Announcement of methodological changes to Statistics on Alcohol: England, 2012 National statistic

Following a public consultation on the Lifestyle compendia publications in 2011, there are a number of methodological issues to bring to the attention of users in relation to the 2012 report, due for publication on 31 May 2012. These relate to the affordability of alcohol index and estimates of alcohol related hospital admissions.

The Health and Social Care Information Centre (HSCIC) has routinely published a series of indices derived from Office for National Statistics (ONS) data in its Statistics on Alcohol: England reports. These include the alcohol price index (API), retail price index (RPI), relative alcohol price index (defined as API / RPI), real households’ disposal income (RHDI) and the affordability of alcohol index (defined as RHDI / relative price index).

Since the publication of Statistics on Alcohol: England 2010, the HSCIC has worked with key customers to investigate the scope for making methodological improvements to the way the affordability of alcohol index is derived. The Institute of Alcohol Studies (IAS) produced a research paper proposing a number of adjustments to the affordability index produced by the HSCIC.

One of these proposed adjustments was implemented in the Statistics on Alcohol: England, 2011 report and as a result, the revised Real Households’ Disposable Income (RHDI) index now tracks, exclusively, changes in real disposable income per capita. Previously, the RHDI index tracked changes in the total disposable income of all households and was not a per capita basis.

This had the implication that changes in the RHDI index over time were, in part, due to changes in the size of the population and not exclusively due to changes in real disposable income per capita. The RHDI index feeds into the affordability of alcohol index, and so this was also affected.

The adjustment was carried out using ONS mid-year population estimates of the adult population aged 18 and over, and was applied to all years in the index (1980 onwards). The adjusted RHDI index was then carried forward to produce an adjusted affordability of alcohol index. Further information on the method is contained in Appendix A of the 2011 report, and will also be included in Appendix A of the 2012 report. 

The unadjusted RHDI index and the unadjusted affordability of alcohol index (as used in Statistics on Alcohol: England 2010 and prior publications) are presented alongside the revised indices for comparability purposes in the 2011 report (Table 2.8 and Figure 2.6).

Further feedback received in relation to the method underpinning the alcohol affordability index is described in the outcomes paper. Feedback was received on the particular revision introduced in the 2011 report described above and also on the remaining proposed amendments contained within the IAS research paper.

All responses received were in favour of the adjustment made in the 2011 report to calculate on a per capita basis. We intend to continue performing this adjustment going forwards. In addition, further information has been included in the report to explain and describe the affordability index as well as how it should and should not be used.

Further work will be necessary in the future to consider whether further amendments to the index are required. This will include continuing to investigate whether an alternative source of data is more appropriate to measure disposable income, or whether to apply the adjustments proposed by IAS to the existing RHDI index used in the calculations. From initial discussions with ONS it appears the adjustments proposed by IAS to the existing RHDI index would be beneficial. An alternative source of data on household disposable income which may not have the limitations of the existing RHDI index was also identified.

We will aim to examine these alternatives more closely in the future, subject to required resources being available. Meantime however, it should be noted that, as in the 2011 report, the method underpinning the affordability index should be regarded as an interim measure that may be refined if deemed necessary upon completion of further enquiry. (Appendix F of the publication will provide further details of the responses received via the public user consultation.)

We received comments relating to both the method, and the presentation and interpretation of the alcohol related admission estimates. Whilst the estimates were recognised as having some merit, a respondent noted the complex nature of the underlying method and its limitations, in particular the methods underpinning the derivation of the alcohol attributable fractions (AAFs).

It was stated that ideally, the estimate would allow comparisons with other countries and the complex nature necessitates careful interpretation and presentation, including reference to the necessary limitations associated with the method.

We also received feedback that the estimates that display alcohol-related admissions based on all diagnoses (i.e. both primary and secondary) have been affected by changes over time in recording practices of secondary diagnoses. Ideally the impact of such factors should be established and explained in the commentary where possible.

We still consider the AAFs (developed by the North West Public Health Observatory (NWPHO) to provide the best estimates currently available. Additional information has been included in Chapter 4 of the publication to describe the method in detail. The measures (and uses) have also been described in greater detail and the benefits and limitations of each explored.

Within the 2012 publication, two main measures of alcohol related admissions are presented: a broad measure and a narrow measure. The broad measure is derived by summing the AAF associated with each admission, where the AAF is based on the diagnosis most strongly associated with alcohol out of all diagnoses (both primary and secondary). The narrow measure is constructed in a similar way but counts only the fraction associated with the diagnosis in the primary position. Within each of these measures, the data can be broken down into admissions that are wholly and partially attributable to alcohol, according to the required purpose. 

It is important to recognise that the AAF represents the likelihood that the condition is the result of alcohol consumption, rather than the likelihood that the admission is the result of alcohol consumption. In some of the cases where an admission episode contains an alcohol-related condition in a secondary diagnosis field but not the primary diagnosis field, the condition may not have been a causal factor leading to the admission. Rather, it may be a complicating factor and affect the care that is given to the patient. The estimates calculated based on the broad measure give an estimate of the number of admissions to hospital caused or affected by alcohol consumption at a particular time or place and hence the pressure put on the health system. Information based on the narrow measure provides a less complicated picture of trends in alcohol-related admissions over time, although give an incomplete picture of admissions resulting from or affected by alcohol consumption. This is because in some cases, the secondary diagnoses will in fact have been a contributing factor to the admission to hospital. This is particularly true of external causes of admission such as accidents and violence, which are never recorded as a primary diagnosis, but some of which can be attributable to alcohol.

To assist in the interpretation of trends in primary and secondary admissions, we have estimated the impact of changes in the recording of secondary diagnoses. We have produced estimates of what the admission numbers in previous years would have been had the current level of recording of secondary diagnoses existed in those years. These figures will be provided in Chapter 4 and a detailed technical Appendix describing the analyses in detail will be provided in Appendix G.

A review of the methods used to estimate alcohol related admissions is taking place in the form of a public consultation led by the NWPHO working with the Department of Health and the HSCIC. The consultation is expected to be launched on 31 May 2012, and will run for 12 weeks. Full details, including how to submit comments, can be found on the NWPHO website. 

A detailed description of the method (and assumptions) developed to quantify the effect of changes in recording practices on secondary diagnoses on national alcohol related admission estimates will be included in Appendix G of the 2012 report. A brief description of the method is below.

All admission episodes will have a primary diagnosis code and may have additional diagnosis codes recorded in up to 19 secondary positions. The overall percentage of episodes (whether including alcohol attributable diagnoses or not) with secondary diagnosis codes recorded in any or all of the secondary diagnosis positions has increased over recent years. This has resulted in a component of the increase in the broad measure being due simply to increased coding. The effect of this is estimated and adjusted for by:

• calculating the overall increase in coding in each secondary position over the years
• modelling the broad measure in earlier years assuming that 2010/11 secondary diagnosis recording conditions applied in those earlier years, by applying various assumptions
• using these modelled values to examine the trends in the broad measure over time