General Practice

Relation of out of hours activity by general practice and accident and emergency services with deprivation in Nottingham: longitudinal survey

BMJ 1998; 316 doi: https://doi.org/10.1136/bmj.316.7130.520 (Published 14 February 1998) Cite this as: BMJ 1998;316:520
  1. Robin Carlisle, lecturera,
  2. Lindsay M Groom, research unit coordinatora,
  3. Anthony J Avery, senior lecturera,
  4. Daphne Boot, research associatea,
  5. Stephen Earwicker, general practitionerb
  1. a Division of General Practice, University of Nottingham, University Hospital, Nottingham NG7 2UH
  2. b Stapleford Health Centre, Nottingham NG9 7AT
  1. Correspondence to: Dr R Carlisle Roundwood Surgery, Wood Street, Mansfield, Nottinghamshire NG18 1QQ
  • Accepted 22 October 1997

Abstract

Objectives: To investigate the relation between out of hours activity of general practice and accident and emergency services with deprivation and distance from accident and emergency department.

Design: Six month longitudinal study.

Setting: Six general practices and the sole accident and emergency department in Nottingham.

Subjects: 4745 out of hours contacts generated by 45 182 patients from 23 electoral wards registered with six practices.

Main outcome measures: Rates of out of hours contacts for general practice and accident and emergency services calculated by electoral ward; Jarman and Townsend deprivation scores and distance from accident and emergency department of electoral wards.

Results: Distances of wards from accident and emergency department ranged from 0.8 to 9 km, and Jarman deprivation scores ranged from −23.4 to 51.8. Out of hours contacts varied by ward from 110 to 350 events/1000 patients/year, and 58% of this variation was explained by the Jarman score. General practice and accident and emergency rates were positively correlated (Pearson coefficient 0.50, P=0.015). Proximity to accident and emergency department was not significantly associated with increased activity when deprivation was included in regression analysis. One practice had substantially higher out of hours activity (B coefficient 124 (95% confidence interval 67 to 181)) even when deprivation was included in regression analysis.

Conclusions: A disproportionate amount of out of hours workload fell on deprived inner city practices. High general practice and high accident and emergency activity occurred in the same areas rather than one service substituting for the other.

Key messages

  • We studied the out of hours activity of six general practices and the local accident and emergency department in Nottingham for six months

  • There were wide variations between electoral wards in both general practice and accident and emergency events

  • Deprivation scores explained more than half of the variation, with out of hours activity being highest in deprived inner city areas

  • Highly deprived areas close to the accident and emergency department generated high levels of work for both general practice and accident and emergency services, with no evidence of one service substituting for the other

Introduction

The extent to which deprivation influences the demand for out of hours medical care is not fully explained.1 Studies of single general practices have shown high workload in deprived areas,2 3 but other studies of practices in affluent areas have also found high workloads.4 Studies of multiple practices have been able to explain only a small percentage of the variation in claims for night visits by socioeconomic factors.5 6

It has been hard to quantify the effect of deprivation on use of accident and emergency departments because they do not serve defined populations.7 8 9 It has been suggested that there is less out of hours activity in general practices in inner cities than would otherwise be expected because of the proximity of accident and emergency departments.10 11 At a time of increasing pressure on both services, it is important to understand the factors that influence demand for out of hours care.

Methods

We recruited six practices with 34 general practitioners which covered an area from inner city Nottingham, close to the sole accident and emergency department, through to suburban areas west of the city centre. Three suburban practices provided their own out of hours cover, while one suburban practice and two inner city practices used a deputising service for some of their calls. Four practices participated in undergraduate teaching. There were no differences between the practices in daytime policy for visiting or accommodating urgent cases. Out of hours was defined as 7 pm to 7 59 am on weekdays and weekends from noon on Saturday.

Our aim was to collect information for all contacts with the general practices and accident and emergency department during the first six months of 1996, and we calculated annual rates by doubling the six months' figures. We collected data on general practices with a specially designed form and collected information on the deputising service and the accident and emergency department from routine data. Rates of activity for general practices and the deputising service included visits and telephone advice, while rates for the accident and emergency service excluded patients referred by general practitioners and telephone advice.

We used patients' postcodes to determine the number of patients registered and the number of out of hours events for each electoral ward. As an indicator of proximity to the accident and emergency service, we calculated the linear distance from the grid reference of each ward's centroid, weighted for population, to the grid reference of the accident and emergency department. We assigned Jarman (UPA 8)12 and Townsend13 deprivation scores and their component variables to the electoral wards using data from the 1991 census, and we calculated Jarman scores for each general practice from the ward scores and the number of patients in each ward registered with the practice.

