Background
Medication side effects also referred to as adverse drug reactions (ADRs) and/or adverse drug events (ADEs) are potentially associated with all medications and include side effects listed in product information, as well as medication-related side effects due to interactions, non-compliance [
1,
2], prescribing errors [
3] and allergic reactions [
4]. Medication side effects are an important cause of morbidity, mortality, costs [
5‐
16], and dissatisfaction with care [
16].
Prevalence of medication side effects depends on the surveillance system and setting [
17]. Surveillance systems vary in case definitions, case finding, and the duration of monitoring [
14]. Mostly, medication side effects have been examined in inpatient hospital settings and to a lesser extent in hospital outpatient or ambulatory clinic settings by chart review and self report. Information on medication side effects in the community is more limited and is mostly provided by national surveillance systems that rely on voluntary reporting and lack denominator data.
Similarly, potential risk factors for medication side effects have been examined mostly in hospital inpatient or outpatient or ambulatory clinic settings. The most consistently identified risk factors include older age, female gender, increasing number of medications, and increasing number of medical conditions/co-morbidities as well as specific co-morbidities [
13,
14,
16,
18]. Other risk factors include those related to compliance, comprehension and general patient knowledge, types of medications (in hospital settings such medications include antibiotics, anti-coagulants, digoxin, diuretics, hypoglycaemic, NSAIDs, chemotherapeutics, other cardiovascular medications and analgesics), hospitalisations, medical knowledge, multiple treating doctors, and multiple filling pharmacies [
18‐
20].
Data from the 1994 Melbourne Longitudinal Study on Healthy Ageing (MELSHA) baseline cohort, offers a unique opportunity to examine prevalence and risk factors for self-reported medication side effects in an older cohort aged 65 years and over living independently in the community.
Discussion
The MELSHA study provided a unique opportunity to examine the prevalence and possible risk factors for self-reported medication side effects in a representative older cohort living independently in the community. The prevalence of self-reported medication side effects in our community based study was comparable with that seen in other studies in the primary care and community setting [
2,
3,
24‐
28]. Our findings were consistent with prevalence estimates from the more recent BEACH reports (from General Practice based in a primary care community setting) 9.3% in 2007-2008 and 10.4% in 2003-2004 [
27,
28]. However, there was a marked disparity with prevalence estimates from hospital based studies through inpatients, outpatients and ambulatory or emergency departments of adults ranging in age from 18 to 75 years [
8,
16,
29]. This disparity may reflect differences in the settings including selection, general health, use of medications and severity of medication side effects, i.e. those hospitalised or attending hospital are different to those living independently in the community attending General Practice. In addition, the disparity may also reflect differences in surveillance practices such as definitions and methods for detectionincluding measurement validity and sensitivity of exposures and outcomes.
The association with increased education of participants may reflect greater awareness and recognition of medication side effects through general knowledge, discussion with their doctor or pharmacist, and reading the product information. Recency of visiting the doctor within a month was protective and having last received advice from their pharmacist more than 5 months ago in contrast to less than 5 months ago was associated with an increased the risk of reporting medication side effects, lending support for the role of education in facilitating advice provided by doctors and pharmacists. In addition, it was suggestive that advice on medications may be forgotten over time by an older cohort, so the importance of regular monitoring and reinforcement by their doctors and pharmacists. The important educative role of doctors and pharmacists is further emphasised by another study that found starting a new medication, cessation of a medication or changes to prescribed and over the counter (OTC) medications were associated with an increased risk of medication side effects [
14]. Therefore, doctors in prescribing, and pharmacists in dispensing, have an important role in detection and education.
The finding that an increased risk of medication side effects was associated with those with a younger treating doctor was more difficult to interpret as it was based on participants' assessment and report of their doctor's age and was associated with a high proportion of missing data. It may be that doctors who were assessed as younger tended to have less experience and make more therapeutic mistakes including dosing and interactions as has been suggested by another study [
5]. Several studies have found that a significant proportion of medication side effects are preventable or ameliorable and include inadequate medical knowledge [
13], flawed prescribing habits including the issuing of inappropriate scripts [
13,
30,
31] and failure to monitor and review [
13,
30,
31]. Alternately, younger doctors may be more likely to use newer medications for which side effects are not fully known or they may be more aware of medication side effects and/or better educate and raise their patient's awareness.
The association of medication side effects with their report of their doctor's awareness of their medications may reflect the use of prescription medication, recency of commencement of treatment, or use and monitoring of medications with low therapeutic/toxicity thresholds. Another possible explanation is that the prescribing doctor is aware, but other doctors involved in the patient's care are not, hence providing an opportunity for errors and interactions. Alternately, it may be that those who report their doctor's awareness of their medications are being monitored more frequently with greater opportunity to detect medication side effects.
