Table 1 Data quality dimensions assessed for deliveries, livebirths and maternal deaths reported in Kampala City*
Data quality metric | Definition† | Computation |
|---|---|---|
Dimension 1: Completeness of reporting | ||
i) Completeness of facility reporting | Number of (monthly) reports received from all health facilities divided by the total number of expected reports. | Overall, we obtained the average percentage of the reporting rates (completeness) of the facilities in the sample as extracted from DHIS2 for each year in the review period. Then, based on the reporting rates (completeness) extracted from DHIS2, we computed the number and percentage of facilities with reporting rates < 75% (i.e., < 9 reports per year). We grouped facilities into three categories of reporting completeness (< 75%, 75-99.9%, 100%). |
ii) Timeliness of facility reporting | The proportion of monthly reports received by the reporting deadline (7th day of the next month). | This was only computed for the whole sample. We obtained this metric as the average percentage of reporting rates (timeliness) as extracted from DHIS2 for each year under review. |
iii) Completeness of data element | The number and % of monthly values for each data element that are (1) not zero; and (2) not missing. | Overall, we computed this metric by adding up the number of monthly values for each data element that were zero from all facilities and dividing this by the total number of expected values. The completeness was then obtained by subtracting the percentage of zero values from 100%. Then, for each facility, we counted the total number of monthly values that were not zero and divided by 12. We then determined the percentage of facilities in which < 90% of the monthly values were non-zero values. We also categorised the facilities into three - <90%, 90-99.9%, 100% completeness of data elements. |
Dimension 2: Internal consistency of reported data | ||
iv) Outliers | The number and % of reported monthly values for the reference year that had an absolute modified z-score > 3.5. | Overall, we obtained the total number of monthly values that were outliers (modified z-score > 3.5) and divided it by the expected number of values. We then computed the number and percentage of health facilities in which at least one monthly value for each indicator was an outlier (modified z-score > 3.5). |
v) Consistency over time** | The average ratio of events for the reference year to the mean events of the three preceding years for selected indicators. | Overall, we determined this metric as the ratio of the total number of events for the reference year to the average number of events reported in the preceding three years for each indicator. We determined the percentage of health facilities with at least a 33% difference between their ratio and the city ratio for each indicator. For each facility, we first computed the average number of events reported in the preceding three years, then obtained the ratio of the total number of events for the reference year to the mean of the preceding three years. We then obtained the percentage difference between the facility and city ratios. Facilities that had a percentage difference > 33% (0.33) were considered to have issues in their reporting. |
vi) Accuracy of facility reporting¥ | The ratio of counted indicator values from facility records to reported values in DHIS2. | Counted number of events for each data element divided by the number of events reported in DHIS2 The ratio of the counted number to the reported number of events should be within ± 10% Counted number refers to the number counted from facility registers and reports; reported number is the number reported in the DHIS2 system |