Statistical analysis

We analysed the data using spss for Windows version 6. To reduce the effects of variations in rates for wards with small numbers and to be more sure that the census data were representative of the study population, we restricted the analysis to the 23 wards with more than 199 patients from the practices in the study. We performed linear regression with the weighted least squares option, using wards' levels of out of hours activity as dependent variables and deprivation scores, distances, or individual census components as separate independent variables. We checked assumptions with normal probability plots and analysis of residuals.

We performed multiple regression to investigate whether the effects of distance from the accident and emergency department and deprivation were independent. Inclusion criterion was a probability associated with the F statistic of ≤0.05. To establish whether practices had differing levels of out of hours activity when deprivation was controlled for, we calculated practice-specific out of hours rates for wards in which individual practices had more than 199 patients. These values were used as dependent variables and analysed by multiple regression forward stepwise entry, with Jarman score as a continuous variable and binary dummy variables comparing five practices against the largest practice used as a constant.

Results

Data analysed

Of the 46 698 patients registered with the six general practices, 45 182 (97%) lived in the 23 wards studied. Of 5057 out of hours events recorded, 4742 (94%) occurred in the 23 wards (2019 contacts with the general practitioners, 1016 with the deputising service, and 1707 with the accident and emergency department). We are preparing a separate paper on substantial differences in case mix between the services: most general practice contacts were for minor illness, while half of accident and emergency contacts were for accidents and injuries.

Distances between wards and the accident and emergency department ranged from 0.83 km to 9 km. Jarman scores ranged from −23.4 to 51.8 (national average = 0 (SD 16), positive values representing relative deprivation). There was a negative correlation between distance and Jarman score (Pearson coefficient 0.59) because of higher deprivation in wards close to the accident and emergency department.

Out of hours activity

By individual practice—Table 1 shows the out of hours activity and Jarman score for each of the general practices. Five practices (B-F) had relatively similar rates of out of hours contacts of between 164 and 227 events/1000 patients/year, whereas the most deprived practice (A) had a rate nearly twice the average.

Table 1

Out of hours activity generated by different general practice populations and estimated Jarman deprivation score against out of hours activity rates

View this table:

By ward—General practice contacts varied between wards from 64 to 229 events/1000 patients/year (SD 42). Accident and emergency rates varied from 21 to 153 (SD 29). As was expected, wards with smaller registered populations showed more variability in rates, but there was no significant correlation between the size of the registered population and out of hours activity (P=0.81): the 12 wards with more than 2000 patients had a mean combined rate of 210 events/1000 patients/year (SD 43, range 157-305), and the 11 wards with 200–2000 patients had a mean of 220 (SD 80, range 110-350). General practice rates and accident and emergency rates were positively correlated with each other by ward (Pearson coefficient 0.50, P=0.015).

By deprivation measures—Higher rates for both services were associated with higher deprivation scores. The Jarman score explained 46% of variation in general practice rates and 43% of variation in accident and emergency rates (1). For combined rates, the Jarman score explained 58% of the variation, and the Townsend score explained 56%. Six of the eight component variables of the Jarman score and all four components of the Townsend score were individually associated with combined rates of out of hours contacts (table 2). Areas with overcrowding, unemployment, more non-owner occupation, low car ownership, and increased ethnicity were all associated with higher rates for both services. Areas with more single parents were also associated with higher rates but with a higher coefficient for general practice contacts than for accident and emergency attendances (95% confidence interval of B coefficient of 8.8 to 17.2 compared with 0.8 to 8.5). The proportion of children aged under 5 was positively related to general practice contacts but not accident and emergency rates, while having moved house within a year was significantly related only to accident and emergency rates. Two variables (unskilled and elderly living alone) had no significant association with out of hours activity.

Figure1

Out of hours activity of general practices and accident and emergency department against Jarman deprivation score for 23 electoral wards

Table 2

Univariate linear regression of deprivation scores, distance from accident and emergency department, and census variables with out of hours activity of general practice and accident and emergency services

View this table:

By distance from accident and emergency department—Greater proximity of wards to the accident and emergency department was associated with increased use of this service. However, when accident and emergency rates or combined rates were analysed by multiple regression with Jarman score and distance together, distance did not add significantly to Jarman score used alone.

Multivariate analysis of practice-specific out of hours rates—We limited the analysis to those wards from which at least 200 patients were registered with a single general practice: in 15 wards two practices met this criterion and in six wards one practice did so, giving 36 practice-specific out of hours rates. Between four and eight wards per practice met the criterion. Multiple regression showed that higher total out of hours activity was independently associated with registration with one particular practice, A (B coefficient 124 (95% confidence interval 67 to 181) P=0.0001) and with Jarman score (B coefficient 1.3 (0.4 to 2.1) P=0.007): these two factors together explained 66% of the variation. None of the other practices differed significantly from the constant.

