Acute upper gastrointestinal haemorrhage in west of Scotland: case ascertainment studyBMJ 1997; 315 doi: https://doi.org/10.1136/bmj.315.7107.510 (Published 30 August 1997) Cite this as: BMJ 1997;315:510
- Oliver Blatchford, audit research fellow ()a,
- Lindsay A Davidson, director of audita,
- William R Murray, consultant surgeonb,
- Mary Blatchford, general practitionerc,
- Jill Pell, senior registrar in public health medicined
- a Royal College of Physicians and Surgeons of Glasgow, Glasgow G2 5RJ
- b Glasgow Royal Infirmary NHS Trust, Glasgow G31 2ER
- c Easterhouse Health Centre, Glasgow G34 9HQ
- d Department of Public Health, Greater Glasgow Health Board, Glasgow G1 1ET
- Correspondence to: Dr O Blatchford Department of Public Health, University of Glasgow, Glasgow G12 8RZ
- Accepted 21 May 1997
Objectives: To determine the incidence and case fatality of acute upper gastrointestinal haemorrhage in the west of Scotland and to identify associated factors.
Design: Case ascertainment study.
Setting: All hospitals treating adults with acute upper gastrointestinal haemorrhage in the west of Scotland.
Subjects: 1882 patients aged 15 years and over treated in hospitals for acute upper gastrointestinal haemorrhage during a six month period.
Main outcome measures: Incidence of acute upper gastrointestinal haemorrhage per 100 000 population per year, and case fatality.
Results: The annual incidence was 172 per 100 000 people aged 15 and over. The annual population mortality was 14.0 per 100 000. Both were higher among elderly people, men, and patients resident in areas of greater social deprivation. Overall case fatality was 8.2%. This was higher among those who bled as inpatients after admission for other reasons (42%) and those admitted as tertiary referrals (16%). Factors associated with increased case fatality were age, uraemia, pre-existing malignancy, hepatic failure, hypotension, cardiac failure, and frank haematemesis or a history of syncope at presentation. Social deprivation, sex, and anaemia were not associated with increased case fatality after adjustment for other factors.
Conclusions: The incidence of acute upper gastrointestinal haemorrhage was 67% greater than the highest previously reported incidence in the United Kingdom, which may be partially attributable to the greater social deprivation in the west of Scotland and may be related to the increased prevalence of Helicobacter pylori. Fatality after acute upper gastrointestinal haemorrhage was associated with age, comorbidity, hypotension, and raised blood urea concentrations on admission. Although deprivation was associated with increased incidence, it was not related to the risk of fatality.
The incidence of upper gastrointestinal haemorrhage in the west of Scotland was 67% higher than the highest incidence previously reported in the United Kingdom
A substantial part of this excess incidence may be attributable to socioeconomic deprivation
The overall population mortality from upper gastrointestinal haemorrhage may increase as the elderly population increases because both incidence and case fatality rise steeply with age
A reduction in the overall case fatality from acute upper gastrointestinal haemorrhage will be best achieved by reducing case fatality among elderly patients
Acute upper gastrointestinal haemorrhage is a common reason for admission to hospitals in the United Kingdom. Until recently, knowledge of the epidemiology of this problem was based on small series with inadequate data on the denominator of the population. These series have reported case fatality (the proportion of cases dying) rather than population based mortality, which better reflects need. A recent large multicentre population based study of acute upper gastrointestinal haemorrhage in four English health regions found an incidence twice as high as expected.1 However, this was lower than the incidence reported in north east Scotland,2 which then had the highest reported incidence in the United Kingdom.
We report the epidemiology of upper gastrointestinal haemorrhage in the west of Scotland and compare it with contemporaneous English data.1 We investigate the effect of social deprivation and identify risk markers of case fatality. Our data were gathered as part of a prospective audit of outcome of treatment for acute upper gastrointestinal haemorrhage in all hospitals admitting adults with acute upper gastrointestinal haemorrhage in the six health boards in the west of Scotland. To our knowledge, this is the only multicentre population based study in the United Kingdom to have included all adult patients treated for acute upper gastrointestinal haemorrhage in a single large geographical area.
Subjects and methods
All adult patients treated for upper gastrointestinal haemorrhage in the acute admitting units of the 19 hospitals in the west of Scotland were prospectively identified over the six months from September 1992 to February 1993 inclusive.
