A nationwide participatory programme to measure person‐centred hospital care in Italy: Results and implications for continuous improvement

Abstract Background Patient‐centredness has been targeted by the Italian government as a key theme for the future development of health services. Objective Measuring patient‐centred health services in partnership with citizens, health professionals and decision makers. Design National participatory survey in a large test set of hospitals at national level. Setting and participants A total of 387 hospital visits conducted in 16 Italian regions by over 1,500 citizens and health professionals during 2017‐2018. Main variables and outcome measures An ad hoc checklist was used to assess person‐centredness in hospital care through 243 items, grouped in 4 main areas, 12 sub‐areas and 29 person‐centred criteria (scored 0‐10). GEE linear multivariate regression was used to explore the relation between hospital characteristics and person‐centredness. Results Person‐centred scores were moderately high, with substantial variation overall (median score: 7.0, range: 3.2‐9.5) and by area (Care Processes: 6.8, 2.0‐9.8; Access: 7.4, 2.7‐9.7; Transparency: 6.7, 3.4‐9.5 and Relationship: 7.3, 0.8‐10.0). Multivariate regression found higher scores for increasing volumes of activity (quartile increase: +0.21; 95% CI: 0.13, 0.29) and lower scores in the south and islands (−1.03; −1.62,‐0.45). Discussion The checklist has been applied successfully by over 1,500 collaborators who assessed hospitals in 16 distinct Regions and Autonomous Provinces of Italy. Despite an overall positive mark, all scores were highly variable by location and hospital characteristics. Conclusion and patient or public contribution A national participatory programme to improve patient‐centredness in Italian hospitals highlighted critical areas with the direct input of citizens.


| INTRODUC TI ON
The goal of person-centredness has become increasingly popular in the organization of health services.
Since its initial definition as one of the six main pillars of quality of care, 1 various concepts and models have been introduced to assess it routinely, [2][3][4][5][6][7][8][9] using specific tools to monitor the personal domains of physical, psychological and social needs in primary, secondary and tertiary clinical settings. 10,11 Relevant experiences addressed the importance of citizen involvement in the direct evaluation of services, embracing the concept of end-user co-design, particularly in hospitals. [12][13][14][15][16][17][18] Following these experimental initiatives, international organizations have recently recognized the role of person-centredness as a key driver for the sustainability of health systems.
In 2016, the WHO called on Member States to promote an approach to care that 'consciously adopts individuals', carers', families' and communities' perspectives as participants in, and beneficiaries of, trusted health systems that are organized around the comprehensive needs of people'. 19,20 At the same time, the OECD released a new framework for performance evaluation including patient-centredness as one of the tree key dimensions of quality, 21 presented as the main theme in the Ministerial Conference where Member States agreed that 'healthcare systems need to engage patients as active players in improving health care'. 22 Relevant developments took place in parallel to strengthen the Italian National Health System (Sistema Sanitario Nazionale, SSN).
In 2012, the Italian Ministry of Health and the Regions and Autonomous Provinces agreed common terms for the accreditation of health-care facilities, incorporating patient-centred care as an essential quality criterion. 23 Two years later, building upon the OECD recommendation of increasing the direct participation of citizens in quality assurance, 24  The agreement included the implementation of specific interventions to foster patient-centred care across the country in a balanced way, calling upon the National Agency for Regional Health Services (AGENAS) to develop a set of core indicators to monitor the results of these interventions. At the same time, relevant activities of patient involvement were carried out by different Regions and Autonomous Provinces. These developments increased the need of improving the comparability of person-centred care at national level. 25 Consequently, AGENAS defined a set of materials and protocols to undertake a national survey on person-centredness in Italian hospitals.
In this paper, we present the results of this activity, focussing on the following research questions: • Can we measure person-centredness at hospital level using the same standardised tool across the country, with the active participation of citizens, health professionals and decision makers?
• Which features of person-centredness are widely applied, and which others deserve increased attention? Is there any significant variation across the country and/or potential association with hospital characteristics?

| MATERIAL S AND ME THODS
The study was carried out between 2016 and 2018 in the context of a multi-year programme financed with infrastructural funds available from the general mandate assigned by the Ministry of Health to AGENAS.

