In this systematic review, the results of 7 publications which met the study inclusion criteria were combined to determine the effect of CPs on the outcomes of in-hospital treatment for heart failure: in-hospital mortality, rate of readmission, LOS and hospitalisation costs.
This study is the first systematic review to show an impact of CPs on the treatment of heart failure.
The primary finding of this systematic review was that the implementation of CPs for heart failure reduced hospital mortality and length of stay without increasing hospitalisation costs. Also readmission rate decreased in the care pathway groups (Figure
). By looking at this data, it is evident that CPs can effectively improve the quality of care provided to patients suffering from heart failure. Lower mortality rates and shorter hospital stays among CP patients are associated with standardisation of the care process. However, we must be careful with the overall conclusions: pathways are not just documents recorded in the patient charts but a way to organize and standardise a multi-disciplinary care for patient groups using well-known quality improvement methodologies
In the included studies, all patients were admitted to hospital with a primary diagnosis of chronic heart failure. The definition of readmission rate was different in the included studies. In two studies, readmission was recorded within 31 days after discharge
]; in one study it was recorded within 90 days of discharge
] and in two studies 6 months after hospital discharge
]. For all included studies cost was calculated for hospital stay per-patient but no additional information was given about the cost calculation.
There are two previous generic meta-analysis reports suggesting that implementation of CPs can achieve a reduction in some of the patient outcomes
]. Although some of the findings were confirmed by our study, there are also several reasons that preclude comparison of our findings with generic meta-analyses. Firstly, pathways are a learning tool both at professional (individual) and organizational (team) level
], and pathways for different patient groups define different tasks to be learned, which affects performance of pathways and also patient outcomes. Also most of the outcomes such as mortality rate, readmission rate, LOS and costs of hospitalisation are related to types of treatments given to patients and show differences among specific patient groups. Different interventions are sometimes classified under the same term. When different clinical topics are included in the meta-analysis, the effect size of the intervention mostly shows huge variations between pathways for different patient groups. Exploring the reasons behind the heterogeneity rather than derivation of a single summary estimate of the effect size has emerged as the main goal of the meta-analysis
]. Some statistics which were estimated for general populations from the generic meta-analysis such as weighted mean difference, OR or RR cannot be extrapolated to any patient group.
This systematic review has some limitations. We initially identified 46 relevant studies among 7981 total search results. However, only 7 studies could be considered as CPs and met our inclusion criteria. The scarcity of publications on CP probably results from the fact that CPs have only recently become a popular tool and the associated benefits for heart failure interventions are still not clear. Therefore, not many studies could be included to evaluate the effect of CPs on in-hospital treatment of heart failure.
In this systematic review, we observed that the findings of controlled trials were always more positive than the RCTs. In the included papers all of the study groups were experimental and the data was collected prospectively. Among included studies, three were randomized controlled clinical trials and one of them was an interrupted time series. In this study, measurements were performed on the same patients before and after the experiment, so patient characteristics are unlikely to have affected the outcomes. But 3 of the included studies were non-randomised; and historical controls were used. In the non-randomised studies, differences in patient characteristics may have affected the outcomes. Although patient characteristics were similar in the study and control groups (Table
), differences in patient characteristics may explain more positive results in CCTs. Moreover, none of the studies mentioned in-hospital complications; thus, the association between hospital complications, costs and LOS could not be addressed. In addition, we assume that combining clinical indicators with a satisfaction survey could have given a more accurate measure of the true level of quality achieved through CPs
]. Concerning included studies, only two studies measured patient satisfaction score
] and reported no significant effect. As studies used different quality assessment tools, combining of results was not possible; so, patient satisfaction scores were not included in the systematic review as an outcome.
Since heart failure is defined in different ways and inclusion criteria can vary from study to study, individual patient data meta-analysis would have been a good option to overcome some of the shortcomings of our analysis. The New York Heart Association classification would have been used to perform subgroup analysis. Unfortunately only four of the seven studies reported NHYA classification and they did not provide separate results for each classification. Another limitation of our study was that despite performing a search from 1985 to 2011, we could not identify any eligible study until 1999. Thus, we cannot generalise our results for the last 26 years but our study results are valid for the last 13 years.
In 6 studies
], pathway and control groups were well-balanced with regard to the number of the patients, and in 5 studies
] pathway and control groups were age- and sex-matched. Moreover, these five studies were similar with respect to the age and sex ratio (female/male) of the patients. One study did not provide patient characteristics
] and one study enrolled only female patients
]. The title of this study suggests that the study sample consists of older women, but the mean age of the patients was similar to the other studies included.
Although the above mentioned patient characteristics of the studies were similar, we observed a high heterogeneity in the statistical analysis by use of the I2
statistic, which was introduced by Higgins et al.
]. Although I2
values of 75% were described as high
], no desired threshold for I2
was determined. Recently it was mentioned that heterogeneity is to be expected in a meta-analysis because studies are performed by different teams in different settings
]. Using a random-effects model to reduce the effect of heterogeneity on a statistical analysis is a widely used approach
]. In addition, it is known that sensitivity analysis should be accompanied by an analysis to show the effect of heterogeneity on statistical analysis
]. A risk of bias check list has been developed according to pre-defined inclusion criteria and papers were ranged according to the risk of bias for each outcome. Results of the sensitivity analysis were given in the Additional file
. In sensitivity analysis, relative risks were similar for in-hospital mortality, re-admission rate and length of stay. A significantly greater mean difference was observed for hospitalization cost which is probably a consequence of the small number of studies included for this outcome.
There is an international controversy on the definition of pathways
]. Vanhaecht et al. defined 17 criteria
] and the paper by De Bleser et al. gave a detailed overview on how a definition of pathways could be built
]. Recently Kinsman et al. introduced a definition in their paper, consisting of five criteria for care pathways based on the aforementioned papers previously published
]. Both criteria are valid and usable. In this study only studies meeting the CPs definition of EPA, which covers Kinsman et al. criteria’s, were included in this systematic review
According to the definition, pathways describe patient processes and in the planning of a pathway, identification, and resolution of process ‘bottlenecks’ frequently occur. However, what works for one organization may not work for another, because of subtle differences in these processes and bottlenecks. Organizations are also different in their readiness for and capacity to change. These are often referred to as context issues, which influence implementation and effectiveness
]. Hawe, Shiell, and Riley (2004) suggested as a possible solution to standardize complex interventions, the function and process of the intervention and not only the components
]. This information on the context and the change process is critical to the ability of others to adapt the findings of a study to their own setting
]. A pathway which works in one place may be ineffective in another without this key knowledge. In our opinion, this important issue did not affect the validity of our findings. In fact, all the papers that have been selected in our review included process and outcome indicators that provided data to understand if pathways worked (Additional file
: Annex 2). Therefore we think that our paper addresses the research question adequately and we also think that our findings are based on consistent data.
A final remark is that the literature suggests that when researchers implement a pathway in a team that is already performing well, one may not identify significant improvements. A poorly performing team may, on the other hand, be greatly improved by the implementation of a new pathway, but these teams may not always be interested in improving the organization of their care process
]. Pathways are one of the tools interdisciplinary teams can use to audit, standardise and improve the organisation of care.