A multi-state Markov model is calibrated to Austrian data on recipients of long-term care payments the amount of which depends on defined frailty state levels. In contrast to a predecessor paper by one of the authors (see Fleischmann in Eur Actuar J 5(2):327–354, 2015), we are able to allow for different mortality intensities for different frailty states. A correction term is introduced in the mortality intensities’ functional representation to deal with observed mortality humps around the retirement age for certain frailty levels. Parameter calibration is done using MCMC methods (adaptive Metropolis–Hastings-within-Gibbs). The results reveal a considerably better fit of refined to raw prevalence units than the original model of Fleischmann (Eur Actuar J 5(2):327–354, 2015). Finally, the results are used to estimate the remaining healthy lifetime for certain ages, indicating slight but significant increases over the last 4 years.

多状态马尔可夫模型根据奥地利长期护理付款接受者的数据进行校准，其金额取决于定义的虚弱状态水平。与其中一位作者的前一篇论文（参见 Fleischmann 在 Eur Actuar J 5(2):327-354, 2015 中的论文）相比，我们能够考虑不同虚弱状态的不同死亡率强度。在死亡率强度的函数表示中引入了一个修正项，以处理在某些虚弱水平下在退休年龄附近观察到的死亡率驼峰。参数校准是使用 MCMC 方法（自适应 Metropolis-Hastings-within-Gibbs）完成的。结果表明，与 Fleischmann 的原始模型（Eur Actuar J 5(2):327–354, 2015）相比，原始流行单位对精炼的拟合要好得多。最后，