Table 5 Rating criteria for the data quality characteristics of the evaluation framework for injury surveillance systems

Data completeness

Data completeness will refer to an assessment of the proportion of: (i) missing; (ii) 'not known'; (iii) 'other specified'; and (iv) 'unspecified' data recorded for key characteristics of the injured population (i.e. WHO's core minimum data set for injury surveillance).

I

There is no missing, not known, other specified or unspecified data and this is considered to be very high.

II-1

The proportion of missing, not known, other specified or unspecified data is less than 5% and this is considered to be high.

II-2

The proportion of missing, not known, other specified or unspecified data is less than 15% and this is considered to be high.

II-3

The proportion of missing, not known, other specified or unspecified data is less than 25% and this is considered to be high.

III

The proportion of missing, not known, other specified or unspecified data is in the range 26 to 50% and this is considered to be low.

IV

The proportion of missing, not known, other specified or unspecified data is in the range 51 to 100% and this is considered to be very low.

Sensitivity

Sensitivity will refer to the ability to correctly detect all cases of true injury events that the data collection intended to detect in the target population.

I

Sensitivity is in the range 90 to 100% and is considered to be very high.

II

Sensitivity is in the range 71 to 89% and is considered to be high.

III

Sensitivity is in the range 51 to 70% and is considered to be low.

IV

Sensitivity is less than 50% and is considered to be very low.

Specificity

Specificity will refer to the ability to correctly detect all non-injury cases that the data collection should not have detected as injury cases in the target population

I

Specificity is in the range 90 to 100% and is considered to be very high.

II

Specificity is in the range 71 to 89% and is considered to be high.

III

Specificity is in the range 51 to 70% and is considered to be low.

IV

Specificity is less than 50% and is considered to be very low.

Positive predictive value

The PPV will refer to the number of correctly identified true injury cases divided by the total number of cases that are identified (correctly and incorrectly) as an injury case from the target population.

I

PPV is in the range 90 to 100% and is considered to be very high.

II

PPV is in the range 71 to 89% and is considered to be high.

III

PPV is in the range 51 to 70% and is considered to be low.

IV

PPV is less than 50% and is considered to be very low.

Representative-ness

Representativeness will refer to the ability of the collection to provide an accurate representation of the distribution of key characteristics of the injured population (i.e. WHO's core minimum data set for injury surveillance) in a sample of the target population.

I

Appropriate statistical tests (e.g. Chi squared test, Fisher's Exact test) confirm there is no significant difference in the distribution of key characteristics of the injured population¹ between data in the surveillance system being evaluated to a gold standard (or other) data collection and the data is considered representative of the target population.

IV

Appropriate statistical tests confirm there is a significant difference in the distribution of key characteristics of the injured population¹ between data in the surveillance system being evaluated to a gold standard (or other) data collection and the data is not considered representative of the target population.