A 10 year (2000–2010) systematic review of interventions to improve quality of care in hospitals

The aim of this review was to retrieve data based articles which implemented interventions that sought to improve quality of care in adult general hospital settings between 2000 and 2010. Relevant articles were retrieved following systematic searches of the following databases: PubMed, PsychInfo, Medline, EmBase and CinNahl (see Additional file 1). Two researchers conducted the initial search by independently examining titles and abstracts. Full texts were retrieved for potentially relevant studies and these were assessed. A third researcher was consulted and reviewed texts in the case of disagreement. An independent review by a fourth researcher was undertaken on all full texts of the final included articles.

As this is the first systematic review undertaken to collate the existing evidence on interventions, the search strategy used a broad brush approach using overarching terms/keywords (Quality of Care’ and ‘Hospital’). Medical Subject Headings (MESH) terms were used in databases where appropriate. The use of overarching terms/keywords ensured that all potentially relevant articles were included in the initial screening. In all databases, the search was restricted to articles where the keywords were the major focus of the article.

This search returned (n = 17,730) articles. Following duplicate removal, (n = 13,195) articles remained for screening. Included articles had to meet the following criteria:

To minimise bias, the above criteria were applied in a structured way to 13,195 articles. This screening process resulted in 120 articles which were examined in detail. Seventeen articles met all of the inclusion criteria and were therefore included in this review. Reference mining of the bibliographies of these articles resulted in a further 3 articles which met the inclusion criteria. The total number of articles included in the review was (n = 20). A PRISMA flow diagram summarises this screening process (Figure 1).

The inclusion criteria permitted the inclusion of studies which were heterogeneous in terms of their design and scientific rigour. The Grades of Recommendation, Assessment, Development and Evaluation Working Group (GRADE Working Group) has developed a system for assessing methodological rigour. This approach is encouraged by BMJ and the Cochrane Collaboration [20] which have adopted the principles of GRADE for the evaluation of evidence in systematic reviews [21]. Included studies were therefore analysed using an adapted version of the GRADE criteria which assessed methodological rigour using five criteria:

The GRADE approach specifies four levels of quality (High, Moderate, Low, Very Low). The highest quality rating is for randomized control trials and observational studies general start with a rating of low. However, if observational studies report large effects and there is no apparent bias, studies can be upgraded to moderate. Similarly, studies can be downgraded if there is evidence of bias or inconsistency.

This systematic review has established that there is a very large volume of literature (n = 13,195) in the area of quality of care interventions published over the last ten years. However, the rigorous application of inclusion criteria in this study has identified a dearth of hospital based interventions at the scientific level. This systematic review identified just (n = 20) studies for inclusion in the review (see Table 1). The selected studies were heterogeneous in terms of their design and scientific rigour. The GRADE approach confirmed that there is a lack of high quality interventions (n = 1) to assess quality of care. Most of the included studies were classified as moderate quality (n = 9), low (n = 8) or very low (n = 2) (see Table 2) according to the GRADE criteria.

Details of all studies (n = 20) included in the review and a summary of the data abstracted are displayed in the data was extracted using the PICO approach. The majority of studies were described as pre/post design (n = 13) and the remaining studies used a phased design (n = 4), observational design (n = 1), time series cohort (n = 1) or randomised controlled trial (n = 1). Included interventions review varied in scale from small scale improvements for specific patient groups in individual settings to large scale quality improvement programmes across multiple settings. As discussed earlier, studies were grouped into two categories: Interpersonal and Technical.

Implementation of interventions was assessed by establishing who carried out the intervention. While the participant groups involved in these interventions varied, interpersonal interventions were predominantly implemented or carried out by nursing staff. Four interventions were led exclusively by nursing staff [23, 24, 26, 28] while one intervention was implemented by midwives and physicians [30]. Quality of care outcomes from interventions implemented by nurses were measured using patient satisfaction questionnaires [23, 24, 26, 28, 30].

