Objectives To assess whether mortality of patients admitted on weekends and public holidays was higher in a district general hospital whose consultants are present more than 6 h per day on the acute medical unit with no other fixed clinical commitments.
Design Cohort study.
Setting Secondary care.
Participants All emergency medical admissions to Dumfries and Galloway Royal Infirmary between 1 January 2008 and 31 December 2010.
Methods We examined 7 and 30 day mortality for all weekend and for all public holiday admissions, using all weekday and non-public holiday admissions, respectively, as comparators. We adjusted mortality for age, gender, comorbidity, deprivation, diagnosis and year of admission.
Results 771 (3.8%) of 20 072 emergency admissions died within 7 days of admission and 1780 (8.9%) within 30 days. Adjusted weekend mortality in the all weekend versus all other days analysis was not significantly higher at 7 days (OR 1.10, 95% CI 0.92 to 1.31; p=0.312) or at 30 days (OR 1.07, 95% CI 0.94 to 1.21; p=0.322). By contrast, adjusted public holiday mortality in the all public holidays versus all other days analysis was 48% higher at 7 days (OR 1.48, 95% CI 1.12 to 1.95; p=0.006) and 27% higher at 30 days (OR 1.27, 95% CI 1.02 to 1.57; p=0.031). Interactions between the weekend variable and the public holiday variable were not statistically significant for mortality at either 7 or 30 days.
Conclusions Patients admitted as emergencies to medicine on public holidays had significantly higher mortality at 7 and 30 days compared with patients admitted on other days of the week.
- emergency department
Statistics from Altmetric.com
Studies have shown consistently higher inhospital and 30 day mortality among patients admitted as emergencies to hospital at weekends.1–4 The magnitude of the excess appears to be approximately 10% which, if correct, amounts to over 3000 additional deaths in England each year.2 The thrust of the discussion in each of the studies, and of the media coverage that has followed, is that these deaths might in some way be preventable or avoidable. The most striking data are from analyses of very large numbers of emergency admissions to hospitals in England,2 ,4 Europe,5 ,6 USA3 ,7 and Canada.1 Similar findings have been reported for the following specific emergency diagnoses: first acute myocardial infarction,8 acute pulmonary embolism,9 acute stroke10 ,11 and acute kidney injury.12 Reports from single centres tend to be underpowered for analyses of this sort.13–15 By contrast, an excess mortality among patients admitted at weekends has not been shown for admissions to the intensive care unit.16
Possible explanations for the adverse effect of weekend admission fall into four main categories: (1) patients admitted at weekends are sicker, (2) patients admitted at weekends have more comorbidity, (3) fewer hospital services are available at the weekend and (4) fewer experienced clinicians work at the weekend. Attempting to control for possible confounders by adjusting for comorbidity at the time of admission does not appear to attenuate the strength of the association between weekend admission and outcome,1 ,3 ,5 ,6 ,8 ,11 ,12 implying that comorbidity is not the explanation. It has only occasionally been possible to adjust for illness severity,9 ,16 ,17 which leaves us to speculate whether reduction in hospital services and/or fewer consultants working at weekends might be responsible.
