General Practice EMRs- What they can tell us- and how

This article is also available as a PDF file.

Abstract
Introduction
Background
Opportunity 1: Indicated Antihypertensive Therapy
Opportunity 2: Treatment Adherence and Persistence
Opportunity 3: Emergency Department “Frequent Flyers”
Discussion
Conclusion
Acknowledgements
References
 
Abstract
The locally-held electronic medical records (EMRs) in General Practice systems represent a tremendous resource for quality improvement. These records offer the potential both to track and to enable progress in areas that include: (1) evidence-based treatment of hypertension; (2) attaining persistence of chronic disease related prescribing and medication adherence in the community and; (3) improved management of high users of Emergency Department services. The New Zealand health sector would benefit from the development of domain-specific reporting tools to exploit more readily and reliably the potential of General Practice EMRs. There is further potential benefit to be had from improved interoperability of systems and establishment of national networks, but we have not yet exploited the full potential of what we already have.
 
Introduction
One of New Zealand’s great health data collections is the data held in the Practice Management Systems (PMSs) of the General Practitioners (GPs). This data resource is not one that comes to mind immediately as a major national data collection. This is in part because it is not a single “collection” – rather, it consists of many separate databases (albeit each using the same National Health Index [NHI] for patient identification). Moreover, it is as much a by-product of activity as a deliberate data collection, or at least the primary collection has not been funded as a project by some particular office. Rather, the data collection emerges as the immediate consequence of care delivery and in direct relation to patient-specific local need. This, in fact, is an asset, at least in terms of the data elements that are direct by-products of clinical workflow. In the setting of New Zealand General Practice this workflow-related data particularly features the prescriptions, for which full and accurate data entry is required to produce the script for the patient, as well as test results which are received by the GP electronically. Many other elements, such as problem coding and clinical observations, are complete to the extent that they are used to support the practice in its ongoing patient management.
 
A 2006 Commonwealth Fund/ Harris Interactive survey^[1] places New Zealand in the top tier globally in terms of GP’s use of electronic medical records (EMRs). Moreover, these GPs are increasingly well connected to networks, and the target of having every New Zealand GP on 10Mb/s broadband within the next five years has been suggested.^[2] This GP population, maintaining high quality EMRs and, increasingly, with ready connectivity to the wider world, provides a rich opportunity for innovation. In this paper we examine the potential for leveraging the New Zealand General Practice systems and their data for the improvement of health care delivery.
 
After a review of use of GP data in Australia and New Zealand, this paper provides suggestions about and discussion on three specific areas of opportunity for more extensive and systematic utilisation of GP EMRs in New Zealand to provide significant benefits for major population health and health care system objectives. From these three areas we move to a discussion of the road ahead in terms of the technical steps necessary to maximise these benefits.
 
Background
In Australia there are a variety of General Practice based data collections in operation:

• The Australian Sentinel Practices Research Network^[3] – this is a bio-surveillance-oriented network based on submission of de-identified patient data on influenza-like illnesses from a self-selected network of GPs.
• Bettering the Evaluation And Care of Health (BEACH)^[4,5] – this is a cross-sectional paper-based data collection that collects data on 100 consecutive consultations from each of 1000 GPs annually.
• The General Practice Research Network^[6,7] – this is based on automated extraction of de-identified information from GPs’ EMRs using HCN’s Medical Director (the lead GP PMS in Australia).

These collections, used individually or all together, have provided the basis for extensive research on General Practice issues.^[8-11] BEACH can also be seen serving as a platform for further data collection^[12] or as a supporting data source^[13] for further research, and BEACH serves as the basis for regular scene-setting opening articles in Australian Family Physician; see, eg, Britt & Fahridin^[14] or Charles, Knox & Britt.^[15]
 
