Background
Over the past two decades, a growing international movement to increase patient involvement in healthcare decisions and development of various decision support interventions has taken shape [
1‐
5]. However, many healthcare decisions are challenging [
6], and negative experiences during or after decision-making can lead patients to have regrets about the choices that were made [
7]. Evidence suggests that decision regret is a common phenomenon in healthcare and that it can reach high levels for some medical decisions [
8]. Decision regret is associated with lower satisfaction with care, negative experiences with the healthcare system, and reduced quality of life [
7,
9‐
11]. As such, it is increasingly viewed as an important patient-reported outcome measure as well as a proxy measure for the quality of healthcare decisions [
12,
13].
Assessing decision regret is particularly important in the context of primary care. For many patients, the clinical encounter with their family physician is the first point of contact with the healthcare system. This is where they learn about health issues, have their problems assessed and diagnosed, and consider steps they can take to preserve or improve their health [
14]. Primary care providers offer care across the lifespan, manage all but the most uncommon or unusual conditions, and ensure continuity and coordination of care provided at other levels of the healthcare system or by other professionals [
14]. Primary care thus encompasses the widest possible spectrum of health conditions and is the forum where the greatest number and diversity of medical decisions take place [
15,
16], making it a very relevant clinical context for the study of decision regret.
Yet surprisingly few studies have investigated decision regret in primary care. In a systematic review of 59 studies examining the extent and predictors of decision regret related to healthcare decisions, we identified only five studies conducted in family medicine practices [
8]. Specifically, authors assessed the extent of decision regret related to decisions about hormone replacement therapy [
17], cardiovascular disease prevention [
18], use of antibiotics for acute respiratory infections [
19,
20], and treatment choices for diabetes [
21]. In three of these studies a low mean level of decision regret was observed among participants [
18‐
20], but in two studies (on hormone replacement therapy and diabetes decisions) authors reported relatively high levels of regret [
17,
21]. None of the five studies examined risk factors contributing to decision regret in their primary care patients, nor did they examine whether decision regret varies across the multiple types of healthcare decisions that take place in primary care. Therefore, we sought to estimate the extent of decision regret experienced by primary care patients and to examine the factors associated with regret.
Discussion
Our study results indicate that the overall extent of decision regret experienced by primary care patients following consultations with family physicians or residents in family medicine is low. To the best of our knowledge, this study is the first to examine factors associated with decision regret in primary care settings. We observed a significant association between patient decision regret and patient decisional conflict in bivariate analyses. In multivariate analyses, we found that higher decision regret was strongly associated with clinically significant decisional conflict, and weakly associated with male physicians and their status as residents, while regret was lower among patients between the ages of 40 and 60 years and those with a university education. Our results lead us to make three main observations.
First, the extent of decision regret found in this study is consistent with the literature. Our systematic review identified 44 studies reporting mean scores on the DRS, leading to an overall unweighted mean ± SD score of 16.5 ± 10.9 out of 100 across all studies [
8]. Furthermore, 14 studies reported the proportion of patients who experienced no regret (DRS score = 0), which ranged from 14 to 98 % across studies, with an unweighted mean ± SD of 59.0 % ± 24.6 %. Our results suggest that primary care patients experience similarly low levels of decision regret as do patients in other clinical contexts. A difficulty that remains, however, is how to interpret these results. Several authors have characterized DRS scores ranging from 5 to 25 on 100 as mild decision regret [
10,
29‐
33,
37], yet no consensus exists and the clinical significance of different scores and categorizations remains to be determined. While it is reassuring that most patients report relatively low levels of decision regret in primary care settings where the bulk of healthcare services are delivered, further research is needed to identify valid and reliable cut-off points for distinguishing clinically significant regret and its impact on patient health.
Second, our results are similarly consistent with the literature regarding the influence of decisional conflict on decision regret. A recent meta-analysis of ten clinical trials found a strong positive association between decisional conflict and decision regret (OR = 5.52; 95 % CI 3.35–9.12) [
36,
38] and our own systematic review has shown that decisional conflict was among the predictors most frequently and significantly associated with decision regret [
8]. Decisional conflict typically occurs when patients face difficult decisions for which multiple reasonable options exist, as well as when patients feel unsupported in the decision-making process. Decisions that are made in a context of uncertainty can lead to decision regret, especially when there is no clearly preferable clinical option. Several studies have shown that regret is also a common consequence of preference-sensitive decisions [
9,
39]. Our results thus justify efforts to identify and reduce patients’ decisional conflict in primary care consultations.
