Factors Contributing to the Incidence of Acute Pancreatitis
Factors Contributing to the Incidence of Acute Pancreatitis
To investigate acute pancreatitis across Wales, UK (population 3.0 million), we used systematic record linkage of national in-patient, mortality and primary care data. These data are incorporated along with other health and social services data in the Secure Anonymised Information Linkage (SAIL) Databank, funded by the Welsh Assembly Government since 2006 and stored on an IBM supercomputer in the College of Medicine, Swansea University. These data have been used as the basis of many published studies in international Medline journals. The hospital in-patient data used, the Patient Episode Database for Wales (PEDW), includes abstracts of in-patient and day case admissions to all National Health Service (NHS) hospitals in Wales. The in-patient data were systematically linked together to enable subsequent admissions for the same people to be traced. They were then record linked to mortality data from the Office for National Statistics (ONS) and the NHS Welsh Administrative Register to identify deaths that occurred following discharge from hospital along with the in-patient deaths. The in-patient and mortality data were also record linked to SAIL primary care data, obtained from 35% of all general practices across Wales during the entire study period (population 1.0 million) to obtain further information on aetiology. The record linkage of these in-patient, mortality and primary care data has been validated previously and has been shown to be >99.8% accurate.
The study included admissions for acute pancreatitis during the 12-year period from 1 January 1999 to 31 December 2010, with 60 day follow-up to 1 March 2011. We selected only those admissions where acute pancreatitis was recorded as the principal diagnosis on the discharge record and included all sources of admission. The International Classification of Diseases tenth revision (ICD-10) code used for acute pancreatitis was K85. We included each person's first admission for acute pancreatitis after the start of the study period on 1 January 1999. As acute pancreatitis is often characterised by subsequent attacks that each require hospitalisation, we included subsequent admissions for acute pancreatitis for each individual – as subsequent attacks – if they occurred at least 60 days following discharge from a previous admission.
We determined the two main aetiologies of acute pancreatitis (gallstone or biliary and alcohol) as follows. Firstly, gallstone acute pancreatitis was defined where there was a diagnosis of cholelithiasis (ICD-10 code = K80) or cholecystitis (K81) recorded in any diagnostic position on the in-patient record during the patient's current admission, previous admissions or from primary care consultations during the previous 5 years. Gallstone acute pancreatitis was also determined if the following surgical procedures were recorded in any position during the current or previous in-patient admissions during the last 5 years: total cholecystectomy and excision of surrounding tissue (OPCS-4 procedure code = J18.1), endoscopic sphincterotomy of sphincter of odi and removal of calculus HFQ (J38.1) or endoscopic retrograde extraction of calculus from bile duct (J41.1). Alcoholic pancreatitis was defined where any one of 21 alcohol-attributable diagnoses or symptoms were similarly recorded in any position during the current admission, previous admissions or primary care consultations. These 21 conditions and their ICD-10 codes are listed in Appendix 1. Other aetiologies of acute pancreatitis were similarly defined from diagnoses recorded in any position on the current or previous health records; including hyperlipidaemia (ICD-10 code = E78), hypercalcaemia (E83.5), malnutrition (E40–E46), abdominal trauma (S30–S39), pancreatic malignancies (C25) and cystic fibrosis (E84).
We measured social deprivation using the Welsh Index of Multiple Deprivation (WIMD), 2005 version as this was at the midpoint of the study period from 1999 to 2010. WIMD is similar to the widely used English Indices of Multiple Deprivation (IMD), and WIMD 2005 comprises seven separate domains of deprivation. These are 'income' (25% contribution), 'employment' (25%), 'education' (15%), 'health' (15%), 'geographical access to services' (10%), 'housing' (5%) and 'physical environment' (5%). The total WIMD deprivation scores were assigned anonymously to 1896 Lower Super Output Areas (LSOAs) across Wales (average LSOA population = 1560). The LSOAs were then ranked according to their social deprivation score and were categorised into quintiles (I = least deprived and V = most deprived quintile).
