General Practice

Identification of patients with atrial fibrillation in general practice: a study of screening methods

BMJ 1998; 317 doi: (Published 01 August 1998) Cite this as: BMJ 1998;317:327
  1. Mark Sudlow, MRC special training fellow in health services research,
  2. Helen Rodgers, senior lecturer in stroke medicine and services,
  3. Rose Anne Kenny, professor of cardiovascular research,
  4. Richard Thomson, senior lecturer in public health medicine. (Mark.Sudlow{at}
  1. School of Clinical Medical Sciences, University of Newcastle upon Tyne,Newcastle upon Tyne NE2 4AA
  1. Correspondence to: Dr M Sudlow, Department of Epidemiology and Public Health, School of Clinical Medical Sciences, University of Newcastle upon Tyne, Newcastle upon Tyne NE2 4AA
  • Accepted 23 April 1998


Atrial fibrillation is common, affecting around 5% of people over 65. 1 2Widespread use of anticoagulants in these patients could greatly reduce the incidence of stroke,3 but many patients are untreated. 2 4Although most people with atrial fibrillation are already recorded as having the condition,4 they may not be easily identifiable from medical records, and this may partly explain the underuse of anticoagulants.

As part of a population survey of elderly people we examined two methods for detecting people with atrial fibrillation or flutter: identification of patients currently taking digoxin, and pulse palpation by a trained nurse. Ethical approval was granted by the Northumberland Local Research Ethics Committee.

Methods and results

We invited an age and sex stratified sample of 1235 subjects aged 65 years and over, registered with nine contiguous general practices in southern Northumberland, for a screening limb lead electrocardiogram. Subjects were asked to bring any medication they were taking, and this was recorded. A nurse palpated the pulse and recorded its character. A pulse that was not “regular” was considered abnormal.

The ability to detect cases of atrial fibrillation or flutter by searching for digoxin prescriptions and by pulse palpation was compared with the results of the electrocardiograms, which were considered the optimal test. We also considered the effect of using both screening methods together. Confidence intervals around the test characteristics were calculated with Confidence Interval Analysis software.

The response rate to the survey was 74% (916/1235). As the predictive values of tests vary with the prevalence of the condition studied, and therefore with age and sex, the table shows test characteristics for each stratum separately. The sensitivity of using digoxin prescriptions as an indicator of atrial fibrillation was around 50% in most strata, and the specificity of this method was over 95% in all strata. The sensitivity of pulse palpation was over 90% in all groups, but the specificity of this method fell to 71% in the more elderly groups. Using both methods together produced similar results to using pulse character screening alone.

Characteristics of methods of detecting people with atrial fibrillation, by age. Values are percentages (95% confidence intervals)

View this table:


This paper reports the sensitivity, specificity, and positive and negative predictive values of two simple methods for detecting patients with atrial fibrillation or flutter. These test characteristics can be greatly affected by the prevalence of the condition of interest. The population we studied was representative of patients in primary care, and our results could be used by general practitioners to estimate the implications of screening in their practices.

Searching for digoxin prescriptions would be relatively simple but would detect only about half of people with atrial fibrillation. Recording the character of the pulse would detect almost all cases, but with a larger number of false positives. This could be done as part of a special screening programme, during routine health checks for elderly people, or opportunistically. Combining both methods provides no advantage over pulse screening alone in terms of test characteristics, but screening using prescriptions could be performed quickly, allowing a proportion of patients needing anticoagulation to be treated earlier than with pulse screening alone.

Atrial fibrillation or flutter fulfils most of the criteria set out by Cuckle and Wald for a worthwhile screening programme,5 but controlled trials of the effect of screening on clinical outcomes are needed. For practices that wish to detect cases of atrial fibrillation or flutter in advance of such trials, a combination of searching for digoxin prescriptions and opportunistic pulse palpation would be a practical approach.


We thank Christine Burridge, Sheena Burton, Ruth Dobson, Caroline Dowell,Jill Robinson, and Dawn Winpenny for their work on this project, and the general practitioners, medical receptionists, practice managers and subjects who were involved with the study for their generous help.

Contributors: MS contributed to the design of the study and analysis, and collected and analysed data. RT was responsible for suggesting and developing the analysis presented in this paper. RT, HR, and RAK developed the initial ideas behind the study, contributed to its design, and supervised data collection and analysis. All authors were involved in interpreting data and writing this paper. MS acts as guarantor.

Funding: The study was funded by the Stroke Association. MS was funded by an MRC special training fellowship in health services research.

Conflict of interest: None.


View Abstract