Article Text

PDF

Exploring the impact of consultants’ experience on hospital mortality by day of the week: a retrospective analysis of hospital episode statistics 
  1. Milagros Ruiz,
  2. Alex Bottle,
  3. Paul P Aylin
  1. Dr Foster Unit, Department of Primary Care and Public Health, Imperial College London, London, UK
  1. Correspondence to Paul Aylin, Dr Foster Unit, Department of Primary Care and Public Health, Imperial College London, 3 Dorset Rise, London EC4Y 8EN, UK; p.aylin{at}imperial.ac.uk

Abstract

Objective To examine the association of consultants’ experience with mortality by day of the week when elective surgery was performed.

Design Retrospective observational study using English hospital administrative data.

Setting All acute and specialist English National Health Service (NHS) hospitals carrying out elective surgery between financial years 2008–2009 and 2010–2011.

Participants Patients undergoing elective surgical procedures.

Main outcome measures Death in or out of hospital within 30 days of the surgical procedure taking place.

Results We examined 3 922 091 (26 409 deaths) elective procedures with valid consultant information between 2008–2009 and 2010–2011 in English hospitals; there were 21 196 consultants in charge of these procedures, which took place in 163 NHS hospitals. Consultant seniority had no significant impact in predicting mortality (p=0.345). Patients undergoing elective surgery under junior consultants had slightly lower odds of 30-day death when compared with patients under more experienced consultants (OR 0.95, 95% CI 0.91 to 0.99). We found significant mortality variation among consultants in charge of elective procedures within hospitals, with only moderate variation between hospitals. The adjusted odds of death remained higher for Friday (OR 1.48, 95% CI 1.42 to 1.54), Saturday (OR 1.97, 95% CI 1.83 to 2.12) and Sunday (OR 1.67, 95% CI 1.50 to 1.85) after adjusting for consultant seniority and patient characteristics. Consultant experience is significantly lower (p<0.0001) on a Friday (median (SD) was 7.9 years (4.4)) than the Monday to Thursday average (median (SD) was 8.5 years (4.3)).

Conclusions Our cohort of patients shows that consultant seniority is not a significant factor in predicting 30-day mortality following elective surgery by day of the week. The end-of-the-week effect remains significant after adjusting for patient, consultant and hospital effects, suggesting that other unobserved factors may be driving the higher mortality towards the end of the week. Consultant's years of experience are lowest on a Friday; however, we do not believe that this small variation has any impact on patient outcomes.

  • Surgery
  • Patient safety
  • Quality measurement
  • Statistics

Statistics from Altmetric.com

Introduction

The ‘weekend effect’, the association of a higher risk of mortality or poorer outcomes, for patients admitted to hospitals on weekends when compared with weekdays, is the subject of numerous publications and a cause for concern among health professionals and public alike.1–6 We reported a ‘Friday effect’ in the elective setting of English hospitals, where patients had a 40% increase in the odds of dying following elective surgery carried out on a Friday when compared with procedures carried out on a Monday.2 Furthermore, a study conducted within the US Veterans Affairs system showed that 30-day mortality increased by 17% after elective surgery on a Friday, as opposed to surgery on Monday to Wednesday.7

It is >30 years since the association of weekend hospitalisation and adverse patient outcomes was first observed in public hospitals,8–11 its complexity reflected by the lack of overall consensus as to the likely causes of this pervasive finding within healthcare providers internationally.12 Speculative contributing factors include inadequate numbers of out-of-hours staff (at night and at the weekend); over-reliance on unregistered support staff and temporary staff; decreased availability of imaging, laboratory tests and specialised treatments; inadequate numbers of experienced doctors on Fridays/weekends and different patient case-mix over weekends when compared with weekdays.2 ,6 ,12 ,13 One further explanation that has been put forward is the possible variability in surgeons’ experience by day of the week.14 We set out to test the hypothesis that junior consultants operate more often towards the end of the week and contribute to the higher mortality outcomes seen in our previous study.2 Furthermore, we hypothesise that patients undergoing surgery by junior consultants would have increased risk of 30-day mortality. For this purpose, we added consultant experience to the model of Aylin et al.2 Consultant experience is defined as the number of years of experience accumulated by each consultant from the year of appointment (entry to the General Medical Council (GMC) specialist register) up to the year of a patient's surgery. Our objective was to investigate whether the addition of consultant experience to the statistical model would explain the higher risk of 30-day mortality observed towards the end of the week.

Methodology

We examined the same dataset of Aylin et al,2 which consists of English National Health Service (NHS) hospital (Hospital Episodes Statistics; HES) administrative data for three financial years from 2008–2009 to 2010–2011. It contains records of all operating room procedures corresponding to elective inpatient admissions to public hospitals. We used case-wise deletion to handle missing data, which in this dataset was due to the lack of recorded consultant information.