In New Zealand, there are both recent investigations and ongoing capabilities for creating an overview of General Practice activity. The National Primary Medical Care Survey (NatMedCa)^[16] provides in-depth information on the content of consultations between primary care providers (notably including GPs) and their patients, based on over 42,000 interactions recorded in two-week blocks six months apart for most providers. This data collection has provided the data for a number of related research articles (eg, see Kairuz, Harrison, Lay-Yee & Davis,^[17] Crampton, Davis, Lay-Yee, Raymont, Forrest & Starfield^[18] and Hider, Lay-Yee, Crampton and Davis.^[19]. The HealthStat Panel, by CBG Health Research Ltd, provides automated data extracts over the HealthLink messaging system for a representative sample of about 100 General Practices across New Zealand.^[20]
 
Other relevant New Zealand collections form around specific activities with goals largely, but not wholly, aligned to General Practice per se. A significant data collection is being acquired largely through General Practice via the use of the PREDICT series of cardiovascular disease (CVD) risk estimation, data collection and risk management technologies. While PREDICT provides interactive decision support for GPs, the collected data is allowing improved estimation of the impact of ethnicity on CVD risk,^[21] feeding back to further research on the use of the adjusted risk factors. ^[22] The New Zealand Intensive Medicines Monitoring Programme (IMMP) undertakes prospective observational cohort studies on selected new drugs in the early post-marketing period, as well as accepting spontaneous adverse drug reaction (ARD) reports from doctors, with the aim of identifying signals of previously unrecognised ADRs and establishing drug risk profiles.^[23]
 
For my own part, I have been involved in analysis of GP EMRs in Australia and New Zealand. This includes examination of GP problem coding practices, ^[24] research on General Practice data entry user interfaces,^[25] and development of high-specificity antihypertensive prescribing alert logic.^[26] Those experiences were in Australia, although in the last case the PMS was from a New Zealand based vendor. More recently, I have been engaged in analysis of antihypertensive prescribing for a largely Pacific patient population in the Auckland area with a particular aim of identification (and ultimately reduction) of adherence issues. ^[27] In addition, in the past couple of years I have been involved in academic forums concerning the potential of health data mining (with a heavy emphasis on use of GP data). The first of these, held in Adelaide in April 2005 formed the basis of the inaugural issue of the Electronic Journal of Health Informatics^[28] (the Australia counterpart to Health Care and Informatics Review Online). Subsequently, as part of Australasian Computer Science Week 2007 (ACSW), there was a workshop on Health Knowledge Management and Discovery, and an ACSW workshop on Health Data and Knowledge Management held in Wollongong in January 2008.
 
Opportunity 1: Indicated Antihypertensive Therapy
Hypertension is a major risk factor in a number of significant medical conditions, including cardiovascular disease (CVD, New Zealand’s number one killer^[29]) and chronic kidney disease (CKD^[30]). Broadly accepted hypertension treatment guidelines exist; notably, JNC7.^[31]. These guidelines identify “compelling indications” for individual drug classes, such as ACEi (angiotensin-converting enzyme inhibitor) or ARB (angiotensin receptor blocker) with diabetes and CKD. ^[31] Although no specific guideline should be expected to apply to 100 percent of individuals in a General Practice setting, by the nature of evidence-based guidelines, we expect that high rates of adherence to these compelling indications will yield the best health outcomes.
 
Thus, our first opportunity, is to use the GP’s EMR to track adherence to compellingly indicated antihypertensive therapies. Such tracking provides two areas of benefit:

1. Benchmarking where a practice stands – this can serve to expose the size of a problem (or to think positively, an opportunity) for quality improvement, and hence can be part of a business case for a change programme. Once a programme has commenced, the benchmark measure can be used to track progress.
2. Following up the specific cases – this could be done in real-time at the point of care as an “alert” or could be done in a batch, as the basis of a patient recall or to attach a note for attention at the patient’s next regular visit. With an adherent patient (ie, who attends regularly and acquires and takes as directed the prescribed medications) improvement should be relatively straightforward once the specific cases are highlighted against evidence-based criteria.

Such statistics are already part of the proposed PHO Performance Management Programme.^[32] In particular, the Second Wave indicator, “Diabetes patients who have had a positive microalbuminuria test and are on ACE inhibitor or A2 receptor agonist at last annual check”^[33] aligns well with the present discussion. (It is worth noting, however, that most of the Performance Management Programme indicators relate to processes that are prerequisite for evidence based treatment, but not to the treatment itself; eg, “Heart Failure ever recorded.”)
 