Information on the consumption of different types of alcoholic beverage were obtained from the official Welsh Health Survey, collected during the three survey years from 2008 to 2010. The Welsh Health Survey is a nationally representative sample of approximately 15 000 adults (0.5% of the population) each year. Respondents reported the numbers of units of wine, beer and spirits that they had consumed on the day in the week prior to the survey on which they had consumed the most alcohol, which were then standardised using the resident population of Wales. The mean alcohol consumption figures were then correlated with acute pancreatitis incidence across the 94 study Upper Super Output Areas (USOAs; average USOA population = 32 000 and average number of survey respondents per USOA each year = 160) during the same 3-year period from 2008 to 2010.
We also compared the incidence of acute pancreatitis for admissions on weekends (00:00 hours on Saturday to 00:00 hours on Monday) on weekdays and on bank holidays. Seasonal effects in the incidence of acute pancreatitis were assessed according to the calendar month of each admission and according to daily admissions in December and January, and weekly admissions during the Christmas and New Year weeks (last week of December and first week of January).
The main study outcome measures were incidence rates for acute pancreatitis per 100 000 population and mortality rates at 60 days following admission. Although mortality at 30 days is also reported, 60 days was chosen as the preferred mortality outcome measure as a 30 day limit would exclude some deaths that occur during prolonged in-patient stays for severe necrotising cases. Incidence rates were calculated using the numbers of hospitalised cases for acute pancreatitis as numerators and the corresponding resident populations as denominators. The incidence rates were then standardised using the direct method and the Welsh resident population during the study period as the standard, and were expressed per 100 000 population. Percentage mortality was calculated by dividing the numbers of deaths (from all causes) by the numbers of hospitalised cases and was standardised directly using the study population hospitalised with acute pancreatitis. Other methods of analysis include mean annual changes over time in incidence and mortality rates, relative risks and their 95% CIs to compare seasonal incidence rates, Spearman's rank correlations to assess possible links between incidence and alcohol consumption, one way analysis of variance (anova) to assess trends over time in durations of in-patient stay, and weighted 5 year moving averages to smooth daily trends in incidence. Significance was measured at the conventional 5% level.
Methods
To investigate acute pancreatitis across Wales, UK (population 3.0 million), we used systematic record linkage of national in-patient, mortality and primary care data. These data are incorporated along with other health and social services data in the Secure Anonymised Information Linkage (SAIL) Databank, funded by the Welsh Assembly Government since 2006 and stored on an IBM supercomputer in the College of Medicine, Swansea University. These data have been used as the basis of many published studies in international Medline journals. The hospital in-patient data used, the Patient Episode Database for Wales (PEDW), includes abstracts of in-patient and day case admissions to all National Health Service (NHS) hospitals in Wales. The in-patient data were systematically linked together to enable subsequent admissions for the same people to be traced. They were then record linked to mortality data from the Office for National Statistics (ONS) and the NHS Welsh Administrative Register to identify deaths that occurred following discharge from hospital along with the in-patient deaths. The in-patient and mortality data were also record linked to SAIL primary care data, obtained from 35% of all general practices across Wales during the entire study period (population 1.0 million) to obtain further information on aetiology. The record linkage of these in-patient, mortality and primary care data has been validated previously and has been shown to be >99.8% accurate.
Study Inclusion and Exclusion Criteria
The study included admissions for acute pancreatitis during the 12-year period from 1 January 1999 to 31 December 2010, with 60 day follow-up to 1 March 2011. We selected only those admissions where acute pancreatitis was recorded as the principal diagnosis on the discharge record and included all sources of admission. The International Classification of Diseases tenth revision (ICD-10) code used for acute pancreatitis was K85. We included each person's first admission for acute pancreatitis after the start of the study period on 1 January 1999. As acute pancreatitis is often characterised by subsequent attacks that each require hospitalisation, we included subsequent admissions for acute pancreatitis for each individual – as subsequent attacks – if they occurred at least 60 days following discharge from a previous admission.