| Governance of the programme
The project stems from a collaboration started in 2011, when AGENAS established a Project Team, an Advisory Board and the Regional Network of experts to agree on a common definition of patient centredness.
The Project Team formed at AGENAS included authors of this paper with a multidisciplinary background in medicine, public health, psychology, sociology and biostatistics. The role of the Project Team was to coordinate the conduct of a national survey to measure person-centredness in Italian hospitals through the use of a standardized assessment tool. For executing the project, the Project Team cooperated with the Regional Network, including 34 representatives of Regions and Autonomous Provinces involved in activities of patient empowerment, and the Advisory Board, including experts in the field of civic evaluation of health services from the non-profit consumer organization 'Cittadinanzattiva'. The lists of members of the Advisory Board and Regional Network are included in the acknowledgements section together with their affiliations.
In August 2011, the Project Team reviewed the scientific and grey literature to identify the main documents reporting on patientcentredness in health care from the point of view of care providers, research organizations, citizens and patient organizations at national and international level. Relevant regional and national Italian regulations were also considered for the scope. The documents were discussed with members of the Advisory Board, who provided guidance on their use in the context of the civic evaluation of quality in hospital care. We found substantial heterogeneity in the terminology and interpretation of the concept of 'centredness', reflecting the background of different professional disciplines, perspectives and clinical settings, in the context of specific regional settings. 26 Consequently, we adopted a holistic vision 10 to define personcentredness as 'the commitment to orient the setting of care, diagnostic and therapeutic programmes as much as possible towards the person, considered in all inherent physical, social and psychological aspects'. 27

| Construction of the survey tool
The structure of the survey tool was defined between October 2011 and June 2012 as a 'checklist' aimed at measuring compliance to the stated principles using multiple items, whose values could be conveniently added up to compute summary scores for specific aspects of interest.
The checklist included four major areas of patient centeredness in acute care: person-oriented organizational and care processes, physical accessibility and comfort, access to information and transparency, citizen-patient professional relationship.
Consensus over the final structure was reached after two rounds of comments received from members of the Regional Network in terms of conceptual coherence, coverage of relevant aspects and overall clarity of the terms utilized. The final 'checklist' adopted for the national survey included the above mentioned 4 areas, subdivided into 12 sub-areas, 29 criteria and 243 items. The majority of items were dichotomous, indicating presence/absence of selected characteristics (coded as 0/10), except for cases where the questions allowed ordinal responses (coded from 0 to the number of levels). Missing data were not allowed, except for items classified as 'not applicable'.
An example of items, scores and explanatory notes included in the checklist is shown in Table 1. Briefly, the value assigned to each item was based on the examination of different type of materials: along with additional explanatory notes.
The details of all components of the checklist, including the description of the single items and the range of their possible values, are attached to the present publication as Appendix S1.
We further assessed the level of correlation between criteria, sub-areas and areas included in the checklist to ensure that no redundant items were included (results not shown).

| Survey design
The protocol of the data collection procedure included a series of steps, underpinned by three guiding principles: empowerment of citizens and health professionals, humanization of care and continuous quality improvement (see Figure 1).
Each participating Region/Autonomous Province was requested to form a Regional Coordination Group, including members of the Regional Network, supported by regional referents of hospital managers, professionals and citizens. The Regional Coordination Group was put in charge of enrolling hospitals and coordinating the survey in each region, while hospitals were requested to form a Local Team, including local referents of health professionals and citizens.
The Project Team provided each Regional Coordination Group with standardized training materials including the programme, slides and guides for implementing the protocol. Subsequently, the

| Data collection and statistical analysis
In All summary scores by criteria, area and sub-area were normalized on a numeric scale ranging between 0 and 10, with 0 corresponding to the lack or absence of requirement and 10 to complete fulfilment.
All analyses were stratified by macroregion (north, centre or south), type of hospital (public, trust, private and academic/research) and volume of activities (determined as a proxy by their quartile in the ordered ranking of number of beds).
Descriptive statistics included the calculation of percentages and medians and ranges for variables that were not normally distributed according to the Shapiro-Wilk test. 31 Results were stratified by area, sub-area and overall, according to the categories above. Histograms were used to examine the distribution of scores within and between classes.
Multivariate linear regression was used to formally test the sta- All analyses were carried out by AGENAS using R. 34

| RE SULTS
The main characteristics of the study sample, compared to all Italian hospitals in year 2017, are presented in Table 2. Therefore, we presented all descriptive results consistently as median scores (range) for criteria, sub-areas and overall (see Table 3).