Obviously, the same pattern of GP data use that is recommended herein for indicated antihypertensive therapy can be applied to other chronic condition treatments (eg, statins); and, in general, it is a special case of our next opportunity.
 
Opportunity 2: Treatment Adherence and Persistence
Little is known specifically about chronic disease medication adherence in New Zealand, but the picture from the international literature indicates an enormous opportunity for improvement. Internationally, long-term adherence with medications for chronic diseases is low, particularly among lower socio-economic groups. ^[34] Poor adherence to antihypertensive medication contributes to inadequate blood pressure control in more than two-thirds of hypertensive patients.^[35] A recent Lancet editorial on the problem of hypertension, reflecting comments in a literature review of hypertension in the same journal, stated that. “The biggest problem, arguably, remains compliance. Despite very effective and cost-effective treatments, target blood pressure levels are rarely reached, even in countries where cost of medication is not an issue for patients.”^[36] A Swedish study found satisfactory refill adherence for thiazides at 55 percent, ACE inhibitors at 59 percent and beta-blocking agents at 66 percent.^[37]
 
At its simplest, the GP’s role in adherence to indicated treatments of chronic conditions is to ensure continuity of prescriptions (whereupon it passes to the patient to fill the prescription and take the medication as directed). From GP EMRs we can discern a reasonable upper bound on the Medication Possession Ratio (MPR) – ie, the proportion of a given time period wherein medication was available to the patient.^[38] The MPR could fall significantly below 1.0, leading to gaps in treatment persistence, due to the patient’s late return for a re-prescription, or due to a lapse on the part of the GP in identifying the need to prescribe or re-prescribe. (Note – in this paper we refer to “persistence” of treatment, as per Andrade et al,^[8] with respect to frequency of patients discontinuing medications.)
 
There are two aspects to the opportunity with respect to what the GP can learn about treatment persistence from their EMR data:

1. Awareness of immediate cases – identification of those patients that, at a particular moment in time, are out of supply of an indicated medication. In the first instance, the action is to treat the non-adherence as inadvertent and recall the patient and/or simply prescribe as indicated at the next opportunity. This includes not just patients with lapsed medications, but also those whose circumstances have changed (eg, due to development of a co-morbidity) and thus require additions to previous therapy.
2. Opportunity for communication with those with poor supply profiles – at some point it becomes logical to look to a lack of concordance between doctor and patient, and/or to the ability of the patient to achieve adherence for other reasons, as the key issue. Low MPR over an extended time period and repeated lapses in medication supply are indicators available from the GP EMR. In this case the EMR is indicating the need for improved communication between GP and patient, possibly to engaged in joint “problem-solving” in relation to underlying adherence barriers.

As a broader research opportunity, GP EMR data is a promising resource for gaining a more detailed understanding of the factors that predispose patients to non-adherence risk and thence for development of targeted intervention strategies for specific clusters of patients (eg, those who are persistently inadvertently non-adherent due to specific lifestyle issues versus those that are intentionally non-adherent due to disagreeing with the doctor’s recommendations or simply due to cost).
 
It is outside a strict consideration of General Practice data collections, however, integration with other data sources is, of course, the road to achieving a fuller picture of patient treatment persistence. Integration of prescribing data over at least a regional level (including specialist and outpatient prescribing) is one link in the chain. Another is integration with dispensing data. Significant information on the latter is possible in theory via the HealthPAC claims data, however, providing such decision support information is a significant stretch of the HealthPAC mission^[39] and it probably not reasonable to expect timely routine information from this route. The ultimate understanding of adherence would come by monitoring of actual patient consumption. This is, in fact, possible already in niche areas – eg, consider the “smart” asthma inhaler that makes an objective measure of patient use.^[40]
 
Opportunity 3: Emergency Department “Frequent Flyers”
With respect to chronic conditions, hospital services have been described as the ambulance that takes the patient back to the top of the cliff. On the surface, the question of high-users of Emergency Department (ED) services seems like a hospital issue; but there is excellent potential for GP engagement.
 
