Background

The rapidly expanding use of health services is challenging and health care expenditures are mounting [1]. This has underscored the need for more efficient use of finite healthcare resources. However, according to the Organization for Economic Co-operation and Development (OECD), approximately 10–34% of health service spending is potentially inappropriate, representing ineffective and wasteful use of health care resources [2]. Such services are referred to as low-value care, which is defined “an intervention in which evidence suggest it confers not or very little benefit for patients, or risk of harm exceeds probable benefit or, more broadly, the added costs of the intervention do not provide proportional added benefits” [3].

While diagnostic imaging provides crucial information for the diagnostics of patients [4], inappropriate or low-value imaging are estimated to constitute 20–50% of radiological examinations internationally [2, 5,6,7,8]. Several interventions to reduce low-value imaging have been evaluated in the literature, including guidelines such as iRefer, iGuide, as well as national and international initiatives such as the National Institute for Health and Care Excellence (NICE) “Do-not-do list,” and the Choosing Wisely campaign [9,10,11,12]. However, the effect of such efforts on low-value diagnostic imaging has been limited due to barriers such as financial incentives, practice behavior, self-referral, lack of feedback, patient expectations, and duplicate imaging examinations [5, 11, 13,14,15,16,17]. Some interventions even seem to increase the use of inappropriate imaging [18, 19].

Several approaches to address the use of inappropriate health services, beyond low-value imaging, have been extensively evaluated. Education or training programs for health care personnel, clinical decision support, feedback, patient education, shared decision making, and economic incentives are but a few examples [5, 20,21,22,23,24]. However, the great quantity and variability of available approaches makes it unclear which measures are most suitable to target low-value imaging and overutilization. While research on interventions to reduce low-value care, in general, recommend implementation of multi-component interventions in complex health care systems [12, 25,26,27,28] there is still uncertainty as to why or when an intervention will be effective in diagnostic imaging specifically and/or in which clinical circumstances they are effective. Earlier systematic reviews on interventions in imaging have addressed specific interventions as image sharing or clinical decision support systems or specific imaging examinations or patient complaints [26, 29,30,31,32]. However, there is no encompassing systematic review assessing the outcome of various types of interventions to reduce low-value imaging. Thus, the objective of this systematic review was to provide an overview and evaluate the outcomes of interventions aimed at reducing low-value imaging.

Methods

This systematic review was conducted based on the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) statement (PROSPERO ID: CRD42020208072). The electronic database search was developed in Medline – Ovid (Table 1) and further adapted to Embase-Ovid, Scopus, and Cochrane Library. The terms used were built from medical subject headings (MESH) for Diagnostic imaging/Radiology, Health service misuse/Medical overuse, and Health planning. Keywords were used for the concept of reduction/avoiding. Also, the search was broadened with text word and keyword synonyms. The complete search strategy is available in Additional file 1. Searches were carried out in September and October 2020; last search made on 13th October 2020. Papers written in English, German, Danish, Norwegian, Dutch, and Swedish were eligible and language filters were used to exclude other languages. Keywords were used to exclude studies on animals, mass screening, and unnecessary care besides imaging services. No other limitations were applied.

Table 1 Search strategy developed in Medline (Ovid)
Full size table

Eligibility criteria

Primary empirical studies assessing interventions to reduce the use of low-value diagnostic imaging examinations were included. Studies designed as randomized controlled trials, non-randomized trials, descriptive studies, mixed-methods studies, and qualitative studies were included. While systematic reviews and meta-analyses were not included, the reference lists of relevant systematic reviews and meta-analyses were hand-searched for additional primary studies for inclusion. Studies published before 2010 were excluded due to the changes in perception on low-value imaging through the preparation and introduction of the Choosing Wisely campaign in 2012. The inclusion and exclusion criteria are provided in Table 2.

Table 2 Inclusion and exclusion criteria for assessing record eligibility
Full size table

Selection of records and methodological quality appraisal

The records were archived using Thomson Reuters EndNote X9.3.3 library and duplicates were removed. All remaining records were transferred to Rayyan QCRI [33] where titles and abstract review (EK and BMH) and full-text review and quality assessment (EK, ERA, LJJS, LvB-V and BMH) were completed by two teams of reviewers. Each study was quality assessed by one reviewer and double checked by EK for consistency. The Mixed Methods Appraisal Tool (MMAT) was used to assess the methodical quality of all included studies as it is considered to be an appropriate tool for appraisal of interventional studies of different methodologies [34]. Any disagreements during abstract or full-text screening were resolved through discussion and consensus. Reference lists of included articles were also hand-searched for relevant articles for inclusion. A grey literature search was also completed (ERA) according to the CADTH Grey Matters checklist [35]. Google Scholar was used for searching for eligible papers that cited the included studies.

