Table 1.
Description of key variables, clinical classification, and outcome measures used in the model.
Fig 1.
Schematic of the first step in the algorithm—the creation and admission of a new patient (blue) to the simulated ward with a high probability of mortality (red), a moderate probability of mortality (orange and yellow), or low probability of mortality (green).
Fig 2.
Schematic of the second step of the algorithm—allocation of resources to patients as prescribed by their respective treatment plans, starting with the patients with the highest probability of mortality (red) and ending with patients with the lowest probability of mortality (green).
Fig 3.
Schematic of the third step in the algorithm—the probabilities of mortality for patients who were treated will decrease as per the efficacy of their assigned treatment plans, while the probabilities of mortality for patients that did not get treated due to insufficient resources will increase based on the deterioration rates of their respective complications (if any).
Fig 4.
Schematic of the fourth and final step in the algorithm, before the four steps are repeated for the next fifteen minute cycle—medications are restocked if necessary, and any doctors or nurses that were treating patients in the current cycle but are not required for treatment in the next cycle are returned to the staff pool to be re-allocated in the following cycle.
Table 2.
Variations in staff capacity and probability of patient admission at Mnazi Mmoja Hospital’s maternity ward over a 24 hour period.
Fig 5.
Comparison of case fatality rates for each complication outputted by the calibrated model and case fatality rates recorded by Herklots et al for the five complications incorporated in the model.
Box plots represent distribution of data over n = 50 identical simulations. Outliers have been omitted from the box plots for clarity. Predicted case fatality rate is depicted as the mean of each data set, inclusive of outliers.
Fig 6.
Comparison of complication incidence rates outputted by the calibrated model and complication incidence rates recorded by Herklots et al for the five complications incorporated in the model.
Box plots represent distribution of data over n = 50 identical simulations. Outliers have been omitted from the box plots for clarity. Predicted case fatality rate is depicted as the mean of each data set, inclusive of outliers.
Fig 7.
Impact of oxytocin inventory and supply frequency on maternal deaths over a 3 month period.
Box plots exclude outliers and whiskers reflect local maxima and minima among n = 50 runs for each data point. Line plots reflect mean values for each data point, including outliers.
Fig 8.
Impact of hydralazine inventory and supply frequency on maternal deaths over a 3-month period.
Box plots exclude outliers and whiskers reflect local maxima and minima among n = 50 runs for each data point. Line plots reflect mean values for each data point, including outliers.
Fig 9.
Predicted effect of staffing combinations for nurses and doctors when staffed consistently across morning, evening, and night shifts.
Data depicts mean maternal deaths for each staff combination over n = 50 identical simulations of over a simulated timeframe of 3 months with 20,000 IU of oxytocin and 2200 mg/mL of hydralazine being supplied each month.
Fig 10.
Effect of different staffing distributions during the morning, evening, and night on maternal mortality.
Inset depicts zoomed version of shift patterns excluding the 1 fewer doctor and nurse condition. All data sets were found to be statistically significant from one another, except for the current staff distribution compared to the 5 nurses for all shifts, with p < 0.05 using a two-tailed test.
Fig 11.
Predicted effect of patient influx rate on mortality rates for patients with complications.