The latter possibility is the one that has captured the imagination of commentators,18 prompting the British Medical Association to recommend that UK consultants undertake more on-call at weekends.19 A Royal College of Physicians of London (RCPL) survey has shown that fewer acute medicine specialists work at weekends than during the week and that admitting consultants are continuously available at weekends for less than 4 h per day in 38% of UK acute hospitals.20 This does not of course prove that mortality rates for weekend admissions would fall if more consultants were able to assess patients for more time at weekends, although in the opinion of many this would be a likely outcome. Indeed, a second RCPL report examining the relation between consultant working practices and patient outcomes has concluded that hospitals should ‘urgently ensure consultants are on-call for more than 1 day at a time, have no other routine duties during on-call periods, undertake two or more ward rounds per day in the acute medical unit (AMU) and are present in the AMU for more than 4 h per day, 7 days a week’.21
The purpose of our study was to determine if we could shed more light on this important area of acute medical care. We hypothesised that the mortality of patients admitted as medical emergencies at the weekend to a district general hospital whose medical unit was run by a team of general physicians, all with considerable experience of acute medical receiving, and all of whom spend a minimum of 6 h per day in hospital assessing patients, would not be greater than that of those admitted during the week. We further hypothesised that any excess mortality, if found, would be a consequence of an increase in mortality among patients admitted during public holidays as holiday periods have previously been shown to be a risk factor for death.6 ,22–24
Dumfries Infirmary is a 350 bed district general hospital serving a population of approximately 150 000 in south west Scotland. It is the only district general hospital in the region and has approximately 6700 emergency medical admissions each year. The AMU in Dumfries is a 30 bed ward led by a single handed acute physician and supported by a team of 16 consultant physicians, all of whom have dual accreditation in general medicine and a specialty. Two consultant ward rounds are undertaken each day, including weekends and public holidays, giving a minimum of 6 h direct clinical contact by a consultant each day. All weekend admissions are reviewed at least once a day while they remain on the AMU. Consultants cancel other clinical commitments when they are on-call for the AMU. Weekend cover is shared between two consultants.
Data on all adult emergency medical admissions to Dumfries Infirmary for the 3 year period from 1 January 2008 to 31 December 2010 were obtained from the Scottish Morbidity Record (SMR) 01 database, accessed via the ACaDMe Business Objects Universe. The SMR01 is an episode based record of all inpatient and day case discharges from general and acute wards in Scottish hospitals, managed by Information Services Division Scotland. The dataset contains both clinical and non-clinical information, such as patient demographics, admission dates, clinical diagnoses and the type of operations and procedures performed. Medical admissions were defined as the following specialties: general medicine (95% of records), cardiology, dermatology, endocrinology and diabetes, gastroenterology, geriatric medicine, neurology, palliative medicine, rehabilitation medicine, respiratory medicine and haematology. We excluded patients who were dead on arrival (DOA) or died in our emergency department (ED).
We conducted our analyses in two stages. We explored the relation between patient characteristics and two distinct variables, comparing all weekend admissions with all weekday admissions, and separately comparing all public holiday with non-public holiday admissions. We then compared mortality for all weekend admissions with any admission that was not at a weekend and for all public holiday admissions with any admission that did not occur on a public holiday, adjusting for possible confounders. All public holiday weekday admissions included a Monday or a Friday during the 3 years of this study, excepting 1 and 2 January 2008 which were on a Tuesday and Wednesday that year. Thus with the exception of these 2 days, all public holidays were part of a 3 or 4 day holiday period. For the purposes of these analyses, we considered the weekends attached to Monday and Friday public holidays to be public holiday weekends. Throughout, all days of admission started from 1 s after midnight.
We provided statistical summaries for all data using numbers and percentages. We used χ2 statistics and corresponding p values to determine whether there was any association between each potential predictor variable and the two independent variables of interest: the weekend or weekday variable and the public holiday or non-public holiday variable. We used multivariate logistic regression models to determine whether weekend or public holiday admission status had a significant effect on mortality by 7 and 30 days after admission, adjusting for other variables known to have an effect on mortality. These variables were age, gender, diagnosis, a measure of comorbidity, a measure of deprivation and year of admission. The measure of deprivation used here is the Scottish Index of Multiple Deprivation quintiles.25 We used the updated Charlson Comorbidity Index26 to weight the comorbidities. We further included an interaction between the weekend and public holiday admission variables to test for any differences between the public holiday status and the weekday or weekend effect.
All tests were two sided with a significance level of 5%. Analyses were performed using SAS V.9.2.