The burden of issues related to poor chronic disease management on EDs can be substantial. Hyperglycaemic emergencies comprised 25.6 percent of high-care unit admissions to a Cape Town hospital, with non-adherence to therapy and new diabetes diagnoses being two of the three predominant reasons for admission in such cases. ^[41] For US urban minority children, visits to a primary care provider and filled prescriptions for controller medications are found to strongly reduce the odds of ED asthma visits. ^[42] Pooled data from systematic literature review shows that chronic obstructive pulmonary disease (COPD) patients who received interventions with two or more Chronic Care Management (CCM) components have lower rates of hospitalisations and emergency/unscheduled visits and a shorter length of stay compared with control groups.^[43] Moreover, survey data shows that patients with common chronic medical disorders have increased prevalence of major depression, and that those patients with such chronic disorders and major depression have increased odds of ED visits.^[44]
 
The opportunity for the GP to disrupt the frequent flyer cycle has at least three components:

1. More aggressive monitoring and treatment – ie, achieving persistence of indicated therapies as per the other two opportunities above, as well as routine monitoring of progress.
2. Education – engagement in CCM-type activities and other less formal communication to probe overall concordance and patient engagement and competence in their own care.
3. Improved identification of additional conditions and risk factors – a reduction in ED visits will follow from timely diagnoses, particularly of likely co-morbidities, such as depression in patients with chronic medical disorders or diabetes in patients with other CVD risk factors.

Most New Zealand GPs receive some form of electronic hospital discharge summary that can, at least in theory, provide a signal that the patient may be advancing to, or already have achieved, frequent flyer status. Beyond this, the EMR provides summary measures of management quality (such as HbA1c) which either indicate the quality of management, or mark the lack of appropriate monitoring by their absence. However, as we take up in the Discussion section below, systems interoperability makes the routine analysis and collation of baseline measure for such information at the General Practice more difficult than it should be.
 
Discussion
The opportunities discussed above are really just a list of many of the things a GP should (and does) do; but the under-exploited opportunity is, with respect to the use of the GP’s EMR, to quantify levels of achievement over time and systematically highlight specific cases for further intervention. In this regard, the high level of computerisation in General Practice, and in the sector generally, provides enormous potential to add greater reliability to care, especially around chronic conditions. The strategic value of General Practice EMRs, and the associated importance of the PMS reporting tools, has been underemphasised.
 
As outlined above, the GP EMR can provide a basis for assessing adherence to evidence-based clinical guidelines and providing interactive alerts, or identifying cases for recall, where there is an opportunity to improve adherence to such guidelines. The GP’s prescribing data can also allow computation of a maximum Medication Possession Ratio (MPR) for any indicated long-term medication. As such, the data forms the foundation for targeted promotion of patient adherence, and adherence research, as well as a touchstone for assessment of evidence based prescribing. Furthermore, the GP’s data collection can serve as a platform for reduction of ED visits. “Frequent flyers” can be identified from electronic hospital discharge summaries, off-target or under-monitored risk factors can be identified from electronic laboratory results, and promising cases for further diagnostic testing and patient education can be highlighted by profiles of co-morbidities and other indicators.
 
It must be acknowledged as a limitation of the present paper that the three opportunities identified have been selected in the absence of a systematic framework. Each was chosen because there is evidence that the issue is major and because of the relatively straightforward potential for GP EMRs to provide a basis for quality improvement in these areas. The choice of hypertension, as compared to overall CVD risk management, may seem particularly limited and/or arbitrary. This is merely a suggestion of a simple but important starting point where there is still outstanding opportunity for improvement – it is in no way meant to detract from the relevance of interactive electronic decision support for holistic CVD risk management, and the opportunities are by no means mutually exclusive. Development of an overall framework (eg, based on return on investment) for systematic prioritisation of EMR-driven quality assurance is outside of the scope of this paper and would be a useful area for future research.
 
