Direct obstetric deaths: those resulting from obstetric complications of

159 • aggregation of data over a longer period, e.g. to take the preceding two years’ data together with those for the current year and produce a “moving average” igure; • using the broadest of the age groupings recommended at 5.6.1 and 5.7.4. What applies for small national populations also holds true in general for subnational segments of larger populations. Investigations of health issues in population subgroups have to take into consideration the effect of the size of each of the subgroups on the type of analysis used. This need is generally recognized when dealing with sample surveys, but often overlooked when the investigation concerns the health problems of special groups in the national population. 5.13 ‘Empty cells’ and cells with low frequencies Whatever list of causes is being used, it may be found that no cases occur in certain cells of a statistical table. Where there are many empty lines in a table, it is worth considering the omission of such lines from a published table or from a computer printout. When only the occasional case of a disease occurs in a country, the line can be regularly omitted from the published table and a footnote added to indicate either that there were no cases or, when sporadic cases do occur, in which cell the case would have appeared. For cells with very low frequencies, especially those relating to diseases that would not be expected to occur, it is important to establish that the cases existed and did not result from a coding or processing error. This should be carried out as part of the general quality control of the data. 5.14 Recommendations 1. Responsibility for medical certiication of cause of death see Section 5.2 The medical certiication of the cause of death is normally the responsibility of the attending physician. In the case of deaths certiied by coroners or other legal authorities, the medical evidence supplied to the certiier should be stated on the certiicate in addition to any legal indings. 2. Form of medical certiicate of cause of death see sections 5.2, 4.1.3 and 4.3.1 The medical certiicate of cause of death should be in line with the international recommendation see Section 4.1.3. Collection of perinatal mortality statistics should be consistent with the recommendations presented in Section 4.3.1. 3. Conidentiality of medical information see Section 5.2 Administrative procedures should ensure the conidentiality of data from the death certiicate or other medical records.