Background

Public reporting (PR) is a mechanism of “providing data about a health care structure, process, or outcome publicly available or available to a broad audience free of charge or at a nominal cost, in order to be able to compare data across providers or to a national/regional data report on performance for which there are accepted standards or best practices” [1]. Public release of quality and clinical performance of the healthcare providers is becoming increasingly common among the healthcare systems worldwide. Policy and decision-makers in a demand-driven healthcare system are becoming more interested in having information about quality performance [2] and thereby PR has been proposed as a mechanism for providing more transparency and accountability of healthcare providers [3]. Constant improvement of the quality of care should be one of the top priorities of healthcare providers [4]. According to the theory of PR, healthcare users are expected to inform themselves about the quality of healthcare system before selecting the particular provider and so those with high performance would be rewarded by selecting, while low performers would be avoided and thereby stimulated to improve their performance [5–8].

There have been suggested several pathways through which quality might be improved after the release of performance data. First pathway, the selection pathway, proposed by Berwick et al., is based on the concern of healthcare providers about their market share, where consumers choosing better performers, motivate providers’ efforts to improve quality in order to attract more patients. The change pathway, also proposed by Berwick et al., is based on the concept that identifying quality deficits is sufficient to stimulate the professional motivation of clinicians and organizations to improve [9]. After observing in their study that these two pathways were a relatively weak stimulus to action, Hibbard et al., introduced the third pathway - the reputation pathway, based on the premise that providers which perform poorly, after being identified through the PR, suffer damage to their reputation and further motivate quality improvements in order to protect or improve reputation [10].

PR of clinical outcomes data is a great tool for increasing the transparency in healthcare which enables patients to make informed choices about their healthcare. The most convincing rationale for the PR of clinical outcomes is the one’s right to be aware of the quality of care that he/she is likely to receive from providers [11]. Cardiac surgery has been a pioneer field for the publication of clinical outcomes, since being among the most frequently performed complex surgical procedures [12]. The collection and publication of standardized military hospital mortality rates by Florence Nightingale in 1863 where highlighting the differences in mortality rates between hospitals is believed to be the earliest attempt of PR of clinical outcome and in general [13]. The modern practice of PR systems started in the late 1980s in the USA with the introduction of Coronary Artery Bypass Graft (CABG) report cards of New York State Department of Health, as the first state wide program where the risk adjusted post-operative mortality rates following CABG surgery are being published at the level of both the hospital and the individual surgeon resulting in a 41% decline in risk-adjusted CABG mortality rate [14]. In Europe, back in 1994, Scotland was the first to adopt PR with the Clinical Resource and Audit Group (CRAG) [15]. After these initiatives, other countries have started to follow their example and implement PR into their healthcare systems.

Several authors, over the years have studied the effects of PR on clinical outcomes and, still nowadays, there are inconsistent results in the literature. The review of healthcare PR by Fung et al. found mixed signals among the studies of the effect of PR on outcomes, with some studies showing no effects and others showing minimal effects. Indeed the authors concluded that a solid evidence is still lacking and the systematic evaluation of many major PR systems is needed [5]. The scarcity of a solid evidence does not necessarily suggests the lack of effect and, since the publication of Fung et al., many studies have been conducted to explore the effect of public release of performance data on clinical outcome, but always with incoherent results. Thus, we took the aim to perform an up-to-date systematic review of scientific literature in order to synthesize, both qualitatively and quantitatively, the impact of PR on clinical outcomes.

Methods

A protocol was developed and a systematic review and a meta-analysis were conducted and reported in accord with PRISMA guidelines for meta-analyses and systematic reviews [16].

Search strategy and study selection

A literature search was performed by accessing Pubmed, Thomson Reuters Web of Science and Scopus databases to identify studies that investigated the relationship between PR and clinical outcomes. The search terms “public reporting”, “quality reporting”, “information dissemination”, “data shar*” and “report card*” were used, by specifying “health” for databases that also covered non-health topics.

Our search was restricted to English language studies published from 1st January 1991 to 31st December 2014. Studies were considered eligible if they comprehensively described the PR mechanism in terms of subjects, setting, location and dissemination way, if the study design adopted was clearly described and if they investigated the relationship between PR and clinical outcomes. Studies not reporting original data as well as studies focusing only on non-clinical effects of PR were excluded.

Two reviewers independently screened titles and identified abstracts of relevant titles. Full texts of potential citations were subsequently obtained and independently screened by the two reviewers for inclusion. Disagreements were resolved through discussion. Additional relevant publications were identified from the references of the initially retrieved articles.

