Background

The stimulus for the NIHR HSDR programme’s call was the increasing difficulty of providing satisfactory emergency health care in the NHS. The background to this problem was multifaceted. The European Working Time Directive in 2003 limited hours of work;² the NHS responded by moving junior doctors to shift-working, introducing non-physician clinicians, and increasing the numbers of consultants on new contracts that explicitly identify and reimburse out-of-hours work. This increased staffing and costs, but did not improve team working, training or continuity of care.³ At the same time, a new contract for family doctors [general practitioners (GPs)] allowed them to withdraw from providing out-of-hours services.⁴ Emergency admissions (EAs) continued to rise every year at a rate similar to that of the increase in consultant staff and faster than that of the background population increase, particularly those involving frail elderly people,⁵ while policies to deliver more health care in the community reduced hospital beds⁶ without a sufficient compensatory expansion in social care funding.⁷ This increased the number of delayed hospital discharges, which adversely affected care quality⁸ and blocked beds, making it even more difficult to accommodate the growing numbers of EAs.⁹ Temporal changes in medical staffing, emergency department (ED) attendances, EAs and hospital beds are shown in Figure 1. We discuss the apparent improvement in reducing delayed discharges of care in Chapter 4.

FIGURE 1

Trends in NHS England emergency care workload, staffing and beds: 2011/12–2018/19. NHS England overall monthly average number of delayed days; number of overnight beds; annual average number of consultants and specialists, and other doctors; ED (more...)

Public concern about the deteriorating quality of hospital care was exacerbated by failings in professional standards identified by the Mid-Staffordshire NHS Foundation Trust Public Inquiry.^10,11 There was also a longstanding perception that hospital care out of hours (at nights and weekends) was unreliable and potentially unsafe because of evidence showing the ‘weekend effect’. The Royal College of Physicians and the Society of Acute Medicine’s survey¹² found that only 20% of hospital specialists were available at weekends for periods exceeding 8 hours, 18% reported never attending hospitals at weekends and 73% of acute physicians were not contractually obliged to provide medical care at weekends; only 19% of responding hospitals reported having a formalised rapid response team for acutely ill patients. Only 39% of specialists working in acute medical units (AMUs) reported having protected time for this work free of other duties, and providing care for blocks of time greater than a single day. Suboptimal specialist input had also been identified in the National Confidential Enquiry into Patient Outcome and Death (NCEPOD) audit of deaths following emergency hospital admission:¹³ at 12 hours following admission, 40% of patients had not been seen by a consultant, and in 12.4% of cases there was no documented evidence of consultant review. Of the 95 cases in which the assessors considered the delay in consultant review to have been unacceptable, the delay was considered to have adversely affected the accuracy or timeliness of diagnosis in 32.6% of cases, and may have contributed to adverse outcomes [intensive care unit (ICU) admission, worsening prognosis or death] in 49.5% of cases.

The profession responded to these findings with proposals for the reconfiguration of hospital services,^14,15 and the development of enhanced standards of consultant practice. The Academy of Medical Royal Colleges convened a working group to develop standards for daily consultant review.¹⁵ In the absence of strong direct evidence, the working group arrived at a consensus recommendation, namely that all hospitalised patients should be reviewed by a consultant every day unless prior review had determined that this was not necessary. The Academy of Medical Royal Colleges’ report concluded that systematic evaluation of the standards was required, within a research framework.

Seven-day services

Initiatives to improve access to emergency care were brought together under the UK government’s policy of ‘7-day services’,¹⁶ with the development of 10 clinical standards for emergency care,¹⁷ six of which directly or indirectly required front-line consultant involvement for delivery. Initially, NHS England required these standards to be introduced at scale and pace within 3–4 years, backed by incentives, rewards and sanctions. This was subsequently diluted to achieving four ‘priority standards’ across the NHS by 2020 (quotations in this paragraph contain public sector information licensed under the Open Government Licence v3.0, URL: http://nationalarchives.gov.uk/doc/non-commercial-government-licence/version/2/).¹⁶ These are:

• standard 2 – EAs assessed by a consultant within 14 hours
• standard 5 – timely access to diagnostic services according to urgency
• standard 6 – access to consultant-directed interventions 24/7
• standard 8 – consultant review of patients in acute medical, surgical and intensive care units twice daily by consultants working blocks of days; daily consultant review of ward patients.

