The Southern Primary Care Research Network 3 years on – reflections from the end of the beginning

Background and context

Clinical and non-clinical data are routinely collected during healthcare provision. Much of this is systematically recorded and stored in administrative databases, electronic health records, and disease registries.^1–3 Data can be linked between repositories to provide insights that can inform health service planning, clinical practice, and innovative research.^1–4 Using these data for research is cost-effective and has high external validity.⁵ Study design and methods can reduce bias and ensure that the research findings are useful for informing health practice and policy.^6–8 Primary care data provide opportunities to identify population-level needs, evaluate healthcare delivery, and improve equity.^9,10 Such data have been used internationally for highly impactful research, including the development of the QRisk score,¹¹ establishing risk factors associated with Covid-19 death,¹² and observing longitudinal population health changes.¹³ Primary care data remain underutilised for research in Aotearoa New Zealand (NZ), at least in part due to a lack of research infrastructure, including non-commercial GP databases.^10,14

Understanding the collection, organisation, and storage of health information will ensure meaningful secondary use of data for research. Health records in NZ are siloed by service, although primary care records should contain correspondence from secondary care and other primary care providers. Primary Health Organisations (PHOs) are responsible for allocating government funding to provide primary care services to enrolled patients within a geographical area, usually through contracted providers such as general practices. PHOs collect primary care data from contracted providers for auditing and reporting purposes.¹⁵

NZ general practices are typically privately owned businesses that choose and pay for their practice management software (PMS).¹⁶ Data are collected for healthcare provision rather than for future research. Each PMS stores data differently, usually with minimal checking or validation once data are entered into a patient record. Coding of long-term conditions and long-term medications is undertaken by individual clinicians and is highly variable.^17–19 Targeted funding initiatives incentivise the correct coding and recording of data, but often only for the life of those initiatives.

Linkage studies combine data that have been collected, organised, and stored in different datasets. Linkage studies hold both potential promise and pitfalls for research. Data quality and variance affect research findings. Poor quality data entry or coding may be detected through personal or institutional knowledge of the dataset, or when analysis finds an implausible discrepancy between the actual and expected results. Including people in the research team who understand every stage of the data lifespan (data collection, recording, and storage) may mitigate the risk of incorrect assumptions: clinicians entering data into these systems are often aware of the limitations of a particular measure, whereas researchers and analysts familiar with the datasets understand the quality of those datasets. Data variance is often unavoidable and difficult to detect, eg if different criteria were used to apply a classification in one system compared to another. Analysts undertaking data aggregation must be alert to possible data deficiencies. Formal assessment tools can help evaluate dataset quality.²⁰ However, any changes in the processes that generate data, collection methods, or funding for clinical activities can invalidate previous assessments of dataset quality.

Primary care research networks (PCRNs) are considered ‘laboratories’ for primary care research, providing infrastructure that can collate, link, and analyse health data.^21–23 PCRNs are well-established internationally and have historically existed in NZ.²⁴ Despite clear need and widespread support, no current national primary care research infrastructure exists.¹⁰ Therefore, the Southern PCRN was inaugurated in 2021.²⁵ This collaboration between the University of Otago and WellSouth (PHO for Otago and Southland) permits researchers with approved projects to access anonymised primary care data held by WellSouth. Practices can opt-out of projects. The Southern PCRN vision is to create the infrastructure that facilitates population-level primary care research using routinely collected data, first regionally, then nationally.²⁵

This paper describes the Southern PCRN infrastructure establishment and foundational research projects.

Southern PCRN infrastructure

The Southern PCRN has a health equity focus and commitment to Te Tiriti o Waitangi. Therefore, designing the network structure entailed consultation with senior Māori academics and healthcare providers affiliated with the University of Otago and WellSouth. Interlinked groups provide oversight and operationalise network processes (Fig. 1). The Governance group has overall responsibility for the network direction, in accordance with the terms of reference (Box 1). This group has 50:50 co-governance with Māori; members include healthcare providers, community, Pasifika, and academics. The Operational group manages communication, data infrastructure, and researcher liaison. This group comprises the principal investigators, researchers, a biostatistician, and a data analyst. There are two Academic Advisory groups – one internal (Otago University), and the other external. Stakeholder groups participated in research prioritisation exercises, and the community consultation will form the foundation for a ‘Peoples Panel’ for ongoing ad hoc consultation. A separate Māori community consultation group was planned, however, that project did not progress due to personnel changes. Further information about the Southern PCRN is available online: https://wellsouth.nz/about-us/southern-primary-care-research-network.

