Project Description
We looked at mental health services as an overall “response” to categorical data e.g. income, education, housing ownership rates etc. We used Queensland PHN (Public Health Networks) data for a proof of concept. A real world working system would use all PHN’s in Australia. To gain a more specific and accurate prediction of health service needs, we would look at local government areas e.g. SA2, SA3, SA4 within PHNs for a finer picture of needs, and ability to aggregate this data to PHN level for summaries and recommendations e.g. PHN funding, PHN mix of services etc. LGAs would be an ideal level of scrutiny to predict more responsively, the unforeseen or sudden need for demand in services, and therefore short and long term planning by relevant authorities.
The second stage was to analyse our chosen summary data to look for mental health trends to then be able (within assumptions) to predict surges in demand for health services based on any number of trigger events (e.g. unforeseen financial pressures from a variety of incidents). Which areas of a geographic region and population distribution variable are more likely to need support can then be predicted more accurately. We accessed 2011 data for mental health statistics - wherein 2 disasters at opposite ends of QLD occurred- Cyclone Yasi in North Queensland and the Brisbane floods. Any dataset year can potentially also correlate trigger events to mental health needs in that year (and eventually in a bigger system, other health needs) and any other events too e.g. financial and social pressure from closure of a major local industry (e.g. Nickel refinery at Townsville). Current issues with our proof of concept:Lack of consistency in metrics – age ranges in datasets; location based data (only PHNs have been consistently accessible); time based data for correlation purposes.There are some (big) assumptions to be made, but consistent data between states at a nationally mandate level would make the concept very usable. Sample assumptions:Even though we are not using comparable years, other stats (e.g. % low income) are assumed to be similar from year to year to allow basic correlations and predictions.We have assumed for the point of the exercise, some population stats can be used to identify a higher need for health (mental health in this case) services more than other measures