Data extraction and analysis

From each study data on the first author’s last name, year of publication, objective, subject, setting, location, PR mechanism, clinical outcome assessed and key findings were extracted. For each clinical outcome assessed, quantitative data were also extracted if available.

Two reviewers conducted all data extraction independently and disagreements were resolved through discussion. The same reviewers evaluated also the quality of the studies using a GRADE derived approach (see Additional file 1) [17].

Among the clinical outcomes evaluated, meta-analysis was performed for overall mortality rate which quantitative data were exhaustively reported in 10 different studies. Because of the significant heterogeneity expected among the studies performed in different settings, the random effects model was employed using the Der Simonian and Laird’s method [18].

Heterogeneity was quantified using the Cochran Q test and I2 statistics [19] and subgroup analyses were performed for different study design and setting.

Sensitivity analyses were conducted by excluding one study at a time from the meta-analysis to determine whether the results of the meta-analysis were influenced by individual studies and whether risk estimates and heterogeneity were substantially modified.

The presence of publication bias was assessed using the Egger’s test [20].

All analyses were carried out using Review Manager, version 5.2.7 for Mac (The Nordic Cochrane Centre, Copenhagen, Denmark) and Stata, version 13.1 for Mac (StataCorp, College Station TX, USA).

Results

Characteristics of the studies

We identified a total number of 22,404 studies through Pubmed, Thomson Reuters Web of Science, Scopus online databases search. After removing the duplicates, 10,578 studies were left, and, carefully reading the titles, 2,145 studies were assessed for eligibility. Further step was reviewing the abstracts, and 254 full text articles were obtained. By not fulfilling the inclusion criteria, 231 full text articles were excluded, and 4 were individuated from the screening of references list of studies that fulfilled inclusion criteria, leaving 27 studies to be included in our analysis [14, 21–46]. Figure 1 depicts the process of literature search and study selection.

Fig. 1
figure 1

Flowchart depicting literature search and study selection

Full size image

The publication years of the studies were ranged from 1994 [21] until the most recent ones, from 2014 [44–46]. Most of the 27 studies included in our review (N = 23) were carried out in the US [14, 21–30, 32–39, 41–44], one in Canada [31] and Italy [40], and the remaining two in China [45, 46]. Twelve were cohort studies in which the control and study cohorts of patients were taken from different facilities over the same period of time [23, 26, 29, 33, 35, 38, 39, 42–46], 14 were cohort studies in which the control and study cohorts of patients were taken from the same facilities before and after the introduction of a PR mechanism [14, 21, 22, 24, 25, 27, 28, 30–32, 34, 36, 37, 41], while the remaining study used both cohort designs [40]. The majority of the studies included in this review presented performance information in the form of “report cards”, while the others used different forms to communicate data to the public (Table 1).

Table 1 Characteristics of the included studies and effects of PR on clinical outcomes
Full size table

There were several clinical outcomes examined throughout the studies. Many investigated the effect of PR on patients’ mortality [14, 21–33, 35–39, 41, 43, 44]. Other aims in these studies included investigation of cardiac readmission to hospital [28, 37], antibiotic use, and waiting times to see a physician [34], injection prescribing rates [45], Percutaneous Coronary Intervention (PCI) rates, hip fractures operated on within 48 h, cesarean deliveries [40], change in CABG operations volume [14], improvements in infection prevention [42] and patient choices [36].

We included a total number of 27 studies in our systematic review that evaluated the effect of PR on clinical outcomes and the results are summarized in Table 1. Mainly, the effect of PR on clinical outcomes was positive. Fourteen studies reported positive results [14, 21–23, 25, 30–33, 36, 37, 40, 44, 45], nine reported not significant results [26, 27, 29, 34, 35, 38, 41–43], three studies reported mixed results [24, 39, 46], some positive and some negative or null, while one study reported a negative effect [28].

We used a GRADE derived approach to assess the quality of the included studies. The body of evidence in our review was characterized by a low quality level (see Additional file 1). Indeed, the study design was, in almost all of the studies, observational, and there were a number of limitations. When a pre-post approach at hospital level was used to assess the performance before and after the release of PR, there was no external control group for comparison. Moreover, there was no information on institutions and participants that were lost-to-follow-up during the study period. Also, the different outcomes, for institutions with and without PR, could be influenced by some characteristics not measured across the studies.

