U.S. flag

An official website of the United States government

NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Bion J, Aldridge C, Beet C, et al. Increasing specialist intensity at weekends to improve outcomes for patients undergoing emergency hospital admission: the HiSLAC two-phase mixed-methods study. Southampton (UK): NIHR Journals Library; 2021 Jul. (Health Services and Delivery Research, No. 9.13.)

Cover of Increasing specialist intensity at weekends to improve outcomes for patients undergoing emergency hospital admission: the HiSLAC two-phase mixed-methods study

Increasing specialist intensity at weekends to improve outcomes for patients undergoing emergency hospital admission: the HiSLAC two-phase mixed-methods study.

Show details

Chapter 4Cross-sectional and longitudinal 5-year study of weekend–weekday specialist intensity and emergency admission mortality

Introduction

As described in Chapter 1, during 2012 and 2013, several initiatives were broadly focused on improving equity of access to quality health care in the NHS across all days of the week, particularly (but not exclusively) for emergency care. There was a degree of consensus between the professions and health policy-makers that weekend services were suboptimal, and that front-line medical leadership (GPs as well as hospital consultants) was central to achieving improvements. There were lower levels of agreement about how this could be funded and delivered, and, in 2015, relationships between politicians and health-care professionals deteriorated when, in the midst of contract negotiations, the Department of Health and Social Care19 and the Secretary of State for Health and Social Care22 made public statements linking the weekend effect to a reduction in specialist availability in hospitals at weekends. There were two problems with implying a causative relationship. The first was that there was no evidence to support the assertion. The second was the absence of any objective measure of specialist availability. The HiSLAC project’s 5-year longitudinal study was under way by this stage, and, given the need for objective evidence, we decided to publish the first year’s data in a cross-sectional study35 while continuing to collect data through to 2019. We describe both the published study and the longitudinal study here.

Methods

We described earlier (see Chapter 2) the establishment of the HiSLAC collaboration, the acquisition of HES and the development of the specialist intensity metric based on specialist hours per 10 EAs obtained from a web-based annual PPS on a Sunday and the following Wednesday in June each year. The parallel directorate-level survey used to provide assurance for the PPS is described in Chapter 2; it also offered the opportunity to obtain information on consultant rotas and vacant posts, which we report below.

Over 5 years, 13 hospital sites were subject to trust mergers. In eight cases, the data collected after a merger refer to a combination of hospitals that had previously provided separate data. In the case of three hospitals subject to trust mergers, it was possible to preserve the integrity of the data by treating these as coming from separate sites, whereas, for the remaining two hospitals, the post-merger data were treated as being derived from a separate independent organisation.

Statistical analysis

Trust size was represented each year by bed numbers, which were acquired by calculating the annual mean of NHS England’s (KH03) quarterly submissions.103 PPS data were summarised using the proportions of specialists present and the average time spent with emergency cases. The specialist intensity measure (specialist hours per 10 EAs) was computed for each trust for both Sunday and Wednesday. Raw estimates were scaled up by the reciprocal of the response rate to correct for the incompleteness of the survey data. The Sunday-to-Wednesday intensity ratio (a quantity unaffected by the scaling correction) was used to quantify the weekend deficit at trust level.

The analysis of trust survey responses over time is based on a weighted analysis of variance of the trust-specific Sunday-to-Wednesday ratio of hours per 10 EAs. Fixed effects were estimated for trust and year. The weights were designed to capture variation generated by the incompleteness of the survey data, together with trust-level variation within the sampling epochs (see Appendix 5). Differences associated with hospital size were analysed using the average number of beds per trust (‘bed-size’) computed across those years during which the trust participated. Quintiles of bed-size were derived from these data.

The analysis of in-hospital mortality of adult EAs was carried out in a logistic regression model with adjustments made for diagnostic category [as represented by the English NHS Summary Hospital-level Mortality Indicator (SHMI) diagnostic grouping],104 patient age (using a restricted cubic spline with five knots), comorbidity and the income deprivation component of The English Indices of Deprivation 2010.34 In the 2013/14 cross-sectional study,35 each diagnostic category was represented by the full Clinical Classifications Software code. This approach is similar to that of Freemantle et al.69 The model was refitted with additional terms for hospital trusts and for trust-specific weekend effects (i.e. the average of Saturday and Sunday effects, minus the average over all weekdays within each trust). Meta-regression was used to analyse the model estimates of the trust-specific weekend effects over time, using standard errors from the logistic regression. Association with the intensity deficit ratios was investigated using correlation methods. The focus on weekend-to-weekday ratios at trust level – rather than absolute levels of weekend mortality and specialist intensity – minimises the impact of unmeasured differences in case mix and other potential trust-level confounders.

