Time series analysis of consumption and short term forecasting of female contraceptives in the Kenyan public health sector
Background: Contraceptive security is crucial in ensuring access to family planning services and improving the contraceptive prevalence rate. This requires proper forecasting and procurement of contraceptives. It is therefore important to study consumption patterns and apply forecasting techniques so as to adjust for any changes in the choice of contraceptives over a given time period.
Objectives: The objectives of this study were to analyze trends, identify any seasonal or cyclic patterns in consumption of contraceptives, compare service point and consumption data, forecast consumption for six months and determine optimal models for forecasting contraceptives.
Methods: Data on consumption of implants, injectables, pills, and Intrauterine Contraceptive Devices (IUCDs) were extracted from the Kenya District Health Information System for the years 2014-2018. An exploratory analysis was done and the data decomposed to evaluate the trends and seasonal components. Service point and consumption data of contraceptives were compared. Short-term forecasting using the Autoregressive Integrated Moving Average (ARIMA) and the Exponential Smoothing (ES) models was done. The optimal model for forecasting was determined and the models validated using actual facility consumption data for 2018.
Results: The consumption of pills, injectables, and IUCDs declined while that of implants increased significantly across the 4 years. There were differences in the data reported for consumption and service point data for injectables, implants, and IUCDs. The ES models recorded the least error when forecasting consumption of all contraceptives except for one-rod implants in which the ARIMA model had the least errors.
Conclusion: There was a general shift towards the use of long-acting reversible methods especially implants in Kenya. The difference in the reporting of consumption and service point data for injectables, implant, and IUCDs showed gaps in the documentation and reporting of contraceptives. The ETS models were generally superior to the ARIMA models for predicting consumption of contraceptives.