Aetiology of Acute Pancreatitis
We determined the two main aetiologies of acute pancreatitis (gallstone or biliary and alcohol) as follows. Firstly, gallstone acute pancreatitis was defined where there was a diagnosis of cholelithiasis (ICD-10 code = K80) or cholecystitis (K81) recorded in any diagnostic position on the in-patient record during the patient's current admission, previous admissions or from primary care consultations during the previous 5 years. Gallstone acute pancreatitis was also determined if the following surgical procedures were recorded in any position during the current or previous in-patient admissions during the last 5 years: total cholecystectomy and excision of surrounding tissue (OPCS-4 procedure code = J18.1), endoscopic sphincterotomy of sphincter of odi and removal of calculus HFQ (J38.1) or endoscopic retrograde extraction of calculus from bile duct (J41.1). Alcoholic pancreatitis was defined where any one of 21 alcohol-attributable diagnoses or symptoms were similarly recorded in any position during the current admission, previous admissions or primary care consultations. These 21 conditions and their ICD-10 codes are listed in Appendix 1. Other aetiologies of acute pancreatitis were similarly defined from diagnoses recorded in any position on the current or previous health records; including hyperlipidaemia (ICD-10 code = E78), hypercalcaemia (E83.5), malnutrition (E40–E46), abdominal trauma (S30–S39), pancreatic malignancies (C25) and cystic fibrosis (E84).
Exposure Measures
We measured social deprivation using the Welsh Index of Multiple Deprivation (WIMD), 2005 version as this was at the midpoint of the study period from 1999 to 2010. WIMD is similar to the widely used English Indices of Multiple Deprivation (IMD), and WIMD 2005 comprises seven separate domains of deprivation. These are 'income' (25% contribution), 'employment' (25%), 'education' (15%), 'health' (15%), 'geographical access to services' (10%), 'housing' (5%) and 'physical environment' (5%). The total WIMD deprivation scores were assigned anonymously to 1896 Lower Super Output Areas (LSOAs) across Wales (average LSOA population = 1560). The LSOAs were then ranked according to their social deprivation score and were categorised into quintiles (I = least deprived and V = most deprived quintile).
Information on the consumption of different types of alcoholic beverage were obtained from the official Welsh Health Survey, collected during the three survey years from 2008 to 2010. The Welsh Health Survey is a nationally representative sample of approximately 15 000 adults (0.5% of the population) each year. Respondents reported the numbers of units of wine, beer and spirits that they had consumed on the day in the week prior to the survey on which they had consumed the most alcohol, which were then standardised using the resident population of Wales. The mean alcohol consumption figures were then correlated with acute pancreatitis incidence across the 94 study Upper Super Output Areas (USOAs; average USOA population = 32 000 and average number of survey respondents per USOA each year = 160) during the same 3-year period from 2008 to 2010.
We also compared the incidence of acute pancreatitis for admissions on weekends (00:00 hours on Saturday to 00:00 hours on Monday) on weekdays and on bank holidays. Seasonal effects in the incidence of acute pancreatitis were assessed according to the calendar month of each admission and according to daily admissions in December and January, and weekly admissions during the Christmas and New Year weeks (last week of December and first week of January).
Outcome Measures and Methods of Analysis
The main study outcome measures were incidence rates for acute pancreatitis per 100 000 population and mortality rates at 60 days following admission. Although mortality at 30 days is also reported, 60 days was chosen as the preferred mortality outcome measure as a 30 day limit would exclude some deaths that occur during prolonged in-patient stays for severe necrotising cases. Incidence rates were calculated using the numbers of hospitalised cases for acute pancreatitis as numerators and the corresponding resident populations as denominators. The incidence rates were then standardised using the direct method and the Welsh resident population during the study period as the standard, and were expressed per 100 000 population. Percentage mortality was calculated by dividing the numbers of deaths (from all causes) by the numbers of hospitalised cases and was standardised directly using the study population hospitalised with acute pancreatitis. Other methods of analysis include mean annual changes over time in incidence and mortality rates, relative risks and their 95% CIs to compare seasonal incidence rates, Spearman's rank correlations to assess possible links between incidence and alcohol consumption, one way analysis of variance (anova) to assess trends over time in durations of in-patient stay, and weighted 5 year moving averages to smooth daily trends in incidence. Significance was measured at the conventional 5% level.