Comparisons of specialist intensity and admission mortality are based on data from the trusts participating in the annual PPS. The analysis of mortality trends since 2007 is carried out using HES data from the slightly larger cohort of all acute non-specialist trusts in England.

Results

The 2013/14 cross-sectional study35

Specialist intensity

Of 141 acute hospital trusts in England receiving unselected emergency medical admissions, 127 agreed to participate and 115 (81.6% of total eligible trusts) contributed data to the survey, providing 15,537 responses. Of these responses, 1003 (6.5%) were from doctors who did not hold a specialist accreditation, and, therefore, these were excluded, leaving 14,534 eligible responses for analysis. The mean response rate was 45% (range 16–79%) and exceeded 40% in two-thirds of trusts.

There were substantially fewer specialists present and providing care to EAs on Sunday (n = 1667, 11.5%) than on Wednesday (n = 6105, 42.0%). This difference was partly offset by the greater average time spent caring for acutely admitted patients on Sunday (5.74 hours) than on Wednesday (3.97 hours). These patterns were consistent across the quintiles of trust size.

For both Sunday and Wednesday, there was a clear relationship between the sum of specialist hours delivered by each trust and the numbers of EAs on that day averaged across the year: as expected, larger hospitals had more specialists and more admissions (Figures 8a and c).

FIGURE 8. (a) Total number of Wednesday specialist hours and average number of Wednesday EAs; (b) Wednesday specialist hours per 10 EAs and average Wednesday EAs; (c) total number of Sunday specialist hours and average Sunday EAs, 2013/14; and (d) Sunday specialist hours per 10 EAs and average Sunday EAs, 2013/14.

FIGURE 8

(a) Total number of Wednesday specialist hours and average number of Wednesday EAs; (b) Wednesday specialist hours per 10 EAs and average Wednesday EAs; (c) total number of Sunday specialist hours and average Sunday EAs, 2013/14; and (d) Sunday specialist (more...)

There was substantial variation between trusts when specialist hours were expressed per 10 EAs (see Figures 8b and d), suggesting that factors other than emergency workload influence the amount of specialist time delivered to EAs in any given trust. The median intensity ratio (Sunday divided by Wednesday) was 48% across all trusts, with similar results across quintiles of trust size, as measured using bed numbers for 2013/14. There was no trust for which the Sunday-to-Wednesday ratio was > 1, and in 90% of trusts the ratio was < 0.7 (Figure 9).

FIGURE 9. Sunday-to-Wednesday specialist hours and ratio per 10 EAs by trust.

FIGURE 9

Sunday-to-Wednesday specialist hours and ratio per 10 EAs by trust.

Mortality

Using 2013/14 HES data, the logistic model for all 141 acute hospital trusts provided a surplus weekend admission mortality risk of 9%. Trust-specific weekend odds ratios (WEORs) for the 115 trusts contributing to the survey ranged from 0.82 to 1.35, with 96 (83%) trusts recording an OR > 1 (Figure 10). There was no systematic relationship between weekend mortality effect and trust size.

FIGURE 10. Trust-specific weekend effect ORs with confidence intervals from the logistic regression model, 2013/14.

FIGURE 10

Trust-specific weekend effect ORs with confidence intervals from the logistic regression model, 2013/14.

Specialist intensity and admission day mortality

The correlation between the weekend mortality ORs and the Sunday-to-Wednesday specialist intensity ratio was very low (r = –0.015) and is not suggestive of a clear relationship (Figure 11).

FIGURE 11. Weekend-to-weekday EA mortality OR and Sunday-to-Wednesday specialist intensity ratio by trust, 2013/14.

FIGURE 11

Weekend-to-weekday EA mortality OR and Sunday-to-Wednesday specialist intensity ratio by trust, 2013/14.

Directorate-level questionnaire

Clinical directors reported that the daily review of all patients was the norm on Sundays (Wednesdays) for 50% (86%) of AMUs, 100% (100%) of ICUs and 27% (58%) of acute wards. Consultant vacancies with gaps in cover were reported by 52% of responding EDs, 71% of AMUs, 65% of acute wards and 30% of ICUs.

