Effectiveness of a Psychosocial Care Quality Improvement Strategy to Address Quality of Life in Patients With Cancer

Key Points Question Does an implementation strategy integrating psychosocial care into cancer centers improve at least 1 of 2 domains (emotional or social function) of health-related quality of life? Findings This stepped-wedge cluster randomized clinical trial included 762 patients with cancer at high risk for reduced quality of life. Health-related quality of life score improved from baseline to 3 months among participants in the Humanization in Cancer Care Quality Improvement Strategy arm vs the usual care arm, showing a statistically significant difference for emotional function, but not for social function. Meaning This study suggests the use of an implementation strategy aiming to provide routine psychosocial care in cancer centers may be beneficial to patients; further investigation is required on factors that can maximize its effects.


Introduction
Individuals with cancer experience a wide range of psychosocial health needs, encompassing mental, emotional, social, and spiritual aspects of health. 1 Frequent practical difficulties and information needs add to the cancer burden. 1,2 These problems are associated with a decline in quality of life over time and thus should be the target of comprehensive care. 3,4 To address these needs, a wide range of psychosocial interventions are available, including any activity aimed at ameliorating or reducing the influence of cancer on mental health and at improving patients' skills to cope with the demands of treatment and uncertainty of the disease outcome across the whole spectrum, from prediagnosis to palliative care and survivorship. 5 Despite numerous clinical practice guidelines, many patients who might benefit from psychosocial interventions do not receive them. 3,5,6 A survey of the International Federation of Psycho-oncology Societies 7 including 25 countries concluded that the development and implementation of psycho-oncology was fragmented and undeveloped. In addition, in Europe, Italy in particular, these services were often regarded as not essential. Barriers to implementation may be related to personal characteristics of health care professionals, as well as to environmental and organizational factors. 8,9 Several of these barriers are modifiable, 8 particularly using implementation strategies. 10-12 These strategies should be tailored to potential obstacles, 13 should be feasible, and should be effective. 14,15 These considerations formed the basis for the Humanization in Cancer Care (HuCare) Quality Improvement Strategy (HQIS) aiming to integrate into practice 6 psychosocial interventions recommended by national and international guidelines 1,16-18 selected by a multidisciplinary task force. 19 Following the process for the development and evaluation of complex interventions of the Medical Research Council, 20 the HuCare Study, 19 conducted from 2008 to 2014, described the feasibility of the HQIS in 28 Italian cancer centers, obtaining a high level of adherence to all interventions (>85%). We then conducted this randomized clinical trial (HuCare2) to evaluate the effectiveness of the HQIS vs standard care in terms of improvement of health-related quality of life (HRQOL). and written informed consent was provided by all patients. This study followed the Consolidated

Standards of Reporting Trials (CONSORT) reporting guideline and its extensions (Patient Reported
Outcomes Extension and Extension to Cluster Randomized Trials). 24 Center recruitment was performed using a brief survey to ascertain motivation and the presence of prerequisites (eg, staff willingness to attend residential training, availability of a room dedicated to encounters with patients to provide information, and availability of a psychologist in the ward). The number of centers was determined according to feasibility, ensuring representativeness of size (number of beds) and geographic location (north, center-south, islands). Centers that had participated in the HuCare feasibility study were not eligible for this trial.
Before trial initiation, a pilot study was performed at the Cremona cancer center on a consecutive sample of 13 patients who accessed the facility over 2 weeks to measure the feasibility and acceptability of questionnaire administration using a computer tablet. The findings of this pilot study led to changes in the protocol, including removal of the upper age limit, extension to hormonal therapy and immunotherapy, and exclusion of patients enrolled in other trials, implying the use of patient-reported outcomes.
Eligible individuals were outpatients older than 18 years with cancer of any type and stage who were consecutively accessed in the center during a 2-week index period, had received the diagnosis within the previous 2 months, were about to start a new medical cancer treatment, had expected survival of more than 3 months, had good comprehension of the Italian language, and had signed the informed consent form. We excluded patients who had received previous chemotherapy or other medical cancer treatment; were recruited in a previous epoch of the study; were currently participating in other trials, implying the completion of patient-reported outcomes; were hospitalized; were receiving psychiatric treatment; were affected by mental or psychiatric disorders due to cancer or coexisting illness that were interfering with the state of consciousness or impeding judgment; or were unable to complete the questionnaire or ensure participation in the 3-month follow-up.

