NISRA’s administrative data research

The Northern Ireland local group of the RSS held an online meeting using MS Teams on Wednesday, 31 March, 2021, at 2pm. This was a double session from the Northern Ireland Statistical Research Agency (NISRA)  statistical staff that afternoon.

The speakers for the first talk were Deborah Lyness and John Hughes, Administrative Research Unit, NISRA.

Deborah Lyness introduced the Administrative Data Research Unit (ARU) and its activities. This was an ESRC-funded research collaboration between Queens University Belfast, Ulster University and NISRA. The core aim was to provide access to de-identified, routinely collected, administrative data in order to facilitate innovative research for the public good.  The activity was underpinned, legally, by the 2017 Digital Economy Act (DEA) and Article 6 of the General Data  Protection Regulations (GDPR). Deborah explained the five basic Safety Principles - safe projects, safe data, safe people, safe environment and safe outputs. The emphasis was firmly on the security of the data. 

Next she discussed how this framework had led to a number of wide-ranging collaborative projects. These  included two projects which were the subject of today’s talk: one on alcohol-specific mortality and the other on  drug-related mortality in the Northern Ireland population.

John Hughes presented the results of these studies. In relation to alcohol-specific mortality there was an  increasing trend in ASMRs (age-standardised mortality rate) per 100,000 pop. aged 15-74 yrs, higher in percentage terms in females than in males, although male ASMRs were higher overall over the period 2001-2019. The acceleration occurred post 2013 in both sexes. John moved on to consider the influence of 13 socio-demographic factors recorded in the 2011 census. Cases were followed until 2017. A Proportional Hazards (PH) survival model was used to assess their influence. Nine of these appeared in the final model and an  interpretation of the adjusted effects was given - most of the effects (age, sex, marital status/living  arrangements, and presence of limiting illness) had hazard ratios between 2 and 3. Not owning a car had a  hazard ratio of c3.5.

Next John introduced the drug-related study. Again the trends in ASMRs were increasing, higher in percentage  terms in males rather than females, with a noticeable upward trend from 2010 over the period 2001-2018  amongst the population aged 10-84 years. As before, the male ASMRs were uniformly higher over this period.  John analysed drug-related mortality in relation to the same 13 socio-demographic factors using the PH model.  All 13 factors appeared in the final model. The adjusted hazard ratios were highest (typically > 2) for young (<25 years), inactive, males, with poor mental health, living alone without a car.

These brief summaries do not fully encompass the depth of the analyses carried out, but suffice to whet the  appetite for more detail.

This was a very good talk, well-received by the audience of c40+ participants attending online. The speakers  were asked several questions. Deborah was asked about the ADR facilitating linking death data to the medical  records of patients (held by researchers) in a historical study. For example, to ascertain long-term survival. Not  much of the safe framework would seem to apply to this situation. She replied that the ADRU would look into  this matter. John was asked several questions about the studies. One participant pointed out that these data  were observational and hence wished to see crude and fully adjusted hazard ratios side by side in order to see  the extent to which one (eg the press) might be misled by the unadjusted data.

The meeting thanked the speakers in the usual way.

Photo shows Queens University Belfast.

Written by Gilbert MacKenzie, RSSNI secretary, on 11 April, 2021

Load more