To identify the consultant overseeing each surgical procedure, we linked the consultant's reference number in our dataset to the GMC List of Registered Medical Practitioners (LRMP, http://www.gmc-uk.org/doctors/register/LRMP.asp). This list contains doctors’ details such as reference number, gender, former names, year and place of primary medical degree, status on the Register including whether doctors hold a licence to practise, date of registration and date of entry into the Specialist Register (we last accessed the LRMP on November 2013).

We calculated consultant experience at the time of surgery by subtracting the year of entry into the GMC specialist register from the year in which the surgical procedure was carried out. It is important to note that records of entry into the Specialist Register started in 1996 and all doctors who earned specialist status (known as consultants from that date) in earlier years will show a Register entry as 1996. We further categorised consultant experience according to the number of years since entry into the specialist register. We refer to this as consultant seniority. Due to the lack of published literature on this topic, the categorisation of consultant seniority was rather arbitrary; however, we tried to represent common views held on consultant ranking while providing ease of interpretation. Consultants having up to 2 years of experience are considered junior-level consultants, those having between 3 and 9 years of experience are classed as mid-level consultants, while those having ≥10 years of experience are classed as the expert-level consultants and considered the most experienced. Another option was to introduce a continuous variable to represent consultant experience, but testing suggested a poorer fit when compared with a categorical variable.

Due to the clustering of patients within consultants and of consultants within hospitals, we used a three-level (patient, consultant, hospital) random intercepts logistical regression model. As we are not interested in identifying variation due to specific hospitals or consultants, these were considered random effects (consultants nested within hospitals); consultant seniority was a fixed effect variable. At patient-level, the model adjusted for patient characteristics (age, gender, Charlson comorbidity score (derived from secondary diagnosis fields with weights specific to England),15 Deprivation Index quintile, ethnic group and number of admissions within previous year), surgical procedure risk and temporal variation (discharge year and day of the week in which procedure took place). At consultant-level, we adjusted for three levels of consultant seniority. As in the paper by Aylin et al,2 surgical procedure risk was calculated by ranking procedures into five equally sized groups based on crude 30-day mortality rate.

Data manipulation was carried out using SAS V.9.2. To carry out the mixed effects logistical regression modeling, we used PROC GLIMMIX. Statistical tests were considered significant when p<0.05.

Results

The original dataset of Aylin et al2 contained a total of 4 133 346 records of elective procedures (27 582 deaths). Due to inaccuracies in the process of recording consultant activity,16 approximately 5% of these (211 255) had an invalid or missing consultant reference number. The remainder consists of 3 922 091 (26 409 deaths) elective procedures with valid consultant information between 2008–2009 and 2010–2011 in English hospitals. There were 21 196 consultants in charge of these procedures, which took place in 163 NHS hospitals. The overall 30-day crude mortality rate for our sample was 0.67%.

To test whether the exclusion of 5% of records is a source of bias, we conducted a sensitivity test by applying the model of Aylin et al to the reduced dataset (excluding the 5% of records with missing consultant information). The results of this analysis revealed a change of <5% in the adjusted odds of death per day of the week, suggesting that the exclusion of these records do not significantly bias our results (data not shown).

Unadjusted analysis revealed that 30-day mortality after elective surgery was higher for men versus women (0.83% vs 0.53%, p<0.0001), for older patients (1.40% for patients aged ≥70 years vs 0.13% for patients aged <30 years, p<0.0001), for those whose procedure risk was the highest (2.45%, p<0.0001) and for those having surgery on a Friday (0.83% vs 0.55% on Monday, p<0.0001). Crude mortality did not significantly differ by consultant's experience (p=0.067). Note that very small variations may result in highly significant p values due to our large sample size (table 1).

Table 1

Unadjusted analysis of death rates by patient and consultant characteristics for patients having elective surgery in 163 NHS hospitals during financial years 2008–2010

Table 2 shows the number of elective procedures carried out during the study period by day of the week. Procedures taking place Monday to Thursday accounted for 80% of the total, Friday procedures accounted for 16%, while weekends accounted for only 4% of the total. We note that the highest crude mortality rate occurs on a Sunday for patients under junior consultants, although on weekdays, patients under more experienced consultants present a similar or higher crude mortality risk when compared with patients under junior consultants.