Whether or not these standards were the correct solution to problems of access, and how they were to be measured, monitored and funded, became a source of tension between the professions on one side¹⁸ and the government and Department of Health and Social Care¹⁹ on the other, with NHS England in the middle. A publication by Freemantle et al.²⁰ in 2015 presented a surplus weekend admission mortality of 16% as justification for 7-day services as the ‘solution’. During negotiations to change NHS doctors’ contracts to facilitate the introduction of 7-day services,²¹ the then Secretary of State for Health and Social Care also linked weekend deaths to doctor numbers,²² stating that ‘. . . someone is 15% more likely to die if admitted on a Sunday than on a Wednesday because we do not have as many doctors in our hospitals at the weekends as we have mid-week’ (contains Parliamentary information licensed under the Open Parliament Licence v3.0, URL: https://parliament.uk/site-information/copyright-parliament/open-parliament-licence/) and referring to the weekend effect as a ‘global scandal’. This precipitated a vigorous antipathetic response to 7-day services on social media (Figure 2) and contributed to a threat of strike action by junior doctors.²³ The ‘weekend effect’ seemed to have evolved from a problem requiring scientific exploration into a political tool for implementing health policy. It was against this background that the High-intensity Specialist-Led Acute Care (HiSLAC) project was established.

FIGURE 2

Frequency of citations of the ‘weekend effect’, 7-day services and integrated care. a, Twitter, Inc., San Francisco, CA, USA (www.twitter.com); b, up to 31 October 2019.

The ‘weekend effect’

The weekend effect was first reported by Bell and Redelmeier in 2001,²⁴ who reported a significantly higher mortality rate associated with EA to hospital at weekends for 23 of the 100 leading causes of death in Canada. The authors were unable to exclude the possibility that patients admitted at the weekend were sicker, but hypothesised that the cause for the surplus mortality was the reduction in medical staffing, particularly senior doctors, in hospital at weekends. In a perceptive accompanying editorial to that paper, Halm and Chassin²⁵ stated that:

Disentangling the potential causal pathways would require painstaking detective work . . . of first accounting for the biologic and social determinants of risk and then identifying the precise differences in processes of clinical care that explain the differences in risk-adjusted outcomes.

Halm and Chassin²⁵

However, few investigators attempted this detective work in the intervening years. Although the literature quantifying the magnitude of the weekend effect continued to accumulate, only one paper had attempted to investigate the cause: a single-centre 8-year study from Dublin that found that patients admitted at weekends were sicker,²⁶ based on adjustment using abnormal biochemistry results. The HiSLAC project therefore chose to focus on Halm and Chassin’s challenge, as the weekend effect was of importance to patients, health professionals, policy-makers and politicians.

How is the weekend effect calculated?

The weekend effect is a complex metric, a ratio of two ratios: the weekend mortality rate divided by the weekday mortality rate. This means that there are four primary mechanisms by which an excess mortality might be generated, as shown in Figure 3.

FIGURE 3

Potential primary mechanisms of the weekend effect.

The definition of ‘weekend’ requires some elaboration. In the Christian tradition, Sunday was a day for worship and rest from labour; the weekend now usually incorporates Saturday and Sunday in the western world. The Jewish Sabbath is from sunset on Friday to sunset on Saturday; the Muslim days of rest in some countries was from Thursday to Friday, but increasingly now spans Friday to Saturday. The most frequent definition of ‘weekend’ in the scientific literature is that period of the week from midnight on Friday until midnight on Sunday. This is the definition we use in the HiSLAC project. In our literature reviews we have used the data as published in the source documents.

The potential causal pathway for the weekend effect

The causal pathway could include a number of independent or interlinked underlying causes that could be community or hospital based, and could be related to case mix, clinical care quality or measurement artefact (Figure 4).

FIGURE 4

The acutely ill patient pathway and factors potentially influencing weekend–weekday admission mortality differences. Reproduced with permission from Chen et al. © Author(s) (or their employer(s)) 2019. Re-use permitted under CC BY. Published (more...)

Rationale for and design of the HiSLAC project

When the Academy of Medical Royal Colleges’ working party produced its recommendations for daily consultant review,²⁸ it did so recognising that this was a consensus opinion across all royal colleges, and that the supporting scientific evidence was largely indirect and observational in nature. Given the estimated cost of implementing 7-day services (including increased consultant presence at weekends) of between £1.07B and £1.43B,²⁹ research using empirical data incorporating health economics modelling was needed. It was self-evidently impossible to conduct a conventional prospective randomised controlled trial of high- compared with low-intensity consultant staffing, but the planned roll-out of 7-day services provided an unique opportunity for a ‘natural experiment’ conducted under real-world conditions to test the hypothesis that increasing the numbers of (senior) doctors would result in better quality of care for patients following EA to hospital, with weekend admission mortality rates as one of the key indicators.

