Up to twenty-six general practices in Coventry and Warwickshire who use EMIS LV software will be invited to participate in the trial. Dr Tim Holt will visit each practice to explain the trial and gain written consent from the general practitioners.
Participating practice patients over 50 years of age will be randomised into two groups – "intervention" and "control." Patients will be consistently allocated to these groups throughout the study using an electronic technique that is concealed to all researchers and practitioners involved. This process will occur electronically during each search, so that those who join the practice during the study will be randomised automatically as soon as they are first provided with electronic notes as a fully registered patient. Temporary residents are not included in the study.
Applying the search strategy
Alerts will be created automatically using patient information that is updated in real time and the search protocol described in Figures , , . For the eight-weekly lists we will apply the same search protocol to the databases of participating practices. This will produce lists for each practice of the high modifiable risk patients in the intervention arm of the study on the day of the search. The groups identified can be summarised as:
Figure 1 Search algorithm to identify those most likely to benefit from cardiovascular prevention based on recent risk variable values. Definitions for terms in inverted commas are given in the appendix along with justification of thresholds and search protocol. (more ...)
GROUP 1: Patients of all ages with existing cardiovascular disease or diabetes, whose blood pressure or cholesterol level is outside the QOF target range at the last estimation, or no "in date" level is recorded.
GROUP 2: Patients who are not known to have cardiovascular disease or diabetes, are under 75 yrs old, and whose risk profile is incomplete – more information is required to perform a risk estimate – but whose cardiovascular risk would be greater than 20% if the "assumed" values of the missing factors are used (see definition in appendix).
GROUP 3: Patients who are not known to have cardiovascular disease or diabetes, are under 75 yrs old, and whose most recent risk variable values indicate that their risk level is raised.
GROUP 4: Patients who are not known to have cardiovascular disease or diabetes, are greater than 75 yrs old and who have persistently elevated blood pressure based on the three most recent consecutive readings.
GROUP 5: Patients with possible undiagnosed diabetes on the basis of at least one previous high blood glucose record.
GROUP 6: Patients with CVD but not diabetes, who have not had a blood glucose measurement in the past three years.
Information on the "intervention" patients identified at each search is revealed to the practices. Information on the control patients including the number identified will be saved but no action will be triggered (Figure ).
Eight-weekly searches on practice databases.
Intervention – the "e-Nudge"
The e-Nudge is an automated feedback system that examines information already contained in practice databases to help practice teams target preventive interventions towards the individuals most likely to benefit. At the same time, the e-Nudge identifies clinically important missing risk variable values and patients with possible undiagnosed diabetes. Designed to run as a series of updated alert messages and searches that use most recent risk variable values, it is able to track practice populations over time as patients enter and leave the area, grow older, and enter practice disease registers, such as those for diabetes, CHD, or stroke. It recognises that risk profiles are dynamic, and that "one-off" estimates of risk in individuals are liable to become outdated [33
The alert messages will arise automatically through EMIS software when a high risk patient's notes are opened, and are continuously updated in real time. To identify patients who may not present to the practice, electronic searches will be undertaken every eight weeks. The purpose of both alerts and the lists is to trigger awareness of individual patients' risk within the practice team, and not to dictate specific treatments. The "e-Nudge" is therefore simply the feedback of this information. The resulting action is at the discretion of the practice team, and can be tailored both to the time available, and to the needs and preferences of the individual patient in the context of the clinician's broader knowledge of co-morbidity, current medication, and past response to treatment. The practice teams will have the following notifications for intervention patients identified in the searches.
1 The eight-weekly search result is presented to a nominated member of the primary care team under the six group headings.
2 Alert messages are displayed automatically on the computer screen each time the patients' electronic notes are opened. These are triggered for those identified in any of the groups:
• Group 1 patients: This CHD/Stroke/Diabetes (state which) patient's (BP) or (serum cholesterol) level (specify which) is out of the target range.
• Group 2 patients: This patient may be at high cardiovascular risk, but values for the following risk variables were either missing or out of date: (specify which variables).
• Group 3 patients: This patient's estimated cardiovascular risk may be elevated, based on the most recent risk variable values.(State assumptions made)
• Group 4 patients: This patient's blood pressure is persistently elevated based on three consecutive values.
• Group 5 patients: This patient may have undiagnosed diabetes, based on a previous raised blood glucose level ≥ 11.1 mmol/L.
• Group 6 patients: This CHD/Stroke patient (state which) has no recorded blood glucose measurement in the past three years.
Control patients at high estimated risk will be identified but the practice teams will not be provided with these extra reminders, although the team will have access to all the clinical information used to assess risk status. Control patients will continue to receive the usual care provided by current general practice under the nGMS contract. Some practices have started to use alerts for CVD or Diabetes patients who are out of the nGMS blood pressure and cholesterol targets since this study was conceived. Where this is now 'usual care,' this part of the intervention (Group 1 alerts) will not be withheld from the control patients, but the rest of the e-Nudge (including identification on the eight-weekly searches) will be. The standard of care is high in the study locality [South Warwickshire Primary Care Trust, QOF data on file], providing a suitable environment to test the e-Nudge. If the study shows a positive effect, this will demonstrate that even good care can be improved, and it is anticipated that the tool will be even more effective in environments where care is of a lower standard.
