Forecasting models for first, return and total attendances at accident and emergency (A&E) departments and yearly forecasts were developed ten years ago for all the health districts in the Trent region in England. The one-yearly forecasts had been checked against the 1986 actual figures and found accurate for first attendances but less accurate for return attendances. The forecasts for 1993 and 1994 were much further from the actual figures than the 1986 forecasts, with an increasing bias towards overestimation, particularly for reattendances. Whether a first attender is reviewed at a further visit may depend on local medical policy, which itself may vary with personnel changes. The one-off original ARIMA forecasts for new attendances for 1994 were no better than the district projections made in 1984, but they were better than the Trent Regional Health Authority guidelines. The ten-year strategic plan for Trent Regional Health Authority overestimated the increase in the number of first attendances at A&E departments in the Trent region. The forecasting methodology on which it was based could be improved by incorporating the ARIMA method into planning at the health district level. New forecasts or updated ones need to be calculated yearly.

Objective-To audit avoidable deaths from stroke and hypertensive disease.Design

Aims-To ascertain the type and relative frequency of major factors associated with deaths from cervical cancer.Methods-Deaths from cervical cancer in Rotherham district for the period [1989][1990][1991] were subjected to multifactorial audit by reviewing laboratory, hospital, and general practitioner records; together with, when appropriate, re-screening of cytology smears. This period represented the three to five years after a computerised National Screening Programme (NSP) Audit of factors associated with deaths from cervical cancer should be a vital part of assessing the effectiveness of the National Screening Programme (NSP). Furthermore, if the Health of the Nation target is to be met (20% reduction of invasive cervical cancer by the year 2000), it is likely that any major problems revealed by audit will require rectification. To date, information on factors contributing to cervical cancer is derived from a relatively small number of studies, each having tended to analyse one specific aspect.'6 Although these have been extremely valuable, surprisingly little is known about the relative frequencies of the different factors in the population. There is currently no national requirement to audit cervical cancer deaths and little attention has been paid to the impact of the NSP on factors associated with death from cervical cancer. For these reasons, a multifactorial district audit of cervical cancer deaths for the period 1989-1991 was undertaken. This period represented the three to five years after a computerised NSP in Rotherham based on a five year recall interval had been introduced.

SUMMARY Over the period 1974-85 the range of mean annual new attendance rates at Accident and Emergency departments among English health districts was 36-673 per 1000 residents. The socio-economic diversity of these districts explained only one-third of the variation. The rates rose significantly (p < 0 05) in 89 per cent ofdistricts over the twelve years. Again, socio-economic variation only partly explained differences in district trends. Increases were greater among districts with higher mean rates. In order to plan first-contact care rationally we need a better understanding of the factors underlying these trends.

The annual new, return and total attendances at Accident and Emergency (A and E) Departments for Trent district and the whole of the Trent region are forecast for the years 1986 to 1994 by using the autoregressive integrated moving average (ARIMA) time series model applied to the SH3 A and E returns for 1974 to 1985. The 1986 forecasts of annual new, return and total attendances in Trent districts are compared with the actual attendances observed; the new attendance forecasts were found accurate, the return attendance forecasts less so. The latter may reflect inability to predict changing policies on return attendances of individual A and E departments. The 1994 ARIMA forecasts of annual A and E new attendances for Trent districts are compared with the 1984 based regional guidelines for 1994 and the projections for individual districts. Both the ARIMA models and the health districts' own projections produce a different forecast to the 1994 regional guideline which seems to overestimate. The forecasting methodology used has other applications in health care planning.