Alex’ description is a bit short. Firstly, yes, the predictors are intended to be used to derive thresholds that can be compared to actual events. Some examples:
for a disease like Cholera, the threshold will normally be 1 (confirmed) case, so no calculated/predicted threshold is required. 1 confirmed case = outbreak.
for a disease like Typhoid Fever, the threshold will higher but dependent on area history. WHO generally use the previous 5 years to determine what is the “normal” number of cases in an area (e.g. the US will have 400 cases of typhoid fever reported per year). T
hat “normal” number is what the DHIS2 predictors will provide, and you can then establish a validation rule saying that if actual number of reported cases in e.g. one week is more than the normal number per week multiplied with factor x, then you flag it as a possible outbreak.
- for diseases that are commonly seasonal like Malaria, the predictor values vary through the year. 100 cases during summer-time is business as usual, 100 cases in winter-time is all alarm-bells ringing.
Secondly, predictor values can be used for on-the-fly calculations and comparisons - where predictor values in theory might change daily due to new data being incorporate into the calculations - or they can be based on historical data only, persisted and regarded as stable for let us say one year. That debate is still on-going, but my guess is we will end up with IDSR systems generally using predictor values with some stability.
Thirdly, predictor methodology will increasingly not only rely on historical data but also predicted environmental factors: So in the case of malaria, if global weather patterns indicate way above normal rains, you would up your predictor values accordingly. From another perspective, this could be regarded as a basic early warning system (note: most useful early warning system would require more sophisticated modeling than such predictor+)
Fourthly, and this goes beyond disease surveillance: predictors can be used for short term forecasting of performance. We did stuff like that in Cape Town already 10-15 years ago: you predict total performance for the financial year by predicting (= extra-polating) from the performance during e.g. the first six months. The idea being that under-performing areas are identified early and in time for mitigating efforts to be made. As should be obvious, you can use the same predictor value calculations to assist with setting short-term and medium-term TARGETS.
Fifthly, predictors might assist with making sense of historical data collected with diverse frequencies and methodologies: Routine data, sample surveys, census, household surveys, case-based data, etc. Predictor values can thus be a result of triangulating multiple disparate data sources.
Finally, and now we are into scenario modelling: Predictors can be used as one component of future “what-if” scenarios, where you have a sophisticated 5-10-20 years model with inputs from demography, disease patterns, climate, human/economic development, global trade patterns, human resources, drugs, technology, and so forth. The DHIS is nowhere near that territory yet, but it is likely to become a focus area in some years.
The main impediment to further development of these predictor values right now seems to be that few if anybody has used them to any extent (yet) - that will hopefully be rectified during 2017 with more widespread implementation of IDSR on the DHIS2 platform.
On 26 December 2016 at 13:15, Alex Tumwesigye firstname.lastname@example.org wrote:
Predictors replaced surveillance rules in data quality app. They have exactly the same functionality as surveillance rules. They are still under development since they are meant to be persisted and uniquely identified. They are used in the idsr for threshold calculations.
On Monday, December 26, 2016, Archana Chillala email@example.com wrote:
We are using latest revision of DHIS 2.25. We find a new metadata entity called “Predictors” in the ‘others’ module of maintenance app. We have also gone through the documentation about Predictors and it tells what it is and how we can configure it, explaining all its properties. But we are not clear as to where they can be used or where could we look for its predicted values on the UI and which particular app to look at for the same. Could you please let us know, how and where can the predictors functionality be leveraged?
While testing 2.25 version, we observed that, we’re able to save a category without assigning any category options to it. We’re able to save a dataset without assigning any data elments to it. So is the case with most of the metadata entities in maintenance app. This wasn’t the behavior in earlier versions of DHIS like 2.20, 2.21. Not sure, but probably, it’s been like this ever since maintenance app was introduced. Is this the expected behavior? Or is it a bug?
Technical Advisor - DHIS2 (Consultant),
Ministry of Health/AFENET | HISP Uganda
+256 774149 775, + 256 759 800161
Skype ID: talexie
IT Consultant (Servers, Networks and Security, Health Information Systems - DHIS2, Disease Outbreak & Surveillance Systems) & Solar Consultant
"I don’t want to be anything other than what I have been - one tree hill "
Mailing list: https://launchpad.net/~dhis2-devs
Post to : firstname.lastname@example.org
Unsubscribe : https://launchpad.net/~dhis2-devs
More help : https://help.launchpad.net/ListHelp
46D Alma Road, 7700 Rosebank, SOUTH AFRICA
Tel/fax (home): +27-21-685-6472
Iridium SatPhone: +8816-315-19119