The effect of computerized decision support systems on cardiovascular risk factors: a systematic review and meta-analysis

The fast paced nature of medical science and practice challenge physicians to keep practicing concurrent to guidelines and the latest evidence. Some state that health care decision making has never been more complex because of multi morbidity and different severities of disease clustered in one individual [1]. Cardiovascular risk management (CVRM) in high-risk patients asks for a comprehensive approach and a lifelong effort of patients that affects lifestyle and dictates adherence to medical treatment of risk factors. CVRM is complex, because it involves a large number of risk factors that may change over time. CVRM guidelines provide support and advocate the use of risk prediction algorithms for the identification of patients at risk for (recurrent) cardiovascular events [2]. Treatment decisions, such as starting or intensifying lipid blood pressure lowering treatment are based on estimated absolute cardiovascular risk in individual patients and their absolute levels of risk factors. Yet, there seems to be a gap between guideline recommendations and daily clinical practice [3]. Adherence to guidelines varies between medical disciplines and between treating physicians, even for similar patients [3]. Completeness of risk factor assessment, pharmacological and non-pharmacological treatment initiation and long term uptake of treatment in patients with a cardiovascular condition can be further optimized, which potentially leads to reduction in to preventable cardiovascular morbidity and mortality [4, 5].

Computerized decision support systems (CDSS) are digital information systems that typically show a summary of patient data in an app, on a webpage, or within the electronic health record (EHR). In CVRM, CDSS can be used for reminders for assessment of risk factor levels, comprehensive presentation and evaluation of risk factor levels and cardiovascular risk estimates and for recommendation of evidence-based treatment modalities. This way, patient data and scientific evidence are incorporated into tailored strategies in daily practice [6]. CDSS have the potential to improve shared decision making, treatment adherence and eventually health outcome, without additional utilization of health-care resources [7].

Multiple apps and decision support systems have been and are being developed: over the last years, at least 16 systematic and 2 narrative reviews on the effectiveness of CDSS on practitioner performance and patient outcomes have been published [8‐25]. But due to the large variation in CDSS functionalities – drug alerts, laboratory test ordering, treatment advice – clinical applications, and patient populations these reviews are restricted in terms of generalizability and applicability. Cardiovascular disease prevention is a multidisciplinary process in a population characterized by multimorbidity and multiple diseases of different severities. Preferably, a CDSS focused on CVRM should fit all these patients. Recently, Njie et al. reported a systematic review on CDSS in cardiovascular risk management, focusing on improvement in guideline adherence by physicians [26]. Complete assessment of vascular risk factors and guideline-adherent clinical testing increased) [26]. It would be even more clinically relevant to know whether use of CDSS would improve cardiovascular risk factor levels and ultimately reduce cardiovascular event rates. This is yet to be investigated.

This systematic review of randomized clinical trials evaluates the effects of CDSS on CVRM and aims to identify CDSS characteristics that are related to effective CVRM.

This review was performed in accordance with the Preferred Reporting Items for Systematic Reviews (PRISMA) guidelines [27].

The systematic literature search (Additional file 1) yielded 14,335 unique articles. Figure 1 shows the number of articles that were retrieved from the searches, were reviewed and were included in the analysis. After selection based on title and abstract, 53 articles were considered potentially eligible for answering the research question. Full text screening followed. Three studies were excluded because of the study design: one study performed a retrospective cross sectional study [33], two studies performed a before and after study [34, 35]. Dorr et al. did not focus solely on CVRM interventions [36] and was excluded. Four studies featured interventions focusing on influencing patient behaviour [37], learning strategies [38], data search queries [39] or prediction models [40] that were non-compatible with our intervention of choice, and were excluded. Three articles focused on patient and physician satisfaction and/or CDSS uptake rather than measurable clinical outcomes and were excluded [41‐43]. Ten studies did not answer our research question [44‐52]. Three studies featured an out-of-date system, three focused on the implementation of an EHR [53‐59]. Lipton et al. investigated glucose control at the intensive care unit, which was considered a too different population [60]. Also a study on a continuous monitoring system was excluded from further assessments [61]. That left us with 25 studies that were critically appraised.

A systematic review and meta-analysis was conducted to provide insight into the effects of computerized decision support systems (CDSS) on cardiovascular risk factor levels and identify characteristics of CDSS related to improved care. A considerable number of CDSS for CVRM were developed, but a clear clinical benefit is absent. Some features of CDSS seem more promising than others. However, the variability in approaches of CDSS and heterogeneity of the results limit stronger conclusions.

