Markovchart: an R package for cost-optimal patient monitoring and treatment using control charts

Control charts originate from industrial statistics, but are constantly seeing new areas of application, for example in health care (Thor et al. in BMJ Qual Saf 16(5):387–399, 2007. https://doi.org/10.1136/qshc.2006.022194 ; Suman and Prajapati in Int J Metrol Qual Eng, 2018. https://doi.org/10.1051/ijmqe/2018003 ). This paper is about the Markovchart package, an R implementation of generalised Markov chain-based control charts with health care applications in mind and with a focus on cost-effectiveness. The methods are based on Zempléni et al. (Appl Stoch Model Bus Ind 20(3):185–200, 2004. https://doi.org/10.1002/asmb.521 ), Dobi and Zempléni (Qual Reliab Eng Int 35(5):1379–1395, 2019a. https://doi.org/10.1002/qre.2518 , Ann Univ Sci Budapestinensis Rolando Eötvös Nomin Sect Comput 49:129–146, 2019b). The implemented ideas in the package were motivated by problems encountered by health care professionals and biostatisticians when assessing the effects and costs of different monitoring schemes and therapeutic regimens. However, the implemented generalisations may be useful in other (e.g., engineering) applications too, as they mainly revolve around the loosening of assumptions seen in traditional control chart theory. The Markovchart package is able to model processes with random shift sizes (i.e., the degradation of the patient’s health), random repair (i.e., treatment) and random time between samplings (i.e., visits) as well. The article highlights the flexibility of the methods through the modelling of different disease progression and treatment scenarios and also through an application on real-world data of diabetic patients.