While the systems and, for the most part, the data are there, achieving highly accurate query results from the practice EMR is far less simple than it may appear. While PMS software such as Medtech32 provides powerful query reporting capabilities, there are several outstanding challenges:  

1. Ontology management – it is seldom straightforward to identify the required scope of concepts and associated clinical codes for significant clinical queries. For instance, the PHO Performance Management Programme^[33] indicates 86 Read Clinical Codes to associate with a record of Diabetes for a patient. The query writer must be careful to include just the desired sub-types of a condition – eg, is it desirable to include gestational diabetes among the diabetes diagnoses when creating baseline reporting statistics around compelling indications on co-morbidities of hypertension? Similar complexity arises around drug ontology; to include just the desired formulations of a medication, and to properly account for combination drugs. When looking at MPR, we may accept some changes among drug classes and adjustment into (but perhaps not out of) combination therapy.
2. Complex/temporal querying – identification of lapses in medication requires consideration of the time period between prescriptions, and, to be maximally accurate, should account for accumulated medication supply over the recent past. Assessing the adherence to evidence-based treatment of a compelling indication over a reporting period requires computation of the MPR since the emergence of the indication (which in some cases will be before the start of the reporting period and in other cases will be during the reporting period). Supporting lab results, observations and other computed values (such as 5 year CVD event risk) should be integrated into the filter for indications of treatment or estimation of the quality of management of a frequent ED user. Furthermore, combination therapy of two or more complementary agents (accomplished with multiple drugs, possibly including combination drugs) must be accurately assessed to achieve a complete picture of the sufficiency of medication supply for adherence. Validity of these complex clinical queries presents a problem, because it is the nature of reporting tools (and the Structured Query Language, SQL, that is the standard underlying query language of relational database technology) to give some answer to almost any query unless it is so poorly specified as to yield a syntax error.
3. Data format and availability – while electronic data collection in New Zealand General Practice is excellent by world standards,^[1] discipline in recording of data into the PMS is not uniformly systematic, and neither lab results nor electronic discharge summaries arrive in an ideal format to support the types of queries we suggest herein. There is a need to have greater consistency and reliability on processes for data to be accessible for quality improvement queries – this includes achieving improved semantic interoperability of health messages. Health messages must fulfil the dual role of being: (1) a document, readable by a human GP, and retaining a sense of the source; as well as (2) serving as a set of data components that can slot seamlessly into query computations.

The first two of these problems, and partially the third, can be addressed to a substantial degree by the development of domain-focused query tools. Such tools would allow the GP, Practice Manager or other quality assurance analyst to formulate queries in terms of high-level concepts such as “anti-hypertensive medications”, “ACE/ARB” or “diabetes mellitus (chronic)” with confidence in their validity and the ability to verify, and change, their subcomponents. One can expect that New Zealand’s participation in the International Health Terminology Standards Development Organisation^[45] will lead to improved ability to query relevant clinical concepts, but this cannot be expected to provide a quick solution or a magic bullet. Domain-focused query tools would also have concepts such as “MPR” and “lapse of medication supply” available for inclusion in queries. Ideally, we should also be able to work with “n-similar-events-per-unit-time” concepts such as a set of ED presentations associated with a high-level problem code grouping, or three successive high blood pressure observations. Many useful queries should fit into common templates such “lapse of therapy after compelling indication” that would be amenable to filling in the blanks with any of a range of therapies and indications to suit the quality improvement agenda.
 
The vision of opportunity described herein is largely one of individual practices conducting their own queries and pursuing individual quality improvement courses, which ignores both the opportunities and the burdens of a larger world. On the opportunity side, it would be ideal to enable a culture of sharing queries, perhaps through professional colleges. Sharing of the query results, and the natural extension to benchmarking of relative performance over a PHO, DHB or region, is a further step that raises the question of managing the confidentiality of individual practice results.
 
