Using aggregated general practice data to evaluate primary care interventions

Aggregated data extracted from computerised general practice records should be used to improve outcomes at patient, health system and population levels

A report released in 2016 by the Primary Health Care Advisory Group (PHCAG), Better outcomes for people with chronic and complex health conditions, highlights the need to use aggregated general practice data to target health resources and interventions.1 The aim of any health program should be to improve outcomes at patient, health system and population levels. These outcomes should be measurable and part of a feedback loop to improve patient care.

To date, much of the data on general practitioner clinical activity has come from surveys such as the Bettering the Evaluation and Care of Health (BEACH) program.2 Following the cessation of data collection by this program in April 2016, there is a need to strategically invest in future data collection systems.3 With the vast majority of Australia’s 32 000 GPs using computers, electronic medical records held in general practice provide a potentially rich data source on the 85% of the Australian population who visit a GP at least once per year. The recommendation to establish a national minimum dataset by the PHCAG supported by a data collection model using a national data warehouse would be a major step toward addressing current data gaps. Drawing on existing resources such as the relational database developed by the BEACH program could expedite this process.2

Need for a coordinated approach to electronic data held in primary care

Key to the success of a sustainable ongoing data collection is the ability to extract information from existing primary care medical record systems, rather than requiring busy practitioners to collect additional data. The United Kingdom research system QResearch (http://www.qresearch.org) is an example of how this can be done by using a centralised data extraction system to collect information. The system contains data from about 1000 general practices with historical records extending back to the early 1990s. There are examples of some similar smaller scale Australian initiatives4,5 supported by academic general practice units that could guide in the development of larger systems.

The use of routinely collected general practice data is problematic in Australia because of the large variation in recording practices among GPs. There are more than ten general practice medical record software packages in use across Australia, although two dominate the market.6 The drivers for GPs to record electronic data for chronic disease management of patients are complex. They include criteria to meet funding arrangements for care,7 doctor computer skills, fitness for purpose of data (particularly around diagnostic data), coding issues, consent, confidentiality, and the relationship of data with decision support and guidelines.8 The completeness of clinical records appears to have improved since the early 2000s,9 although some concerns remain regarding the quality of data extracted.5 Consequently, there is a need to implement strategies to ensure data quality if the full potential of clinical data extraction is to be realised. Fundamental to the process of data collection is GP engagement and, as such, ongoing practical support of general practices is essential.10

The 2016 MJA Supplement “Value co-creation: a methodology to drive primary health care reform” (https://www.mja.com.au/journal/2016/204/7/supplement) discusses a value co-creation approach to deliver health care reform in Australia. Coming out of this debate has been the increasing recognition that the better use of existing health data is an area that must be addressed to fully realise the potential of such an approach.11 Chronic and complex care management is a good example of where better data management that crosses all health levels of patient care from the individual patient through to the development of government policy and funding is critical. Primary care data are key in this process, with potential users including not only general practices but also Primary Health Networks, allied health professionals, specialist practitioners, the hospital sector, all levels of government, researchers and the private sector.

Potential for better evaluation

To date, the evaluation of some primary care programs in Australia has focused on process rather than outcomes. An example is the Practice Incentives Program, whose evaluation was the subject of an Australian National Audit Office audit in 2010.12 The audit noted that the key performance indicators used to monitor the program primarily measured take-up or participation rates rather than assessing effectiveness. To measure effectiveness, there is a need for evaluators to have ready access to data on clinically important outcomes and, at present, this is an information gap in Australian general practice.1

Even when outcome measures are used to evaluate chronic care programs, there is a temptation to use short term measures (eg, changes in clinical parameters) simply because outcomes such as complications and death have a long lead time. This can be seen clearly in the evaluation of many diabetes programs where changes in glycated haemoglobin, blood pressure and lipids are often used in isolation to measure effectiveness. These measures may be appropriate for assessing quality assurance programs for the implementation of established interventions but are likely to be inadequate for clinical trials where interventions are being tested.

But how can long term impacts be measured or at least estimated in a time frame that is useful for decision makers? In the field of diabetes, there is now a well established body of work that uses computer modelling to predict long term outcomes based on changes in clinical factors such as glycated haemoglobin, blood pressure, serum lipids and smoking status.13 The models developed use risk equations derived from cohort studies but unfortunately none are based upon Australian data. This raises the concern about the applicability of current models to assist in the evaluation of chronic care programs in Australia. Consequently, it has been argued that we need to develop country-specific models.13 Data collected by Australian GPs could be used for this task, provided high quality data management systems are established and maintained. The need for a “fresh approach” to research and data alluded to by the federal government14 provides an opportunity to invest in general practice data systems that allow better targeting of health resources and interventions. To do so requires the impact of initiatives to be measured in terms of patient-focused outcomes, including long term clinical sequelae.

Conclusion

There is little doubt that routinely collected data stored in Australian general practices provides an opportunity not only to assist practitioners to review their practice but also to facilitate research initiatives, evaluate the effectiveness of interventions and better target health resources. The questions that need urgent answers are how best to implement this vision and how to put in place the necessary infrastructure. Building a consensus around data sharing in this space is fundamental to any co-creation approach aimed at delivering health care reform in Australia.