Table 2

Number of elective procedures, crude mortality rates and adjusted odds of 30-day death by day of the week by consultant seniority during the study period

Consultant years of experience and seniority

We found a significant difference (p<0.0001) in the number of years of experience since entry in the specialist register per day of the week for consultants in charge of elective surgery (table 2). Early in the week, Monday to Thursday, the median (SD) experience is 8.5 years (4.3). Those consultants working on a Friday averaged the lowest with a median (SD) of 7.9 years (4.4), while weekend consultants averaged a median (SD) of 8.2 years (4.3). The maximum difference in consultant experience was 0.6 years (95% CI 0.59 to 0.63) or 7.2 months, between those consultants working on a Friday and those working early in the week.

There was a significant difference between the number of consultants, by seniority level, working during a weekday and weekends (p<0.001). From Monday to Friday, the most experienced consultants outnumber the junior by 1.5:1 and by almost 2:1 on weekends (see online supplementary table A1 and figure A1). Over 85% of procedures were carried out by senior and mid-level consultants throughout the week (table 2). On a Friday, however, senior consultants perform significantly fewer procedures when compared with the rest of the week (a drop of 3.1% from the weekly average); conversely, junior and mid-level consultants take on a higher burden of surgery on Fridays (2.8% increase; table 2). At consultant level, the mean number of procedures per hospital consultant and per year is significantly different on a Friday, Saturday and Sunday (p<0.0001) when compared with the Monday to Thursday mean (table 2).

Procedure-risk by day of the week and consultant seniority

Procedure-risk varied by day of the week (see online supplementary figure A2). There were significantly fewer high-risk procedures carried out on weekends and proportionally more low-risk procedures carried out on Saturdays and Sundays, but no variation during the normal working week. Junior, mid-level and senior-level consultants performed 12.6%, 45.7% and 41.7% of all procedures within the highest risk quintile respectively.

Multilevel modelling analysis

Our multilevel analysis showed that the adjusted odds of 30-day death for elective surgery patients under junior consultants is lower than in patients under senior-level consultants (OR 0.95, 95% CI 0.91 to 0.99) and for mid-level consultants when compared with senior consultants (OR 0.97, 95% CI 0.94 to 0.99). Table 2 presents the adjusted odds of 30-day death after elective surgery by day of the week showing overall lower odds of death for patients under junior and mid-level consultants when compared with patients under senior consultants. Accounting for seniority did not modify the effect of weekday on the odds of 30-day death as originally found by the study by Aylin et al2 (table 3). Furthermore, Consultant seniority was not a significant contributing factor (p=0.345) to day-of-the-week mortality patterns as revealed by type 3 tests of fixed effects (table 3).

Table 3

Adjusted odds of death per day of elective procedure for our multilevel model and the model of Aylin et al2

Adjusted mortality variation at hospital and consultant level is shown in table 3, under covariance parameters estimates. The model showed significant outcome variation among consultants as given by the median OR (MOR=1.58, 95% CI 1.72 to 1.99, p<0.0001). After the large consultant variation is taken into account, we observed a non-significant variation of patient mortality between different hospitals in our study (MOR=1.01, 95% CI 0.99 to 1.03).

We explored the possibility of interactions for day of the week and consultant seniority. Our results indicate that this interaction is not strong (p=0.02). We also tested for an interaction between day-of-the-week and procedure-risk quintile, which we showed to be significant (p<0.0001), with the highest risk procedures (top two risk ranks) taking place earlier on the week (Monday to Thursday). However, procedure-risk and seniority revealed a non-significant interaction (p=0.687).

Discussion

Using a sample of patients that showed higher adjusted odds of death towards the end of the week in the 30 days following elective surgery,2 we found that the observed end-of-the-week effect remains significant and was not altered after accounting for consultant seniority. Junior consultants in charge of elective surgery showed lower adjusted odds of death when compared with senior consultants. Because our model fails to adjust for disease severity, this result might reflect the care of lower case-mix severity patients by junior consultants (only 12.6% of the highest risk quintile procedures were overseen by junior consultants). In agreement with the study by Aylin et al,2 we found no significant clustering effect by hospital, with non-significant variation in 30-day mortality (after surgery) among hospitals once the large mortality variation among consultants was taken into account.

In the past, consultant activity has been examined by gender, specialty and by type of contracts held by consultants,17–19 but there is limited information on NHS consultant activity in relation to patient outcomes.20 We have not found any other piece of literature that compares patient outcomes with respect to consultant activity or expertise by day of the week. A recent study of the UK National Adult Cardiac Surgery Audit Registry focused on the association between surgeons’ length of service and risk-adjusted outcomes.20 In their study, a surgeon (consultant) experience was defined as the time elapsed since medical qualification (the award of first clinical degree), thus overestimating the true extent of consultant experience as periods of absence are more likely to occur before specialist appointment. We believe that a more reliable proxy figure representing the expertise of a surgeon (consultant) is obtained by considering the year of entry into the GMC specialist register. Nevertheless, their study found a positive correlation between risk-adjusted mortality and surgeon experience for cardiac surgery patients, but no analysis was carried out to reveal a day-of-the-week effect.