Conclusion

This cross-sectional study showed a substantial weekend–weekday difference in specialist involvement in the care of patients admitted as emergencies to acute hospitals in England. There was no sign that the variation in the weekend-to-weekday specialist hours ratio was associated with the variation in the weekend-to-weekday admission mortality ratio. This suggests the need for caution in attributing the weekend effect primarily to a lack of consultants at weekends.

Longitudinal study results

Point prevalence survey responses

A total of 116 participating sites contributed 548 sets of survey results during the 5-year period, of which 96 sites contributed data in every year (thus five times in all). The primary analysis was conducted using all available data from the 116 trusts. A total of 66,425 analysable responses were received. The response rate by trust varied from 7% to 93% and declined by 9.4% during the study period, from 45.2% in 2014 to 35.8% in 2018. The average number of specialist responses from which relevant data could be drawn was 121.2 per trust (Table 2).

TABLE 2

TABLE 2

Survey responses over 5 years

Emergency admissions

Between 2013/14 and 2017/18, EA rates increased by 14.7% (all acute trusts in England), continuing the background trend (see Appendix 6). The increase in EAs per day was greater on weekdays (15.7%) than at weekends (11.3%) (mean weekend-to-weekday ratio 0.78).

Specialist intensity

Between 2013/14 and 2017/18, there was an increase in whole-time equivalent consultant staff of 14.6% across the NHS in England and of 4% among non-consultant doctors (see Figure 1).

The PPS data from 116 trusts are summarised in Table 3, alongside bed numbers and EAs. The results for specialist hours are shown both with and without correction for response rate. Importantly, no adjustment is necessary for the Sunday-to-Wednesday intensity ratio, as the numerator and denominator are affected equally by any such correction. Therefore, any change in the intensity ratio over time is likely to represent a true effect.

TABLE 3

TABLE 3

Specialist intensity over time

There was a significant change in the intensity ratio over time (p = 0.0081). The response curve is flat to begin with (2014–16) and then rises over the next 2 years (2017 and 2018), as shown in Figure 12 (data are provided in Appendix 7). The variation over time was equivalent to an average annual increase in the intensity ratio of approximately 1.5% (standard error 0.46; p = 0.0011) per year, or about six percentage points during the period of the study. There is a significant difference (p < 0.0001) between trusts in the (time-averaged) intensity ratio (Figure 13). Furthermore, there is some evidence (p = 0.0035) of differences in rates of improvement over time. However, it is not clear how, if at all, these differences are related to the size of the trusts and we were unable to identify any other potential causative mechanisms.

FIGURE 12. Sunday-to-Wednesday intensity ratios and adjusted weekend-to-weekday mortality ORs, 2013/14 to 2017/18.

FIGURE 12

Sunday-to-Wednesday intensity ratios and adjusted weekend-to-weekday mortality ORs, 2013/14 to 2017/18.

FIGURE 13. Sunday-to-Wednesday intensity ratios by trust (mean and 95% confidence interval).

FIGURE 13

Sunday-to-Wednesday intensity ratios by trust (mean and 95% confidence interval).

How was the ‘improvement’ in the Sunday-to-Wednesday specialist intensity ratio achieved? If the 7-day services policy initiative were to have had its desired effect, it would have stimulated trusts to increase specialist numbers and, hence, specialist hours on Sundays, either through employing more specialists committed to service delivery at weekends or by redirecting specialist activities from weekdays to weekends. Specialist hours per 10 EAs would increase if more specialists were delivering care to the same number of patients, remain stable if the increase in specialist hours were accompanied by a parallel increase in EAs, or diminish if the increase in EAs outstripped the increase in specialist input or if specialist input actually fell.

Figure 14 shows that, over the 5 years, there was only a very modest overall increase in specialist hours per 10 EAs on Sundays, and a marked reduction on Wednesdays. As shown above, between 2013/14 and 2017/18, EAs increased by 14.7% overall, by 11.3% at weekends and by 15.7% on weekdays (see Appendix 6), and the increase in consultant staff during the same period was 14.6% (see Figure 1). If additional consultant input had been distributed uniformly across all days of the week, it would, therefore, seem likely that, although the increase in consultant staffing would have exceeded the increase in EAs at weekends, it would have been outstripped by the larger increase in EAs on weekdays.