Effects of public reporting on mortality

The effect of PR on mortality, as isolated clinical outcome, was evaluated throughout 22 studies [14, 21–33, 35–39, 41, 43, 44]. All studies were set in hospitals, in the US and Canada, and 19 of them were on cardiac patients and used PR data mostly from specific CSRS. A positive effect of PR on mortality was reported in 12 studies [14, 21–23, 25, 30–33, 36, 37, 44].

Two studies reported mixed effect of the PR on patients’ mortality. Baker et al. [24] demonstrated that, for most conditions, after the release of PR, risk-adjusted in-hospital mortality declined and mortality rate in the early post discharge period rose, while the 30-day mortality rate declined for heart failure and obstructive pulmonary disease and increased for stroke, while Joynt et al. [39] stated no differences in reporting versus non reporting states for overall mortality among patients with acute myocardial infarction (AMI), except among Medicare beneficiaries with AMI [39].

The only study in our review that reported a negative effect of PR was a cohort study from Dranove et al. [28], where the authors showed a statistically marginal evidence that after releasing the report cards, the average mortality rate in New York and Pennsylvania hospitals performing CABG increased by 0.45 percentage point on a base of 33 %. The remaining 7 studies reported non-significant effect of PR on mortality [26, 27, 29, 35, 38, 41, 43].

Ten out of the 22 studies investigating mortality as an outcome [22, 23, 26, 28–31, 38, 39, 41] reported sufficient quantitative data to be pooled through meta-analysis. Overall, this analysis included a total of 1,840,401 experimental events and 3,670,446 control events. The meta-analysis resulted in a RR of 0.85 (95 % CI, 0.79-0.92) in a contest of high heterogeneity (p ≤ 0.0001; I2 = 99.1 %) (Fig. 2). Publication bias was not evident using the Egger’s test (p = 0.91). We also performed a one-way sensitivity analysis, where one by one study was omitted from the overall meta-analysis, but no significant change in risk estimates was noticed.

Fig. 2
figure 2

Meta-analysis of the PR effect on mortality as clinical outcome by facilities

Full size image

A subgroup analysis on mortality by study design was also carried out. The six publications [22, 23, 28, 30, 31, 41] reporting mortality rates in the same facilities during different periods showed a RR of 0.85 (95 % CI, 0.76-0.94) in a context of high heterogeneity (p < 0.0001; I2 = 100 %). Whilst, the four included studies [26, 29, 38, 39] recording mortality rates during the same period in different facilities showed a RR of 0.91 (95 % CI, 0.85-0.97) in a context of high heterogeneity I2 = 95 % (p < 0.0001) (Fig. 2). Test for subgroup differences resulted negative with a p value of 0.28.

Another subgroup analysis was performed by studies considering different mortality causes. Pooling the results from studies focused on mortality from cardiovascular disease, six studies were included [23, 28–31, 39] and a RR of 0.83 (95 % CI, 0.77-0.91) was calculated, with high heterogeneity (p < 0.0001; I2 = 95 %). For the subgroup of studies that included patients with a wide range of conditions [22, 26, 38, 41], a RR of 0.91 (95 % CI, 0.83-0.99) was obtained, with heterogeneity I2 = 99 % (p < 0.0001) (Fig. 3). Test for subgroup differences resulted negative with a p value of 0.18.

Fig. 3
figure 3

Meta-analysis of the PR effect on different mortality causes as clinical outcomes

Full size image

Effects of public reporting on other clinical outcomes

Other clinical outcomes, beside mortality, were explored in 8 studies [28, 34, 37, 39, 40, 42, 45, 46]. In an Italian study performed by Renzi et al. [40], a pre-post evaluation of clinical outcomes in Lazio Region was done and also a comparative evaluation versus Italian Regions without comparable PR programs. The study demonstrated that in Lazio the PCI within 48 h from admission increased from 22.49 to 29.43 % following the reporting of the Regional Outcome Evaluation Program (P.Re. Val.E) results. In the other regions without comparable programs, during the same period, this proportion increased from 22.48 to 27.09 %. Hip fractures operated on within 48 h from admission increased in Lazio, while not in other regions with no PR. Cesarean deliveries did not decrease in Lazio (from 34.57 to 35.30 %), while only slightly decreased in the other regions (from 30.49 to 28.11 %) [40]. Joynt et al. [39] reported that after implementation of PR, odds of undergoing PCI in Massachusetts where PR was present decreased comparing with non-reporting states (41.1 % vs 45.6 %; OR = 0.81; 95 % CI = 0.47- 1.38; P = 0.03). Marsteller et al. [44] in one cohort study on patients from 1,046 adult intensive care units found reductions in central line-associated bloodstream infections (CLABSIs) rates in the first 6 months from the release of PR, compared to the units from the states with no reporting mechanisms. During months 13–18, groups with mandatory PR of data about CLABSI showed a trend toward greater reduction in CLABSI compared with states with no PR [44]. Hospital readmissions were evaluated in two studies. Dranove et al. [28] reported a negative effect of PR, with a significantly increased average rate of readmissions with heart failure by approximately 0.5 percentage point. On the other hand, more recently, Werner et al. [37] reported performance improvements for AMI in terms of declines in readmission rates beside lengths-of-stay and mortality rates.

