Data from: Improving access to essential medicines via decision-aware machine learning

A critical challenge in healthcare systems in Low- and Middle-Income Countries (LMICs) is the efficient and equitable allocation of scarce resources, particularly essential medicines. This problem is complicated by limited high-quality data, which restricts the applicability of traditional data-driven techniques. We propose a novel decision-aware machine learning framework for essential medicines allocation, which additionally leverages multi-task learning to ensure sample efficiency and catalytic priors to ensure equitable allocation. In collaboration with the Sierra Leone national government, we performed a staggered, nationwide deployment of our system as a decision support tool and evaluated its impact using synthetic difference-in-differences. We find an estimated 19% increased consumption of allocated products in treated districts, demonstrating its efficacy at improving access to essential medicines. Our tool was subsequently scaled nationwide, covering an estimated 2 million women and children under five. Our work demonstrates how machine learning methods can improve efficiency at very low cost in resource-constrained global health settings. # Data from: Improving access to essential medicines via decision-aware machine learning [doi.org](doi.org) ## Description of the data and file structure * Data S1: list of facilities * Data S2: consumption data for evaluation - Data S3: supply data (added random noises to comply data privacy agreement) * Data S4: same as Data S2. Consumption data for evaluation but include control products - Data S5: population based demand for each facility across products ### Files and variables #### File: S1.csv **Description:** facility list ##### Variables * facility_type: categorizing facilities as Community Health Centre (CHC), Community Health Post (CHP), Maternal and Child Health Post (MCHP), or Clinic.  * hf_pk: facility unique ID * district: larger administrative regions, comprising a total of 16 districts #### File: S2.csv.zip **Description:** consumption data for evaluation.  **Variables** * **hf_pk**: unique facility ID * **name1**: product name * **date**: record date * **Inventory balance**: stockout, received, consumption, openBalance, closeBalance * **Facility information**: facility type, latitude (lat), longitude (long), district * **Quarter**: time period * **productID**: product ID * **Consumption statistics**: normAvg (average consumption by product), normStd (standard deviation of consumption by product) #### File: S3.csv **Description:** This data includes stock information across different time periods. Note that random noise was added to the data to comply with data privacy agreements. ##### Variables * Item: medication name with dosage and unit specification * Stock: stock quantity  * Quarter: time period #### File: S4_AlternativeData.csv.zip **Description:** Same as S2.csv.zip but include control products' information. #### File: S5_dfImp_popbased.csv **Description:** Population based demand for each facility across products, which is used to run population-based imputation **Variables** * **quarterID**: time period ID * **hf_pk**: unique facility ID * **productID**: product ID * **name1**: product name * **Q3AIalloc**: machine-learning-based allocation decision in Q3 * **popDQ3**: estimated demand proportional to population size in Q3 * **Q2AIalloc**: machine-learning-based allocation decision in Q2 * **popDQ2**: estimated demand proportional to population size in Q2 * **ExcelAlloc**: allocation decision based on the Excel tool * **popD**: estimated demand proportional to population size ## Code/software All data are in csv format. ## Access information Data was derived from the following sources: * DHIS2  * Grid3