Study design and population
We performed a prospective cohort study in the southwest of the Netherlands. The echocardiography service was made available by SHL-Groep in Etten-Leur, a diagnostic centre which provides support services to primary care. Information on the service and an invitation to join a symposium in which the service was presented was sent to 316 GPs from 181 family practices. GPs were informed of the application procedure and restrictions for referral; patients younger than 18 years and patients with suspected acute cardiac conditions were not allowed to participate.
Patients referred to SHL-Groep for echocardiography between April 2011 and April 2012 were eligible. The Medical Ethics Committee of the Maastricht University Medical Centre approved the study.
Measurements and variables
We provided GPs with standardised request forms with tick boxes for the indication, relevant medical history, signs and symptoms, and results of diagnostic tests, i.e. electrocardiogram (ECG), N-terminal pro-brain natriuretic peptide (NT-proBNP), and chest X-ray. On the application form we asked what the GP would have done with this patient if open access echocardiography had not been available. In case of incomplete forms, we contacted the GP by telephone before echocardiography results were known.
The telephonists of the SHL-Groep filled out a list every Monday, describing the waiting time for an appointment for echocardiography at that time.
Echocardiography was performed every Thursday by one of two ultrasound technicians using a Philips CX50 device. The images were parasternal long axis (PLAX), parasternal short axis (PSAX), and apical two, three, four, and five chamber views (2CH, 3CH, 4CH, 5CH). The inferior vena cava (IVC) was visualised subcostally. The technician posted the echocardiogram in a portal on a secured website.
One of two participating cardiologists from Erasmus Medical Centre evaluated the images within three working days. The cardiologists had access to the indication and patient reported length and weight. They classified systolic, diastolic and valve function according to the criteria of the American Society of Echocardiography and the European Association of Echocardiography [9–11]. Systolic function was determined with eyeballing, which was less time consuming than calculating ejection fraction . Diastolic function was measured using mitral inflow (E/A ratio), tissue Doppler imaging of the mitral annulus (E/E’ ratio), left atrial diameter, and left ventricle wall thickness. Colour flow Doppler was used to visualise the flow through the valves; the severity of valve disease was determined by eyeballing. The sniff test was used to measure IVC collapse. The echocardiography results with the conclusion and advice of the evaluating cardiologist were sent to the GP.
After the GP had received the results and contacted the patient, we asked in a telephone interview what management was initiated. We also inquired whether the GP had followed the evaluating cardiologist’s advice, and whether they thought the echocardiogram had been of benefit. Finally, we asked GPs to estimate the waiting time for receiving an echocardiogram if patients would have been referred to a regional cardiologist.
To significantly demonstrate the expected difference between hypothetical referral to a regional cardiologist and actual referral after echocardiography, 26 patients were required. We calculated the sample size for McNemar’s testing. We aimed at a power of 0.90 and a level of statistical significance of 0.05 for double-sided testing (α = 0.05). We expected a 0.75 probability of referral before echocardiography, and a 0.25 probability of referral after echocardiography. Since patients were not burdened because of our study, we included more patients for the descriptive analyses.
We used SPSS version 19.0 for Windows. Missing values were assumed to be missing completely at random. We checked outliers with the original data and corrected them if they proved to be wrongly copied. We used descriptive statistics to calculate frequencies. Confidence intervals are shown for continuous variables. With McNemar’s test for discrete paired variables we investigated whether GPs would refer less to cardiologists with open access echocardiography available and manage more patients by themselves than they otherwise would have done.
Referring more than one patient per GP could cause a bias (‘nesting’) because patient management after echocardiography is GP-dependent. We assessed nesting by including only the first referred patient of each GP and compared this with the analysis including all eligible patients.
To calculate the GP participation rate the denominator was determined as all family practices within the range of 38 km from the diagnostic centre (N = 172). This was the distance to the remotest referring practice ‘as the crow flies’. This was done because not every family practice informed of the existence of the service (N = 181) could reasonably be expected to refer to the echocardiography service because of the long distance.