Intervention and Outcomes
The HQIS has been described elsewhere. 19 It consists of the provision of support activities to the centers aiming to introduce 6 evidence-based psychosocial recommendations-1 targeting clinicians and 5 targeting patients. 21 The strategy lasts 16 weeks and comprises 3 phases (Supplement 1). In phase 1, medical and nursing staff attend communication skills training designed according to literature-reported indications (recommendation 1). In phase 2, center support is provided by the improvement team, composed of external personnel (sociologist, psychologist, and research nurse), during on-site visits. According to the plan-do-study-act cycle, 25  The primary outcome was the difference between the means of changes of individual scores in emotional function (EF) or social function (SF) of HRQOL detected at baseline and 3-month follow-up (within each group) and during the postintervention epoch compared with control periods (between groups). These 2 functions were chosen because they are mostly affected by psychosocial interventions, following evidence-based guidelines. 26 The corresponding timing (ie, time of effect measurement) was selected because, during periods of active treatment, improvement from baseline is more likely to be observed. 26  of Life Questionnaire, version 3 (EORTC QLQ-C30). 27 The questionnaire was self-administered using a touch-screen tablet at baseline and 3 months after enrollment.
The secondary aim was to investigate whether the strategy had an effect on mood disorders, defined as a cutoff score greater than 7 on the Hospital Anxiety and Depression Scale-Depression (HADS-D) scale 28,29 over the long-term (up to 1 year), overall HRQOL, and specific domains.
Subgroup analyses were performed to obtain information about treatment effect modifiers and subsequently define how to implement more targeted interventions. Process indicators were also measured: frequency of clinical staff (oncologists and nurses) who completed training and proportion of patients who, at 3-month follow-up, experienced a reduction in their baseline unmet needs detected with the Needs Evaluation Questionnaire, a self-administered instrument with 23 dichotomous items divided into 5 areas. [30][31][32][33] Statistical Analysis The number of participants was defined following the methods for incomplete, cross-sectional, stepped-wedge cluster randomized trials, [34][35][36][37] considering (1)  Patients entered data anonymously using tablet computers that included control checks. For the principal analysis of effectiveness, we considered an intention-to-treat population, composed of all clusters according to randomization and all eligible patients with HRQOL assessments at baseline and 3-month follow-up. Differences of HRQOL values between the 2 groups were analyzed using a binomial β regression model, as suggested by different authors, 42,43 owing to the asymmetric value distribution. This model also enabled an estimate of the strategy's effect in terms of odds ratio (OR), the preferred measure by oncologists for its more immediate interpretation and greater usefulness in clinical practice compared with absolute values. 42 For the binomial β model analysis, responses were transformed into a scale (0, 1) by using the formula Y-a / b-a, where a is the lowest and b is the highest possible score and Y is the observed response. Precision coefficient was measured using the Log-phi parameter. The demographic and clinical variables that correlated with the outcome at P < .20 in the univariate analysis were included in the regression model.
Because the study was a cross-sectional, stepped-wedge cluster randomized clinical trial (with an implementation period), the final model was adjusted extending the basic Hussey and Hughes model 34,44 to include a fixed interaction between intervention status and implementation period (first, second, or third).
Concerning the secondary outcomes, intention-to-treat analyses were performed. First, longterm variables were entered into the binomial β model as covariates in the model for the principal analysis. Second, the EORTC QLQ-C30 global scales and the other domain scales were represented as observed response Y in the binomial β model, and the covariates were those used in the primary analysis. Third, to examine whether the HQIS had any effect on mood disorders, general linear modeling with findings significant at P < .05 was used.

JAMA Network Open | Oncology
To assess whether the intervention modified associations of the primary outcome measure with candidate variables (subgroup analysis), the binomial β model was performed separately for each predictor. In these analyses, the difference in HRQOL values between preintervention and postintervention scores served as the dependent or outcome variables. The interaction terms of each candidate predictor with the randomized treatment groups dummy (HQIS intervention vs control) served as the independent predictor, along with constituent main effects.
Concerning reduction of unmet needs (process indicator), the χ2 test was used. Data were processed with SAS, software STATA/SE version 11.0 (StataCorp LLC) and R-cran, version 3.5.3 (R Foundation for Statistical Computing).

