The levels of use of antithrombotic drugs, beta-blockers and statins by Finnish patients with established CHD were largely appropriate, but some interesting subgroup differences emerged. The levels of drug use in the present study are similar to earlier findings in Finland , although somewhat lower than those reported in EUROASPIRE II and somewhat higher than in the study of home-dwelling elderly in Helsinki. There was also an especially markable difference in the use of statins compared to the study by Strandberg and colleagues . These differences are probably due to differences in the study populations and time periods. Our analyses were based on a randomly sampled cohort of coronary patients whereas EUROASPIRE II analysed hospital patient (MI, coronary revascularisation, or hospital care due to myocardial ischaemia) populations. However, it should be noted that patients in our study had established disease according to at least some objective criteria; those with symptoms but without objective evidence of CHD were not included. The population studied by Strandberg and colleagues  was much older than ours. Moreover, coronary patients first qualified for extra reimbursement for statins in 2000, after the Strandberg study.
Drug use decreased with age among men, but the reverse was found among women. Whether these findings indicate less than appropriate health care among older men and excess care among older women is difficult to interpret. The former interpretation may imply the need for more efficient care for elderly men. In line with earlier results [e.g. [12, 18]] the likelihood of drug use tended to grow with increased disease severity, although the differences were statistically significant in only a few subgroups. Thus patients with past MI and history of coronary revascularisation, and women with exertional chest pain were more likely to use drugs. Patients with known diabetes also used somewhat more drugs than non-diabetic patients, but the differences were not statistically significant. The fate of coronary patients with diabetes is known to be more serious than that of patients without diabetes . If the present results indicate that Finnish doctors do not consider that diabetes justifies more effective drug treatment of coronary patients, then more physician education is needed.
There were no marked disparities in use of antithrombotic drugs or beta-blockers by socioeconomic indicators. However, after controlling for age, disease history and severity, and comorbity, socioeconomic differences persisted among men in statin use. There could be several reasons for this: whereas aspirin and beta-blockers are relatively inexpensive drugs, statins – even after an increased reimbursement rate – remain more costly for the individual patient, which may explain the lower statin use among less affluent groups. Moreover, the tradition of using antithrombotic drugs and beta-blockers among coronary patients is much longer than for statins. On the other hand, earlier studies both in Finland and elsewhere have reported more effective treatment of CHD among higher socioeconomic groups, including thrombolytic treatment after MI  and revascularisation operations after MI [15, 20, 21] and among coronary patients in general [22, 23]. As our analysis was based on questionnaire information, no laboratory examination data were available on possible differences in serum cholesterol levels between socioeconomic groups. However, we were unable to find any Finnish studies reporting higher serum cholesterol levels among more affluent men. Instead, Finnish research has consistently reported higher serum cholesterol levels among lower socioeconomic groups .
Our results nevertheless fail to clarify whether socioeconomic differences in statin use are due to physicians' uneven prescription practices or variation in patients' compliance with drug treatment. Several Finnish studies have reported healthier behaviours among higher socioeconomic groups both in terms of dietary fat use and leisure time physical exercise .
Earlier studies on hormone replacement therapy and antihypertensive medication have suggested area variation in drug use [26, 27]. Area differences especially in secondary prevention practices could have an effect on socioeconomic differences found in our CHD cohort. We carried out additional analysis (random effects model, SAS 8e Mixed procedure) to control the potential effect of regional differences between 20 hospital districts in our results. However, the analysis showed no area effect.
A strength of our study is that it was based on a random sample from a cohort of coronary patients and not merely hospitalised patients. Additionally, the specific criteria used for defining CHD and the approval procedure for special entitlements would have minimised false positive cases. There are some factors, however, that might potentially bias our findings on socioeconomic differences in drug use. First, a possible source of bias is multiple comparisons which could have produced false positive results by chance. Since our results were systematic by all the socioeconomic variables, this is unlikely to be the case. Another issue is selective mortality which is likely to have resulted in fewer respondents from lower socioeconomic groups due to higher mortality before hospitalisation and higher case-fatality from first coronary events . Moreover, in our data those with lower level of education were more likely than others to be non-respondents. To correct the potential bias introduced by selective non-response, weights were calculated to our data. However, weighting assumes that the drug use of respondents and non-respondents is similar within these groups, which may not be the case. Due to selective mortality and the distribution of non-responders, it seems obvious that our results give a conservative estimate of socioeconomic differences rather than an overestimate. This conclusion is supported by an additional analysis of non-response using SII drug register data which revealed that non-respondents were less likely to have had medicine costs for statins (OR = 0.66, 95% CI 0.59 to 0.75) and beta-blockers (OR = 0.82, 95% CI 0.71 to 0.94) reimbursed during 2000 than respondents. No differences were found between respondents and non-respondents in reimbursements for nitrates (OR = 0.92, 95% CI 0.81 to 1.04). This selective non-response means that our estimates for prevalence of beta-blocker and statin use are probably somewhat optimistic. A third potential source of bias is that other than CHD related comorbidity – which is likely to be more prevalent among lower socioeconomic groups – was not controlled for. Since statins are not reported to have major interaction effects with other drugs, the possible mechanism could be either the need to control the total amount of drugs used or high total drug costs. The effect of non-CHD related morbidity on statin use remains an important issue for further analysis. A fourth potential source of bias is recall bias in reporting drug use. We were not able to analyse the impact of recall bias in detail because we did not have register information on medicine use in our data. However, we do not expect there to be systematic socioeconomic bias, since earlier studies have reported high reliability and validity in survey self-reports of medicine use .