Table 4 Addressing data quality challenges

Missing/incomplete reporting

Maina et al. proposed a method to use an adjustment factor known as the k-factor to correct for incomplete reporting by making an assumption on the number of people served by the facilities that did not report compared to the number of people served by facilities that reported the defined intervention to the local health administration. The k-factor is heavily influenced by the extent to which large health facilities as well as private sector facilities are reporting and engaged in the provision of service for the intervention of interest in the given location[5]. This method was used to address numerator challenges for preventive interventions.

The k-factor is often arbitrarily determined from HMIS officers and may not reflect the truth

Missing monthly report from a reporting facility

Assegaai et al. used defined rules to decide how to handle such data depending on the level of missingness [41]. If a facility missed more than 8 monthly reports for a given year, they were removed from estimation of coverage at the subnational level. However, if a facility missed less than 9 monthly reports, the average value of the months where data were reported was assigned to the months with missing data. The authors did not indicate how they arrived at the defined cut offs or provide any justification for the selected actions on missing data[41]. This method was used to address numerator challenges for preventive interventions.

Threshold for missingness and decisions on how to handle missingness is subjective making it difficult to compare estimates across different subnational units with different missingness threshold and decisions on how to handle such missingness.

Inconsistent reported numerators

Maina et al. defined outliers as reported numerators that are more than 2 standard deviations of the mean reported numerator for the multi-year period and such outliers were adjusted if there were no reasonable justifications. The authors did not state how the adjustment was done [5]. The authors did not indicate what was considered as a “reasonable justification” for outliers. Assegaai et al. handled inconsistencies in reported numerators differently. For district or county level RMNCH coverage estimation, if the reported value from a facility in a given month was extremely higher (> 3 times) or lower than the yearly average numerator reported for the given indicator by the facility, the outlier was replaced with the average yearly reported value of the selected indicator from the given facility. This methods were used to address numerator challenges for preventive interventions.

The definition of outliers and how to handle such data varies making it difficult to compare coverage overtime and between subnational units