Data extraction and synthesis

Data extraction was completed independently by EK, LJJS, LvB-V, BMH and ERA using a standardized summary table consisting of the following categories: author, title and year, country, design, population, clinical setting, outcome measures, low-value practice, intervention, targeted personnel or roles, control or comparator, use of low-value practice before or after intervention, and change in use of low-value practice. Data extraction was discussed in the research team for quality assurance purposes.

The findings from included studies were narratively synthesized. This synthesis was performed due to the variety of study designs among included studies and thus a meta-analysis was not feasible [36]. The narrative synthesis included familiarization, the development of a preliminary synthesis by organizing findings in tables. Then, relationships, patterns, and connections in the data were explored [36]. In addition, a subgroup analysis was done for interventions done in the USA separately.

Results

Search of the literature

As shown in Fig. 1, the electronic database searches resulted in 15,659 records. After the removal of 7468 duplicates, 8191 unique records were screened through title and abstract screening and 8108 records were excluded. An additional 103 records were identified through snowballing techniques and from the grey literature. A total of 186 articles were reviewed in full-text and 91 articles were excluded. Thus, 95 studies were included in the narrative synthesis. An overview of excluded studies with reason for exclusion is provided in Additional file 2.

Fig. 1
figure 1

Flow diagram for record selection

Full size image

Quality of included studies

The 95 included studies [37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107,108,109,110,111,112,113,114,115,116,117,118,119,120,121,122,123,124,125,126,127,128,129,130,131] are summarized in Table 3. All included papers fulfill the screening questions in MMAT. Forty-six papers were given a full score in the MMAT appraisal. These are marked with a * in Table 3 (the full MMAT report is available in Additional file 3), while the others had one criterion unfulfilled, or lacked sufficient information in the report.

Table 3 Characteristics of the included studies, outcome of the intervention and quality assessment result
Full size table

Characteristics of included studies

A majority of the included studies applied quantitative study designs. Retrospective chart reviews (n = 26) and uncontrolled before-after studies (n = 14) were the most common. Seventy-eight of the studies were conducted in the USA (82%). The setting of the studies included hospital (n = 40), emergency department (n = 24), or outpatient medical center (n = 18). Musculoskeletal (n = 26), neurological (n = 23) and vascular (n = 16) imaging were most commonly evaluated. The most targeted imaging examinations were chest CTA (n = 15) and head CT (n = 12). Further, pulmonary embolism (n = 15), lower back pain (n = 14), and minor head injury (n = 12) were the most commonly explored medical conditions.

Interventions

Guidelines (n = 28) and education (n = 28), either alone or in combination with other measures, were the most common interventions evaluated to reduce low-value imaging. The outcome measures reported in the included studies varied, with the number or rate of imaging examinations (n = 75) most frequently reported primary outcomes. A majority of studies (n = 61) used a single component intervention and most studies (n = 90) targeted referring physicians. An overview of participants exposed to the intervention (referring physicians, imaging staff, patients and/or family members), types of interventions, and combinations of components in multi-component interventions are presented in Fig. 2.

Fig. 2
figure 2

Overview of participants, interventions, and combinations in multi-component interventions in the included studies. Figure legend: To the left an overview of single interventions used for each participant group. To the right an illustration for how multi-component interventions were combined. Overlapping circles illustrate different combinations of two of more components

Full size image

A variety of imaging modalities or patient diagnoses were targeted, and the primary outcomes varied among studies that reported improvements post-intervention. Among studies targeting several imaging modalities or diagnoses, 74–79% of the studies showed a reduction in use of low-value imaging. In contrast, studies targeting one specific modality only showed that targeting X-ray [37, 47, 48, 63, 64, 83, 86, 90, 92, 98, 104, 112, 119, 124, 130], CT [41, 46, 49, 53, 54, 58, 60, 62, 67, 69, 70, 77, 79, 82, 87, 88, 94,95,96, 101, 102, 106, 108, 121,122,123, 125,126,127, 129] or MRI [45, 68, 75, 105, 118] led to a 87, 86, and 83% reduction in low-value imaging, respectively. Few studies included other imaging modalities.

