Repository logo

The epidemiology of hospital admissions in a general population: record linkage of hospital episode statistics to the European Prospective Investigation of Cancer (EPIC-Norfolk) cohort



Change log


Luben, Robert Neil 


The UK National Health Service (NHS) is primarily funded by taxation free at the point of delivery. Hospitals account for approximately 50% of overall NHS spending. Two-thirds of people admitted to hospital are over 65 with those over 85 accounting for 25% of bed days. This thesis aims to quantify hospital usage in a general population of middle-aged and older men and women over a 20-year follow-up period and to examine related demographic and behavioural factors. Patterns of hospital usage are described using two main hospital usage measures: admission numbers and length of stay. Socioeconomic factors such as education, occupational social class and residential area deprivation that may predict future hospital usage are examined. I assess the relationships between potentially modifiable factors such as cigarette smoking, the consumption of alcohol, body mass index and physical activity and future hospital usage while the implications for clinical and public health planning, policy and practice are also considered.

The thesis is based on the European Prospective Investigation of Cancer in Norfolk (EPIC-Norfolk), a community-based cohort of 25,639 men and women aged 40-79 at recruitment between 1993-1997 and followed up to the present. Participants completed a lifestyle questionnaire and attended a clinic where measurements and blood were taken at baseline and again at a second time-point after 12 years. All participants were linked to hospital records using their unique NHS numbers and to census data using their postcode. Episode statistics including admission and discharge dates were used to create numbers of admissions and length of stay outcomes. ICD-10 diagnosis codes were used to construct a hospital multimorbidity outcome using the Charlson Comorbidity Index above the level of 3. Logistic regression was the primary statistical model used throughout the analyses. Exposures were examined prospectively, prior to any hospital admission.

The current analyses were conducted on 25,014 participants in the cohort still alive in 1999 when hospital admission data were first available. Over the first 10 years of follow-up, 73% of study participants had at least one admission to hospital, 14% with ≥7 admissions and 20% with >20 hospital days. After 20 years, 90% of participants had a hospital admission, 65% had ≥7 admissions and 59% had >20 hospital days. High numbers of admissions and hospital days were positively associated with male sex, age, manual social class, current cigarette smoking and body mass index (BMI) >30 kg/m². The thesis examined levels of deprivation both at individual level, using education and occupational social class, and residential area level using the Townsend Area Deprivation Index. Compared with those having Townsend Index lower than the average for England and Wales, those with a higher than average deprivation index had a higher likelihood of spending >20 days in hospital. Occupational social class and educational attainment modified the association between area deprivation and hospitalisation; those with manual social class and lower education level were at greater risk of hospitalisation when living in an area with higher deprivation index.

The thesis also examined potentially modifiable behavioural factors. Compared with current non-drinkers, men and women who reported any alcohol drinking had a lower risk of spending >20 days in hospital. Participants with a baseline physical activity score of at least moderately inactive had fewer hospital admissions and fewer days in hospital over 10 years, than those who were inactive. Similar associations were observed over 10 years from time-point two (TP2) and similar but attenuated results were observed for 20-year follow-up. Participants who remained physically active or became active between baseline and TP2 had lower risk of subsequent hospital usage than those who remained inactive or became inactive.

An additional hospital-based outcome measure, hospital admission with multimorbidity (HAWM), was used to examine incident multimorbidity for participants free of the condition at baseline. Baseline 5-year and 10-year incident HAWM were observed in 11% and 21% of participants, respectively. More men had incident HAWM than women and those aged >75 years had the highest proportion of multimorbid conditions with 29% at 5 years and 47% at 10 years. HAWM rates at TP2 were similar to baseline. Longer duration of hospital stay and number of admissions, age, male sex and prevalent diseases, smoking, physical inactivity, high BMI and low fruit and vegetable intake were associated with incident HAWM.

Simple demographic and behavioural indicators are related to the future probability of cumulative hospital admissions, length of stay and hospital admissions with multimorbidity. Increasing age, male sex and modifiable factors such as smoking, body mass index and usual physical activity are all strongly associated with subsequent hospital usage. Modest feasible differences in lifestyles in the general population may potentially mitigate the future impact of long hospital stay and multimorbidity and have a substantial impact on hospital usage and costs. The social determinants of health are well recognised. While some of the socioeconomic gradient in ill health has been attributed to differences in behavioural factors, there is also a socioeconomic gradient in hospital usage for individually measured social class and education and for area level deprivation apparently independently of measured behavioural factors and reported prevalent disease which warrant further exploration.





Pharoah, Paul
Khaw, KT


epidemiology, hospital episode statistics, cohort study, linkage, sociodemographic and lifestyle factors, hospitalisation and hospital usage, middle aged and older participants, hospital admissions and length of stay


Doctor of Philosophy (PhD)

Awarding Institution

University of Cambridge
Medical Research Council (G1000143)