Cost analysis provides cost per output and CEA provides cost per outcome. The distinction between these two analyses is essential for making good decisions about value. A hypothetical example of a cost analysis compares the cost per trainee of a computer-based anti-retroviral therapy (ART) training to a more intensive ART training. In a CEA of the same example, cost per trainee who met ART clinical performance standards is compared. The cost analysis is misleading when the effectiveness differs across trainings. Two additional hypothetical examples progress from simple to more complex costs and from a narrow to a broader scope: 1) CEA of the cost per ART patient with 95% adherence that compares the performance of doctors to counselors who attend additional training, and 2) CEA of the cost per infant HIV infection averted for a Prevention of Mother to Child Transmission program that compares the current program to one with additional training. To create an evidence base on CEA of training, more well-designed analyses and data on the cost of training are needed. Analysts should understand more about how capacity is built, how quality is improved within a health facility, and the costs associated with them. Considering the life of an investment in training, evaluations are needed on how many trainees apply the skills taught, how long trainees continue to apply them, and how long the content of the training conforms to national or international guidelines. Better data on effectiveness of training is also needed. It is feasible to measure effectiveness by clinical performance standards, or intermediate outcomes and coverage. Intermediate outcomes and coverage can also be combined with published estimates on health outcomes.