Effect of patients seeing a general practitioner in accident and emergency on their subsequent reattendance: cohort study

BMJ 2000; 320 doi: (Published 01 April 2000) Cite this as: BMJ 2000;320:903
  1. Andrew W Murphy, professor of general practice (andrew.murphy{at},
  2. Patrick K Plunkett, consultantb,
  3. Gerard Bury, professor of general practicec,
  4. Conor Leonard, researchera,
  5. Jane Walsh, lecturerd,
  6. Finian Lynam, database administratore,
  7. Zachary Johnson, public health medicine specialistf
  1. a Department of General Practice, National University of Ireland, Galway, Republic of Ireland
  2. b Department of Accident and Emergency Medicine, St James's Hospital, Dublin
  3. c Department of General Practice, National University of Ireland, Dublin
  4. d Department of Psychology, National University of Ireland, Galway
  5. e Information and Management Systems Department, St James's Hospital, Dublin
  6. f Health Information Unit, Eastern Health Board, Dublin
  1. Correspondence to: A W Murphy
  • Accepted 13 January 2000

General practitioners working in an accident and emergency department manage non-emergency patients safely and use fewer resources than do usual accident and emergency staff.1 2 In our previous study we speculated that this intervention might have the potential to break the cycle of “inappropriate attendance” at accident and emergency, use of hospital resources, and perceived confirmation of need for a visit.2 We now report the results of a review of the reattendance rates of our original study group.

Subjects, methods, and results

The setting and methodology of our original study have been described.2 In short, patients who had attended St James's Hospital accident and emergency department between 1 August 1993 and 1 October 1994 were triaged using a validated system into four categories—“life threatening” (1), “urgent” (2), “semi-urgent” (3), and “delay acceptable” (4). Local general practitioners were employed on a sessional basis to manage patients only from categories 3 and 4. Randomisation of patients to general practitioners or usual accident and emergency staff depended on time of registration.

We identified the patients included in our original study and, using their unique identifying numbers, determined the number of times that they had reattended the accident and emergency department within two years of their index visit. The date of each reattendance was not recorded. With this information, we classified patients as reattenders or non-reattenders. Only subsequent visits categorised as 3 or 4 in the triage system were included in this analysis. We excluded visits that patients had been asked to make for the purpose of review, dressings, etc. We assessed the effects of six variables (see table) on reattendance and, using SPSS, performed a direct logistic regression analysis to test the power of these variables to predict reattendance. Socioeconomic status was determined by eligibility for General Medical Services (access to free primary care and drugs). Roughly a third of the Irish population are eligible for General Medical Services and represent the poorest section of the community.

Effect of study variables on reattendance of 4459 patients to an accident and emergency department within two years of index visit

View this table:

Of the 4684 patients in our original study, 225 could not be identified for inclusion in this study. Of the remaining 4459 patients, 1890 (42%) reattended at least once to the accident and emergency department within two years of their index visit for management of an unrelated complaint (median number of visits 3 (range 1-293, interquartile range 2-5)). The table shows the effects of the study variables on reattendance. Eligibility for General Medical Services, registration with a general practitioner, male sex, and having an index visit categorised as 3 increased the likelihood of reattendance. The median age of those who reattended was 49, not significantly different from the median age of 45 of those who did not reattend (Kruskal-Wallis test). A test of the full model, with all six predictors, against a model with a constant value only was statistically reliable (χ2=164.45 (df=6, n=4356), P<0.01). The model was more effective in predicting those who did not reattend: 74% of non-reattenders and 43% of reattenders were correctly predicted, with an overall success rate of 61%. However, the variance in reattendance accounted for overall was small (Cox and Snell test R2=0.04).


The decision of patients to attend an accident and emergency department is complex and involves social, psychological, and medical factors.3 Attempts by health services to decrease the numbers of patients attending accident and emergency departments have generally failed.4 Our hypothesis that a single contact with a general practitioner working in accident and emergency would have a longlasting effect on health service use has not been supported, although brief, focused interventions by general practitioners have been shown to have lasting effects in other settings.5


We dedicate this paper to the memory of our coauthor Zachary Johnson.

Contributors: AWM had the original idea for the study, participated in interpreting data and writing the article, and is the guarantor for the study. PKP discussed core ideas, organised data analysis, and participated in interpreting data and writing the article. GB discussed core ideas and participated in interpreting data and writing the article. CL, FL, and ZJ participated in data retrieval and analysis. JW participated in interpreting data and writing the article.


  • Zachary Johnson died in November 1999

  • Funding None.

  • Competing interests AWM, PKP, GB, and ZJ were authors of the previous article that suggested the present study hypothesis.2

  • A table outlining the logistic regression analysis appears on the BMJ's website


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
View Abstract