Clinicians implemented the interventions in two studies [25, 27]. In these studies, quality of care outcomes were measured using nurse and family satisfaction questionnaires [27] and a Diabetes Attitude Scale (DAS-3) [25]. In contrast to the majority of interpersonal interventions, two studies differed in terms of participants with one study involving patients and childbirth companions [29] while the second study assessed the impact of building restructuring on quality of care [22]. Despite having the differences, both studies [22][29] measured quality outcomes used patient satisfaction questionnaires.

Interpersonal interventions were described as having either a multifaceted structure involving several components or having one central component i.e. an educational intervention. However, two intervention studies [22, 23] were unique in that they sought to improve quality of care by altering the physical structure of the hospitals. Both studies sought to improve quality of care by integrating separate oncology clinics into one unit within the hospital.

Three studies implemented interventions which were multifaceted in design [27, 28, 30]. These interventions sought to improve various aspects of maternity care [30], postnatal care [28] and palliative care [27]. Similar approaches were adopted by the interventions which sought to improve maternity care [28] and postnatal care [30]. In both studies, baseline data was collected to identify patient needs and a medical team then redesigned care processes based on those needs.

The study examining postnatal care encouraged parental self-efficacy by providing ‘one to one’ time with the midwife each day so that the woman could discuss her concerns and gain knowledge [28]. The study [30] which sought to improve childbirth implemented a care model based on five criteria (availability of resources, the physical environment of the maternity ward, clinical experience and culture and correspondence with women’s needs and requirements). Similarly, a multifaceted interdisciplinary intervention to improve palliative care identified five key components of an effective intervention-clinician education, local champions, academic detailing, feedback to clinicians and system support [27]. This intervention was based on the theory of self efficacy and it was hypothesised that changes in attitudes, behaviour and knowledge of clinicians would improve palliative care.

Educational workshops or training for staff were a central component in four of the interpersonal intervention studies [24‐26, 29]. The aim of the training/education in all cases was to increase knowledge so that the delivery of care and care processes could be improved. Staff who received training/education included maternity staff [29], emergency department staff [26], nursing staff [24] and doctors [25].

Reported quality of care outcomes or improvements were varied across the nine interpersonal studies [22‐30]. An educational intervention targeted at doctors to improve diabetes care reported improvements in the knowledge, attitude and clinical management of diabetic patients [25] while the improvement in patient satisfaction was statistically significant in only one of the educational interventions [26]. In contrast, patient satisfaction approached but did not reach statistical significant in an intervention which sought to improve teamwork and staff engagement although, reduced staff turnover, improved teamwork and lower patient fall rates were reported. [24].

Multifaceted interventions reported improvements in care for women during childbirth [30]. In a study which aimed to improve palliative care, nurse satisfaction improved but family satisfaction did not reach statistical significance [27]. A multifaceted approach to improve postnatal care reported no significant differences post intervention in perceived quality of care [28] however it was reported that, the key strategy of ‘one to one time’ for patients had not been implemented consistently. Similarly, an educational intervention to encourage the presence of childbirth companions found no significant difference in patient satisfaction or humanity of care based on whether a companion was allowed by nursing staff [29]. In summary, interpersonal interventions tended to be focused on patient satisfaction and were implemented by nursing staff. They also tended to be multifaceted or involve education/training. Most reported some improvements in patient satisfaction but not all findings reached statistical significance.

Five technical interventions set up a team or panel of experts prior to the intervention [31‐34, 38]. Teams tended to be multidisciplinary and had the task of setting goals and reaching consensus on quality indicators prior to intervention. Five technical interventions were implemented by physicians [36, 37, 39‐41], while in one study [35] the intervention was implemented by both nurses and physicians.