FIGURE 14. Specialist hours per 10 EAs adjusted for response rates and Sunday-to-Wednesday specialist intensity ratio.

FIGURE 14

Specialist hours per 10 EAs adjusted for response rates and Sunday-to-Wednesday specialist intensity ratio. Specialist hours are corrected for response rate using the formula n × (1/response rate).

To explore this relationship between specialist hours and EAs in more detail, we present each trust’s mean annual percentage change in specialist intensity for Sundays against Wednesdays in Figure 15. The majority of trusts (n = 71, 62%) achieved an increase in the Sunday-to-Wednesday intensity ratio (i.e. a positive correlation between intensity ratio and year). In this group, 23 trusts increased specialist intensity for both Sundays and Wednesdays, 22 increased Sunday intensity while reducing intensity on Wednesdays, and a further 26 showed a reduction in both, but which was more marked for Wednesdays than for Sundays. Forty-four trusts showed a reduction in specialist intensity ratios over time; of these, 10 had marginally increased Sunday and Wednesday intensity, 15 had increased Wednesday intensity only, and 19 showed a reduction in both Sunday and Wednesday intensity. However, it would be misleading to conclude that actual specialist hours declined during the period. Indeed, there is evidence of a general increase in hours in the majority of trusts for both Wednesdays (63%) and Sundays (61%) (see Figure 15).

FIGURE 15. Trends in specialist involvement in EAs, 2014–18.

FIGURE 15

Trends in specialist involvement in EAs, 2014–18. (a) Sunday and Wednesday specialist intensity (specialist hours per 10 EAs); and (b) Sunday and Wednesday specialist hours. The average proportional changes per year are computed from trust-level (more...)

The final step in elucidating the causes for the change in specialist intensity ratios is to examine the ratio of supply (specialist hours) to demand (EAs) across the 115 trusts. Here, we use the alternative formulation of the weekend-to-weekday specialist intensity ratio:

Sunday specialist hoursWednesday specialist hours×Wednesday EAsSunday EAs.
(3)

During the study, each component of this expression increased by about 7%. Using data from Table 3, the ratio of Sunday to Wednesday specialist hours increased from 0.39 (= 85.4 ÷ 216.6) in 2013/14 to 0.42 (= 80.0 ÷ 189.9) in 2017/18, and the ratio of Wednesday to Sunday EAs increased from 1.23 (= 106.9 ÷ 87.1) to 1.31 (= 128.0 ÷ 97.7) during the same period. Together, these changes produced an overall increase of 14% in the intensity ratio, from 0.484 (2013/14) to 0.552 (2017/18). The general increase in EAs was more pronounced for Wednesdays (19.7%) than for Sundays (12.1%), and this change in the distribution of EAs across the week accounts for half of the improvement in the intensity ratio. It reflects a change in the pattern of demand on the service rather than a redirection of specialist resources.

In summary, the increase in the specialist intensity ratio appears to have been achieved by a general increase in specialist hours, which has been more pronounced at weekends, and a general rise in EAs, which has been more pronounced during the week.

Mortality and associations with specialist intensity

Hospital and 30-day post-admission EA mortality rates for all acute trusts in England reduced each year (30-day post-admission mortality reduced from 6.43% in 2007/8 to 5.47% in 2013/14), with weekend and weekday admission mortality rates falling in parallel. From 2014, overall mortality rates stabilised, but, in 2017/18, the weekend 30-day post-admission mortality rate increased from 5.8 to 6.06 and the unadjusted mortality ratio increased from 1.08 to 1.15. Adjustment for baseline characteristics abolishes this apparent increase (see Appendix 8), indicating that the rise in mortality rates is attributable to case-mix differences. The difference between hospital and 30-day post-admission mortality rates increased over these 11 years (Figure 16) for both weekend and weekday admissions, so that the ratio of hospital to 30-day mortality rates diminished. Figure 1 shows that, in the last 2 years, the progressive rise in delayed transfers of care has been reversed, indicating that patients are being discharged more rapidly back to the community. Taken together, these data suggest that an increasing proportion of the mortality risk is being transferred to the community post discharge (see Figure 16 and Appendix 8).

FIGURE 16. Ratio of in-hospital to 30-day post-admission mortality and overall 30-day post-admission mortality.