Doing this required the development of measures of specialist (consultant grade) doctors’ input across the health service, and metrics for care quality that would allow us to map 7-day service penetration and impact over time. At the time of implementation of 7-day services, it was not clear how the uptake of the standards would be measured and monitored. In the event, NHS Improvement required trusts to perform board-assured audits based on local case record reviews. However, the methodology changed in subsequent years, making analysis of secular trends difficult, and the data did not allow for the calculation of the ‘dose’ of specialist input. The 7-day service standard of greatest relevance to potential consultant impact on the weekend effect is standard 2, ‘emergency admissions assessed by a consultant within 14 hours’. For 2017 and 2018, the methodology for standard 2 was based on local case record review of 20 weekday and 10 weekend admissions; the results were presented as a single figure of percentage compliance across the 7 days of the week.³⁰ In 2019, the methodology was changed to a binary ‘yes’ or ‘no’ response based on each trust’s self-assessment of whether or not 90% of their audited patients had been reviewed by a consultant on weekends and weekdays combined.³¹ Neither method permitted the separation of weekend from weekday performance without access to trust-specific data, which we were unable to obtain from NHS Improvement. Moreover, the definition of consultant review is not specified; it could mean direct bedside review by the consultant, remote review (e.g. a ‘board round’), or that a patient had been reviewed by a junior doctor working under the supervision of a consultant present in the hospital but not at the patient’s bedside.

Consultant numbers were available by trust, but trusts provided no information on how those consultants were deployed during the week, in hours or out of hours, or in caring for EAs.

The first phase of the HiSLAC project, therefore, required us to develop measures of the ‘dose’ of specialist input. We also had to acquire sufficient preliminary data (phase 1) to demonstrate that a project extending over several years was achievable (phase 2). Although HiSLAC was initially conceived as a 3-year project, the HSDR programme board proposed that we extend this to 5 years to allow sufficient duration to determine secular change. Appropriate stewardship of public funds was assured by establishing an independent Governance and Oversight Board to control ‘decision gates’ based on meeting specific progress goals.

An overview of the study is presented in Figure 5 and the initial research plan design is shown in Figure 6. We chose two key methods to strengthen the observational nature of the proposed research: triangulation and difference-in-difference analyses. Triangulation employs concurrent use of different (and preferably independent) measures of a particular phenomenon; if similar observations are made using different methods, this increases the confidence in the credibility of those observations. We did this by using quantitative and qualitative (mixed-methods) approaches, and by combining whole-system measurements across NHS England with in-depth examinations of a subgroup of 20 hospitals. Difference-in-difference analyses strengthen causal inferences by reducing the impact of confounding variables (which indirectly influence associations between dependent and independent variables) in studies where random assignment to intervention or control is infeasible.³² We used this approach in comparing trust-level differences between weekends and weekdays (specialist intensity, admission mortality, error and adverse event rates, and judgements of care quality) and then comparing those differences with change over time (a ‘triple’ difference).

FIGURE 5

Overview of the HiSLAC project.

FIGURE 6

The initial research plan flowsheet for the HiSLAC project. CPR, cardiopulmonary resuscitation; LoSLAC, Low-intensity Specialist-Led Acute Care; ICNARC-CMP, Intensive care national audit & research centre – case mix programme; NICE, National (more...)

The research plan was broadly retained over the 5 years of the project, with two modifications. The first modification was that, despite repeated efforts, we were unable to realise the proposed collaboration with NHS Improving Quality [renamed NHS Improvement] when it assumed the regulatory roles previously held by the Trust Development Authority (for NHS trusts) and Monitor (for NHS foundation trusts) in 2016. Now merged with NHS England, NHS Improvement holds the annual self-reported audit data returns from each trust for implementing 7-day services. These data are reported by individual trusts and are aggregated by NHS Digital for each standard in binary form as having been attained or not, but the original raw data were not made available to the HiSLAC project; therefore, we were unable to compare HiSLAC data on specialist intensity with each trust’s self-reported 7-day service standards. In addition, NHS Improvement’s methodology for collecting 7-day service data and the definitions employed have changed over time, making secular changes difficult to assess.

The second modification was the decision to perform an additional study to test the hypothesis that patients admitted to hospital at weekends were sicker. As this required detailed information, including physiology (vital signs measurements), that is not collected nationally in a standardised format, we performed a single-centre study using the clinical information system available at a large urban university teaching hospital.

Aims and objectives

The aim of the HiSLAC project was to determine whether or not increasing the intensity of specialist-led care at weekends improves outcomes for patients admitted to hospital as emergencies at weekends.

The objectives of the HiSLAC project were to:

• quantify specialist input into the care of EAs and map changes in provision over time
• compare the quality of care in hospitals with high levels of weekend specialist cover with that of hospitals with lower levels, using mixed methods
• determine whether or not the case mix of patients admitted at weekends differs from that of patients admitted on weekdays
• develop a health economics model to estimate the costs and health outcomes [quality-adjusted life-years (QALYs)] associated with increased intensity of specialist provision.