Studies set up outside hospital settings [45, 46] addressed different clinical outcomes, such as antibiotic prescription and usage as well as injection prescribing rates, in primary healthcare institutions. Yang et al. [46] evaluated the impact of PR on antibiotic prescribing for upper respiratory tract infections (URTIs) demonstrating that PR interventions reduced the incidence of oral antibiotic prescription and slowed down the increase of combined use of antibiotics for URTIs in primary healthcare setting. According to Wang et al. [45], PR led to a reduction of approximately 4 % in the injection prescribing rate four months after intervention (OR = 0.96; 95 % CI: 0.94, 0.97), although with an inconsistent effect in each month after intervention.

Discussion

The public release of hospital performance data has been recommended as one key strategy for stimulating improvement of quality of care by putting the focus on transparency and accountability of healthcare providers. Also, PR is expected to stimulate active patients/citizens participation by helping them make informed choices when choosing health care providers [47].

According to Berwick, PR can improve performance through 2 pathways. In the Improvement through Selection pathway, patients and providers could shift care from low-quality to high-quality hospitals by using the publicly disclosed reports of hospital performance quality and thereby stimulating quality improvement efforts for the benefits of market share. In the Improvement Through Changes in Care (or quality improvement) pathway, published performance data can identify areas in which providers had low accomplishment and help them to focus on improving performance, by appealing to their professionalism or their concern about reputation or direct market position. As well, it is more likely that quality improvement happens in combination of these two pathways [9].

Nowadays, PR is a more and more common health policy tool to stimulate and maintain quality improvement of health care. There is a growing international interest in providing the necessary information of clinical quality and performance of healthcare providers [48]. Many studies of PR performance data have been published so far, but available reviews of the association between clinical outcomes and public disclosure of performance data are limited.

In our systematic review on impact of PR on clinical outcomes, we identified 27 new articles published since 1994. These studies, which were mostly hospital-level and had medium global ratings, focused primarily on mortality rates and cardiac procedures. Ten studies found that mortality rates decreased after PR, while nine studies did not find a significant link between PR and improvement.

Most of the studies examined the impact of USA and Canada PR of mortality rates for cardiac surgery (CABG and PCI). Different reporting systems were evaluated across the publications. Many studies that we found, were focusing their research on the same group of reporting systems like NYS CSRS which is considered as a pioneer among PR systems, despite the fact that nowadays many PR systems are available on the market [14]. Also studies comparing effectiveness among different PR systems are lacking. Many studies used reporting systems based on cardiovascular outcomes, including surgical interventions, hospitalization and mortality. In our review, we found seven studies based on the NYS CSRS [14, 21, 23, 25, 28, 32, 39], four studies on Cleveland Health Quality Choice Program (CHQC) [22, 24, 26, 27], as well as Medicare’s PR initiative [35, 37, 41, 42]. Other studies used various state-level PR data sources. The majority of the studies (N = 25) were placed in hospital settings [14, 21–23, 26, 27, 29–38, 40–43], and two in primary healthcare institutions [45, 46].

As opposed to the studies performed in the hospital setting, little information is available from the literature regarding the effectiveness of PR in primary care settings. Translating knowledge and experience from one healthcare setting to others is a reasonable method for building the high-quality healthcare service everywhere [49]. Two studies included in this review [45, 46] demonstrated that PR can have an impact on medication prescription and usage in primary healthcare settings. The authors concluded that even though further clinical outcomes of such an impact were not investigated in their studies, a positive effect might be expected. To the best of our knowledge, our review is the first to tackle the effects of PR in primary care setting, so additional research is still needed to further investigate the mechanism by which public reporting takes effect in outside-hospital settings.

