sim_markov("actt1") generates synthetic daily health states for a
population like that of the
Adaptive COVID-19 Treatment Trial 1 (ACTT-1),
which compared Remdesivir with placebo in adults hospitalized with COVID-19
(Beigel et al., 2020). It uses the ACTT-2 transition model described in
actt2_preset, with the same eight states, and changes two things:
the patients' starting states follow the distribution reported for ACTT-1, which had more patients on ventilation than ACTT-2; and
the threshold for death is raised so that 18% of control patients have died by day 28.
The output has 28 rows per patient by default, and dead patients stay in state 8.
Details
Patients. Each patient's state on the day before follow-up (yprev) is
drawn from states 4 to 7 in proportion to the 138, 435, 193, and 285
ACTT-1 patients reported in these states at enrollment (13.1%, 41.4%,
18.4%, and 27.1% of the 1,051 patients with a reported state). In ACTT-2
these proportions were 13.7%, 54.6%, 20.9%, and 10.7%. Treatment is
assigned independently with probability treatment_prob: tx = 1 for
treated and tx = 0 for control patients.
Death threshold. With these starting states and the unchanged ACTT-2 model, 11.1% of control patients would die by day 28. The threshold for death, \(\alpha_8\), is therefore raised from -22.0693 to -21.4071. On each day, this multiplies every patient's odds of dying by \(\exp(0.6622) = 1.94\) and leaves their probability of being in state 7 or worse unchanged, so the additional deaths come from patients who would otherwise have needed invasive ventilation (state 7). The value was found by calculating the control arm's state probabilities exactly, multiplying the 28 daily transition matrices, and solving for 18% in state 8 on day 28. All other thresholds and coefficients are those of actt2_preset.
The resulting control-arm state probabilities, in percent, are
| Day | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 |
| 7 | 18.81 | 7.22 | 1.11 | 10.80 | 19.76 | 10.83 | 22.79 | 8.68 |
| 14 | 42.78 | 8.03 | 0.53 | 5.26 | 10.58 | 5.93 | 14.53 | 12.36 |
| 28 | 55.70 | 9.79 | 0.21 | 2.02 | 4.44 | 2.68 | 7.17 | 18.00 |
Treatment effect. The default treatment_effect of -0.0730 is the
coefficient fitted to the ACTT-2 data for Baricitinib, not an estimate from
ACTT-1. Choose the effect you want to study with treatment_effect; the
calibration to 18% applies to the control arm only.
Changing the model. The thresholds, from the best cutoff to death, are
-0.6465, -4.9314, -5.2230, -7.4169, -11.5321, -15.2262, and -21.4071.
Supply them reversed as intercepts in sim_markov() after changing one,
as shown in actt2_preset.
References
Beigel JH, Tomashek KM, Dodd LE, et al. (2020). "Remdesivir for the Treatment of Covid-19 - Final Report." New England Journal of Medicine, 383(19), 1813-1826. doi:10.1056/NEJMoa2007764
See also
sim_markov() to simulate, and actt2_preset for the transition
model.
Examples
actt1 <- sim_markov("actt1", n_patients = 20, seed = 123)
head(actt1)
#> id time y yprev tx
#> 1 1 1 5 5 1
#> 2 1 2 5 5 1
#> 3 1 3 5 5 1
#> 4 1 4 5 5 1
#> 5 1 5 5 5 1
#> 6 1 6 4 5 1