Our study showed that consultants working on Fridays have the lowest consultant experience (number of years), but we do not believe that this small deviation is sufficient to cause a negative impact on patient's outcomes. Moreover, we found no evidence for the suggested lack of senior consultants working towards the end of the week, as revealed by the number of consultants per level of seniority in charge of surgical procedures throughout the week. Given that surgical patients under junior consultants have lower risk of death when compared with patients under senior consultants and coupled with the fact that on weekends senior consultants outnumber the juniors by 2:1, we believe this is strong evidence that consultant experience is not as significant as patient case-mix when determining mortality risks.

Strengths of this study

To the best of our knowledge, this is the first study to explore the contribution of surgeon's working experience to the end-of-the-week effect in elective surgery. Using an administrative data set which is gathered regularly and consistently throughout NHS hospitals, we are taking advantage of large sample sizes at relatively low cost, but caveats remain as to their suitability in healthcare research (see limitations).

As noted by Aylin et al,2 this study included linkage to out of hospital deaths (after discharge) thus removing potential bias of only counting hospital deaths. Furthermore, our analysis takes into consideration outcome variation due to the clustering of patients within consultants and in turn the clustering of these within hospitals, a more realistic representation of health data.

Limitations

We have used administrative data which are subject to change in administrative procedures, may contain missing and/or erroneous data and ultimately are subject to miscoding issues. In spite of this, their use has been justified in a number of studies.21 We recognise that the named consultant on English HES records may not be that of the operating surgeon, but most likely represents the surgery team leader. Thus, it is difficult to assess the proportion of surgery actually performed by the named consultant;16 nonetheless, this is the person responsible for the patients’ outcome. This could be a major limitation in our study and we encourage the introduction of further consultant fields within HES to indicate the actual operating surgeon. It has been highlighted that HES data do not accurately record consultant information, particularly in emergency patients; however, this process is improving due to stricter coding practices.16 ,21 We have tried to account for surgery risk and case-mix complexity using procedure-specific predicted mortality as a proxy for surgery and patient complexity. Ideally, we would adjust our models using a clinical severity factor, but administrative data lack the necessary information to do so. In addition, postoperative risks inherent to specific operations are unaccounted for. Furthermore, our estimated procedure risk model does not use any variable that relates to surgery delays or to adequacy of staffing levels in hospital wards. Similarly, we are unable to account for delays or postponements of surgery due to patients waiting to be operated on by more senior consultants. To avoid loss of power in our analysis, we did not account for different medical–surgical specialties to further characterise consultant activity.

To understand the weekday effect, there is need for more focused research into specific medical specialties, whose varying working practices might be at the heart of weekday effects. In examining emergency stroke care at the weekend versus weekdays, Palmer et al22 found lower rates of CT scans on the day of admission and lower thrombolysis rates associated with higher mortality at the weekend. There is also some evidence that lower stroke mortality at the weekend is associated with higher intensity of weekend nursing staff.23 There is clearly greater control over the scheduling of elective surgery when compared with emergency care, but further research is required around interventions to reduce weekday variation in outcomes.

Conclusions

In this study, we find that consultant seniority, derived from consultant years of experience, is not a significant factor in predicting 30-day mortality following elective surgery by day of the week. The end-of-the-week effect remains significant after adjusting for patient, consultant and hospital effects, suggesting that other factors might account for the higher risk of mortality towards the end of the week. We observed that consultants’ experience is lowest on a Friday; however, we do not find that this small variation has a significant impact on patient outcomes. There is small but significant decrease in the proportion of elective surgery carried out on Fridays and Saturdays by senior consultants. These findings are highly relevant in the context of moving to a 7-day-a-week hospital service24 and the implications for consultant staffing.25 Further research is needed to understand the causes of the Friday and weekend mortality effect.

References 

View Abstract

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

Footnotes

  • Contributors We confirm that all named authors meet the three criteria (1) substantial contributions to conception and design and/or analysis and interpretation of data, (2) drafting the article or revising it critically for important intellectual content and (3) final approval of the version to be published.

  • Funding The Dr Foster Unit at Imperial is affiliated with the National Institute for Health Research (NIHR) Imperial Patient Safety Translational Research Centre. The NIHR Imperial Patient Safety Translational Research Centre is a partnership between the Imperial College Healthcare NHS Trust and Imperial College London. The Dr Foster Unit at Imperial is grateful for support from the NIHR Biomedical Research Centre funding scheme. The Unit is largely funded by a research grant from Dr Foster Intelligence (an independent health service research organisation ).

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.