Discussion

Although there may have been some underrecording of general practice telephone contacts, we believe that the data give an acceptable picture of out of hours activity for the period described. The fact that the practices were willing to record detailed information suggests that they may have been less varied than average in terms of their practice organisation. Nevertheless, they covered an area with broad socioeconomic variability.

Out of hours activity

Out of hours activity for general practices and the accident and emergency department varied substantially between areas. For wards with smaller numbers, some of the variation would have been due to chance,14 but even when the analysis is restricted to the 12 wards containing more than 2000 patients activity varies nearly twofold.

Whether considered by individual general practice or by area, the general practice rates and accident and emergency activity were positively correlated. We found no evidence of accident and emergency activity substituting for general practice contacts. Our results suggest that if some accident and emergency work is diverted to general practice some practices will be affected more than others and those most affected would already have the highest workload.

Although distance from the accident and emergency department had a significant negative association with out of hours rates, the association disappeared when deprivation was taken into account. This result has to be interpreted with caution because the range of distances was small and distance and deprivation were correlated. However, it suggests that in Nottingham the high activity from areas close to the accident and emergency department was because of deprivation rather than proximity. This contradicts a study from Northern Ireland,15 where distance was more important than socioeconomic characteristics in explaining accident and emergency attendances over a larger area.

Deprivation

Deprivation score could explain 58% of the variation in out of hours activity between wards. This is a high figure for a regression analysis of a general practice variable and suggests that out of hours activity is more strongly associated with deprivation than either daytime consultation rates16 or referral rates.17 A difference in Jarman score of 60 between two wards would predict a difference of 144 out of hour events/1000 patients/year (95% confidence interval 90 to 198), which is more than half the overall average.

The Jarman score, which is used to determine general practice deprivation payments, performed slightly better than the alternative Townsend score in explaining out of hours activity. The analysis of individual census components must be interpreted with caution. Because a variable is associated with out of hours activity on an area basis does not necessarily mean that individuals with that characteristic consult more (the ecological fallacy).18 19 It has also been shown for general practice consultation rates that some socioeconomic factors such as single parents and ethnicity have different effects in different parts of the country.20 Nevertheless, it is interesting that the variables of children under 5 and single parents were more important for general practice contacts while a change of address within the previous year was associated with accident and emergency rates. It should be borne in mind that the accident and emergency rates for each ward were practice specific and so took no account of deprived patients who were not registered with a general practitioner.21

Multivariate analysis showed that one practice (which was the most deprived) had significantly higher out of hours rates even when deprivation was included in the regression equation. This was apparent from the raw data, which showed markedly different rates for different practices for the same wards. Similar substantial practice effects on out of hours rates have been described before22; they could be mediated either by practice factors influencing patient demand or by differences within the patient populations not detected by census variables at ward level.

Conclusions

Out of hours activity is currently undergoing dramatic change.23 Out of hours payments now have a flat rate component with less emphasis on item of service payment. There is increased use of cooperatives, which usually involves a payment proportional to the number of calls delegated. There have recently been proposals to try to redirect some accident and emergency attendances to primary care.24 Those involved in changing out of hours services should take into account the wide variations in demand we have described. A disproportionate amount of activity falls on deprived inner city practices; these practices already have the most difficulty recruiting new partners. The challenge is to find ways of supporting such practices to provide good quality out of hours care.

Acknowledgments

We thank the participating general practitioners.

Funding: Grant from Nottingham Health Authority.

Conflict of interest: None.

Notes

Contributors: AJA and SE conceived of the idea for the study and prepared the grant application. The project team, led by AJA, consisted of LMG, DB, SE, and RC. Clive Richards attended early project meetings and gave helpful advice. DB collected the data and, with Edna Gibson, entered the data, with advice from LMG and AJA. LMG and AJA performed data analysis for the overall study, and RC performed specific analysis for this paper with statistical advice from Carol Coupland. RC wrote the paper with input from all members of the project team. Brian Thompson-Bialy provided data on the deputising service; Andrew Dove and James Scott provided data on the accident and emergency service; Chris Kerry, Jean Robinson, Christine Edwards, and Saalije Howard provided data on practice populations; Dave Ebdon calculated distance and census information; and Professor Jarman and Debbie Hart provided deprivation scores and census information.

References

  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
View Abstract