The most commonly targeted patient diagnosis was bronchiolitis [48, 90, 104, 112, 119], pulmonary embolism [49, 58, 60, 82, 100, 102, 121, 122, 125], and head injuries [67, 70, 77, 79, 87, 94, 101, 123, 126, 127]. In studies targeting these complaints, a reduction in use of low-value imaging were reported in 78–80% of the studies, while imaging in lower back pain [38, 45, 51, 53, 54, 65, 68, 71, 75, 89, 115, 118] were reduced in 58% of studies.

Among the 77 studies that reported improvements following the intervention [37,38,39,40,41,42,43,44,45, 47,48,49, 52, 53, 56, 57, 59, 60, 62,63,64, 67, 69, 71,72,73,74,75,76,77,78,79,80,81, 83, 85,86,87,88,89,90, 92,93,94,95,96,97,98,99,100,101,102,103,104,105,106, 108,109,110,111,112,113,114,115,116,117, 120,121,122,123, 125, 128,129,130,131], decreases in low-value imaging varied largely from < 1 to 62%. Of the remaining studies, three studies reported mixed results, where only some of the targeted low-value imaging examinations were reduced [46, 54, 68], and 16 studies showed a non-significant change or increase in the use of low-value imaging post-intervention [50, 51, 55, 58, 61, 65, 66, 70, 82, 84, 91, 107, 118, 119, 124, 127].

Implementation of multi-component interventions (2 or more components in combination) reportedly reduced the use of low-value imaging among 94% of the included studies [40,41,42,43,44, 47, 48, 51, 53, 56, 59, 63, 72, 78, 79, 81, 83, 86, 87, 94, 101, 104, 105, 111, 115, 116, 118, 120, 122, 123, 126, 128]. Multi-component interventions were found to be more effective when education was one of the components. Following implementation of a single component intervention, 74% of included studies reported decreases in low-value imaging [37,38,39, 45, 46, 49, 50, 52, 54, 55, 57, 58, 60,61,62, 64,65,66,67,68,69,70,71, 74,75,76,77, 80, 82, 84, 85, 88,89,90,91,92,93, 95,96,97,98,99,100, 102, 103, 106,107,108,109,110, 113, 114, 117, 119, 121, 124, 125, 127, 129,130,131]. Data analyses based on the USA studies demonstrated similar results as 96% of multi-component and 68% of single-component interventions showed reduction in the use of low-value imaging. Thus, county of intervention does not affect the result alone. Implementation of guidelines or clinical decision support systems were the most effective single-component interventions [37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107,108,109,110,111,112,113,114,115,116,117,118,119,120,121, 126, 128, 130]. Furthermore, 23% of single-component interventions compared to 6% of multi-component interventions showed no statistically significant difference or an increase in the use of low-value imaging. In Fig. 3, the green bars represent studies with a reduction in the use of low-value imaging, red bars represent no significant change or increase, and orange bars represent mixed results.

Fig. 3
figure 3

The number of studies and outcome of different types of interventions to reduce low-value imaging

Full size image

Single-component interventions such as education, shared-decision-making, and financial measures alone often had no effect on use of low-value imaging [55, 61, 70, 82, 107]. Studies with more than 30% reduction in use of low-value imaging were both single-component (n = 11) and multi-component (n = 8) interventions [37,38,39, 41, 42, 44, 69, 73, 75, 80, 81, 83, 85, 86, 96, 108, 116, 117, 131]. All these studies targeted refereeing physicians while one also targeted imaging staff. Another targeted referrer, imaging staff, and patients. Of these studies, 16% were from countries other than the USA. Table 4 provides an overview of the type of interventions that resulted in more than a 30% reduction in low-value imaging.

