Australian National Diabetes Information Audit & Benchmarking (ANDIAB) 2011 Final Report

Missing Data

A discussion on how missing data was taken into account when analysing the results

Page last updated: December 2011

The missing data were [where relevant] adjusted and reported conditionally [for example, pregnancy for female only]. Overall, data completeness varied considerably. In contrast to previous years where the majority of fields were over 80% complete, in 2011 a significant percentage of Fields were missing. Whilst indeed, some data items were almost 100% collected, overall Missing Data ranged from [n=24] 0.5% [Sex of Individual] to [n=2674] 57.8% [Retinal Camera], thence [n=2382] 51.5 % [Visual Acuity-Left] and [n=2381] 51.4% [Visual Acuity-Right]. There were 13.9% of the data items less than 20% missing {[0-5] 5.2% / [5-10] 3.5% / [10-15] 2.6% / [15-20] 2.6%]}, 27.8% were missing from 20-40% of records and 58.3% were missing from >40% of records [Table 24].

Several considerations need to be made when addressing these findings:

  • Firstly, a significant amount, [1300 of the 4629 individuals of the data provided], came from established in-house databases [ie Electronic data]. Some data fields not recorded in these databases have contributed to the increased amount of Missing Data overall.
  • Many multiple item fields were left empty when completing forms – eg – Management Method, where Diet Therapy was Yes, this would often result in the remaining options being left empty rather than being marked No. These must be counted as missing.
  • Where relevant data are reported as conditional [eg only recording a BP therapy option as missing in those on BP therapy], nevertheless, this cannot be done for all fields.
  • We have contributed to this in some ways, by application of the Validation Rules by:
  • increasing missing data in some fields, by removing incorrect / impossible results;
  • decreasing missing data in some fields by other data manipulations [eg the eGFR and eGFRresult fields - where in cases when an eGFR result was provided, but the eGFR Yes/No field was empty – we have changed this to Yes or No appropriately].
  • Finally also, as in previous years – Eye data items were poorly recorded (although less so than previously) and this remains an ongoing problem presumed to result from the fact that an eye data report may not be available in initial visit individuals [14.2% in 2011]. It is unlikely however that this information is not sought by Diabetes Centres and specialists, but most likely in fact that the information is unknown to them.
Taking these facts into account means that it is difficult to draw conclusions from a comparison of ‘ANDIAB Process performance’ to previous years.

Whilst the amount of Missing Data remains less than ideal, attempts to address missing ‘vital’ data fields [by seeking review by sites] were again successful in improving data completeness this year, as shown in Table 24a [section 2.5].

Document download

This publication is available as a downloadable document.

Australian National Diabetes Information Audit & Benchmarking (ANDIAB) 2011 Final Report(PDF 559 KB)