TA B L E 2 General characteristics of the study sample
Overall, hospitals achieved moderate levels of personcentredness, with a median score equal to 7.0 (3.2-9.5).
In terms of criteria, the maximum median score of 10 was achieved by 'personal anonymity', 'signposting and internal pathways', 'child-friendly wards', 'simplified access to services' and 'reception'. A high mark was also reached for 'taking on commitments towards citizens'.
However, scores for all the above criteria were also quite variable, ranging from 0 to 10. Scores were less variable at the macrolevel of sub-area, which could be explained by the fact that hospitals did not consistently achieve systematically higher or lower scores across different criteria. As the sums involved a higher number of items, results over multiple criteria had a tendency to regress towards the mean. The maximum median score was achieved by 'overall comfort'.
The sub-areas of 'person-oriented wards' and 'relationship with citizens' presented a median at best one mark below the overall median.
The level of correlation between sub-areas showed levels of correlation between 0.09 and 0.64, confirming the relevance of all items included in the checklist (detailed results not shown). Table 4 shows the distribution of area scores for specific categories of hospitals and overall. Median scores achieved for single areas did not differ substantially from the overall median, although 'physical accessibility' achieved a clearly higher score (7.4, 2.7-9.7), immediately followed by 'relationship', which was also very variable (7.3, 0.8-10.0).
For specific categories, we found that northern regions presented consistently higher scores, while hospital trusts and public hospitals were consistently lower. The only exceptions were 'relationships for hospital trusts' (7.9, 0.8-9.6) and 'transparency for public hospitals' (6.8, 4.0-9.5).
Hospitals in the lower quartiles of the distribution of number of beds showed invariably lower scores for all areas and overall. The examination of histograms highlighted a potentially significant linear relation (see Figure 2).
The statistical significance of this relation was formally tested in a series of multivariate linear regression models, whose compliance with fundamental assumptions showed to be problematic (see Table 5). Consistently with the non-normality of the outcome variable, we found that the residuals were almost in all cases not normally distributed and heteroskedastic. Given the large sample, we considered non-normality of the residuals as a minor problem and focussed more on heteroskedasticity. 36 Sensitivity analysis showed that heteroskedasticity could be resolved by excluding the macroregions from the set of covariates. Nevertheless, there was still a moderate degree of autocorrelation between residuals, ranging between 0.27 and 0.45.
Therefore, we used generalized estimating equations as an appropriate means to take into account the intra-regional cluster effect caused by the same Local Team being involved in multiple hospital visits. The values of exchangeable correlation ranging between 0.21 and 0.30 in the GEE models confirmed this assumption. The GEE confidence intervals were considerably wider than those obtained by multivariate linear regression, turning several terms into non-significant.
The results obtained from the application of GEE models are shown in Table 6.
Briefly In terms of volumes of activity, after adjusting for all the above characteristics, we found that the average difference between adjacent quartiles was significantly associated with increased levels of               Overall, an increase of one quartile in terms of hospital volume was associated with an average increase in the score of patientcentredness equal to +0.21 (0.13 to 0.29). In other terms, regardless of the region or type of hospital, the average difference between hospitals in the highest vs lowest quartile of volume of activity is equal to 60% of one mark of patient-centredness out of a scale of ten.

| D ISCUSS I ON
The results of the first nationwide survey coordinated by AGENAS allowed responding to our initial set of research questions.
Regarding measurement of the person-centredness of hospital care, our results suggest the applicability of a novel tool that is appropriate to use in similar settings, for the same purpose.
The checklist was applied successfully by over 1,500 collaborators who assessed hospitals in 16 distinct Regions and Autonomous TA B L E 4 Scores achieved (median, range) for specific categories of hospitals, by centeredness area and overall  Average Scores Achieved Therefore, the derived measures of person-centredness can be also actionable. 21 However, we did not include the details of service improvement plans in the scope of the project. AGENAS is currently engaged in research aimed at defining methods and tools to monitor progress of co-designed improvement plans. We are confident that this step will complete the creation of a National Monitoring System for Person-Centred Hospital Care that will incorporate the participatory process as a key component of the continuous improvement cycle.
Regarding our second question on the differences observed across the country, despite an overall positive mark, we found that all scores were highly variable. Single criteria clearly showed that In 2020, while preparing the present report, the outbreak of

<.001
Results for statistically significant covariates are shown in bold.
Our approach was aimed at testing the significance of any relation, for example the independent association between hospital size and total scores of patient centredness, taking the main potential confounders into account. Further work will be required to collect all relevant characteristics, within improved information infrastructure 60 to collect data in compliance with the most current legislation on privacy and data protection. 61

| CON CLUS ION
We conducted a nationwide participatory programme for the evalu- in the complex scenario of decentralized governance. 52,62 The necessary link with patient safety and co-designed planning may represent a useful learning process for other countries experiencing the same problems.

ACK N OWLED G EM ENTS
The present paper would have not been produced without the Vitale. We thank the following representatives of the R&AP from the Regional Network (RN) and the Regional Coordination Group questionnaire and led the study on behalf of AGENAS. GD and FC conducted the statistical analysis for the paper. FC led the production of the paper, drafting its initial version. All authors revised and completed the production of the manuscript, revising and agreeing on its present contents.

PATI E NT A N D PU B LI C I N VO LV E M E NT
This study has been conducted with the direct participation of policymakers, patients and the public, as described in the section on materials and methods and the acknowledgement section.

D I SS EM I N ATI O N
The results of this study have been included in extended format in a series of interim and final national report, coordinated by AGENAS, which has been circulated among all participants, including relevant patient and public communities, as fully documented in the acknowledgement section.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from AGENAS upon reasonable request.