How to minimize inappropriate utilization of Accident and Emergency Departments: improve the validity of classifying the general practice cases amongst the A&E attendees

Health Policy. 2003 Nov;66(2):159-68. doi: 10.1016/s0168-8510(03)00023-x.


Studies have found that one-third to two-thirds of all patients attending Accident and Emergency (A and E) Departments could be managed appropriately by general practitioners (GPs). There is also evidence that referral to GPs can be acceptable to patients. The question of primary concern is screening non-urgent cases with high degrees of sensitivity (S), specificity (SP), and positive predictive value (PPV). This paper reports the findings of the validity (S, SP and PPV) of nurses and patients in triaging A and E visitors. A cross sectional study was conducted over a 1 year period and subjects were randomly selected from four A and E Departments located across the four principle geographic regions of Hong Kong by stratified, two-stage sampling. S, SP and PPVs were computed for both non-weighted and weighted conditions. The gold standard for defining the true urgency status of each selected patient was based on a review of the patient's record 3-21 days (or longer if necessary) following the A and E visit. The record review in each A and E was blinded and done independently by a panel of two (and if disagreement existed, three) senior emergency physicians who did not practice in the same hospital. The greatest weights would be for incorrect decisions with greatest impact on patients' well being. The most accurate unweighted nurses' triage classification had an average sensitivity of 87.8%, specificity of 83.9%, and a PPV of 70.1%. When weighted, the average sensitivity reduced to 75%, specificity to 65.7%, and PPV to 54%. The most accurate unweighted patients' self-triage classification yielded a sensitivity of 62.5%, specificity of 69.2%, and a PPV of 58.1%, and correspondingly reduced to 43.3, 49.2 and 38.6% if weights were applied. Validity of the derived patients' self-classifications was too inaccurate for practical use. Hong Kong's current use of a five-point urgency scale by nurses would be further refined for identifying non-urgent visitors. If a mechanism was put in place for additional screening on visitors with a borderline semi-urgent or non-urgent status, the nurses could safely reassign non-urgent patients to GP care. If implemented, a significant impact on hospital costs could be realized.

Publication types

  • Research Support, Non-U.S. Gov't
  • Validation Study

MeSH terms

  • Concurrent Review
  • Cross-Sectional Studies
  • Decision Making
  • Emergencies / classification*
  • Emergency Service, Hospital / organization & administration
  • Emergency Service, Hospital / statistics & numerical data*
  • Health Services Misuse / statistics & numerical data*
  • Health Services Research
  • Hong Kong
  • Humans
  • Nursing Assessment
  • Primary Health Care / classification*
  • Self-Assessment
  • Triage / standards*