Table 4 Overview of interventions with more than a 30% reduction in use of low-value imaging examinations
Full size table

Discussion

A large body of literature evaluating the outcome of interventions aimed at reducing low-value imaging was identified through this systematic review. Broadly, most interventions were found to be effective, with multi-component interventions more frequently reported to be effective compared to single-component interventions. All studies evaluating multi-component interventions with an education component reported reductions in low-value imaging. Multi-component interventions targeting the participants on several points providing education and then feedback and reminders over a longer period seems to be effective as change takes knowledge, motivation, and time [132]. Single-component interventions, particularly guideline implementation, clinical decision support systems, feedback, or actions required from the referrers, showed reduction in use of low-value imaging in several studies but not in all. This might be caused by organizational differences, differences in the clinical setting, or participants motivation [132]. Shared decision-making, new referring procedures, and financial measures demonstrated no effect; however, these interventions were only evaluated in a limited number of studies. Targeting specific examinations for specific conditions (e.g., bronchiolitis), targeting referrers, and only targeting one imaging modality seemed to be more effective than targeting several modalities or examinations referred from a variety of referrer groups (e.g., lower back pain). There was also a variety of outcome measure used among included studies. The number or rate of low-value imaging was the most common. Others included appropriateness and diagnostic yield. This warrants caution when comparing outcomes between different types of interventions.

The present results are in line with previous systematic reviews on interventions to reduce low-value services in general [12, 26, 29,30,31,32] and with a previous scoping review on unnecessary imaging, diagnostic tests, and procedures in hospitals [28]. The results indicate great variation in outcomes for many interventions. This is in accordance with research on innovation and interventions suggesting that the formal and informal network in the organization, motivation, flexibility, and fitness to the internal culture and core values in the organization where interventions are implemented, were key factors for a successful and long-lasting change of clinical practice [132,133,134,135]. Whether or not an intervention is successful in reducing the use of low-value imaging would thus depend on a variety of factors. Comparing studies conducted in the USA to those from other countries showed no difference in type of interventions used or in the rate of studies demonstrating > 30% reduction in the use of low-value imaging. Thus, the effect of intervention seems to be dependent on local culture and health care organizations rather than the national health system alone. In addition, only a few interventions were directed against patients, which is somewhat surprising as patients are also identified as drivers in the use of low-value imaging [136, 137]. Further research should include the patient perspective and the role of the radiology department in interventions to reduce low-value imaging in addition to a review on cost-effectiveness of interventions to reduce low-value imaging. Further investigation should focus on how interventions can be adapted to the culture and core values of the providers of health services in different contexts.

Our study has several limitations. Publication bias may have been introduced as articles with negative or nonsignificant findings are less likely to be published. Among the included studies, few reported null or low effect. Most studies had an uncontrolled before-after design not considering that there may be a secular downtrend in the use of the low-value imaging examinations due to the attention in campaigns, such as Choosing Wisely. Thus, the outcome may be overestimated. Further, the review may be subject to contextual bias and have limited generalizability, as most of the studies were conducted in the USA. Accordingly, caution is warranted when inferring from and applying the results in different settings. The proportion of single-center studies and observational studies may enhance the overall positive effect of the interventions [26]. Yet another limitation is related to indirect outcome measures, as many publications focus on interventions’ impact on volume and not on value. This is understandable as the change of low-value utilization is a warranted measure, but one should notice that the value of these services is not assessed. Moreover, it may be argued that the spectrum of imaging that is targeted by the interventions is biased by the methods to assess intervention outcomes. Nonetheless, we report a wide variety of interventions targeting many examinations. There are reasons to believe that the interventions are targeted strategically. For example, interventions that are believed to be effective may be targeted towards examinations documented to be of low value.

It may also be argued that retrospective chart reviews are not proper intervention studies, but as they are used systematically to assess change in practice, we have included them in this review. Additionally, the included studies afford providers’ perspective and as indicated, few studies elicited patient preferences or included patient-reported outcome measures. In addition, the snowballing uncovered a few studies published in 2021 not included in our analysis due to the inclusion criteria. The results of these studies are in line with the studies included in the analysis and thus not including these did not reduce the strengths of the analysis in this review [138,139,140,141,142].

Conclusions

This systematic review demonstrates that interventions to reduce low-value imaging can be very effective, but that there is a large variation in types of interventions and their outcomes. We found that multi-component interventions reported reduction in low-value imaging or increased diagnostic yield more frequently compared to single-component interventions. The context in which the intervention is introduced seems to be of vital importance for successful reduction of low-value imaging. Thus, in the future multi-component interventions that are adapted to the local context are more likely to be successful. Further research is needed to assess how interventions to reduce low-value imaging can best be adapted to specific contexts and how to reduce the use of low-value imaging cost-effectively.