Fig. 1.

Interrelationship between governance, operational and stakeholder groups in the Southern PCRN.

Southern PCRN Research – data linkage projects

Scoping review

A limited scoping review was undertaken to understand the research potential of routinely collected primary care data.²⁹ Peer-reviewed English-language publications from 1990 to 2022 that described research using routinely collected data from primary care settings in high-income countries with similar health policies to NZ were included. A total of 267 articles met the strict inclusion criteria (Table 2).

Eighty-five percent of all research consisted of observational studies. General practice data were used most often (255/267, 95.5%), followed by regional datasets including research networks and insurance companies (189/267, 70.8%), then prescribing data (71/267, 26.6%). Health delivery research was the top research category (81/267, 30.3%), then pharmaceutical (66/267, 24.7%) and epidemiological research (58/267, 21.7%). Less commonly, clinical prediction studies used routinely collected data to model or predict specific conditions or disease severity (22/267, 8.2%). Social determinants of health studies investigated the effects of non-medical factors such as socioeconomic status, deprivation, or housing on health outcomes or service access (14/267, 5.2%). Intervention evaluation was the least common research type (12/267, 4.5%). Other studies typically related to dataset quality evaluation and research that didn’t fit within any other category (14/267, 5.2%).

This scoping review demonstrated that routinely collected data can evaluate health services, identify risk factors for specific conditions, monitor public health, and assess medication safety.

Research prioritisation exercises

Stakeholder engagement in research prioritisation improves research legitimacy, reduces waste, and improves the implementation of findings into policy and practice.^30,31 Two stakeholder engagement projects were undertaken to identify research priorities for the Southern PCRN.

A modified Delphi survey was conducted to understand the research priorities of primary care clinicians in Otago and Southland and researchers in NZ.³² Fifty-eight participants completed both rounds of the survey. The top five priority research areas identified were access to primary care, health workforce, health services, mental health, and models of primary care. The top priority research question queries the impact of annual checks on people with intellectual disability, followed by questions regarding the impact of embedding allied health professionals and social workers in primary care.

Focus groups were conducted to identify the research priorities of people in Otago and Southland. Seven focus groups were held with 50 people from underserved health populations, including Māori, Pasifika, tāngata whaikaha (disabled), neurodiverse, rainbow, migrant, and rural communities. Participants discussed issues relating to health equity to determine their research needs. An inductive reflexive thematic approach was used to analyse the dataset. The key areas of potential research identified by community members to improve health equity were patient–doctor communication, health literacy, and improving primary care access.

The results from the prioritisation exercises are currently being synthesised to enable the Southern PCRN to establish a fit-for-purpose research agenda to support health equity. This approach to research prioritisation has resulted in a strong health service delivery focus. It is recognised that other research foci may be equally useful, eg identifying health risk, improving diagnostic and prediction accuracy, and evaluating treatment efficacy.

Ethics

One major question that remains unresolved is how to open large primary care data sets to researchers in an ethically responsible way. The current safeguard in place is the review of study protocols by ethics committees. However, these committees may have limited expertise in assessing the ethical risks of large dataset research. Moreover, anonymised data studies may be fast-tracked or undergo limited review, meaning potential negative consequences (eg group harm) may be missed. Although several ethical values have been highlighted as relevant to general data governance, we currently lack an ethical framework that deals specifically with the secondary use of primary health care data. Therefore, we set up a satellite study to produce an ethical framework for data governance groups to use when making decisions about the access to and appropriate use of primary health care data. This study will conduct a separate scoping literature review to develop a preliminary framework consisting of ethical themes and practical questions. Specific attention will be paid to Māori data sovereignty and indigenous values through the literature review methodology. This framework will then be presented to expert stakeholders in a (modified) Delphi study to test its validity and identify any missing values or themes.

Coding quality

A project is planned to audit the quality of common long-term condition codes in primary care records. This project will determine coding reliability, which will directly inform future research validity.

Discussion

This paper describes the development of the Southern PCRN governance and oversight processes, which developed congruently alongside foundational research projects. Data linkage studies have been trialled; lessons from earlier projects contributed to the success of later projects. A scoping review describes the range of research possible with routinely collected data. An ethical guideline is under development, and a coding quality evaluation project is planned.