'It is not possible to divorce the practice of medicine from the society in ...

VOLUME 4 • NUMBER 3 • SEPTEMBER 2012

OF PRIMARY HEALTH CARE

‘It is not possible to divorce the practice of medicine from the society in which it is practised.’ See page 223

Original Scientific Paper Calculating cardiovascular risk in people with Type 2 diabetes See page 181

Original Scientific Paper Pre-calling babies to improve immunisation timeliness See page 189

Original Scientific Paper PSA testing in asymptomatic males See page 199

Original Scientific Paper Acute otitis media in the under-fives See page 205

Original Scientific Paper Brief mental health intervention for Maori See page 231

Viewpoint The platform model of pain management See page 254

CONTENTS VOLUME 4 • NUMBER 3 • SEPTEMBER 2012 ISSN 1172-6164 (Print) ISSN 1172-6156 (Online)

OF PRIMARY HEALTH CARE

178 Editorials

231

From the Editor 178

Guest Editorial 180

Fiona Mathieson, Kara Mihaere, Sunny Collings, Anthony Dowell, James Stanley

A practical issue Felicity Goodyear-Smith

Maori cultural adaptation of a brief mental health intervention in primary care

Short Report 239

Commentary: risk prediction models for people with Type 2 diabetes

The anatomical placement of body organs by Australian and New Zealand patients and health professionals in general practice Marjan Kljakovic

Kamlesh Khunti

Improving Performance

181 Original Scientific Papers

242

Quantitative Research 181

New Zealand Diabetes Cohort Study cardiovascular risk score for people with Type 2 diabetes: validation in the PREDICT cohort Tom Robinson, C Raina Elley, Sue Wells, Elizabeth Robinson, Tim Kenealy, Romana Pylypchuk, Dale Bramley, Bruce Arroll, Sue Crengle, Tania Riddell, Shanthi Ameratunga, Patricia Metcalf, Paul Drury

189

Early connections: effectiveness of a pre-call intervention to improve immunisation coverage and timeliness Felicity Goodyear-Smith, Cameron Grant, Tracey Poole, Helen Petousis-Harris, Nikki Turner, Rafael Perera, Anthony Harnden

199

205

The incidence of acute otitis media in New Zealand children under five years of age in the primary care setting Barry Gribben, Lesley Salkeld, Simon Hoare, Hannah Jones

213

Sally Abel, Bob Marshall, Donny Riki, Tania Luscombe

249 Continuing Professional Development 249

Does the order of presentation and number of online resources affect the frequency of access by learners? Steven Lillis, Samantha Murton

249

String of PEARLS about musculoskeletal conditions

250

Vaikoloa: Hearing loss among Pacific peoples Ofa Dewes

251

Addressing patient alcohol use: a view from general practice Thomas Mules; Jennifer Taylor; Rachel Price; Logan Walker; Baneet Singh; Patrick Newsam; Thenmoli Palaniyappan; Toby Snook; Mahfuzah Ruselan; John Ryan; Jaishree Santhirasegaran; Phoebe Shearman; Petronella Watson; Richard Zino; Louise Signal; Geoff Fougere; Helen Moriarty; Gabrielle Jenkin

Mixed Method Research 223

Factors influencing diagnostic decision-making Kathleen Callaghan

Potion or Poison? Colloidal silver David Woods

253

Nuggets of Knowledge: Statins and memory loss Linda Bryant

254 Viewpoint 254

From ladder to platform: a new concept for pain management Lawrence Leung

259

Frequently asked questions on measurement of bone mineral densitometry Joseph Lee, Nelson Loh

Qualitative Research 217

Cochrane Corner: Topical antibiotics are probably better than placebo for acute conjunctivitis but most get better anyway Bruce Arroll

PSA testing in general practice Fraser Hodgson, Zuzana Obertová, Charis Brown, Ross Lawrenson

Evaluation of Tu Meke PHO’s Wairua Tangata Programme: a primary mental health initiative for underserved communities

262 Letters to the Editor 263 Film Review 263

A good death: a film about end-of-life care and advance care planning—produced by Prof. D Robin Taylor and Paul Trotman Reviewed by Prof. Rod MacLeod

264 About the Journal of Primary Health Care

VOLUME 4 • NUMBER 3 • SEPTEMBER 2012 J OURNAL OF PRIMARY HEALTH CARE 177

EDITORIALS FROM THE EDITOR

A practical issue

Felicity GoodyearSmith MBChB, MD, FRNZCGP, Editor

J PRIM HEALTH CARE 2012;4(3):178–179.