Local, regional and national health information networks provide the opportunity to enhance the accuracy of queries by integrating problems, observations and prescriptions (ideally both prescribing and dispensing) from multiple sources. While it is obvious that this integration will provide some improvement in accuracy (eg, to know about a diabetes diagnosis from another provider), there is a need for further study to quantify just how much benefit data integration will yield for specific quality improvement objectives and to prioritise technical requirements accordingly. The Health Information Strategy for New Zealand 2005^[46] is squarely aimed at enabling these capabilities, particularly with respect to Action Zones 4 to 7 (ePharmacy, eLabs, Discharge Summaries, and Chronic Care and Disease Management). Active and iterative dialogue must be maintained with respect to the quality improvement enabling power of specific technical implementation aims within these Action Zones – ie, it would be perfectly possible to have otherwise excellent capabilities in these four Action Zones and yet not support the query capabilities of the GP data collections.
 
A final issue is one of funding. Luckily for us all, health care providers are intrinsically motivated to deliver excellent care; but this still begs the question of who takes professional time crafting and validating queries, analysing their implications, and implementing changes of patient management to exploit the opportunities identified. The quality improvement potential of GP data collections will remain a latent capability, exploited only partially and mostly on an ad hoc basis, until there is a clear financial model supporting such activity. The concept of practice incentives is one clear strategy; however, it is distinctly double-edged. Any specific set of incentives invites even the most altruistic of providers to “game” the system; at the worst it could discourage providers from engaging with disadvantaged populations that are less able to meet certain performance targets (or undermine the viability of those providers that do so anyway).
 
Conclusion
The data in the Practice Management Systems of General Practices has the potential to address major health care issues in terms of both tracking the extent and progress on areas for quality improvement, and in providing information directly relevant to implementation of quality improvement per se. Unlocking this potential would be much easier with improved querying tools that are tailored to the opportunities available in the data. Further implementation of health information networks would strengthen the potential for accurate quality improvement queries, but we should start learning to exploit more fully the substantial potential that is already there – both for its own sake, and so we more accurately understand what we require in terms of further health systems interoperability.
 
Acknowledgements
I would like to acknowledge my colleagues in my present ongoing work with analysis of General Practice prescribing, including Tim Kenealy, Rekha Gaikwad, Thusitha Mabotuwana, John Kenelly, Raina Elley and Jeff Harrison.
 
References
1. Schoen C, Osborn R, Huynh PT, Doty M, Peugh J, Zapert K. On the front lines of care: primary care doctors’ office systems, experiences, and views in seven countries. Health Aff (Millwood). 2006 Nov–Dec;25(6):w555-71.

2. Bell, S. Health pushes for broadband, Computerworld. 21 May 2007. http://computerworld.co.nz/news.nsf/news/7D12CDCDE93160A3CC2572DF0005DCFB. Accessed on 9 December 2007.

3. Royal Australian College of General Practitioners. The Australian sentinel practices research network. August 2006. http://www.racgp.org.au/aspren. Accessed on 9 December 2007.

4. Australian General Practice Statistics and Classification Centre. The BEACH project: bettering the evaluation and care of health. October 2007. http://www.fmrc.org.au/beach.htm. Accessed on 9 December 2007.

5. Britt H, Miller G, Bayram C. The quality of data on general practice - a discussion of BEACH reliability and validity. Aust Fam Physician 2007;36(1–2):36–40.

6. Health Communications Network. “General Practice Research Network”. 2007. http://www.hcn.net.au/doctors/gprn.asp. Accessed on 9 December 2007.

7. Sayer GP, McGeechan K, Kemp A, Bhasale A, Horn F, Hendrie L, et al. The General Practice Research Network: the capabilities of an electronic patient management system for longitudinal patient data. Pharmacoepidemiol Drug Saf. 2003 Sep;12(6):483–9.

8. O’Halloran J, Britt H, Valenti L. General practitioner consultations at residential aged-care facilities. Med J Aust. 2007 Jul 16;187(2):88–91.

9. Roughead EE, McGeechan K, Sayer GP. Bisphosphonate use and subsequent prescription of acid suppressants. Br J Clin Pharmacol. 2004 Jun;57(6):813–6.