The association of medication side effects with diabetes may reflect specific disease related issues around management including patient self management and collaboration of treating health professionals as well as therapeutic safety and/or greater opportunity for detection. In contrast for hypertension and osteoarthritis, there was a protective association with medication side effects. The difference in the findings for diabetes, hypertension and osteoarthritis, all chronic conditions likely to require regular follow up and review and hence offering the opportunity for detection, suggests that the opportunity for detection is an unlikely explanation. It may be that the difference in their findings reflects greater therapeutic safety of medications used in hypertension and osteoarthritis; more experience in their use by their usual doctor as management of hypertension and osteoarthritis is primarily managed by GPs in the primary care setting, whereas diabetes care may be more fragmented with management by specialists, GPs and other health professionals; or differences in patterns of associated co-morbidities and other factors not measured.
Self-reported poor to fair health was associated with an increased risk of medication side effects but it did not persist in the multivariate analysis when specific medical conditions were included. This association in the univariate analysis may have reflected significant medical conditions or alternately a focus on health including medication side effects. An increased risk of medication side effects was associated with particular medical conditions that may mimick medication side effects such as skin and liver problems; reflect changes to medication metabolism such as with liver problems especially in older patients; or use of medications with low therapeutic/toxicity thresholds [
4,
13,
18]. The number with liver problems was too small to have sufficient power in the reported multivariate analysis. Cataracts were associated with an increased risk of medication side effects, possibly reflecting ability to read labels and instructions and comply. However, use of glasses or contacts was not associated with an increased risk, perhaps reflecting the use of a strategy to cope with restricted vision and hence ability to read labels. Other studies have found that medical conditions and patient related factors including impaired eye sight such as cataracts, hearing and cognition may impinge on compliance [
13,
32].
A lack of association with age and gender was consistent with the prevalence findings previously discussed; with a study that reported minimal changes in pharmaco-kinetics, pharmaco-dynamics and interactions with age [
32]; and studies that have found that increasing age may be associated with increased numbers of medications [
18,
30,
33,
34] and multiple co-morbidities [
17,
18,
33,
34] which are independent risk factors. Those studies that have found an association between age and medication side effects [
6,
30,
33,
34], have been in young to middle aged patient groups [
29] perhaps suggestive of an age threshold. As genetics may account for 20-95% variability in medication side effects [
35], those who are long-lived may have favourable/protective genetics. Likewise, the lack of association with gender was consistent with those of another study in an older cohort with a mean age of 81.4 years in which no gender association was observed [
34]. Again, those studies that have found an association with female gender [
31,
34,
36] have been in young to middle aged patient cohorts [
29].
There were several potential limitations of the study findings. The MELSHA data were collected in 1994 and may not reflect the current situation. However, the prevalence of self-reported medication side effects in the MELSHA cohort in the community in 1994 was/is similar to more recent and contemporary prevalence estimates of medication side effects in the BEACH reports. We have not focused on specific medications for which indications and use may have changed. Our analysis focused on independent stable risk factors including socio-demographics, health status and medical conditions, and health service factors. For example the measurement of age or classification of gender or marital status or medical conditions or definitions and scaling of self reported health status would be consistent now with 1994. The examined risk factors predicted a significant percentage (45%) of the risk of self-reported medication side effects.
The reliance on recall and self report may have potentially introduced systematic error or bias affecting the internal validity of the study findings as well as random error affecting the reliability and precision of the study findings. The internal validity of the study may have been affected by information bias due to the reliance on recall and self report. Such a possibility can not be definitively excluded. However, previous studies have found good levels of agreement between recall and self report of medication usage with pharmaceutical claims data as well as home visit verification [
37,
38]. It is possible that the prevalence of medication side effects may have been underestimated as only more recent or severe or persistent medication side effects may be recalled. However, our findings with respect to the prevalence of medication side effects appeared to be consistent with those from other studies including the more contemporary studies in the ongoing BEACH report as discussed previously [
27,
28]. It is also possible that those with medication side effects may have reflected more on possible associated risk factors. However, socio-demographic factors would not be affected and those of medical conditions and medications were verified independently by the interviewer. The survey was administered by a trained interviewer and so there was the opportunity for clarification and probing. In addition, the interviewer provided a standard written list of medical conditions and checked medications, providing some additional verification to self-report alone with respect to medical conditions and medications. Therefore, although recall bias is a possibility, it would seem an unlikely explanation for our findings. In addition, the effect of potential bias from potential confounders has been adjusted for in the multivariate analysis. However, it is not possible to adjust for unknown confounders or confounders on which data were not collected.