Cases were defined as patients aged 15 years or over who had been admitted with an initial diagnosis of acute upper gastrointestinal haemorrhage. Cases occurring in inpatients admitted for other illness were also included. Cases subsequently shown not to have sustained an upper gastrointestinal haemorrhage were excluded. Standard definitions of haematemesis and melaena were used when abstracting data from clinical records.3
Cases were identified using different identification methods at each site appropriate to local record keeping systems. These included reference to admission records, ward records, records at operation and endoscopy, discharge letters and records, and patient administration systems. Because of known problems with the use of routine data sources for identifying cases4 at least two different identification methods were used at each site to ensure complete case ascertainment. The case records of at least 95% of identified cases were reviewed, but at one hospital only 60% of case records were retrievable owing to misfiling of records.
The Carstairs deprivation score, which is based on four census variables (unemployment among men, domestic overcrowding, car ownership, and low social class), was used as a measure of socioeconomic deprivation.5 Higher Carstairs scores indicate reduced affluence. The scores used in the study were derived from the 1991 census.6 Deprivation scores were assigned to cases according to their postcode sectors. These were divided into quarters instead of Carstairs' deprivation categories5 because of the large number of patients at the most deprived end of the scale.
Factors associated with case fatality were determined by multiple forward stepwise logistic regression using spss.7 Cases with missing data were excluded from analyses.
A total of 1882 adult cases were coded. This includes 61 patients whose acute upper gastrointestinal haemorrhage occurred while they were inpatients for other reasons and 61 patients who were transferred from other hospitals. Table 1 shows summary details of the patients' ages, sex distribution, and diagnoses.
In eight cases the final outcome of the episode was not recorded owing to incompleteness of the clinical record. No postcode could be traced in 24 cases, and the recorded postcodes were not valid in 52.
The adult population of the area from which these patients were admitted was 2 184 285 at the 1991 census, giving an overall incidence of 172 per 100 000 people aged 15 and over per annum (95% confidence interval 165 to 180).
The incidence rose sharply with age, being 5.7 times higher among those over 75 than among those aged 15 to 29 (P<0.00001), and it was twice as high among men as among women (P<0.00001). Table 1 also shows that the incidence of acute upper gastrointestinal haemorrhage also rose with increasing Carstairs deprivation score, being 2.2 times greater in the most deprived quarter than in the most affluent quarter (P<0.00001).
There were 153 deaths among the 1882 patients, resulting in an overall population mortality of 14.0 per 100 000 per annum (11.9 to 16.4). The population mortality increased sharply with increasing age (113 times greater among those over 75 than among those aged 15 to 29; P<0.00001) (table 1), and mortality among men was 1.6 times greater than among women (P=0.005). The population mortality in the most deprived quarter was double that in the least deprived quarter (P<0.0002).
The proportion of case fatalities was 8.1% (6.9% to 9.6%). Case fatality among the oldest age group was 20 times that among the youngest (P<0.00001) (table 1), but there was no significant difference in case fatality between the sexes or in association with the Carstairs deprivation score.
The case fatality of patients admitted after acute bleeds was 6.7% (5.6% to 8.1%). Deaths were significantly more frequent among those who bled as inpatients (42%, P<0.00001) or after having been transferred from another hospital (16.4%, P<0.00001).
Table 2 shows diagnoses by age, sex, and death rate. No diagnosis was recorded in 542 cases; 308 cases had two recorded diagnoses, 60 had three, and 10 had four. The commonest recorded diagnoses for cases with multiple diagnoses were oesophagitis (153, or 41% of these cases), gastritis (112, 30%), duodenitis (107, 28%), duodenal ulcer (106, 28%), and gastric ulcer (73, 19%). Gastric ulcers were more common in women. Duodenal ulcers and duodenitis were more common in men. Oesophagitis, oesophageal ulcer, gastric ulcer, and malignancy were more common in elderly people, while the Mallory-Weiss syndrome, gastritis, and duodenal ulcer were more common in young people. Oesophageal varices, oesophageal ulcer, and gastrointestinal malignancy were associated with increased risks of death. The Mallory-Weiss syndrome, oesophagitis, gastritis, and duodenitis had significantly low risks of death.
Risk factors for case fatality
Table 3 shows the case fatality odds ratios for each clinical risk factor recorded at the time of admission of emergency cases. Among the factors not associated with increased case fatality were a history of previous upper gastrointestinal haemorrhage, upper gastrointestinal surgery, peptic ulceration, and dyspepsia. Non-steroidal anti-inflammatory drugs, systemic steroids, acid blocking drugs, and anticoagulants were also not risk factors for case fatality; neither was there a relation with Carstairs deprivation score.
