CHARGE-AF in a national routine primary care electronic health records database in the Netherlands: validation for 5-year risk of atrial fibrillation and implications for patient selection in atrial fibrillation screening

Aims To validate a multivariable risk prediction model (Cohorts for Heart and Aging Research in Genomic Epidemiology model for atrial fibrillation (CHARGE-AF)) for 5-year risk of atrial fibrillation (AF) in routinely collected primary care data and to assess CHARGE-AF’s potential for automated, low-cost selection of patients at high risk for AF based on routine primary care data. Methods We included patients aged ≥40 years, free of AF and with complete CHARGE-AF variables at baseline, 1 January 2014, in a representative, nationwide routine primary care database in the Netherlands (Nivel-PCD). We validated CHARGE-AF for 5-year observed AF incidence using the C-statistic for discrimination, and calibration plot and stratified Kaplan-Meier plot for calibration. We compared CHARGE-AF with other predictors and assessed implications of using different CHARGE-AF cut-offs to select high-risk patients. Results Among 111 475 patients free of AF and with complete CHARGE-AF variables at baseline (17.2% of all patients aged ≥40 years and free of AF), mean age was 65.5 years, and 53% were female. Complete CHARGE-AF cases were older and had higher AF incidence and cardiovascular comorbidity rate than incomplete cases. There were 5264 (4.7%) new AF cases during 5-year follow-up among complete cases. CHARGE-AF’s C-statistic for new AF was 0.74 (95% CI 0.73 to 0.74). The calibration plot showed slight risk underestimation in low-risk deciles and overestimation of absolute AF risk in those with highest predicted risk. The Kaplan-Meier plot with categories <2.5%, 2.5%–5% and >5% predicted 5-year risk was highly accurate. CHARGE-AF outperformed CHA2DS2-VASc (Cardiac failure or dysfunction, Hypertension, Age >=75 [Doubled], Diabetes, Stroke [Doubled]-Vascular disease, Age 65-74, and Sex category [Female]) and age alone as predictors for AF. Dichotomisation at cut-offs of 2.5%, 5% and 10% baseline CHARGE-AF risk all showed merits for patient selection in AF screening efforts. Conclusion In patients with complete baseline CHARGE-AF data through routine Dutch primary care, CHARGE-AF accurately assessed AF risk among older primary care patients, outperformed both CHA2DS2-VASc and age alone as predictors for AF and showed potential for automated, low-cost patient selection in AF screening.

-Height: latest recording in 2013 of data code 560 (height). When an entry for data code 560 was absent, but data codes for weight and BMI were both present in 2013, we calculated height in centimeters as 100*√(weight/BMI) and used the latest recordings in 2013. In order to include only realistic values, and to prevent inclusion of values erroneously entered by GP personnel, we included only height values 130-230cm. Values below 130 were multiplied by 100 in order to include data entered as meters instead of centimeters. We subsequently applied the same limits of 130-230cm; -Antihypertensive medication: ATC subcodes for C02 (antihypertensives) and/or C03 (diuretics), C04 (peripheral vasodilators), C05 (vasoprotectives), C07 (beta blocking agents), C08 (calcium channel blockers), or C9 (agents acting on the renin-angiotensin system); -Hypertension: entry of ICPC-1 codes K86 (uncomplicated hypertension) and/or K87 (hypertension with involvement target organs) or data code 1694 (hypertension comorbidity); -Diabetes mellitus (DM): entry of ICPC-1 code T90 (DM) and/or data code 2206 (treating physician for DM); -Heart failure (HF): entry of ICPC-1 code K77 (HF) and/or data codes 3016 (treating physician for HF), 2722 (NYHA severity of HF symptoms) or 1643 (HF comorbidity); -Myocardial infarction (MI): entry of ICPC-1 code K75 (acute MI) and/or data code 1693 (MI comorbidity); -Current smoking: classified as current smoker when indicated as smoker as per data codes 1739 (smoking) and/or 1992 (number of (rolling tobacco) cigarettes per day), 1993 (number of cigarettes per day), 1996 (wants to quit smoking in short term) or 2405 (motivation to quit smoking), and not followed in time (but before 01-01-2014) by an indication of having quit smoking as per data codes 1739 (smoking) and/or 2003 (quit smoking since); -Stroke: entry of ICPC-1 code K90 (stroke/cerebrovascular accident) and/or lab code 2132 (cerebral ischaemia history comorbidity); -Transient ischemic attack (TIA): entry of ICPC-1 code K89 (transient cerebral ischaemia); -Pulmonary embolism (PE): entry of ICPC-1 code K93 (PE); -Angina pectoris: entry of ICPC-1 code K74 (angina pectoris); -Vascular disease: entry of ICPC-1 codes K74 (angina pectoris) and/or K91 (atherosclerosis), K92 (other arterial obstruction/peripheral vascular disease) or MI as defined above; -Congestive heart failure, Hypertension, Age, Diabetes and previous Stroke or Transient Ischaemic Attack, Vascular disease and female Sex category (CHA2DS2-VASc): 1 point for each of female sex, HF, hypertension, DM, vascular disease or age 65-74 years, plus 2 points for each of (stroke, TIA or PE) or age ≥75 years; -Asthma: entry of ICPC-1 code R96 (asthma) and/or indication for asthma as per data codes 1598 (asthma diagnosed by) and/or 1599 (asthma goals attained), 1618 (medication adherence asthma), 1621 (avoids provoking factors asthma), 1716 (reason for failure to achieve asthma goals), 1776 (asthma management), 1806 (change asthma medication), 1822 (asthma severity), 1824 (asthma self-management), 1826 (appointment for asthma self-management), 1877 (asthma comorbidity), 2406 (treating physician for asthma), 3018 (adverse effects asthma medication), 3608 (degree of control in asthma management), 3338 (ACQ question 1), 3339 (ACQ question 2), 3340 (ACQ question 3), 3341 (ACQ question 4), 3345 (C-ACT question 1), 3346 (C-ACT question 2), 3347 (C-ACT question 3), 3348 (C-ACT question 4), 3349 (C-ACT question 5), 3828 (enrolment in care program for asthma); -Chronic obstructive pulmonary disease (COPD): entry of ICPC-1 code R95 (COPD) and/or indication for COPD as per data codes 1779 (medication adherence COPD) and/or 1785 (COPD management), 1786 (causes for COPD exacerbation), 1807 (change COPD medication), 1818 (reason not to enrol in COPD care program), 1909 (reasons for not attaining COPD goals), 1911 (COPD diagnosed by), 2209 (GOLD classification COPD), 2399 (mean symptom score CCQ COPD), 2400 (mean function score CCQ COPD), 2401 (mean psychological score CCQ COPD), 2402 (mean limitations score CCQ COPD), 2407 (treating physician COPD), 2676 (cachexia COPD), 3013 (COPD disease burden), 3019 (adverse effects COPD medication); -Atherosclerosis: entry of ICPC-1 code K91 (atherosclerosis); -Hypercholesterolaemia: entry of data code 2053 (hypercholesterolaemia comorbidity) and/or value for data code 181 (cholesterol/HDL ratio) ≥5 mmol/L; -Gout: entry of ICPC-1 code T92 (gout); -Enrolment in care program for asthma: indication for enrolment in care program for asthma as per data codes 2406 (treating physician for asthma) and/or 3828 (enrolment in care program for asthma); -Enrolment in care program for COPD: indication for enrolment in care program for COPD as per data codes 2407 (treating physician for COPD) and/or 3829 (enrolment in care program for COPD); -Enrolment in care program for DM: Enrolment in care program for COPD: indication for enrolment in care program for DM as per data codes 2206 (treating physician for DM) and/or 3827 (enrolment in care program for DM); -Enrolment in care program for any care program: indication for enrolment in one or more care programs of asthma, COPD or DM as defined above, or for indication for enrolment in care program for HF as per data codes 3016 (treating physician for HF) and/or 3833 (enrolment in care program for HF), or for indication for enrolment in care program for thyroid disease as per data codes 3040 (treating physician for thyroid disease) and/or 3835 (enrolment in care program for thyroid disease).
BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s)