CORRESPONDENCE TO: Felicity Goodyear-Smith Professor and Goodfellow Postgraduate Chair, Department of General Practice and Primary Health Care, The University of Auckland, PB 92019 Auckland, New Zealand f.goodyear-smith@ auckland.ac.nz

178

T

his issue has a wide range of research on a variety of topics, but the common thread is that these studies address practical issues relevant to New Zealand primary health care. The lead paper by Robinson et al. reports on the locally developed PREDICT-based risk model which is more accurate in predicting cardiovascular risk than the adjusted Framingham equation.1 In his accompanying commentary, Dr Kamlesh Khunti, Professor of Primary Care Diabetes and Vascular Medicine at the University of Leicester, points to the need for studies on the impact of use of risk prediction models with outcomes such as patient adherence to medication or increased understanding before the use of these models becomes routine practice.2 A study of routine pre-call of infants at four weeks to alert parents to the need to present their babies to general practice at six weeks to start their immunisation schedule has found that this increases the coverage and timeliness of the immunisation series.3 However, the most significant finding of this study is the importance of enrolment of newborn babies with a general practice. Making an early connection with a practice means that an infant is much more likely to be immunised fully and on time, as well as the other likely health care gains that may result from engagement with their general practitioner (GP) or practice nurse. Controversy around prostate-specific antigen (PSA) screening continues. A study of GPs in the Waikato found a high likelihood of them PSA testing asymptomatic men including those aged 70 years or older, despite there being no evidence of benefit in this activity.4 A qualitative study suggests that, in Wellington at least, GPs may not be screening and intervening sufficiently with patients around alcohol misuse.5 While some GPs may be ignoring Ministry of Health

recommendations regarding prostate and alcohol screening, it appears that best practice guidance on management of acute otitis media in children is being followed. A cohort study of nearly 20 000 New Zealand children shows a significant decline in the use of antibiotics in treatment, in concordance with accepted best practice.6 There are two studies on primary mental health initiatives for Maori and other underserved populations. Mathieson and colleagues report on a Maori adaptation of a brief intervention involving cognitive behavioural therapy and guided self-management,7 and a research team in Hawkes Bay describe an integrated, holistic tikanga Maori–based programme targeting Maori, Pacific and quintile 5 populations aimed at reducing mental health inequalities.8 Callaghan explores factors that might influence GPs’ decision making and finds that clinical information and the probability of disease are rated as highly important and desirable by ‘standard setters’ in general practice academic departments and The Royal New Zealand College of General Practitioners.9 Lillis and Murton explore the provision of online resources to GP registrars in training and conclude that they are only likely to access the top four in the list, underlining the importance of prioritising and possibly limiting provided resources.10 A study by Kljakovic finds that patients from both Australian and New Zealand general practices performed poorly in correctly locating body organs in line drawings, and health professionals achieved this only moderately better than their patients.11 On an even more practical bent, two viewpoint papers offer specific clinical tips. Leung extends the concept of the World Health Organization analgesic ladder to that of a platform, providing a broad range of pain relief interventions in a

VOLUME 4 • NUMBER 3 • SEPTEMBER 2012 J OURNAL OF PRIMARY HEALTH CARE

EDITORIALS FROM THE EDITOR

stepped-up manner,12 and Lee and Loh review the facts around bone mineral densitometry.13 Our regular columns provide practical tips on the use of topical antibiotics in acute conjunctivitis (Cochrane Corner), best practice evidence for managing a number of musculoskeletal conditions (String of PEARLS), the potential harm and lack of evidence of benefit for use of colloidal silver (Potion or Poison?), strategies to identify cognitive impairment from statin use (Nuggets of Knowledge) and possible actions to address hearing loss among Pacific peoples (Vaikoloa).14 Callaghan writes, ‘it is not possible to divorce the practice of medicine from the society in which it is practised’.9 The papers in this issue are about primary health care practice in our New Zealand communities with our own patient populations. No longer do we need to rely on international research conducted in secondary care settings to inform our practice—the breadth and depth of New Zealand primary care research means our discipline has come of age.