Table 3 shows that many risk factors on admission were associated with increased case fatality when they were considered in isolation. Logistic regression was used to determine the risk factors that best predicted death among these emergency admissions. When these factors were considered together in this multifactorial analysis many of the associations with risk lost significance. Factors selected were age, uraemia, diastolic hypotension, cardiac failure, hepatic failure, pre-existing malignancy or other major pre-existing disease (such as rheumatoid arthritis or inflammatory bowel disease), and presentation with frank haematemesis or syncope (table 3). Among the factors not selected in the regression model were haemoglobin concentration, creatinine concentration, pulse, and a history of oesophageal varices.
The incidence of acute upper gastrointestinal haemorrhage in the west of Scotland in this study was much higher than that previously reported in the United Kingdom. The areas covered by the health boards in the west of Scotland are geographically distinct from the rest of Scotland, which minimises cross boundary flow of patients. This resulted in good matching between cases and data on the denominator of the population, which improves the reliability of the estimates of incidence. Forty per cent of case records were not available for review at one hospital, so the true incidence would have been higher.
Deprivation and Helicobacter pylori infection
We believe that this is the first study in the United Kingdom to show a link between socioeconomic deprivation and an increased incidence of upper gastrointestinal bleeding. Peptic ulceration is the commonest disease underlying acute upper gastrointestinal haemorrhage in the United Kingdom.2 8 The incidence of peptic ulceration is higher in Scotland and the north of England than further south.9 10 This difference has been the subject of speculation for some time9 10 but remains unexplained. There is now a recognised association between Helicobacter pylori infection, socioeconomic group,11 and childhood living conditions,12 and an increased prevalence of Helicobacter pylori infection has recently been found in the north of England and Scotland.13 Peptic ulcer bleeding has been found to be more common among less affluent patients in Germany14 and Africa,15 but these studies lacked adequate data on the denominator of the population.
Although this study is subject to the usual cautions about applying population characteristics such as deprivation scores to individual people (the so called ecological fallacy), such an association is clinically plausible. This relation between Carstairs deprivation scores and incidence of upper gastrointestinal haemorrhage may be related to the reported greater prevalence of Helicobacter pylori associated with reduced socioeconomic group or crowded childhood living conditions. Although socioeconomic deprivation was associated with an increased risk of developing upper gastrointestinal haemorrhage, it did not seem to be associated with differences in the underlying disease processes or with clinical outcome indicated by case fatality.
The findings of a greater incidence among men and elderly people and of a greater case fatality among elderly people and patients who bled after having been admitted for other disease are compatible with other studies.1 2 16 The fivefold increase in incidence from the youngest to the oldest age groups coupled with the 20-fold increase in case fatality gives a dramatic increase in age specific population mortality among elderly people of more than 100 times that among young people. With an increasing proportion of elderly people in the population, there may be an apparent increase in crude death rates from acute upper gastrointestinal haemorrhage even if improved clinical management results in an improved crude case fatality. This has implications for resources in planning clinical services for acute upper gastrointestinal haemorrhage. Reductions in crude case fatality may be achieved by targeting clinical resources (such as endoscopic intervention) towards elderly patients.
The relative proportions of individual diagnoses and case fatalities are similar to those of Rockall et al,1 except for gastrointestinal malignancy, which was less common and had a lower case fatality in our study. The findings of this study are broadly compatible with other studies that have shown an increased risk of case fatality associated with coexisting disease, hypotension or shock, and the onset of haemorrhage in inpatients.1 2 16 17 18 19 20 21 22 The finding that anaemia at the time of admission might not be strongly associated with case fatality after adjustment for other factors including blood urea concentration has not, to our knowledge, previously been reported.
We thank the Scottish Office Clinical Resource and Audit Group and the Royal College of Physicians and Surgeons of Glasgow for their support. The views contained in this paper are ours. We thank the college's upper gastrointestinal haemorrhage steering group for its advice, and Mr Robert Dunsmuir, Mrs Kathleen Lindsay, and Mrs Fiona Marjoribanks for their help. We thank all of the clinicians in the west of Scotland for allowing us access to their patients' records.
Funding: Scottish Office Clinical Resource and Audit Group (grant No CA91/24F) and the Royal College of Physicians and Surgeons of Glasgow.
Conflict of interest: None.