Participants and Centers
Cluster, center, and participant characteristics are provided in Figure 1

Outcomes
Data analysis was conducted on 647 patients (332 HQIS and 315 usual care) without missing data.
The HRQOL score improved by any amount from baseline to 3 months among more participants in the HQIS arm than in the usual care arm both for EF (45% vs 37%), and for SF (22% vs 20%), but the score worsened among fewer patients for EF (HQIS, 35% vs usual care, 41%) (Figure 2, eTable 1 in Results for the primary outcome are reported in ) and a greater reduction in the HQIS arm at 3-month follow-up was recorded in all 5 need areas ( Table 3). Regarding the latter indicator, the difference was always statistically significant, with the OR at least 2 (ie, patients in the intervention group were twice as likely to have their needs met with compared with patients in the usual care group) ( Table 3).

Discussion
To our knowledge, this randomized clinical trial is the first report that a quality improvement strategy aimed at integrating evidence-based psychosocial care interventions into practice significantly  These positive results support the need for strategies to introduce psychosocial care capable of addressing the multiple obstacles and barriers that may hinder implementation. 3,9 In particular, our study demonstrates the effectiveness of a system-based approach, which implies organizational change requiring collaboration and commitment across hospital departments, disciplines, and individual clinicians. 45 Despite the potential advantages of system-based interventions, the review by Sanson-Fisher et al 45 of publications in 5 top-ranking journals in the field found that only the HuCare feasibility study 19 met all 4 criteria for evaluating system-based change. Furthermore, we confirm the conclusions of a recent survey of 102 German oncologists 46 in which physicians' personal commitment to psycho-oncology was related to the integration of psycho-oncological aspects into patient treatment. The authors emphasize the importance of specific communication skills training, which was a central part of the HQIS. In the same survey, the perceived reluctance of patients to discuss their distress or patients' assertion of not being distressed was seen by oncologists as a reason not to further pursue the topic. In this view, the HQIS comprised interventions aimed at encouraging patients to communicate openly with clinicians, build a trust relationship with them, and obtain needed information.

Limitations
The trial has limitations. Although statistically significant, the difference in the mean change in HRQOL between the 2 groups was trivial-lower than the 3-point value that has been defined as the minimal important difference. 38  A noteworthy finding of our trial is the lack of a statistically significant difference for SF.
However, this domain is investigated by only 2 questions on the EORTC QLQ-C30. A general tool was preferred for this trial because it applied to patients with any cancer type. 21 However, the use of an illness-specific instrument may have yielded a different result, as suggested by the findings of the aforementioned Cochrane review, 49 in which a small but significant positive result was observed in the data using illness-specific measures.
Two additional elements emerging from our findings deserve consideration. First, the striking association between HQIS and the reduction of patient needs compared with usual care, although only a secondary outcome reported descriptively, suggests that the strategy may prove to be potent for helping patients cope with cancer and its treatment, with important implications for clinicians.
Second, the statistically significant lower effect on EF in cluster 3 with respect to the other clusters may be due to local (cultural, organizational, and social) barriers, which should be further explored.

Conclusions
Although our findings are positive, further research is needed. First, participation of different institution types nationwide ensures generalizability to Italian hospitals, but the effect of the HQIS in other contexts should be tested, because health care system characteristics are potential predictors of psychosocial outcomes. 45 Second, it would be important to evaluate the effect in the longer term (3-5 years after implementation), investigating the need for strategy refreshes, as well as the longer-term effect on patient EF with a 6-and 12-month follow-up. Third, inclusion of in-depth interviews with participants (health professionals and patients) in studies of this kind may assist in defining core aspects of the intervention, which are considered beneficial, and in understanding factors underpinning acceptability.
Evidence-based psychosocial care is important for all patients. The strategy tested in this trial is feasible and has the potential to improve patient outcomes. Further research will allow refinement of the HQIS, enabling us to understand the factors that can optimize its effects.