The trial has been developed in accordance with the Declaration of Helsinki, and approved by Warwickshire Local Research Ethics Committee (Ref: 05/Q2803/85).
The searches and alerts will continue for a period of two years, at the end of which the data will be examined. We will continue to collect and analyse data on the primary and secondary outcomes of the study for a further year after this. Outcomes will be measured using searches on practice databases. Analysis will be undertaken on an "Intention To Treat" basis within practices. Practices that withdraw will have their data censored from the date of withdrawal from the trial.
Difference in the annual incidence rate of cardiovascular events (see definition in the appendix) in the intervention and control populations during the two years of the study, and for a third year following the end of the e-Nudge intervention.
• Difference in the proportion of high risk patients (Groups 1, 3 and 4) identified in the control and intervention populations averaged over the last three searches in the two year intervention period, and in the third year following the end of the intervention.
• Difference in the proportion of patients in each population identified with missing data (Groups 2 and 6) averaged over the last three searches in the two year intervention period, and in the third year following the end of the intervention.
• Difference in the proportion of patients with undefined diabetes status (i.e. raised blood glucose levels with no diagnosis of diabetes and no FBG or OGTT results to confirm status) (Group 5) in the intervention and control populations averaged over the last three searches in the two year intervention period, and in the third year following the end of the intervention.
Analysis of the data will be carried out in STATA. The principle analyses will be on an intention-to-treat basis and will be performed using the CONSORT guidelines (2001) [34
Data monitoring committee
Outcomes will be assessed annually during the study by an independent data monitoring committee, who will inform the trial investigators if the trial should terminate early on ethical grounds due to a 20% difference in mortality or morbidity between the intervention and control groups.
Data quality assurance measures
We will examine the cause of death of every patient in the practices over age 50 years who dies during the study, to ensure that all cardiovascular deaths are recorded appropriately in searchable form prior to outcome data extraction. Any patient recorded as having more than one cardiovascular event during the study will have their clinical record examined, to identify patients who have had the same event recorded twice (which may happen when a consultation for a stroke, TIA or myocardial infarction is mistakenly labelled as a "new episode" rather than a "review"). This process will be carried out both on controls and intervention patients. In addition, we will examine the notes of any patient who has a record of an event dated within 4 months of registration with a practice, in case this event occurred in the past but was incorrectly dated when the patient registered.
Sample size calculation
Our study defines a cardiovascular event as a new diagnosis of CVD, a new myocardial infarction, a new stroke, a new transient ischaemic attack, or sudden death from CVD. A new stroke in someone with a previous stroke will count as a new event. An acute myocardial infarction in a patient previously diagnosed with angina will be recorded as a new event, but a new onset of angina in a patient who already had a diagnosis of acute myocardial infarction might not be recorded as a new diagnosis, as the patient will already be on the CHD register.
The British Heart Foundation [35
] has compiled an estimate of the number of cardiac events in the UK population in 2002 from several available data sources. The number of myocardial infarctions (all ages) was estimated to be 268,000, while the number of new cases of angina (all ages) was estimated to be 338,000
The UK population was 59,321,700 in 2002 [Sources: Office for National Statistics, General Register Office for Scotland, Northern Ireland Statistics and Research Agency], so estimated incidence rates for coronary heart disease are
Incidence of myocardial infarction 451.77 per 100,000
Incidence of new case of angina 569.77 per 100,000
For cerebrovascular disease
, the OXVASC study [36
] provides a local source of information drawn from an Oxfordshire population. The incidence rates were:
Incidence of stroke 187 per 100,000
Incidence of TIA 51 per 100,000
Incidence of all cardiovascular events 1260 per 100,000
We aim to demonstrate at least a 10% reduction in the cardiovascular event rate. This means that for a positive outcome, the event rate in the intervention population must be ≤90% of the event rate in the control population. We therefore estimated the necessary sample size for this reduction to be detected at the 5% level with 80% power.
Estimating population size needed
A Poisson distribution model is appropriate for events that are rare on an individual level, occurring randomly and independently at a constant rate in a population [37
Assuming a Poisson distribution, the formula for the sample size is:
λ0 = the expected incidence of cardiovascular events (i.e. 1260/100,000)
δ = new incidence in the intervention group
z1-α = standardised normal distribution value based on 0.05 significance level
z1-β = standardised normal distribution values for 80% power
N = total number of patients required in the study
Nw = total number of patients required in the study + 10% to account for practice withdrawal
For 80% power and 0.05 significance level (2-tailed) [38
] (see ):
The practice population required to detect both statistically and clinically significant changes in the cardiovascular event rate is therefore estimated to be approximately 70,000, the combined list size of all age groups in participating practices.