intervention to improve immunisation coverage and timeliness. J Prim Health Care. 2012;4(3):189–198. 4. Hodgson F, Obertová Z, Brown C, Lawrenson R. PSA testing in general practice. J Prim Health Care. 2012;4(3):199–205. 5. Mules T, Taylor J, Price R, Walker L, Singh B, Newsam P, et al. Addressing patient alcohol use: a view from general practice. J Prim Health Care. 2012;4(3):217–222. 6. Gribben B, Salkeld L, Hoare S, Jones H. The incidence of acute otitis media in New Zealand children under five years of age in the primary care setting. J Prim Health Care. 2012;4(3):205–212. 7. Mathieson F, Mihaere K, Collings S, Dowell A, Stanley J. Maori cultural adaptation of a brief mental health intervention in primary care. J Prim Health Care. 2012;4(3):231–238. 8. Abel S, Marshall B, Rikki D, Luscombe T. Evaluation of Tu Meke PHO’s Wairua Tangata Programme: a primary mental health initiative for underserved communities. J Prim Health Care. 2012;4(3):242–248. 9. Callaghan K. Factors influencing diagnostic decision-making. J Prim Health Care. 2012;4(3):223–230. 10. Lillis S, Murton S. Does the order of presentation and number of online resources affect the frequency of access by learners? J Prim Health Care. 2012;4(3):213–216. 11. Kljakovic M. The anatomical placement of body organs by Australian and New Zealand patients and health professionals in general practice. J Prim Health Care. 2012;4(3):239–241. 12. Leung L. From ladder to platform: a new concept for pain management. J Prim Health Care. 2012;4(3):254–258. 13. Lee J, Loh N. Frequently asked questions on measurement of bone mineral densitometry. J Prim Health Care. 2012;4(3):259–261. 14. Dewes O. Hearing loss among Pacific peoples. J Prim Health Care. 2012;4(3):250–251.

References 1. Robinson T, Elley R, Wells S, Robinson E, Kenealy T, Pylypchuk R, et al. New Zealand Diabetes Cohort Study cardiovascular risk score for people with type 2 diabetes: validation In the PREDICT COHORT. J Prim Health Care. 2012;4(3):181–188. 2. Khunti K. Commentary: risk prediction models for people with Type 2 diabetes. J Prim Health Care. 2012;4(3):180. 3. Goodyear-Smith F, Grant C, Poole T, Petousis-Harris H, Turner N, Perera R, et al. Early connections: effectiveness of a pre-call

A tribute to Professor Marjan Kljakovic Just as this issue is going to press we have received the very sad news that Marjan, the author of one of the papers in this issue (Kljakovic M. The anatomical placement of body organs by Australian and New Zealand patients and health professionals in general practice. J Prim Health Care. 2012;4(3):239–241), died today, having suffered a major myocardial infarction on 29 July. A New Zealand academic general practitioner, Marjan relocated across the ditch seven years ago as Professor at the Academic Unit of General Practice and Community Health, Australian National University Medical School, Canberra. However, he remained a Kiwi at heart. Marjan was highly regarded as an inspirational thinker and an innovative teacher who contributed extensively to philosophical debate in the general practice arena. He will be sadly missed. Editor, 14 August 2012


EDITORIALS GUEST EDITORIAL

Commentary: risk prediction models for people with Type 2 diabetes Kamlesh Khunti PhD, MD, FRCGP, FRCP

Department of Health Sciences, University of Leicester, United Kingdom

J PRIM HEALTH CARE 2012;4(3):180.

CORRESPONDENCE TO: Kamlesh Khunti Professor of Primary Care Diabetes and Vascular Medicine, Department of Health Sciences, University of Leicester, Gwendolen Road, Leicester LE5 4PW, UK [email protected]

180

P

eople with Type 2 diabetes are at increased risk of cardiovascular disease, the determinants of which are multifactorial.1 A number of international guidelines recommend calculating future cardiovascular disease risk for management of patients with Type 2 diabetes. There has been a plethora of cardiovascular disease risk prediction models for Type 2 diabetes and a recent systematic review identified 45 prediction models, of which 12 were developed for patients with Type 2 diabetes.2 Less than onethird of these were externally validated in a diabetes population and overall the discriminative value for most prediction models was moderate.2 Another systematic review confirmed limited evidence of impact on patient management and outcomes with the use of prediction models.3

References 1. Sarwar N, Gao P, Seshasai SR, Gobin R, Kaptoge S. Diabetes mellitus, fasting blood glucose concentration, and risk of vascular disease: a collaborative meta-analysis of 102 prospective studies. Lancet. 2010;375:2215–22. 2. van Dieren S, Beulens JWJ, Kengne AP, Peelen LM. Prediction models for the risk of cardiovascular disease in patients with Type 2 diabetes: a systematic review. Heart. 2012;98:360–9. 3. Willis A, Davies MJ, Yates T, Khunti K. Primary prevention of cardiovascular disease using validated risk scores: a systematic review. Journal of the Royal Society of Medicine. 2012. 4. Robinson T, Elley C, Wells S, et al. New Zealand Diabetes Cohort Study cardiovascular risk score for people with Type 2 diabetes: validation in the PREDICT cohort. J Prim Health Care. 2012;4(3): 181–189.