It has been suggested that self report of medication side effects may be less accurate and less reliable than chart review that is possible in clinical settings [
32,
34]. A lack of reliability in the use of self-report, may result in lack of precision and loss of power to detect an effect/association. However, a study of older Australians' medication use that examined self report by phone compared with a home visit and inventory found self report was accurate (high agreement as measured by kappa) for all prescribed medication categories [
37].
Other potential issues were missing data and the small numbers with medication side effects. Missing data may have reduced the power to detect an association or introduced potential bias. However, sensitivity analyses around missing data found a similar pattern of results. The small numbers with medication side effects may have meant that the study had insufficient power to examine risk factors of smaller magnitude of effect or rarer/less common risk factors, but we have only commented on those associations detected by this study. The small numbers also meant that some categories had to be collapsed which may have further affected the findings, although this was limited by an apriori approach based on similarity and logic.
Overall, given the random sampling and selection of the MELSHA cohort, the findings are likely to be generalisable to older people living independently in the community. It is important to reflect on strategies to address some of the potential risk factors identified. From a review and framework proposed by the National Prescribing Service (NPS) [
39] in conjunction with our findings, key approaches would include: targeting specific diseases identified by our study; addressing the awareness and knowledge of health professionals in particular doctors and pharmacists as well as patients; and communication between health professionals and between health professionals and patients [
39].
In targeting specific diseases diabetes would be a useful starting point, and other diseases that may be considered include osteoporosis, heart problems such as arrhythmias and cardiac failure, emotional or psychiatric conditions and other arthritis. In addition to existing strategies for improved diabetes management, it may be that the management of hypertension and osteoarthritis may provide insights into disease focused management strategies to minimize medication side effects. As well, the need for additional care in using medications in those with liver problems needs emphasising.
Greater awareness and knowledge of health professionals including doctors and pharmacists as well as patients is also important. Further education during medical training and scaffolding in the workplace may support doctors in developing safe rational prescribing practices that minimize medication side effects. Specific education strategies for health professionals in particular doctors and pharmacists may include integration into undergraduate curriculae as well as post graduate/professional ongoing continuing medical education programs; supportive systems such as provision of information through professional colleges, support software, and emphasis on continuity of care over fragmentation; and quality control through participation in audits. In particular areas needing highlighting include the need for increased awareness and follow up and review in relation to the timing of medication side effects such as when introducing new medications and in particular for newly licensed medications as well as over the counter medications; similarly with the cessation and changing of medications; increased awareness of the range of presentations of medication side effects and the need to consider as a differential diagnosis; importance of participation in national surveillance and reporting systems for adverse drug reactions to increase the shared information available and the ability to detect a signal earlier. Improved health education of patients may enable early recognition and prevention of potential medication side effects.
Communication between health professionals including between GPs and specialist as well as pharmacists and other health professionals and with patients may facilitate management as well as surveillance and detection. Given the pivotal role of treating doctors and pharmacists in providing patient education and monitoring/surveillance of medication side effects, factors such as continuity or systems for information sharing may facilitate this role for doctors and pharmacists as well as other health professionals. As suggested by the NPS e-health may be one way in which such communication and information sharing may occur.
Finally in recognising and better managing medication side effects, it is important that medication side effects are not simply attributed to increased age. This requires increased awareness and consideration of other potential risk factors in older people, to better prevent and manage medication side effects. Prevention may require greater use of strategies by patients and health professionals to address potential sensory, physical and cognitive deficits with aging that may be associated with compliance problems: ongoing supervision and use of dosette boxes; regular medication reviews; provision of readily understood written instructions on medications including how to take and potential side effects in large readily read print; establishment of hotlines for patients to obtain additional information this would offer the potential to provide for those with hearing or language difficulties.
Competing interests
All authors declare that there are no competing interests and therefore have nothing to declare.
The Corresponding Author has the right to grant on behalf of all authors and does grant on behalf of all authors, an exclusive licence.
Authors' contributions
JAT contributed to the conceptualization and design of this specific sub-study, data extraction and analysis and interpretation and drafting the manuscript. WCW contributed to data extraction and analysis and commented on the manuscript. CB was and is the joint leader of the MELSHA project and participated in its design and co-ordination and funding acquisition, provided data and commented on the manuscript. HLK was and is the joint leader of the MELSHA project and participated in its design and co-ordination and funding acquisition, provided data and commented on the manuscript. All authors have read and approved the final manuscript.