HPP Occasional Paper Series
Confidence intervals are not routinely estimated or published with national health accounting studies. However, the availability of these become more important as the results of NHA estimations are more widely used by policy-makers, and are used increasingly in cross-national analyses of health expenditure variations.

Errors in most NHA studies arise from similar origins. The impact of sampling errors on survey data are well known, but non-sampling errors are often not fully accounted for. While the literature on non-sampling error is less extensive than for sampling error, it is still possible to analyse sources of non-sampling error and to develop methods of approximating estimators of variance for resulting statistics.

It is recommended that health accountants should routinely estimate vacancies and confidence intervals in NHA estimates. It is feasible to do, despite the frequent limitations of the data sources used. Such an effort would not only assist users of the statistics, but would also help analysts’ better target efforts at quality improvement.

By Ravi P. Rannan-Eliya, Peter Berman, Eltigani E. Eltigani, Aparnaa Somanathan and Varuni Sumathiratne