When analyses are invalidated by erroneous assumptions

To the Editor:

In a recent publication in the “International Journal of Technology Assessment in Health Care” (7), Kildemoes and Kristiansen claim to address “Cost-effectiveness of interventions to reduce the thrombolytic delay for acute myocardial infarction.” Their study is based on a “Master of Public Health Assessment” thesis published by Kildemoes in the year 2001 (6). Three years ago, the author was informed that several of her assumptions were incorrect. In this letter, we will address six of the erroneous assumptions made by Kildemoes and Kristiansen.

§1. Kildemoes and Kristiansen assume that 0 percent of patients are treated within 1 hour of symptom onset if a strategy of prehospital diagnosis with the use of telemedicine is implemented. Kildemoes and Kristiansen claim that the short transport distances make prehospital thrombolysis impossible in Denmark. Furthermore, Kildemoes assumes in her thesis that “due to short distances to hospital in Denmark, the prehospital diagnostic procedure with the use of telemedicine will not be completed in the majority of patients until arrival at the hospital.” These assumptions are in clear contrast to previous international and Danish findings. Data from studies in the Netherlands indicate that implementation of a prehospital strategy of diagnosis and thrombolysis results in a 1-hour reduction in treatment delay, and results in 25% of patients being treated within 1 hour of symptom onset (8;9). Notably, the transport time to hospital in the study region was comparable to Danish findings (15;16). Furthermore, previous Danish findings report that, even in urban areas, the majority of patients can be diagnosed before hospital admission with the use of telemedicine (15).

§2. When estimating the benefit of a reduction in treatment delay before initiation of thrombolysis, Kildemoes and Kristiansen used a curve by Boersma and colleges. This curve describes an inverse relationship between treatment delay and the number of extra lives saved per 1,000 treated with thrombolysis instead of placebo (3). However, the Boersma curve was based on the assumption that thrombolysis is initiated promptly after admission or randomization in a study. More likely, 45–60 minutes elapse from admission to initiation of thrombolysis in regions not covered by a prehospital diagnostic strategy (18). Thus, a right-shifted Boersma-curve may be more appropriate to describe the relationship between mortality and time to initiation of thrombolysis (14). In addition, the trials giving rise to the Boersma curve were not designed to evaluate the benefit of thrombolysis according to different treatment delays. The trials were designed to estimate the benefit of thrombolysis compared with placebo. The only trials comparing the beneficial effect of thrombolysis according to different treatment delays are the trials comparing prehospital thrombolysis with in-hospital thrombolysis. Meta-analyses including these trials report that a reduction in treatment delay from 2.7 hours to 1.7 hours, achieved by initiating thrombolysis before hospital admission rather than at the hospital, results in 15–22 extra lives saved per 1,000 treated (3;11;17). Similar results are obtained by a modified Boersma curve (14). In Denmark, the median treatment delay among patients diagnosed and treated with reperfusion therapy after hospital admission is close to 2.7 hours (12). If 75 percent (n=3,000) of the latter patients can be diagnosed and treated before hospital admission, then 45–66 extra lives could be saved per year in Denmark, that is, seven to nine times the number expected by Kildemoes and Kristiansen.

§3. In their study, Kildemoes and Kristiansen do not specify in detail the cost of a prehospital diagnostic strategy with the use of telemedicine. However, detailed information is available in her original thesis. To establish a prehospital diagnosis of acute myocardial infarction (AMI) with the use of telemedicine, it is necessary that ambulances have the equipment for electrocardiogram (ECG) acquisition and transmission. For a physician at the hospital to evaluate the ECGs, the transmitted ECGs must be received at a hospital-based “receiving station.”

The first major miscalculation in the thesis by Kildemoes was to introduce 500 extra receiving stations in the model, that is, one receiving station for each ambulance, equal to an extra cost of 34,400,000 DKK. In fact, five receiving stations are sufficient to cover the country of Denmark.

In her thesis, Kildemoes stated that all ambulance staff was to receive additional training in acquiring and transmitting ECGs. The extra cost was estimated at 62,447,400 DKK. Kildemoes ignored the fact that a Departmental order was given in Denmark in the year 2000 by the National Board of Health to improve the prehospital management of patients. Thus, as a part of their basic education, all ambulance staff are scheduled to receive additional training in prehospital management of acute patients, irrespective of whether a prehospital diagnostic strategy is implemented or not. During this extra education, the ambulance staff will automatically obtain skills in acquiring and transmitting ECGs.