In this issue of the journal, Robinson and colleagues have conducted a validation study of the Diabetes Cohort Study (DCS) CVD Risk Predictive model in people with Type 2 diabetes in New Zealand.4 The strengths of this study are the large numbers of people included, the long follow-up with 12.8% of people having a cardiovascular outcome, and the validation being conducted in a population in which the score was derived. The study found that the DCS model had marginally better discrimination than the currently used New Zealand Framingham risk equation. Overall the discriminative value was moderate. Many of the previous new scores have been compared with the well-established United Kingdom Prospective Diabetes Study risk score and, therefore, this is one limitation of this study. In addition, robust evidence on the impact of use of risk prediction models on patient outcomes in terms such as adherence to medications, patient understanding or improvements in harder outcomes is lacking. Until such evidence is available, the use of risk prediction models in routine clinical practice will not be adopted.


ORIGINAL SCIENTIFIC PAPERS QUANTITATIVE RESEARCH

New Zealand Diabetes Cohort Study cardiovascular risk score for people with Type 2 diabetes: validation in the PREDICT cohort Tom Robinson MBChB, MPH, FNZCPHM, FRNZCGP;1,3 C Raina Elley MBChB, FRNZCGP, PhD;1 Sue Wells MBChB, FRNZCGP, FNZCPHM, PhD;2 Elizabeth Robinson MSc;2 Tim Kenealy MBChB, FRNZCGP, PhD;1 Romana Pylypchuk MPH, MSc;2 Dale Bramley MBChB, MPH, FNZCPHM;3 Bruce Arroll MBChB, FRNZCGP, PhD;1 Sue Crengle MBChB, FNZCPHM, FRNZCGP, PhD;4 Tania Riddell MBChB, MPH, FNZCPHM;2 Shanthi Ameratunga MBChB, FRACP, FNZCPHM, PhD;2 Patricia Metcalf PhD;5 Paul L Drury MA, MB BChir, FRCP, FRACP6

Department of General Practice and Primary Health Care, The University of Auckland, Auckland, New Zealand 1

Section of Epidemiology and Biostatistics, The University of Auckland 2

Waitemata District Health Board, Auckland 3

ABSTRACT INTRODUCTION: New Zealand (NZ) guidelines recommend treating people for cardiovascular disease (CVD) risk on the basis of five-year absolute risk using a NZ adaptation of the Framingham risk equation. A diabetes-specific Diabetes Cohort Study (DCS) CVD predictive risk model has been developed and validated using NZ Get Checked data. AIM: To revalidate the DCS model with an independent cohort of people routinely assessed using PREDICT, a web-based CVD risk assessment and management programme. METHODS: People with Type 2 diabetes without pre-existing CVD were identified amongst people who had a PREDICT risk assessment between 2002 and 2005. From this group we identified those with sufficient data to allow estimation of CVD risk with the DCS models. We compared the DCS models with the NZ Framingham risk equation in terms of discrimination, calibration, and reclassification implications.

Te Kupenga Hauora Maori, The University of Auckland 4

Department of Statistics, The University of Auckland 5

Auckland Diabetes Centre, Auckland District Health Board 6

RESULTS: Of 3044 people in our study cohort, 1829 people had complete data and therefore had CVD risks calculated. Of this group, 12.8% (235) had a cardiovascular event during the five-year follow-up. The DCS models had better discrimination than the currently used equation, with C-statistics being 0.68 for the two DCS models and 0.65 for the NZ Framingham model. DISCUSSION: The DCS models were superior to the NZ Framingham equation at discriminating people with diabetes who will have a cardiovascular event. The adoption of a DCS model would lead to a small increase in the number of people with diabetes who are treated with medication, but potentially more CVD events would be avoided. KEYWORDS: Cardiovascular disease; diabetes; prevention; risk assessment; reliability and validity