Our research, including 27 studies carried out in North America, Europe and China, was mainly conclusive with the previously published reviews of public reports, whereas the effect of the PR on clinical outcomes was in our qualitative and also, through meta-analysis technique, quantitative assessment positive for patients in the hospital setting and primary healthcare but with a low quality of the evaluated studies. In one of the first reviews about PR, Marshall et al. [50] showed improvement in health outcomes among 7 PR systems based in the USA. In 2008, a systematic review of 11 studies by Fung et al. [5], showed inconsistent association between PR and effectiveness on patient outcomes. They reported that studies of the effect of PR on outcomes provided mixed signals and that most of the evaluated studies were descriptive and had low global ratings thus limited strength of evidence. They also concluded that PR stimulated quality improvement activity in hospitals and yet did not affect hospital selection by patients and generally was limited to a few clinical areas, like cardiac surgery. In one recent report, Specchia et al. [51] performed a review on the use of publicly released performance data and concluded that the introduction of standard set of evidence based outcomes and performance measures at national level could reduce the pressure from the selection of performance measures to be disseminated through the PR and make health care providers aware of the importance of transparency and accountability. They also concluded that it would be very effective to link PR to Pay-for-Performance (P4P) systems by basing payments on outcome results and thereby supporting the quality improvement process.

Strengths and limitations

This review has some strengths and limitations. The comprehensiveness, the focus on clinical outcomes and the attempt to provide a quantitative synthesis of the available evidence are strengths of this review. Our search, in fact, covered a wide time interval and a variety of settings including both hospitals and primary care institutions and results has been synthetized both qualitative and, for mortality outcome, quantitative through meta-analysis technique. However, because of the resulting heterogeneity, caution should be placed in interpreting the results of the quantitative synthesis made for mortality outcome.

Some studies [14, 21, 43] have also noted how, after the introduction of PR, structures starting from a lower quality level have a greater propensity to improve their quality compared to those starting from an already high level of quality. Difference in starting levels of quality between the different hospitals and institutions when introducing PR is also a source of heterogeneity among the studies included in our review.

Limitations are also present due to the observational nature of the included studies. Indeed, the studies which control and study cohorts were taken from the same facilities before and after the introduction of a PR mechanism may have overestimated the effect of PR, because of a technological improvement trend running parallel to the adoption of the PR [41].

Search strategy was also limited to English language studies published from 1991 and most of published evidence concern cardiovascular disease in hospital setting.

Policy implications

PR can be considered as one of the key drivers for transparency and accountability in the public health field. Far from providing a quantitative estimate of the current use of PR, the experiences described in this paper can represent a framework of opportunities for changing the relationship between patients/customers and healthcare providers and as a tool to support policy makers in addressing and allocating resources according to the assessment of providers’ performances and the publication of their results which are accessible and understandable to all of the stakeholders, including patients/customers.

Several successful examples of stakeholder involvement in the processes deriving from PR have been described both in the United States [50] and in Europe, as in Netherlands [52] or Germany [53]. In particular WHO Regional Office for Europe suggested key considerations for a successful strategy to encourage providers in improving the quality of services, the accountability of processes, the identification of failures, and providing with the use of publicly reported quality information strengthening communication tools, supporting public health professionals and making clearer consequent decisions [54].

The decision-making is always a difficult process for patients. To obtain informed choices high levels of numeracy and literacy are needed and PR can be considered a useful instrument to face this issue. Therefore, policy makers should support good practices in Public Health especially those, such as PR, focused on increased patient awareness.

Moreover, health policies should be promoted with the aim to integrate different strategies pursuing quality and excellence in healthcare. For example, linking PR to P4P, and therefore to remuneration for incentives schemes based on performance data, would strengthen the competitiveness of the whole system, triggering at the same time a virtuous circle oriented to the increase of both appropriateness and continuous quality improvement of healthcare [51].

Conclusions

The introduction of PR programs at different levels of the healthcare sector is a challenging but rewarding public health strategy. Existing research covering different clinical outcomes supports the idea that PR could, in fact, stimulate providers to improve healthcare quality.

Transparency and accountability resulting from PR implementation not only give patients those information tools customers commonly are able to access in many other sectors, but are key points in the process that make patients and citizens empowered protagonist of their care.

Abbreviations

AHRQ, Agency for Healthcare Research and Quality; AMI, acute myocardial infarction; CABG, Coronary Artery Bypass Graft (CABG); CAHPS, Consumer Assessment of Healthcare Providers and Systems; CHQC, Cleveland Health Quality Choice Program; CLABSIs, central line-associated bloodstream infections; CRAG, Clinical Resource and Audit Group; HQID, Hospital Quality Incentive Demonstration program; NYS CSRS, New York State Cardiac Surgery Reporting System; P4P, pay-for-performance; PCI, Percutaneous Coronary Intervention; PR, Public Reporting; URTIs, upper respiratory tract infections (URTIs)