Kildemoes and Kristiansen assume that all ambulances need to be equipped with new defibrillators to acquire and transmit ECGs, with an additional cost of 85,520,000 DKK. In fact, more than 90 percent of ambulances in Denmark have defibrillators suitable for ECG acquisition. To transmit the ECGs, only an extra communication module is necessary in the ambulances, corresponding to an extra cost of 13,907,500 DKK.

In her original thesis, Kildemoes ended up with a total cost of 351,913,149 DKK when implementing a 5-year strategy of prehospital diagnosis with the use of telemedicine. If selling the 500 extra receiving stations, if you accept that ambulance staff as a part of their basic education obtain skills in acquiring and transmitting ECGs, and if you accept that a nationwide strategy of transmitting ECGs only necessitate that ambulances are equipped with a communication module, then the actual extra cost of a 5-year strategy of telemedicine is at maximum 183,453,249 DKK or less than two thirds of the amount mentioned in the study by Kildemoes and Kristiansen and only half that mentioned in the thesis by Kildemoes.

§4. Kildemoes assumes that media campaigns have a long-lasting effect. This is not consistent with previous findings. Most trials studying the effect of media campaigns have concluded that, if any effect at all, it is of short duration. It is surprising that Kildemoes and Kristiansen ignore the result of the largest study so far addressing this issue. Thus, the REACT trial, performed in twenty cities in ten states in the United States, and with a more prolonged intervention phase than the trial by Blohm and colleges (2), documented no effect of a media campaign on treatment delay, no effect whatsoever (10).

§5. Kildemoes and Kristiansen conclude that “programs aimed at reducing delay of thrombolysis in patients with AMI are likely to have a limited impact on AMI fatality.” This statement indicates that they are unaware of the fact that implementation of prehospital thrombolysis instead of hospital thrombolysis is reported to result in 15–22 extra lives saved per 1,000 treated (3;11;17). In comparison, when introducing treatment with thrombolysis instead of placebo in the 1980s, this strategy resulted in 10 extra lives saved per 1,000 treated (5).

§6. Kildemoes and Kristiansen ignore that primary percutaneous coronary intervention (primary PCI) is accepted as the preferred reperfusion strategy in most regions in Denmark after the publication of the DANAMI-2 trial (1). A recent study has documented that a prehospital diagnostic strategy results in an 81-minute earlier initiation of primary PCI if combined with direct referral of patients to an interventional center (13). If combining these findings with estimates of the beneficial effect of primary PCI according to treatment delay (4), and with current mortality rates among patients treated with reperfusion therapy (12), then a widespread implementation of a prehospital diagnostic strategy with referral of patients directly to an interventional center would result in 14–20 extra lives saved per 1,000 treated. Assuming that 75% (n=3,000) of patients are eligible for prehospital diagnosis, then 42–60 extra lives would be saved per year in Denmark.

Kildemoes and Kristiansen also ignore the significance of earlier imitation of reperfusion therapy on comorbidity. In addition to the lives saved by earlier initiation of therapy, one would also expect an improved outcome among the remaining patients, that is, improved left ventricular function, less need of medication, fewer readmissions with heart failure, fewer reinfarctions, and so on.

In summary, the number of lives saved by earlier initiation of thrombolysis or primary PCI may be seven to nine times higher than reported by Kildemoes and Kristiansen, and the costs when implementing a strategy of prehospital diagnosis with the use of telemedicine may be less than two thirds of that reported by Kildemoes and Kristiansen. Thus, Kildemoes and Kristiansen may overestimate the cost per life year gained after implementation of a prehospital diagnostic strategy by at least a factor of eleven to fourteen. There is no substantial evidence indicating that the cost-effectiveness analysis by Kildemoes and Kristiansen reflects real-life. With this viewpoint, it is hoped that we have corrected some of the erroneous assumptions made by Kildemoes and Kristiansen. We hope that any pessimism caused by their study will be replaced by optimism and inspire the readers to establish a prehospital diagnostic strategy to reduce treatment delay in patients with AMI. Reappraisal of a prehospital diagnostic strategy may, in fact, be one of the new successes in modern cardiology resulting in improved patient outcome. Hopefully, proper cost-effectiveness analyses will address this issue in the future.