There were 20 072 emergency admissions to the medical unit in Dumfries during the 3 years from 1 January 2008 to 31 December 2010. The number (percentage) of admissions were as follows: all weekdays 15 469 (77.1% of all admissions) and all weekends 4603 (22.9%); all non-public holidays 18 944 (94.4%) and all public holidays 1128 (5.6%). The characteristics of patients admitted as emergencies were broadly similar across both variables. Patients admitted at weekends were marginally older (p<0.001), less likely to have a cancer diagnosis and more likely to have a respiratory diagnosis (p<0.001). Patients admitted on public holidays were also more likely to have a respiratory diagnosis (p=0.017). There were no statistically significant differences in gender, deprivation or comorbidity with either variable of interest (table 1).
The relation between day of admission and outcome is explored in table 2 which gives the number of admissions, and number and percentage of deaths at 7 and 30 days by day type before examining the risk associated with all weekend and all public holiday admissions separately. Seven hundred and seventy-one (3.8%) of 20 072 emergency admissions died within 7 days of admission and 1780 (8.9%) within 30 days. Percentage mortality was slightly higher at weekends (4.1% vs 3.8% at 7 days and 9.0% vs 8.8% at 30 days) but this was not statistically significant at 7 days (OR 1.10, 95% CI 0.92 to 1.31; p=0.312) or at 30 days (OR 1.07, 95% CI 0.94 to 1.21; p=0.322) after adjusting for possible confounders (table 2, figure 1).
By contrast, percentage mortality was higher for public holiday admissions (5.8% vs 3.7% at 7 days and 11.3% vs 8.7% at 30 days) and was statistically significant at both 7 days (OR 1.48, 95% CI 1.12 to 1.95; p=0.006) and 30 days (OR 1.27, 95% CI 1.02 to 1.57; p=0.031) (table 2). Interactions between the weekend variable and the public holiday variable were statistically non-significant for mortality at both 7 days (p=0.769) and 30 days (p=0.822), indicating that the increased risk of death associated with public holidays was not simply a consequence of admission on a public holiday weekend. These results indicate an increased risk of mortality if admitted on a public holiday, which may explain, at least in part, the previously observed adverse effect of weekend admissions.
Our study has shown that patients admitted as emergencies to medicine at weekends have a slightly but not significantly higher mortality at 7 and 30 days compared with patients admitted during the week. Higher percentage mortality among patients admitted at weekends has been a consistent finding in previous studies, and our results, which may have reached significance with a larger number of admissions, are broadly in line with reports from much larger datasets,1–4 despite a consultant presence on the AMU that was the same on weekends as on weekdays. Indeed, our working practices met all of the RCPL recommendations bar one: new admissions in Dumfries are seen once daily while on the AMU unless their clinical condition merits more frequent review. The belief that a lack of consultants at weekends is responsible for the ‘weekend effect’ has been the subject of much recent media interest and has also contributed to an RCPL recommendation that consultants should spend more time on the AMU at weekends.21 It remains uncertain, however, to what extent this would reduce variations in mortality.
When designing our study we hypothesised that if mortality was higher on any particular day or days of the week and that if this was a consequence of a lack of staff or services then this should be most apparent during public holidays. By considering mutually exclusive groups of admissions—those admitted on a weekend or not, and separately those admitted on a public holiday or not—we have been able to show an excess mortality that is indeed associated with public holiday admissions. While there is little here to suggest that a lack of services or lack of medical staff at normal weekends is in any way harmful, the same reassurance cannot be given to patients admitted as emergencies on public holidays. If we assume that patients with severe illnesses are no more likely to be admitted on any one day of the week than any other, then it becomes difficult to escape the view that a cumulative effect of lack of services and/or lack of doctors on public holidays must have a part to play in the higher public holiday mortality demonstrated in this study.
