Patient-centered care in Swiss acute care hospitals: addressing challenges in patient experience measurement and provider profiling

Hospitals are under continuous pressure to enable and provide care that is safe, effective, timely, efficient, equitable and patient-centered. To assess the overall quality of the care they deliver, patient experience measures are commonly used across health care settings, countries and patient populations. While these measures are widely employed, though, their impact on quality improvement remains questionable. The analyses of the dissertation indicates that, in addition to lacking a clear conceptualization of the measured construct (e.g., patient-centered care (PCC) or patient satisfaction), the psychometric properties of the most widely-used and influential quality of care instruments are inadequate to ensure reliable assessment of the target criteria. For example, patient experience questionnaires commonly suffer from ceiling effects, resulting in their failure to differentiate between providers scoring above a certain level. Still, the resulting data are incorporated in the steering mechanisms intended to improve quality of care, and weigh heavily on hospital ranking and profiling systems. Based largely on inadequate rating models, then, health care administrators identify hospitals as positive and negative deviants, i.e., high or low performers. Following the logical principle that quality-based selection will lead to long-term quality improvement, higher-ranked hospitals are selected to provide more services and receive more funding. However, at the questionnaire development level, the basic methodological weaknesses noted above preclude accurate quality measurement. If the instruments used in provider profiling lack the capacity to distinguish meaningfully between providers, it follows that the fairness of decisions based on those instruments’ data is dubious at best.
Besides quality improvement through selection, improvement through change is discussed. Effective measurement of organizational processes and structures illuminated which areas worked well and which could benefit from improvement. Still, studies found that individual staff factors such as communication and relationship building skills clearly improve PCC, the effects of structures and processes at the unit and hospital levels were less clear.
One vital task in quality assessment and improvement is to determine how instruments can be improved to fully assess their underlying constructs. Moreover, with consideration for individual patient preferences, every instrument requires both the sensitivity and the reliability to differentiate meaningfully between levels of quality of care.
This dissertation aims to assess PCC and its association with institution-level structures and processes in Swiss acute care hospitals. Beyond that, its target is to improve PCC measurement by including patient preference ratings. By providing a much-needed frame of reference regarding patient care ratings, thereby increasing between-provider variances to usable levels, this addition is intended to improve the care quality measurement process.
The included studies are embedded in the Matching Registered Nurse Services with Changing Care Demands (MatchRN) study.
The dissertation is organized in seven chapters.
Chapter 1 gives an overall introduction to PCC. Focusing on the definition and conceptualization of PCC, it provides a conceptual model for PCC provision in the hospital setting. With the description of challenges in PCC provision, two improvement pathways – selection and change – are discussed. An overview of the current state of research on PCC in acute care hospitals, its measurement and its associations not only with patient-related, clinical and economic outcomes, but also with health care policy, is presented. The chapter ends by summarizing the gaps in the literature, alongside this dissertation’s contribution to bridging those gaps.
Following this introduction, Chapter 2 describes the dissertation’s aims. The findings of the four component studies are reported Chapter 3 to Chapter 6.
Chapter 3 explains the MatchRN study protocol. It provides a general introduction to the MatchRN study’s background, rationale and aims, design and methodology, including measurement techniques.
Chapter 4 reports on the first of the dissertation’s studies, which describes the level of PCC in Swiss acute care hospitals and its associations with the nursing work environment and rationing of nursing care. In its sample of 2073 patients and 1810 registered nurses in 23 Swiss acute care hospitals with 123 units, patients reported generally high levels of PCC. Based on four items assessing PCC, the large majority reported that they easily understood the nurses (90%) and felt the treatment and care they received were adapted to their situations (91%), four-fifths (82%) received sufficient information, whereas one-third (30%) felt insufficiently involved in treatment and care decisions. Further, the analysis identified PCC-associated structural and process factors. Generalized Linear Mixed Models for analysis, including individual-level patient and nurse data aggregated to the unit level, identified positive associations between PCC and the nurse work environment: higher staffing and resource adequacy was associated with higher levels of all four items, with sufficient information (β 0.638 [95%-CI: 0.30 – 0.98]) and adapted treatment and care (β 0.456 [95%-CI: 0.04 – 0.87]) yielding the highest correlations. Higher leadership ratings were associated both with sufficient information (β 0.403 [95%-CI: 0.03 – 0.77]) and with adapted treatment and care (β 0.462 [95%-CI: 0.04 – 0.88]). Negative associations were found between implicit rationing of nursing care and three PCC dimensions: adapted treatment and care (β -0.912 [95%-CI: -1.50 – -0.33]), sufficient information (β -0.764 [95%-CI: -1.27 – -0.26]) and easy understanding (β -0.781 [95%-CI: -1.41 – -0.15). No associations were found between PCC and adjusted staffing. To improve PCC, the nurses’ work environment and the level of implicit rationing of nursing care should be taken into consideration.
