vignettes/pseudo-code.md

In n8thangreen/ltbiScreenLite: ltbiScreenLite

title: "Pseudo-code for LTBI screening model" author: "Nathan Green" date: "2019-01-22" output: html_document: keep_md: yes vignette: | %\VignetteIndexEntry{Vignette Title} %\VignetteEncoding{UTF-8} %\VignetteEngine{knitr::rmarkdown}

Introduction

Describes a type of microsimulation and cost-effectiveness analysis using a synthetic cohort of recent immigrants to the UK. Much of the work is in setting up the cohort.

Pre-simulation setup

• Load in ETS/pre-screening dataset, denote by $X$
• Define simulation constants
• Number of Monte-Carlo iterations, $N$
• Age range for screening
• Year cohort
• Screen delay range, 5 years
• Time horizon (default life-time 100 years)
• Define cost-effectiveness distributions
• Willingness to pay
• Secondary infections
• Costs
• QALY
• Test performance
• Effectiveness
• Current year
• Create policy
• Define which subsets of cohort to target for screening
• WHO incidence
• All QALY?
• All costs?
• Treatment regimen
• LTBI test
• Define scenario
• Screening pathway probability distributions
• Cost distributions
• Sample screening year $~unif[0,5]$ for $X$
• Remove individuals from $X$ according to
• Simulation constants
• Screening year
• Join probability LTBI by incidence in country of origin and year
• Sample realisations
• Calculate time to events from dates
• Create probability incidence curve
• Append Sutherland and Lancet
• Generate TB progression times for LTBI individuals, consistent with other dates
• Join CFR for given age
• Sample case fatality realisations, $I_{cf}$
• Calculate QALYs for TB cases from progression date to death for
• Disease-free
• Cured
• Case fatality
• Status-quo i.e. either fatality or cured depending on $I_{cf}$
• Calculate future discounts at
• Time of notification
• Time of secondary infection (i.e. 1 year afterwards)
• Define decision tree object structures, denote $d_{health}$ and $d_{cost}$

Simulation

• Set policy $i = 1$
• Remove individuals from $X$ according to $f(X, i) = X'$
• Calculate mean screening delay from entry to screening and associated discount
• Calculate proportion in each incidence group from $X'$
• Get treatment regimen from policy, $t = treatment(i)$
• Get cost of treatment, $cost(t)$
• Get effectiveness of treatment, $eff(t)$
• Substitute cost and probabilities in to $d$
• Incidence groups
• LTBI status
• $eff(t)$
• $cost(t)$
• ~~GP incentives~~

Screening model

• Set scenario $j = 1$
• Assign branch value for $j$ to $d_{health}$ and $d_{cost}$
• For $d_{health}$ and $d_{cost}$ and each iteration $n = 1, \ldots, N$
• Sample
• Branch probabilities
• Costs
• Utilities/QALYs
• Calculate
• Total expected values for cost $c_1^s$ and QALYs $q_1^s$
• Subpopulation probabilities, $P$, including LTBI to cured

TB model

• Set scenario $j = 1$
• For each iteration $n = 1, \ldots, N$
• Sample TB treatment cost
• Calculate status-quo, with discounting
• Cost, $c_0^{tb}$
• QALYs, $q_0^{tb}$
• Get $p = P(\mbox{LTBI to cured})[n]$
• Replace first $p$ proportion of TB cases with disease-free individuals
• Calculate screened outputs, with discounting
• Cost, $c_1^{tb}$
• QALYs, $q_1^{tb}$
• Sum screening and TB costs and QALYs
• $c_1^s+c_1^{tb}$
• $q_1^s+q_1^{tb}$

n8thangreen/ltbiScreenLite documentation built on May 28, 2020, 9:37 p.m.