Introduction Globally there is an epidemic of Type 2 diabetes.1,2 It was estimated that in 2010 there were over 195 000 people in New Zealand (NZ) with diabetes—5.6% of the adult population.3 People with diabetes are at increased risk of dying of cardiovascular disease (CVD) which accounts for almost 50% of all deaths amongst people with diabetes.4,5

There is considerable evidence that energetic management of risk factors such as blood pressure, dyslipidaemia, and glycaemia reduces the risk of CVD in people with diabetes.6–11 However, it is accepted that rather than treating risk factors separately, clinicians should use absolute CVD risk to guide patient management.12,13 NZ guidelines for cardiovascular risk assessment use a predictive risk equation adapted from

J PRIM HEALTH CARE 2012;4(3):181–188.

CORRESPONDENCE TO: Tom Robinson Department of General Practice and Primary Health Care, The University of Auckland, PB 92019 Auckland, New Zealand [email protected]



the United States Framingham Heart Study.14,15 This equation has a number of disadvantages for predicting risk of CVD amongst people with diabetes in NZ. In particular, the Framingham cohort was from the United States, did not include ethnic groups that are important in NZ, and only included a small number of people with diabetes.16 In addition, the equation does not include a number of diabetes-specific variables— such as duration of diabetes, glycaemic control, and albuminuria—that are predictive of cardiovascular outcomes.17–20 The NZ adaptation of the Framingham model does include adding a single additional five-year risk of 5% for these factors.14 In 2010, Elley et al. reported two predictive CVD equations based on the New Zealand Diabetes Cohort Study (DCS). This was a prospective open cohort that used data from a national primary care diabetes programme (Diabetes Get Checked), which commenced in 2000.21 Full details of the derivation and validation of the equation are described in the original article.21 Briefly, data from 36 127 people with Type 2 diabetes, but without pre-existing CVD, were matched to national hospitalisation and mortality databases. Predictor variables for the first equation (DCS-A) included age at diagnosis, duration of diabetes, sex, ethnicity, smoking status, systolic blood pressure, HbA1c, total cholesterol: HDL cholesterol ratio (TC/HDL), and the presence of microor macroalbuminuria. A second equation (DCS-B) also included current antihypertensive treatment. The performance of both equations was tested on 10 030 individuals from a different geographic area in NZ with discrimination and calibration superior to the original Framingham equation.21 Before using a prognostic model in clinical practice it is important to validate it using data from other independent populations of patients.22 This study aimed to validate the DCS models using data from a cohort of people routinely assessed in NZ general practice with PREDICT, a CVD risk assessment and management programme.

Methods Design This validation study uses data from primary care to assess the discrimination, calibration and

182

reclassification implications of the DCS equations in predicting CVD events, compared with actual events over five years.

Study population PREDICT is a web-based, real-time decision support programme that has been integrated with most practice management software in use in NZ primary care.23 General practitioners and practice nurses enter required clinical data to create a risk profile. This profile is sent by a secure internet connection to a central server that returns the patient’s NZ Framingham five-year cardiovascular risk score with management recommendations. At the same time, an electronic profile is stored and linked to an encrypted National Health Index (NHI) number. These were anonymously linked to national hospitalisation, pharmaceutical dispensing and mortality outcomes and also to regional laboratory data. Individuals identified as having Type 2 diabetes and no known pre-existing CVD with a PREDICT assessment between 27 August 2002 and 31 December 2005 were included. Individuals were said to have diabetes if they were identified by their primary care physician as having diabetes at first risk assessment, or if they had been identified as having diabetes in the national hospitalisation database, or had been prescribed insulin or an oral hypoglycaemic agent prior to or on their first PREDICT assessment date. If the type of diabetes was unclear, we assigned them as Type 2 if they were never on insulin, if they had been on an oral hypoglycaemic agent, or if their age of onset was over 30 years in Maori and Pacific or over 50 in other ethnic groups. Preexisting CVD was identified from the primary care physician’s risk assessment record.

Risk variables Risk factor variables required for the DCS equations were extracted for each individual. Data on some of the variables were missing from early PREDICT risk assessments. Duration of diabetes was included if it could be calculated from any subsequent PREDICT risk assessment record. Missing laboratory data were obtained from laboratory records where results from up to five years prior to the baseline assessment or two



weeks after the assessment were available. After the addition of these data, only individuals with a complete minimum dataset were included in the final study cohort. Ethnicity was derived from both the primary care practitioner records and the encrypted NHI database and was prioritised in the order: Maori, Pacific, South Asian, East Asian, ‘Other’ and European.