Chapter 5 discusses the need for to consider intra-class correlations (ICCs), i.e., ICC1 (levels of random variation) and ICC2 (measurement error due to “noise”) as prerequisites for provider profiling. For the measurement and comparison of performance (e.g., PCC levels) between providers (e.g., hospitals), “noise” (also referred to as statistical uncertainty, chance or random variation) has to be filtered out to assess “true” variation.
To apply provider profiling, patient survey data (n=1716–1863) assessing patient hospital stay experiences from the MatchRN 2015/2016 data collection were used. To gauge variations between providers and the reliability of current profiling methods, this study used mixed effect models to calculate ICC1 and ICC2 at the unit (n=123) and hospital (n=23) levels. Via analytical approaches including plots, permutation tests, and the application of a 95% confidence interval to the ICC1 value, between-provider variance was examined for all nine patient experience items. While ICC1 values for both unit (0.013 to 0.059 [mean: 0.03]) and hospital levels (0.009 to 0.035 [mean 0.023]) indicated little to no between-provider variability, the ICC2 indicated moderate to good reliability on the unit (0.62 to 0.885 [mean 0.691]) and hospital (0.176 to 0.454 [mean 0.345]) levels. In addition to the low ICC1 values providing a compelling argument against the use of patient experience data as a quality indicator, this analysis emphasizes the benefits of the applied analytical approaches for provider profiling.
As described in Chapter 5‘s study, provider profiling measurements need to be improved regarding their between-profiler variances. Chapter 6 presents the results of an explorative study examining patient preferences as predictor variables of between-provider variance in hospital profiling, while also examining the extent to which hospitals are able to meet patient preferences and needs. This study used data from the second MatchRN data collection (2017/2018), which included a sample of 2159 patients in 142 units in 25 Swiss hospitals. The findings indicate an imbalance between patients’ perceptions of PCC levels and their preferences in all 13 assessed care aspects, i.e., for every tested aspect of care, overall patients’ ratings of their perceived care levels where considerable lower than their overall preference levels. The greatest differences concerned whether patients received detailed information about the side effects of prescribed medications: while 87% of the patients reported this as very important, slightly more than one-third (31.4%) reported always receiving sufficient information. With results such as these, growth targets can be defined and improvement initiatives designed accordingly. Likewise, resources can be optimized to develop and implement improvement strategies where they are most needed. Further, the inclusion of patient preferences yielded readily discernible inter-provider differences regarding PCC performance. Between-provider variances increased in all 13 models incorporating patient preferences as predictor variables: of those 13, the 9 adjusted to test preference variables yielded the highest between-provider variances. This study concludes that patient preferences are important predictor variables, and should be included in assessments both of patient hospital stay experiences and of provider profiling analyses.
Finally, Chapter 7 both synthesizes the major findings of the dissertation’s individual studies and discusses the methodological strengths and limitations of the dissertation as a whole. Moreover, implications for further research, clinical practice and policy are recommended.
Overall, this dissertation highlights six major findings. First, high levels of PCC are apparent in the studied sample. Still, the data highlighted potential for improvement regarding patient involvement in decisions regarding their treatment and care could be identified. As a key element of PCC, patient involvement builds the basis for equal partnerships between clinicians and patients. Second, structure- and process-related factors need to be considered in PCC improvement. While this applies especially strongly to PCC’s correlations first with increased staffing and resource adequacy’s and second with reduced levels of implicit rationing of nursing care, it was also significant regarding leadership’s associations with patient perceptions of PCC delivery. Surprisingly, though, adjusted staffing was not associated in any of the four tested PCC dimensions. Third, provider profiling based on patient experience items such as PCC perceptions currently fail due to lack of between-provider variance. Fourth, to fully assess and examine between-provider variance, both calculated (i.e. ICC1 values) and visualized (i.e. empirical Bayes) analytical approaches should be used. Fifth, to improve measurement of patients’ PCC perceptions, preference ratings have to be included in provider profiling analyses, as they increase between-provider variance to usable levels. Sixth, to assess the extent to which hospitals meet the need for patient-specific care, patient preferences need to be assessed and included in analyses.
This dissertation contributes to the existing literature by providing primary evidence regarding the influence of patient preferences on the measurement of quality of care. Future research should explore further opportunities to improve PCC measurement.