The possibility that holiday periods, including Christmas, New Year, public holidays and annual leave, might adversely affect outcome has been explored by others. A Spanish analysis of 429 880 emergency medical admissions reported that public holiday admissions were independently associated with inhospital mortality while noting that this was not any greater than that shown for weekend admissions generally (OR 1.07 in both instances).6 A report from Wales found a 41% increase in case fatality rate for upper gastrointestinal bleeding if patients were admitted on public holidays. The authors felt their findings could not be explained by case mix and might reflect reduced staffing levels and delays in endoscopy in some hospitals.24 A Danish study has shown that patients with a hip fracture had an independently increased risk of early postoperative mortality when admitted during holiday periods, which they defined as the 10 weeks in the year when staffing in the hip fracture unit was reduced because of planned holidays.23 A North American analysis of over 57 million deaths from 1997 to 2004 showed mortality spikes in DOA and ED settings over Christmas and New Year but not during other holiday periods, together with a more diffuse and slightly longer lasting increase in mortality in other settings (not further specified) during these two holiday periods. DOA and ED admissions accounted for approximately 10% of all deaths in this study. The authors considered nine possible explanations for the DOA/ED spikes, one of which was that fewer doctors were available in the ED, but felt unable to draw definite conclusions.22
Our study has strengths and limitations. Our main strength is that nearly all patients with medical emergencies in south west Scotland are admitted to the AMU in Dumfries and Galloway Royal Infirmary which has been served for the past 10 years by consultant physicians whose working practices meet nearly all of the recommendations made by the RCPL in their recent report.21 Further strengths are that we used a single coding team, thus ensuring consistent coding of comorbidities, and that we were able to analyse the characteristics and outcome of patients admitted on public holidays separately. The main limitations of our study are that we were unable to record illness severity and that as a single centre documenting admissions and deaths over a period of only 3 years we inevitably have fewer admissions and deaths to analyse than those reported by national registries. In particular, there were fewer public holiday admissions than we would have wished. Despite these limitations, we have shown an excess mortality for patients admitted on public holidays but not for those admitted at weekends. Incidentally, we know of only three other studies in which the impact of illness severity was assessed. Weekend admission no longer predicted a worse outcome after a biochemical illness severity score was included in the regression model in one study17 while patients admitted to an intensive care unit on a weekend were no more likely to die in hospital after adjusting for factors associated with mortality, such as the APACHE Score in another.18 In a study of patients admitted with pulmonary embolism, only those who were most severely affected had a worse outcome if admitted at weekends.9
In conclusion, our study of emergency admissions to a district general hospital in south west Scotland has shown statistically higher mortality associated with admission on both public holiday weekdays and public holiday weekends but not with weekend admissions generally. These results may explain, at least in part, the previously observed adverse effect of weekend admissions. Our data do not allow us to determine the cause of the excess public holiday mortality with certainty. Consultant physicians in Dumfries spend as much time on the AMU during public holidays as they do on normal days and weekends, but it is also true that fewer consultants and fewer junior doctors cover the other medical wards during holiday periods. We might speculate that higher mortality among patients admitted on public holidays reflects a cumulative lack of services and/or doctors during these 3–4 day periods. It goes without saying that all patients admitted as medical emergencies should have access to early consultant review and to an equally high standard of ongoing care, regardless of the day of admission. More research is warranted to clarify the precise nature of these mortality effects and especially to confirm whether there is a causal relationship between senior presence on the AMU and patient outcome.
Previous studies have suggested that patients admitted as emergencies at weekends are more likely to die than those admitted during the week.
Patients admitted as emergencies to medicine on public holidays had significantly higher mortality at 7 and 30 days compared with patients admitted on other days of the week.
Higher mortality was not seen in patients admitted at weekends
We speculate that the higher mortality among patients admitted on public holidays may relate to a cumulative lack of services and/or doctors during these 3–4 day periods.
Strengths and limitations of this study
Consultant physicians in our hospital spend as much time on the acute medical unit at weekends and on public holidays as they do on weekdays.
We used a single coding team and were able to analyse the characteristics and outcome of patients admitted on public holidays separately.
We were able to adjust our mortality data for comorbidity but not for illness severity.
Contributors CI, SS and AA had the idea for the paper. NG and AA provided statistical support. CI, SS and SF wrote the first draft. All authors contributed to the final version.
Competing interests None.
Ethical approval Because this study was a survey and did not involve identifiable patient data, we did not seek ethics approval, in keeping with our Health Board policy.
Provenance and peer review Not commissioned; externally peer reviewed.
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.