IJPDS Semantic-based Privacy-preserving Record Linkage. IJPDS 2399-4908 Swansea University 10.23889/ijpds.v7i3.2006 7:03:231 Using linked administrative data to evaluate and improve the quality of end-of-life care in nursing homes. Webber Colleen 1 Milani Christina 1 Clarke Anna "1" Isenberg Sarina R 2 Downar James 4 Kobewka Daniel 1 Hsu Amy 3 Lau Jenny 4 Sinnarajah Aynharan 6 Simon Jessica 7 Boese Kaitlyn 8 Arya Amit 9 Hannon Breffni 3 Roberts Rhiannon 3 Turcotte Luke 3 Howard Michelle 12 Maxwell Colleen 3 Tanuseputro Peter 3 Ottawa Hospital Research Institute ICES Bruyere Research Institute University of Ottawa University of Toronto Lakeridge Health Alberta Health Services The Ottawa Hospital North York General Hospital University Health Network University of Waterloo McMaster University 092022 2022 7 3 2006 This work is licenced under a Creative Commons Attribution 4.0 International License. This article is available from the IJPDS website at: https://ijpds.org/article/view/2006 Objectives

Prescribing of symptom management medications may reflect the quality of end-of-life care provided to nursing home residents who are nearing death. The objective of this study was to examine variations in the prescribing of end-of-life symptom management medications in nursing home residents in the last 14 days of life.

 Approach

TREs provide secure facilities to analyse confidential personal data, with staff checking outputs for disclosure risk before publication. Artificial intelligence (AI) has high potential to improve the linking and analysis of population data, and TREs are well suited to supporting AI modelling. However, TRE governance focuses on classical statistical data analysis. The size and complexity of AI models presents significant challenges for the disclosure-checking process. Models may be susceptible to external hacking: complicated methods to reverse engineer the learning process to find out about the data used for training, with more potential to lead to re-identification than conventional statistical methods.

 Results

There were 55,029 deaths across 626 nursing homes. Overall, 64.8% of residents received at least one end-of-life symptom management medication. The proportion of dying residents who received ≥1 end-of-life medication ranged from 37.6% in quintile 1, 59.8% in quintile 2, 69.1% in quintile 3, 74.8% in quintile 4, and 82.9% in quintile 5. Opioids were the most commonly prescribed medications, with an average of 62.2% of residents receiving a prescription (35.9% to 81.2% across the quintiles). Nursing home residents that resided in homes in the lowest prescribing quintile were older and more likely to be Allophones (first language not English or French). Low prescribing homes were also larger, with a higher number of beds, and were more likely to be in rural areas.

 Conclusion

Records can be linked successfully, but estimated performance of record linkage depends on the validation set used. Manually reviewed data contain noise and may underestimate performance, while national ID numbers may overestimate performance due to non-random patterns of missingness.