Comparison of patients with and without complete baseline CHARGE-AF data
Supplementary Table 1 shows a comparison between those free of AF at baseline with complete baseline CHARGE-AF data and those free of AF at baseline without complete baseline CHARGE-AF data (n=538,308). Five-year AF incidence was significantly lower among incomplete CHARGE-AF cases (2.10%, p<0.001). Patients with complete CHARGE-AF baseline data were significantly older and had significantly higher burden of cardiovascular comorbidities than patients with incomplete CHARGE-AF variables at baseline. The percentage of missing CHARGE-AF measurements varied from 69.3% (SBP) to 81.3% (height). Patients with at least 1 but not all 4 CHARGE-AF measurements recorded in the EHR in 2013 had a higher mean SBP, DBP and height, but lower weight, than patients with complete baseline CHARGE-AF measurements.

Additional CHARGE-AF validation analyses
In the stratified analyses on CHARGE-AF, discrimination was consistently higher in the lower risk groups (women, age <65 years and CHA2DS2-VASc <2), with highest C-statistic in the subgroup of women (0.751; 95%CI: 0.740-0.763). Calibration of CHARGE-AF was insufficient in all subgroups as assessed by the Nam-D'Agostino χ 2 , and the calibration slope significantly deviated from 1 in all subgroups except in patients younger than 65 and in patients with CHA2DS2-VASc <2 (see Table 2 in main text).
Calibration plots for the stratified CHARGE-AF analyses were similar to that of the overall analysis, except in the subgroups age <65 years and CHA2DS2-VASc <2. In these lower risk strata, risk prediction was accurate for all deciles, without overestimation in the highest deciles seen in the other analyses (Supplementary Figure 4).

Supplementary Figure 2. Baseline CHARGE-AF risk distribution in the sample and relative contribution of CHARGE-AF risk factors to increments in baseline risk (n = 111,475 with complete baseline CHARGE-AF data)
AHM, antihypertensive medication use; CHARGE-AF, Cohorts for Heart and Aging Research in Genomic Epidemiology model for AF; DBP, diastolic blood pressure; DM, diabetes mellitus; HF, heart failure; MI, myocardial infarction; SBP, systolic blood pressure. Panel A, Baseline CHARGE-AF risk distribution; Panel B, Relative contribution of CHARGE-AF risk factors to mean baseline CHARGE-AF risk score in successive strata of increased CHARGE-AF risk. Since DBP has a negative coefficient in the CHARGE-AF formula, DBP is depicted as such in this graph.  Open Heart , et al.

Himmelreich JCL
The points indicate intersects of observed and expected for each decile of baseline CHARGE-AF risk, with brackets indicating the 95% confidence intervals of observed AF probability during 5-year follow-up in each decile. The red line indicates the trend for CHARGE-AF calibration in the sample. The spikes on the x axis indicate the distribution of AF-free survivors by CHARGE-AF risk.
BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s)