Due to differences in technical basis, content and comparison group, comparability of the studies included in this review is limited. CDSS vary greatly in technical approaches, as well as to how and to which extent they support the physician. In addition, or as a result, there was large heterogeneity in the results of the studies, such that definite conclusions cannot be drawn from the pooling of findings. Moreover, the usual care group was ill defined such that it was impossible to understand what care was delivered as standard. Visual inspection of funnel plots investigating the relationship between effect and sample size does not show a distinct publication bias (Additional file 3). But it is still possible that positive results on CDSS are more likely to get published. These aspects limit strong recommendations on success factors in use of CDSS. Also, our analyses were restricted by what was reported: some studies reported raw means, prevalence or odds ratios, others reported adjusted and/ or imputed effect measures [72, 76, 78, 84]. We were unable to incorporate these differences into our analyses. An easy, but major step forward would be to apply the existing guidelines for reporting trial results using drug or devices in the publications and registration in national or international trial registries.

Although very relevant in CVRM, effects on lifestyle factors such as smoking and physical inactivity were outside the scope of this review. CDSS focus on guideline adherent management measured by change in pharmacological treatment and risk factor profiles without registering meta-information on the decisions: how did the CDSS affect counselling by the physician and the shared decision making process? Helping the physician remember to address smoking can be achieved by CDSS, but hów to address this has proven to be an important factor in the uptake [92]. Lifestyle interventions in particular need a more personal approach that is directed towards coaching and long-term engagement, rather than incidental support of guideline adherent management (the goal of CDSS). Current literature is very positive about such smoking cessation interventions: there is consistent evidence that web-based and mHealth smoking cessation interventions may increase abstinence moderately [92].

Though the results are conflicting and strength of conclusions is limited, the findings from this study can perhaps be used to give direction to future developments. More evidence is needed on long-term evaluations including assessment of the effect on treatment adherence and vascular event rate [93]. This could the also include other relevant cardiovascular risks such as anti-coagulation therapy for patients with atrial fibrillation to prevent stroke, cardio protective medicine in heart failure patients to prevent death or cardiac function replacement therapy, and fluid balance in patients witch chronic kidney disease to prevent renal replacement therapy. The paucity of good quality studies, with sufficient sample size and follow up, on clinical outcomes hinders interpretation and restricts transposing these results into clinical practice. This again emphasizes that guidance for generating relevant evidence needs to be followed and taken up as requirement for funding and publication of novel developments [94]. In designing a study investigating the effect of CDSS, the CDSS’s life cycle should be taken into account including the type of device, the intended use and users, and working mechanism. The device should be investigated in its natural habitat: applied to the same target individuals and in the same setting as in usual care. The complex interplay between device performance, user skills and learning curves has to be incorporated methodologically [95]. The regulatory environment, including guidance on scientific evaluation, regulations and legislation and privacy issues needs to evolve together with this emerging field of health technology.

CDSS in principle enable the physician to integrate evidence and patient information into tailored strategies for daily practice and increase guideline adherence [6, 26]. In this review, it seems that the technical basis, prompting and the type of information provided influence the effect of the CDSS on cardiovascular risk factor improvement. Also, patient involvement in the CDSS process seems to increase the effects. Roumie et al. added patient education in a third study arm: 59.5% of the patients in this group were on target regarding their blood pressure (compared to usual care: RR 1.31 (95%CI 1.06; 1.62)). This significant improvement could be driven by increased patient empowerment. Patient empowerment can be defined as a personal disposition (patient’s control over medical strategies) and as a relational concept (collaborative patient-doctor relation) and has been shown to significantly increase the compliance to therapy [96‐98]. The Lancet Commission of Hypertension also suggests patient empowerment as one of the strategies to address the global burden of hypertension in the future [99].

In conclusion, we did not find a clear clinical benefit from CDSS in cardiovascular risk factor levels and target attainment. Some features of CDSS seem more promising than others. However, the variability in CDSS characteristics and heterogeneity of the results – emphasizing the immaturity of this research area - limit stronger conclusions. Clinical relevance of CDSS in CVRM might additionally be sought in the improvement of shared decision making and patient empowerment.