Analysis of the effect of adjustments on the forecast accuracy of antiretroviral medicines in Uganda
MetadataShow full item record
Background: Outputs from the quantifications are used to advocate and allocate funding, as well as develop supply plans with the aim of ensuring availability of medicines. However, the actual consumption is not always equal to the forecast (Akhlaghi et al, 2011) and thus forecasting teams often adjust further to improve the forecast accuracy. However, this could introduce bias (Davydenko & Fildes, 2013) affecting downstream processes. Objectives: The study aimed at determining forecast accuracy of the final forecast, the effect of adjustments, and the factors that affect the forecast accuracy of ARVs in Uganda. Methods: A descriptive study using both quantitative and qualitative data collection methods were conducted. The statistical forecasts, final forecasts, and actual consumption data from January 2016 to December 2019 was collected retrospectively from the Ministry of Health Pharmacy Department, and the Web based ordering and reporting system. Data cleaning and analysis was done using Microsoft excel. The mean average percentage error (MAPE) was used to measure forecast accuracy and student-t test was used for determining the significance of adjustments. Key informant interviews were held to determine the factors that affect the forecast accuracy. Results: The national final forecast MAPE ranged from 2.33% to 5.15% (Jan 2016 -Jan 2019) while the statistical MAPE ranged from 2.65% to 4.31% (Aug 2017 - Jan 2019). The MAPE for adult ARVs were generally lower than those of pediatric ARVs and a variation in forecast accuracy with one of the central warehouses (n=3) was noted. Adjustments generally lowered the MAPE for adult ARVs but increased the MAPE for most pediatric ARVs. Adjustments were significant for 53.8% (n=13, p=0.05) of the ARVs. Factors including dispensing practices, complexity of the quantification, adherence to treatment guidelines, data quality, product availability and quantification team attributes were identified to affect the forecast accuracy. Conclusion: The final and statistical forecast have a good accuracy and adjustments made improved the accuracy of adult ARV forecasts however decreased the accuracy for pediatric ARVs. Accuracy of the forecast is affected by the quantification complexity, product usage and availability and team attributes. Recommendations: More precautions are required when adjusting for pediatric ARVs. There is need to develop ARV use guides especially to aid manage coping habits in case of drug shortages. Forecast accuracy of new drugs and the degree of significance of the factors affecting forecasting should be further studied.