Use of a coronary heart disease simulation model to evaluate the costs and effectiveness of drugs for the prevention of heart disease

Abstract

A discrete event simulation model of the patient pathways in the treatment of coronary heart disease (CHD) was used to quantify the health gains and costs associated with increasing secondary prevention drugs prescription for patients with CHD based on the level recommended in the National Service Framework for the UK. A Gompertz distribution was sampled for time to failure (death or non-fatal heart attack). The time to failure was modified in relation to the reduced risk of failure for those on the relevant drugs. The results from the model were validated against national data. Increasing the levels of prescription of secondary prevention drugs to those patients with CHD might prevent 100 deaths per million population per year and cost an additional £4 million per million population per year. With cost per life year saved of £5520, this appears good value for money compared with other health technologies.

Appendix. Derivation of MI/death rates

Data used

The EMMACE data set (n=2196) (Lawrance et al, 2001; Robinson, 2001) was a prospective cohort that provided post-MI mortality rates over 5 years. The British Regional Heart Study (BRHS) (Lampe et al, 2000) was a prospective cohort which followed a group of 7735 men aged 40 to 59 years, at baseline; including some who had CHD at the start.

Although the BRHS data only related to men, we used it to derive long-term survival (after 1 year after an MI), assuming that men and women would have similar survival probabilities. BRHS did not provide survival in the first year after an MI or the age gradient, for either the first year or subsequent years.

EMMACE provided mortality data following an MI. The data were adjusted to estimate cardiac mortality, by age group and by year following the MI. Further adjustments were made, using UKHAS data (Norris, 1998), to estimate surviving MIs as well as deaths. The age groups provided the gradients we needed.

Survival probabilities in the first year after an MI were lower than subsequently and therefore we looked for different distribution parameters for the first year than for subsequent years. More details on the derivations are available in the project report (Davies et al, 2003) or on request from the authors.

The steps in the calculation

• 1) The EMMACE data were analysed to provide the sudden CHD death/non-fatal MI rate in the first year after an MI and to find the gradient of event rates between different age groups. The death rates by year since the last MI, were grouped together in 10-year age bands. Males and females were grouped together. It was clear that the death rate in the first year, for each age band, was higher than in the subsequent years. There were small numbers of deaths in the subsequent years, and in some cases none at all. In order to overcome the problem of small numbers, the probability of dying in each of years 2 to 5, in each age band, was assumed to be the same. It was thus possible to derive the death rate in the first year and the expected death rate in each subsequent year.

• 2) The EMMACE data were for all cause mortality. Office for National Statistics (1999) non-CHD death rate data were subtracted from the total death rate, by age for both sexes to give the CHD death rate.

• 3) The CHD death rate was adjusted to take account of the effects of the secondary prevention drugs to give the ‘natural history’ of CHD patients who are not on any medication. The prevalence of secondary prevention prescribing in EMMACE was as follows: 42% received beta blockers, 38% ACE inhibitors, 86% aspirin and 8% statins and the hazard ratios of the individual drugs were assumed to be 0.77, 0.80, 0.75 and 0.73, respectively.

• 4) The UKHAS data (Norris, 1998) giving deaths in and out of hospital were used to estimate the number of MIs to survive to discharge in addition to the CHD deaths, using conditional probability calculations.

• 5) The estimated MI/sudden death rates in the first year and in the subsequent years were fitted by exponential distributions, across the age groups. The MI/sudden death rate in the first year was used in the model. The age gradients between ages 55 and 70 were calculated for the MI/sudden death rate in subsequent years.

• 6) The long-term MI/sudden death rate and for those with angina only and history of MI was estimated using the age gradient for the MI/sudden death rate in subsequent years and a point estimate from BRHS. BRHS gives major heart disease events (deaths and MIs) related to patient years survived per year of those who had a history of MI and those with angina only at baseline. The event rate was adjusted by the proportions who were on secondary prevention medication at that time to give the natural history of CHD patients. The prevalence of secondary prevention prescribing in BRHS was as follows: 40% beta blockers, 0% ace-inhibitors, 32% aspirin and 0% statins.