FIGURE 16

Ratio of in-hospital to 30-day post-admission mortality and overall 30-day post-admission mortality.

The weekend mortality effects (i.e. the adjusted weekend-to-weekday mortality ORs) are shown in Appendix 9 and have been plotted alongside the specialist intensity ratios for each year of the study in Figure 12.

There is no evidence for any change in the weekend admission mortality ratios during this period (p = 0.1971). The weekend effect may differ between trusts (p = 0.0034), but these differences do not appear to relate to hospital size. Details are given in Appendix 10. We found no apparent correlation (r = 0.077) between the weekend mortality effect and the Sunday-to-Wednesday intensity ratio.

Discussion

We have shown that patients admitted as emergencies to acute hospital trusts in England on a Sunday collectively receive, on average, half as much specialist time (hours per 10 EAs) as patients admitted on a Wednesday. There is considerable variation between trusts in specialist intensity, which is independent of hospital size. Over the 5 years of the study, the Sunday-to-Wednesday specialist intensity ratio increased, and this appears to have occurred through several mechanisms: a modest increase in specialist input on Sundays; some redistribution of specialist hours in favour of Sundays; and an increase in the number of EAs, which was more marked for Wednesdays than for Sundays. It seems likely that the 7-day service policy had some effect by transferring specialist time from weekdays to weekends, and increasing consultant numbers, but the impact has been obscured by the increase in EAs during this period.

We have been unable to identify a relationship between weekend-to-weekday specialist intensity ratios and trust-level weekend admission mortality effects in either the cross-sectional analysis or the longitudinal study. Trusts with a narrower Sunday-to-Wednesday specialist intensity ratio do not seem to have a correspondingly smaller weekend effect.

Hospital admissions have continued to increase during the past 11 years. However, there was a progressive reduction in mortality rates until 2013/14, when the improvement in survival appeared to have plateaued and the 30-day mortality rate for weekend admissions increased (see Figure 16). Adjustment for baseline characteristics shows that this increase in mortality is attributable to case-mix differences (see Appendix 8). However, the increasing gap between hospital and 30-day mortality rates suggests that there has been a progressive transition in the place of death from hospital to the community. The reduction in length of hospital stay during the same period would be consistent with this interpretation. Whether this represents a desirable effort to allow those destined to die to do so in the peace of their own homes, or a failure to prevent avoidable deaths through, for example, effective rehabilitation of frail elderly patients who are admitted to hospital at weekends, requires further research at the intersection of community and hospital care at discharge, not just at admission.

The conventional narrative about the perceived – and now measured – gap in specialist input between weekends and weekdays is that patients admitted at weekends are being ‘short-changed’ in some way: that half as much specialist input means half as good care. An alternative interpretation might be that patients newly admitted at weekends and on weekdays are receiving the urgent care they need for this acute phase of their illness, while on weekdays specialists are additionally providing the less time-critical components of acute care targeted at refining treatments and promoting recovery. The decrement in specialist input at weekends might result in delays in progressing patients along the therapeutic pathway, but without contributing additional mortality risk.

Limitations of our study relate to the inherent weaknesses of surveys, namely variable response rates and subjective data recall. We mitigated the response rate issue by comparing weekends with weekdays in each trust. Other factors unrelated to specialist intensity might include unmeasured case-mix differences;26 variation in support services in hospital, including ‘failure to rescue’ (the inability of the system to respond promptly to patient deterioration), which has been shown to explain the difference in surgical outcomes between high- and low-volume centres;89 or a difference in outcomes from patient safety incidents at weekends.105 We explore these factors in the following chapters, starting with an examination of patient case mix (see Chapter 5) and following with an evaluation of hospital care quality (see Chapter 6).

Image 12-128-17-fig1
Copyright © Queen’s Printer and Controller of HMSO 2021. This work was produced by Bion et al. under the terms of a commissioning contract issued by the Secretary of State for Health and Social Care. This issue may be freely reproduced for the purposes of private research and study and extracts (or indeed, the full report) may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NIHR Journals Library, National Institute for Health Research, Evaluation, Trials and Studies Coordinating Centre, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.
Bookshelf ID: NBK571873
Contents

Views

  • PubReader
  • Print View
  • Cite this Page
  • PDF version of this title (3.5M)

In this Page

Other titles in this collection

Recent Activity

ClearTurn OffTurn On

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...