Scholarlink

<p>In Australia, residential aged care accounts for a significant share of government expenditure, with financial sustainability and changing population profiles emerging as critical concerns. This study utilizes a longitudinal dataset and develops a statistical simulation model to project the profiles of individuals entering residential aged care, the number of admissions, and the associated government spending. Our approach integrates ensemble learning methods, including eXtreme Gradient Boosting (XGBoost), with the Hamilton–Perry demographic projection method to provide precise predictions of residents’ length of stay. This integration enables more reliable forecasts of the future residential aged care landscape. The analysis was conducted at the state level to capture regional differences. We examined both demographic features, such as age, marital status, and gender, and medical features, including dementia status and care needs as measured by the Aged Care Funding Instrument (ACFI). We find that XGBoost outperforms previously used survival models for predicting residents’ length of stay. Our simulation model projects that both the number of residents and aggregate government expenditure in residential aged care will continue increasing until 2041. The model also provides groundwork for future research by offering micro-level insights into usage patterns and financial sustainability challenges.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Forecasting residential aged care expenditure with learning algorithm and demographic simulation

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月22日

查看全文
  • Hengzhe Zhao,
  • Jinhui Zhang,
  • Han Lin Shang,
  • Evelyn Lee

摘要

In Australia, residential aged care accounts for a significant share of government expenditure, with financial sustainability and changing population profiles emerging as critical concerns. This study utilizes a longitudinal dataset and develops a statistical simulation model to project the profiles of individuals entering residential aged care, the number of admissions, and the associated government spending. Our approach integrates ensemble learning methods, including eXtreme Gradient Boosting (XGBoost), with the Hamilton–Perry demographic projection method to provide precise predictions of residents’ length of stay. This integration enables more reliable forecasts of the future residential aged care landscape. The analysis was conducted at the state level to capture regional differences. We examined both demographic features, such as age, marital status, and gender, and medical features, including dementia status and care needs as measured by the Aged Care Funding Instrument (ACFI). We find that XGBoost outperforms previously used survival models for predicting residents’ length of stay. Our simulation model projects that both the number of residents and aggregate government expenditure in residential aged care will continue increasing until 2041. The model also provides groundwork for future research by offering micro-level insights into usage patterns and financial sustainability challenges.