The costs of human Campylobacter infections and sequelae in the Netherlands: A DALY and cost-of-illness approach

Abstract

Campylobacter infections and sequelae pose an important public health problem for the Netherlands. With the help of a second order stochastic simulation model (using @Risk), confidence intervals (CI) for the associated disease burden (summing up morbidity and mortality) and the associated costs-of-illness were estimated. Approximately 80,000 persons per year (90% CI 30,000–160,000) are estimated to experience symptoms of acute gastro-enteritis, of which 30 are fatal cases. Around 18,000 patients visit a doctor and 500 are hospitalized each year. Additionally, each year some 1,400 cases of reactive arthritis, 59 cases of Guillain-Barré syndrome and 10 cases of inflammatory bowel disease are associated with a previous Campylobacter infection. The disease burden is expressed in Disability Adjusted Life Years (DALYs) and was estimated at 1,200 DALYs (90% CI 900–1,600 DALYs) per year. The associated costs for the Dutch society, using cost estimates for the year 2000, included direct health-care costs, direct non-health-care costs and productivity losses from missed work and were estimated to total €21 million (90% CI €11 million –€36 million) per year.

Keywords:

• Disease burden
• DALYs
• cost-of-illness
• direct health care costs
• direct non-health care costs
• indirect non-health care costs
• productivity losses
• friction-cost-method

Acknowledgments

The authors would like to acknowledge many people for their contributions to this study. There is, first of all some of the members of the Department of Neurology at the Erasmus Medical Centre Rotterdam, especially Pieter van Doorn, Marcel Garssen and Bart Jacobs; and there is Maarten Postma from the University of Groningen, all from the Netherlands. Further we would like to thank Margrit Ebinger from the University of Ulm, Germany, Timo Hannu from the Helsinki University Central Hospital, Finland, and Morten Helms and Henning Locht, both from the Statens Serum Institute, Copenhagen, Denmark, for their contributions. Finally, we would like to thank our colleagues from the National Institute of Public Health and Environment for their various contributions to this study, mainly Winette van den Brandhof, Yvonne van Duijnhoven, Wilfrid van Pelt and Matty de Wit from the Centre for Infectious Disease Epidemiology, Peter Teunis from the Centre for Information and Technology and Methodology and Nancy Hoeymans from the Centre for Public Health Forecasting. Comments from two anonymous referees are acknowledged.

Notes

Medical data, for example the duration of symptoms and the length of hospitalization, are often not normally distributed. By giving both, the mean and the median, the reader can get an idea of the possible skewness of the underlying distribution of the estimated value. For a symmetric distribution the median and the mean are equal. But if the distribution is right (positively) skewed, the mean is larger than the median, and vice-versa.

De Wit et al. (2001c) detected only 9 Campylobacter isolates in the community-cohort study, but used 24 different age and sex classes when standardizing. In order to avoid subgroups with no observed cases, we performed a re-standardization with only the 6 age classes. Details on the re-estimation procedure can be found in Havelaar et al. (2003), or are available on request.

De Wit et al. (2001a) detected Campylobacter as triggering agent in 10.5% of the gastrointestinal cases visiting a GP. We performed a re-standardization similar as for the community-cohort study, only the 6 age classes.

More details are available on request from first author.

Further details are given in the IBD-specific part of this paper.

Other non-Dutch studies estimating productivity losses, using both methods, are for example Goeree et al. (1999), Lopez-Bastida et al. (2003) and Marcotte and Wilcox-Gök (2001).

It was assumed that no additional travelling was required in order to buy over-the-counter medication.

It was assumed that medicines on prescription will be bought in a pharmacy on the way back from the GP.

According to Oostenbrink et al. (2000) the average cost of travelling by car or public transport in order to visit a GP is estimated to be _0.44 /visit. For patients visiting their GP, we did assume that most likely half of the patients would bike or walk, whereas the other half would take a car or public transport. Given the uncertainty of this assumption we included a Pert distribution with 50% being the most likely, and 10 and 90% the minimum and maximum estimates.

Based on the total number of Dutch hospital beds, we assumed that approximately 14% of the patients would be admitted to a university hospital and 86% would be admitted to a general hospital (Oostenbrink et al., 2000).

Apart from their consultation fee and their medical services specialists in the Netherlands charge their patients a so-called “subscription fee”. This could be either a “short subscription fee” or a “yearly subscription fee”. The short subscription fee is mostly only valid for about 2 months. Patients returning after those 2 months are charged an “additional subscription fee”, which is then valid for the rest of the year. Yearly subscription fees are charged if patients need treatments over a longer period. These “subscription fees” vary from specialist to specialist. However, only one specialist is allowed to charge these subscription fees per illness. A second specialist involved in the treatment of the illness can only charge what is referred to as a “clinical subscription fee”, which is less than the normal subscription fee. For more details see CTG (2000b).

According to Oostenbrink et al. (2000) the average cost of travelling by car or public transport to a hospital/specialist is estimated to cost _1.70/visit. Given the severity and given the longer distance, we assumed that hospitalized patients would always use either a car or public transport.

Given that fatal GE cases were already considered in one of the other sub-groups of GE patients and given that they do not die on the first day of the onset of symptoms, the productivity losses for this group might be slightly overestimated.

No study provided information on the duration of absence from work. In order to consult their GP, we assumed that patients might spend 0–0.25 days/consultation off work. Using a uniform distribution, we modelled the uncertainty of this assumption.

According to chorus (2001) the average length of sickness leave for patients with inflammation of the joints in the Netherlands (ReA is one of the illnesses within this category) was 40 days. For the general population this was only 14 days. However, the percentage of people who had to take days off due to illness was similar in both groups. In order to determine the length of sickness leave related to ReA only, we therefore used the difference between these two groups, which was 26 days.

We assumed that a specialist consultation or an outpatient hospital visit would result in a loss of 0.25–1 days/consultation.