Fig 1.
Main paper forms supporting inventory control system in Zambia.
(A) Stock Control Card completed for every product in each storage facility. (B) Request and Requisition Form transmitted upstream to calculate and communicate monthly replenishment orders. Populating data shown for illustration purposes only. Only some of the pre-printed product descriptions are shown.
Fig 2.
Estimated mean, 5th and 95th percentiles of combined weekly demand for all artemether-lumefantrine (AL) products at an average health clinic.
Statistics computed from stock card data of 17 clinics with at least nine months of consecutive demand data in stock card dataset, assuming no annual trend.
Fig 3.
Estimated mean, 5th and 95th percentiles of delivery lead-times between the central warehouse and an average health clinic.
Estimates based on record of lead-times to 212 clinics between May 2009 and June 2010 and surveys of subjective probabilities of shipment delays due to weather-related access problems, assuming no annual trend.
Fig 4.
Number of health clinics for which artemether-lumefantrine (AL) products stock data was obtained from stock card collection process, January 2009 –May 2010.
Fig 5.
Actual and simulated stock-out probabilities for artemether-lumefantrine (AL) products, observed through field stock cards (mean), and as predicted using simulation model (mean, 5th and 95th percentiles).
Actual stock-out probabilities estimated as the fraction of surveyed health facilities without stock. Mean simulated stock-out probabilities estimated as the fraction of 100,000 replications without stock. Plotted 5th and 95th percentiles correspond to a binomial distribution with success probability equal to the mean simulated stock-out probability and number of trials equal to the clinic sample size shown in Fig 4.
Fig 6.
Randomly selected simulation sample path of historical max-min inventory control policy.
Simulated data for policy 4 x I[-3,0] shown after warm-up period of one hundred years chosen to eliminate impact of initial conditions. Demand mean data from Fig 2.
Fig 7.
Simulated average service level, average and maximum inventory levels when monthly replenishment targets are calculated as different multiples of average monthly consumption.
Each simulation replication includes a warm-up period of two years and a data collection period of three years. Vertical whiskers around each data point show 95% confidence intervals around mean simulation estimates, but are barely visible because they are calculated from 100,000 simulation replications. Estimated baseline values are indicated with a vertical line.
Fig 8.
Simulated average service level, average and maximum inventory levels when monthly replenishment targets are calculated with different methods for estimating historical average monthly consumption and demand.
Each simulation replication includes a warm-up period of two years and a data collection period of three years. Vertical whiskers around each data point show 95% confidence intervals around mean simulation estimates, but are barely visible because they are calculated from 100,000 simulation replications. Estimated baseline values are indicated with a vertical line.