10. Kerr SJ, Mant A, Horn FE, McGeechan K, Sayer GP. Lessons from early large-scale adoption of celecoxib and rofecoxib by Australian general practitioners. Med J Aust. 2003Oct 20;179(8):403–7.

11. Stocks NP, McElroy H, Sayer GP, Duszynski K. Acute bronchitis in Australian general practice. A prescription too far? Aust Fam Physician. 2004 Jan-Feb;33(1–2):91-3.

12. Larkins SL, Geia LK, Panaretto KS. Consultations in general practice and at an Aboriginal community controlled health service: do they differ? Rural Remote Health 2006 Jul–Sep;6(3):560. Epub 2006 Jul 19.

13. Miller GC, Britth HC, Valenti L. Adverse drug events in general practice patients in Australia. Med J Aust. 2006 Apr 3;184(7):321–4.

14. Britt H, Fahridin S. Presentations of nausea and vomiting. Aust Fam Physician 2007 Sep;36(9):682–3.

15. Charles J, Knox S, Britt H. Postnatal depression in Australian general practice. Aust Fam Physician 2006 Sep;35(9):668–9.

16. Ministry of Health, The National Primary Care Survey (NatMedCa) 2001/02 Reports. http://www.moh.govt.nz/natmedca. Accessed 26 November 2007.

17. Kairuz T, Harrison J, Lay-Yee R, Davis P. A snapshot of general practitioner encounters with diabetic patients. New Zealand Pharmacy. 2006;26(8):20–21, 28.

18. Crampton P, Davis P, Lay-Yee R, Raymont A, Forrest CB, Starfield B. Does community-governed nonprofit primary care improve access to services? Cross-sectional survey of practice characteristics. Int J Health Serv. 2005;35(3):465–78.

19. Hider P, Lay-Yee R, Crampton P, Davis P. Comparison of services provided by urban commercial, community-governed and traditional primary care practices in New Zealand. J Health Serv Res Policy. 2007 Oct;12(4):215–22.

20. Cumming J, Gribben B. “Appendix 2: HealthStat Sample”. In Evaluation of the primary care strategy: practice data analysis 2001–2005. September 2007. http://www.victoria.ac.nz/hsrc/reports/downloads/PHCSE%20FINAL.pdf. Accessed 9 December 2007.

21. Riddell T, Jackson RT, Wells S, Broad J, Bannink L. Assessing Maori/non-Maori differences in cardiovascular disease risk and risk management in routine primary care practice using web-based clinical decision support: (PREDICT CVD-2).” N Z Med J. 2007 Mar 2;120(1250):U2445.

22. Wells S, Kerr A, Broad J, Riddell T, Kenealy T, Jackson R. The impact of New Zealand CVD risk chart adjustments for family history and ethnicity on eligibility for treatment (PREDICT CVD-5). N Z Med J. 2007 Sep 7;120(1261):U2712.

23. Coulter D. Signal generation in the New Zealand Intensive Medicines Monitoring Programme: a combined clinical and statistical approach. Drug Saf. 2002;25(6):433-9.

24. Bolton P, Warren J. Toward a model for the evaluation of clinical coding systems. Aust Fam Physician. 2001 Aug;30(8):814-6.

25. Warren J, Bolton P. Empirical norms as a lever for on-line support of general practice. . IEEE Computer Society Press; January 2001.

26. Gadzhanova S, Iankov I.I, Warren JR, Stanek J, Misan GM, Baig Z, Ponte L. Developing high-specificity anti-hypertensive alerts by therapeutic state analysis of electronic prescribing records. J Am Med Inform Assoc. 2007 Jan-Feb;14(1):100–9.

27. Warren J, Gaikwad R, Mabotuwana T, Kenelly J. Kenealy T. Developing a quality audit report for general practice prescribing for hypertension: methodology. HCIRO Online. October 2007. /journal/index.cfm?fuseaction=articledisplay&FeatureID=021007. Accessed on 9 December 2007. Also in Proceedings, HINZ 2007, Rotorua, October.