Technical interventions tended to involve a number of interconnecting components [31‐34, 38]. A multifaceted intervention [31] sought to improve pneumonia care took place in multiple centres although the data collection was predominantly hospital based. Three studies implemented quality improvement programs which aimed to improve hypertension care [38] and care of patients with heart diseases [34, 35]. Similarly, two studies implemented multifaceted interventions but these interventions were part of state-wide initiatives including the ‘Pneumonia Pathway Project’ [32] and the ‘Guidelines Applied in Practice’ GAP initiative to improve care of patients with myocardial infarction [33]. Four of the technical interventions had structural similarities in that they were all implemented by physicians and sought to alter care processes [36, 37, 39, 41]. One intervention altered care processes for diabetes patients by implementing a diabetes education workshop for doctors [39]. Another [40] intervention sought to improve evidence based practice for women during labour by implementing workshops for obstetric practices for staff. Interventions implemented by physicians included the addition of a tobacco order set to an existing computerized order entry [41], the appointment of a new neurointensivist team to an intensive care unit [37] and the adoption of myocardial infarction guidelines [36].

Multifaceted interventions reported improvements in quality of care with an absolute improvement in blood pressure control in a study to improve hypertension [38]. Three multifaceted interventions [33‐35] aimed to improve quality of care for patients with heart disease and reported improved medical outcomes including hospital administration of key treatments such as aspirin at admission [33] and improvement in angiotensin [35]. An intervention [34] to improve acute coronary care reported improvement of key quality indicators including timeliness of treatment but found no significant change in the proportion of patients accessing treatments such as antiplatlet agents or undergoing coronary angiography.

Technical interventions [31, 32] which sought to improve pneumonia care reported some quality of care improvements with an increase in the use of guideline recommended antimicrobial therapy [31] and antibiotic administration within eight hours [32]. However, it was reported that there was no significant improvement in indicators such as timeliness and patient education in one study [31] and no significant improvement in indicators such as thirty day mortality and thirty day readmission in the other study [32].

Interventions implemented solely by physicians reported quality of care improvements. An increased number of patients accessed NRT or smoking counsellors after a computerised order entry form introduced for use by doctors [41]. Care improved for diabetes patients as a result of a physician education programme [39] and medical outcomes for women during childbirth improved as a result of an educational programme on obstetric practices for staff [40]. Mortality outcomes for patients in intensive care improved following the appointment of a neurointensivist [37] and quality measures for heart diseases improved after a multidisciplinary programme was implemented [36]. In summary, technical interventions were mainly implemented by physicians and concentrated on improving care for patients with specific conditions such as heart disease or pneumonia. Multidisciplinary panels of experts were formed to set goals and reach consensus on quality indicators prior to intervention. Technical interventions tended to achieve improved medical outcomes for patients with specific illnesses.

The results of this review must be interpreted with caution. As this was the first systematic review of its kind, a broad reaching search strategy was necessary in order to capture all potentially relevant studies. One of the disadvantages of this search strategy was that studies of heterogeneous design were included which resulted in the use of a modified version of the GRADE criteria for quality assessment.

The inclusion of studies of varied design and scientific restricted us to presenting a broad assessment or overview of studies. Different approaches were explored for presenting the studies in a meaningful way. While interventions in the main focused on improving either technical or interpersonal aspects of care, there was overlap with some interventions seeking to improve patient satisfaction along with medical outcomes. However, interventions categorised as technical reported primary medical outcomes and interpersonal interventions reported interpersonal primary outcomes.

The majority of interventions included in this review were pre/post design. Results of any before and after study must be interpreted with caution. In hospital based pre-post interventions, it is often the case that participants at time one and time two differ and this can have the effect of diluting the intervention effects. Improved outcomes reported in the pre/post designed intervention studies may have several possible explanations including secular or temporal trends. The most effective method of overcoming this possibility is to use a randomized controlled trial (RCT). However, using RCTs is difficult when implementing complex interventions involving multiple components since it is not possible to ‘blind’ providers or recipients to the control and intervention groups and it is also difficult to establish which components of a complex intervention worked and which did not.