One ideal measure for identifying decisional conflict in clinical settings is the SURE (Sure of myself; Understanding information; Risk-benefit ratio; Encouragement) tool [
40]. SURE is a short, four-item clinically oriented screening tool for decisional conflict that has been validated in primary care settings [
34]. Identifying patients experiencing uncertainty about the choices they face and the consequences of these choices may prompt physicians to increase supports for decision-making. These supports may take the form of SDM facilitated by decision aids. High-quality evidence suggests that decision aids reduce patients’ decisional conflict and also increase decision-specific knowledge and overall satisfaction with care [
41]. Yet, despite the evidence for their benefits, decision aids and SDM generally are not widely implemented in clinical practice [
42,
43]. Broader adoption of SDM strategies to address factors contributing to decisional conflict [
44], such as feeling uninformed about options, feeling uncertain about risks and benefits of options, or feeling unsure about values and preferences related to choices or consequences, should help reduce the decision regret that some patients experience [
45].
Third, some of our findings differ from those of previous studies. For instance, our multivariate analyses revealed that patient age and education were associated with decision regret. Yet our systematic review, albeit mostly of studies conducted mostly in non-primary care settings, indicated that these factors rarely predict regret [
8]. We observed higher regret in older patients, which may be a consequence of higher rates of adverse outcomes in these patients or to their preferences for more paternalistic decision-making approaches with physicians [
46]. Higher regret was also observed in patients with lower education levels, possibly due to lower rates of health literacy in these populations [
47,
48]. Low health literacy can limit patients’ capacity to understand basic health information and acts as a barrier to their participation in healthcare decisions [
47,
48]. Interestingly, our study also found that regret was higher when physicians were male and were residents. Why patients of male physicians had higher regret is unclear, as a recent systematic review indicated no clear impact of physician gender on a number of patient-doctor communication characteristics [
49]. But other studies have suggested that female physicians adopt more of a partnership style and a more patient-centered approach than their male colleagues [
50‐
52]. As there is no consensus on the level of decision regret that is clinically significant and the association between physician gender and regret did not affect the highest scores, it is likely that this statistical association was not clinically significant. Residents, however, have been shown to lack some communication and SDM skills [
43], which may explain why their patients had higher decision regret scores. These latter results should be interpreted cautiously, however, as the strength of association with decision regret was variable and associations had confidence intervals with values close to 1. They cannot be interpreted with the same confidence as our findings regarding decisional conflict.
This study has some limitations. First, as this was a secondary analysis of an observational descriptive study we were not able to explore the effect of time on decision regret scores. It could be that levels of decision regret were underestimated because of the short time period (two weeks) between the initial consultation and the assessment of decision regret. Patients may not have had sufficient time to develop feelings of regret about their decisions or experience adverse outcomes. Unfortunately we cannot compare this with the literature because the intervals between decision and regret most frequently used have been one, three, six and 12 months [
8]. Longitudinal and prospective research is needed to examine the emergence of regret and its subsequent trajectory to explore the possible conceptual division of regret into two categories, immediate and delayed [
13]. A potential desirability bias may also have contributed to low regret scores, particularly related to the phone call follow-up questionnaires. Patients may not have wanted to compromise their relationships with their physicians by reporting high levels of regret about their decisions. This bias was not evaluated in the context of decision regret but one study has shown that mail administration of SF-12 scored lower total averages than telephone administration [
53]. In addition, the study was conducted on a convenience sample, so a selection bias is possible especially as we had no data about those individuals who refused to participate in the study and because recruited physicians were members of practice based networks and not randomly selected. As the current state of knowledge does not permit accurate interpretation of the DRS scores regarding the clinical impact of decision regret on physical and mental health, future research is needed to examine this impact and the types of decisions that lead to more regret in patients. Also, we may have overinterpreted our findings due to lack of sufficient information. For example, the difference in decision regret between patients with secondary and lower levels of education and a University education was 8 %, yet its significance was only
p = 0.11. Future research is needed with larger samples to re-evaluate our interpretation. Finally, although we found no impact of the cluster effect on the final results, this information can be used now to calculate a sample size for a future cluster trial in Canadian primary care settings with decision regret as the primary outcome.
Acknowledgements
The authors wish to thank Louisa Blair for editing this manuscript.