There has been a substantial increase in published research evaluating interventions targeting medication use in managed care settings since our previous review . This review identified 151 intervention studies in a six-year publication period using a narrower inclusion criteria (that is, disease management studies only included where pharmacotherapy was a primary focus) compared with 105 studies over 35 years in our previous review (70 of which were reported since 1996). Despite the increase in research on this topic, a smaller proportion of published studies met our criteria for methodological adequacy (31% versus 46%). This failure to increase the degree of methodological rigor in published research is of concern because significant resources are used conducting studies of poor methodological quality and such research may produce scientifically invalid conclusions that influence future research and policy. Although not a focus of this review, we noted additional methodological issues apart from research design. For example, the proxy measures of adherence varied between studies (n = 10). Three studies used the medication possession ratios [20, 27, 59], while others measured the proportion of patients continuing treatment [33, 38, 42, 43, 67, 68, 70]. The variability in measures of adherence and infrequent reporting of this important outcome in part reflects the well-known methodological challenges in the measurement of adherence .
Several findings from this report are consistent with our previous review and the results of other systematic reviews on the impact of interventions designed to change medication use. We once again confirmed that dissemination of educational materials alone is ineffective and one-to-one educational outreach visits are effective in increasing adherence to prescribing guidelines [6, 9, 10, 15, 78]. Further, this review confirms that multifaceted interventions are more likely to be successful in changing medication use than those using single strategies [6, 10, 78].
The current analysis, like other reviews, suggests group education interventions may be beneficial but the findings are inconclusive [78, 79]. We found that peer leader group education was effective in changing medication use but was less effective than one-to-one academic detailing [16, 31, 32]. The absence of incremental impact of group education when combined with computer-based alerts, as reported by Simon et al  and Feldstein et al , may reflect the weakness of the approach or the superior effectiveness of computerized alerts in changing the particular outcomes targeted in these interventions.
Intervention approaches involving monitoring and feedback have been shown previously to produce small to moderate effects on medication use [6, 23, 80], and this review confirms these findings. Further, our review also supports the existing literature regarding the short-term effectiveness of real-time computer-based alerts in changing prescribing and test ordering [28, 81]. However, most studies in our review evaluated electronic medical record systems with particular characteristics in a single HMO, thus limiting the generalizability of findings. Only one study  reported patient-relevant outcomes (namely, quality of life measures).
Our review provides evidence that tiered formulary and patient copayment interventions decrease non-preferred drug use, reduce overall insurer costs, and increase patient out-of-pocket expenses as intended. The use of prior authorization policies resulted in similar effects. However, our findings concur with previous reviews in managed care and other settings that these interventions may also be associated with undesirable effects such as increased rates of switching to other medications or discontinuation of essential and cost-effective medications [3, 48, 82, 83]. In addition, some studies reported changes in measures such as hospitalization or medical resource utilization (e.g. office visits) resulting from formulary-related interventions [44, 45, 49, 50, 70].
Evidence from the literature suggests that coordinating pharmacist services as a component of patient care improves quality of care [84–86]. This review documents positive results for collaborative care led by pharmacists, including improved drug adherence and clinical outcomes. Only some data on costs and resource use (e.g. cost per mmHg of blood pressure decrease, average provider visit cost per patient) have been reported [59, 60, 63].
Finally, our findings confirm that disease management interventions are associated with improvements in the process and quality of care, both short and longer term . However, these disease management interventions are by their nature multifaceted, and the type and number of features within each disease management program vary widely. Therefore, it is difficult to draw firm conclusions about the contribution of individual components to the overall impact.
There are several limitations to this review which need to be acknowledged. Despite our intensive efforts to identify all of the published literature on interventions to improve the quality and efficiency of medication use in US managed care settings, our search strategy may have missed published articles. The full range of interventions currently used by managed care programs to influence use of medicines is not covered by the studies in our review. For example, no studies evaluated the impact of physician incentives on prescribing, despite that fact that this strategy is commonly used by MCOs. Further, our findings are likely to be subject to under-reporting bias with interventions showing no or negative effects, which are less likely to be reported in the peer-reviewed literature. Even when studies are published, many are methodologically inadequate; 69% of the studies identified in this review were not included in the detailed analyses because of design flaws. There are also fundamental difficulties in comparing interventions with diverse objectives, intervention targets, measurement methods, and outcomes. Finally, it is difficult to reach definitive conclusions about the contribution of individual components to the overall effect of multifaceted interventions.
Despite the substantial number of interventions to improve drug use in managed care, our understanding of the impacts of these interventions is still limited.
First, a large proportion (40%) of the reviewed studies did not detail the specifics of the managed care setting within which the study was conducted. We found few studies conducted in PPOs despite their prominent role in the US managed care industry or in other "lightly" managed care settings. The majority of studies in our review were conducted in staff, group, or mixed-model HMOs, which represent more heavily managed settings. Future research should establish whether intervention approaches that are successful in such settings are also effective in more lightly managed settings. Second, most studies in this review used common surrogate outcome measures, such as prescribing rates or proportions of patients achieving specific medication use practices. Few studies reported medication adherence or patient-relevant outcomes such as clinical status, hospitalization rates, or quality of life measures. Furthermore, evidence about the cost-effectiveness of most interventions is still quite limited. Cost-effectiveness data are needed to select among many possible strategies for improving medication use and associated costs. Third, there is still a lack of publicly available, high quality evidence concerning the effects of interventions involving formulary changes or financial incentives, which are commonly used to influence medication use. Although the increased reporting of formulary-related interventions is encouraging, the high prevalence of inadequate designs (e.g. pre-post without comparison group, post-only) among studies of formulary-related interventions (77%) is striking. An important conclusion to be drawn from this review is that improvements in research methodology will be essential in order to produce valid and reliable study results about cost, quality of care, and patient outcomes to inform managed care policy decisions. It is highly likely these interventions do reduce costs to the payer as they continue to be dominant in the managed care environment. Evaluation of the patient-relevant outcomes of formulary-related interventions is a research priority. Finally, the durability of most of the intervention effects reported in this review is uncertain because few studies extend beyond two years follow-up and many cover even shorter periods. In particular, further research is needed to determine if computerized clinical support systems are associated with improved patient outcomes over the long-term.