28. Warren J, Roddick J, Webb G, Williams G. Electronic Journal of Health Informatics. Special Issue on Health Data Mining (Guest Editors). 2006;1(1):e1. http://ejhi.net/ojs/index.php/ejhi/issue/view/1. Accessed on 9 December 2007.

29. Turley M, Stefanogiannis N, Tobias M, van der Hoorn S, Lawes C, Ni Mhurchu C, Rodgers A. Nutrition and the burden of disease: New Zealand 1997–2011. Wellington: Ministry of Health; August 2003. Public Health Intelligence Occasional Bulletin Number 17.

30. Kalra PA. Renal specific secondary hypertension. EDTNA ERCA J. 2007 Jan–Mar;33(1):4–10.

31. Chobanian AV, Bakris GL, Black HR, Cushman WC, Green LA, Izzo JL Jr, Jones DW, Materson BJ, Oparil S, Wright JT Jr, Roccella EJ. Seventh report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure. Hypertension. 2003 Dec;42(6):1206-52, 1221.

32. PHO Performance Management Operational Framework: Including Referred Services Management (RSM) and Clinical Performance Indicators (CPI): Version 8. http://www.dhbnz.org.nz/Site/SIG/pho/PHO-Framework-Appendices.aspx. Accessed 27 November 2007.

33. Ministry of Health. Performance monitoring framework: data format standard: clinical performance indicators: HealthPAC version 1.7. October 2006. http://www.moh.govt.nz/moh.nsf/indexmh/phcs-projects-perfmgmt#clinical. Accessed on 9 December 2007.

34. Booth, G L, J E Hux. Relationship between avoidable hospitalizations for diabetes mellitus and income level. Arch Intern Med 2003 Jan;163(1):101–6.

35. Miller N H, M Hill, Kottke K, Ockene IS. The multilevel compliance challenge: recommendations for a call to action. A statement for healthcare professionals. Circulation. 1997 Feb 18;95(4):1085–90.

36. Messerli F H, B Williams Ritz E. Essential hypertension. Lancet. 2007 Aug 18; 370(9587): 591–603.

37. Andersson K, Melander A, Svensson C, Lond O, Nilsson JL. Repeat prescriptions: refill adherence in relation to patient and prescriber characteristics, reimbursement level and type of medication." Eur J Public Health. 2005 Dec;15(6): 621–6.

38. Andrade S E, Kahler KH, Frech F, Chan KA. . Methods for evaluation of medication adherence and persistence using automated databases. Pharmacoepidemiol Drug Saf. 2006 Aug;15(8):565-74.

39. Ministry of Health. HealthPAC: about us. October 2007. http://www.moh.govt.nz/moh.nsf/indexmh/healthpac-about. Accessed on 9 December 2007.

40. Burgess SW, Wilson SS, Cooper DM, Sly PD, Devadason SG. In vitro evaluation of an asthma dosing device: the smart-inhaler. Respir Med. 2006 May;100(5):841-5.

41. Pepper DJ, Levitt NS, Cleary S, Burch VC. Hyperglycaemic emergency admissions to a secondary-level hospital – an unnecessary financial burden. S Afr Med J. 2007 Oct;97(10):963–7.

42. Smith SR, Wakefield DB, Cloutier MM. Relationship between pediatric primary provider visits and acute asthma ED visits. Pediatr Pulmonol. 2007 Nov;42(11):1041–7.

43. Adams SG, Smith PK, Allan PF, Anzueto A, Pugh JA, Cornell JE. Systematic review of the chronic care model in chronic obstructive pulmonary disease prevention and management. Arch Intern Med. 2007 Mar 26;167(6):551–61.

44. Egede LE. Major depression in individuals with chronic medical disorders: prevalence, correlates and association with health resource utilization, lost productivity and functional disability. Gen Hosp Psychiatry. 2007 Sep–Oct;29(5):409–16.

45. International Health Terminology Standards Development Organisation. New Zealand Ministry of Health Representatives. http://www.ihtsdo.org/members/nz/. Accessed on 17 December 2007.

46. Health Information Strategy Steering Committee. Health Information Strategy for New Zealand. Wellington: Ministry of Health; 2005.