The validity of elderly persons' self-reported drug use and reasons for use may be questioned. However, asking which drugs were taken the previous day (point prevalence) probably minimized recall bias . Asking which drugs were actually taken also bypassed the problem of non-compliance, as we know that around 25% of prescribed drugs may not be used . Data quality in our study was considered good, as drug names were generally correctly written and only few questionnaires had to be rejected because of inconsistencies. The high response rate (77%) also contributes to strengthen the validity of the results. We may assume, however, that the 23% who did not participate, would have been found to use more drugs than the respondents, as people with poor health, poor social status and unhealthy lifestyle are known to be generally over-represented among non-respondents in health surveys .
75% of those who had taken any drug gave an explicit reason for using it. Probably, the remaining quarter did not know or was not sure about the indications (low dose ASA, beta-blockers and estrogens were the main drug groups taken by those who gave no reason for taking them). Our figures, however, correspond well with the results in a comprehensive British interview study among people aged 65 +, where 76% knew the reason for taking their medications .
In our study, 28% (men: 33%, women: 25%) reported no use of ATC-drugs on the day previous to the survey. Direct comparison with other studies is complicated by differences in data collection, population sample and kind of drugs included. The previously cited British interview study, gave results almost identical with ours, as roughly one third had taken no drugs during the last 24 hours, and mean number of drugs for those on treatment was 2.8 . In a Swedish study among 70-year old individuals performed in 2000, only one in five men and one in ten women reported to use no drugs at all . Average number of drugs among those on treatment in Sweden was 3.3 in men and 4.0 in women, which is substantially higher than our figures. This is, however, not surprising, as the overall drug consumption in Norway (measured in daily doses per inhabitants) only reached 60% of the Swedish level in 2000 . A corresponding study carried out by the Swedish authors back in 1972, showed that 40% of men and 24% of women did not use any drug. The proportion who took four or more drugs increased from 21% (1972) to 28% (2000) in men and from 30% to 41% in women .
A Danish study among 75-year old persons living in the community revealed that only 3% of the subjects did not take any drug, and that the average total number of drugs – prescribed and OTC – was as high as 5.4 . The higher level was probably partly caused by examination of drug storages in patients' homes and thereby including drugs that were in fact not used, as total drug consumption in Denmark is slightly lower compared to Norway (close to 90%) . In the Danish study, CNS drugs were the most commonly used category, compared to cardiovascular drugs in our and other studies [1, 2, 16, 19]. 26% of the elderly Danes had taken a benzodiazepine the previous day, reflecting the relatively high use of benzodiazepines in Denmark in the late 1990s . In our study, only 5% reported having taken a benzodiazepine, corresponding to the average level of use in the population in Hordaland County . Other studies have shown that use of benzodiazepines increases by age , and we can only speculate if our results are influenced by selection bias or possible underreporting.
The lower use of drugs in our study compared to neighbouring countries probably reflects habits and preferences in the population as well as among physicians. In a population traditionally founded on fishing and agriculture like on the Western coast of Norway, it is often considered a benefit to be independent of regular medication, and to cope with occasional illnesses without drugs. Norwegian family physicians may also be more cautious than colleagues elsewhere in prescribing regular medications for elderly people, especially drugs with potential CNS side-effects. On the other hand, we may simply lag behind our neighbours in an ongoing process towards increased drug use. As the Swedish study shows, drug use among 70 year old persons increased remarkably during three decades, despite generally improved health status . However, over that period, a range of new drugs have been introduced, especially for the treatment of cardiovascular risk factors [22, 23]. Recently, focus has switched from concerns of unnecessary treatment to possible under-prescribing of potentially beneficial therapies to seniors . Low dose aspirin, beta-blockers and lipid lowering agents were already the dominant drug groups in our sample (13% used a statin). The sales of lipid reducing agents have doubled in Norway from 1999 to 2004 , and recently, 13% of presumed healthy, low-risk 75-year old women in Oslo were found to use statins .
In our study, 31.7% had used at least three drugs, while 11.5% had taken five or more daily drugs. This is less compared to the Swedish  and the Danish  studies cited above, but equals the results of the British interview study . Based on data from a Danish prescription database, Bjerrum found that on a random day, close to 5% of 70–74 year old persons used five or more prescribed drugs . He also found a six-fold variation between general practices regarding the prevalence of major polypharmacy (5+ drugs), correlating – among other things – with the doctors' workload, as increasing workload implied increased prescription.
Potentially influencing factors
Many of the factors which correlated positively with drug use in our study Tables (4 and 5) have also been identified by others: female sex, a relatively low education (especially for respiratory and cardiovascular drugs)  depressed mood, and a poor self-reported general health . Not surprisingly, we found (in bivariate analyses) that reporting poor general health, feeling depressed, and performing no regular exercise were all strongly correlated with drug use, both with the use of any drug, with an increasing number of drugs used, as well as with the concomitant use of five or more different drugs.
It may be surprising that being a non-smoker and being abstinent from alcohol both correlated with increased use of drugs. However, having quit smoking strongly correlated to increased drug use. In the group who had smoked more than 10 cigarettes a day for 30 years or more, and who now had quit smoking, (n = 237), one out of six used five or more drugs. That former smokers were most likely to be on medication, corresponds to what has been found in other studies [25, 26]. But how can we explain that heavy present smokers (more than 10 cigarettes a day for 30 years or more, n = 114) in fact used less drugs than the average? One explanation may be that people who had fallen ill and were put on regular medication, quitted smoking, while those who remained healthy in spite of their smoking habit, were more likely to continue smoking. Another probable explanation may be that a substantial proportion of heavy smokers died before reaching the age of 70–74 years .
A corresponding relationship may also be valid for alcohol use, but because we did not ask about quitting drinking, we could not explore this. The negative correlation between use of alcohol and medications was nevertheless weaker than for smoking. Even regular drinkers (> 10 drinking days a month or > 20 glasses of alcohol a month) used (non-significantly) less drugs compared to the rest of the sample. This corresponds with epidemiological data supporting that regular moderate alcohol consumption is associated with less morbidity and mortality . An American study found no significant relationship between current alcohol use and current medication use among persons 65 and over . However, in our study under-reporting of actual drinking and a selection effect implying that heavy smokers and drinkers with poor health may be overrepresented among the non-responders can not be ruled out.