Outcome measures Primary care data were linked to national hospitalisation and mortality databases by NHI number to identify all CVD events over the five years following baseline for each individual. CVD events included hospital admission or death from ischaemic heart disease, cerebrovascular disease or peripheral vascular disease. These were identified from national hospital and mortality databases coded according to ICD-9 and ICD-10 (see the appendix in the web version of this paper).21 Five-year risk was calculated for each individual according to both the NZ Framingham and the DCS equations.21

Analyses We compared predicted risk with observed outcomes. To assess discrimination, the ability of the models to distinguish between individuals who do or do not have a subsequent CVD event, we calculated the area under the receiver operating characteristic (ROC) curve (C statistic).13,22,24 Calibration was assessed by comparing the observed and predicted probabilities of CVD events in the pre-specified deciles of DCS model risk, and performing a Hosmer–Lemeshow test for equivalence. The effect of reclassification of risk from the NZ Framingham model to the DCS models was measured using a 15% five-year cardiovascular risk threshold. NZ guidelines recommend drug treatment with five-year risks above 15%. A scatter plot of risks predicted by the two models with these pre-determined risk categories was also produced.24 All analyses were undertaken using Stata® 11.2.

Ethical approval This validation study was approved by the Multiregion Ethics Committee (WGT/04/09/077) as part of the Diabetes Cohort Study. The PREDICT cohort study and research process was ap-

WHAT GAP THIS FILLS What we already know: People with Type 2 diabetes are at high risk of a cardiovascular event. A locally derived Diabetes Cohort Study CVD risk equation—http://www.nzssd.org.nz/cvd/—has been found to be more valid for those with diabetes in New Zealand than the currently used Framingham equation. What this study adds: Before incorporating the new equation into national recommendations for management, further validation was required using an independent cohort. The Diabetes Cohort Study CVD risk equation predicted risk more accurately than the currently used adjusted Framingham equation among people with diabetes in the New Zealand PREDICT cohort.

proved by the Northern Region Ethics Committee Y in 2003 (AKY /03/12/314) with subsequent annual approval by the National Multi-region Ethics Committee since 2007 (MEC/07/19/EXP).

Results Study population The derivation of the study cohort and subsequent CVD events is shown in Figure 1. We classified 3044 (13.3%) people on the database as people with Type 2 diabetes without pre-existing CVD. Of these, 1829 (60.1%) had the minimum dataset of risk variables required and formed the final cohort for the study. About two-thirds of these individuals (65.9%) were included after having data added from sources other than the first risk assessment record. These data were diabetesspecific variables (HbA1c, urinary albumin/creatinine ratio, diabetes duration, and age of onset of diabetes). All individuals were followed for five years from their initial CVD risk assessment. During that time, 235 had first CVD events (12.8%), in which 45 (2.5%) were fatal and 190 (10.4%) were non-fatal events. Baseline characteristics of participants are presented and compared with those of the 1215 people excluded due to missing variables in Table 1. Compared with those included, a higher proportion of the excluded group were European and a lower proportion Pacific. Excluded participants had slightly higher systolic blood pressures and TC/HDL ratios and were less likely to be recorded as being past smokers. Although the two groups had similar risks of CVD events



Discrimination

Figure 1. Flow diagram of participants through study

24 446 people with first PREDICT risk assessment between 27 August 2002 and 31 December 2005

851 people with Type 2 diabetes and pre-existing CVD

20 551 people with Type 2 diabetes

3044 people with identified Type 2 diabetes

1215 people without minimum dataset

1829 (60.1%) people with minimum* datasets 34.1% complete data from first PREDICT assessment 65.9% duration of diabetes derived from subsequent assessments

Calibration

65.4% HbA1c from urinary ACRs variables derived from other laboratory data 0.5% derived from subsequent PREDICT assessments within 1 year

26 died from non-CVD causes

235 (12.8%) people had first CVD event within five years of date of their first PREDICT risk assessment (45 fatal and 190 non-fatal)

Figure 3 compares the mean predicted risk with the mean observed five-year event rate for each decile of predicted risk for DCS-B and NZ Framingham equations. The DCS equations predicted higher risks than the NZ Framingham equation for people in the higher deciles of risk. The Hosmer–Lemeshow test showed that estimated risks based on the baseline risk profile tended to be higher than the real event rate for all equations (p