Validation of Psychometric Properties Of The Partners in Health Scale in Heart Failure

1. Introduction

Congestive heart failure (CHF) is a chronic disease. Chronic disease self-management (CDSM) is important for CHF patients for the following two reasons. Firstly, in Australia, at least 50% of population suffer one of these eight chronic conditions namely, heart disease, diabetes mellitus, arthritis and back pains, mental health conditions, asthma, chronic obstructive pulmonary diseases and cancer [1]. CHF in many cases is associated with at least one of the chronic conditions above [2]. Secondly, the prevalence, morbidity, mortality, readmission and resource cost of CHF are on the rise. CHF is resource intensive, it’s disease loads and hospital readmissions continue to rise [3,4]. While empowering patients through chronic disease self-management has sound theoretical basis, the CHF guidelines are complex, there are unanswered questions and evidence gaps [5].

An important theoretical framework Wagners ‘Chronic Care Model’ [6] published several decades ago, identified two key domains and six key elements: the health system domain requires organised health services, delivery system redesign, clinical information systems and decision support; and a community domain requiring community linkages resources and policies to enhance self-management support. Cameron et.al [7] in 2009 established from 14 CHF specific instruments, that were promoted in best practice, for the provision of disease-specific measures of CHF self-care behaviours; psychometric analysis revealed that only 2 tools were reliable and valid to specifically measure CHF self-management. In the ensuing decades publications have covered a range of topics from theory, practical management, and consensus guidelines, with optimism [8,9,10,11,12]. Nonetheless, trial results were not forthcoming, and the American Heart Foundation demoted self-management in CHF from a performance to quality measure [10], which has filtered into the guidelines [3,4,5]. Thus, the purpose of this paper is to create a focus on generic CDSM programs and build the evidence from this platform. The program in review is the Flinders Program of Chronic condition Self-management (CFPI), a gold standard tool for CDSM. The CFPI has been validated, published data support good validity and internal consistency, and it is also utilised clinically across a range of conditions [1,13,14,15,16,17,18,19,20,21,22,23] however not CHF. The basis of its theoretical foundation was published in 2003, where the Partners in Health Scale (PIH), a patient reported outcome tool, using 12 questions and over 4 domains, assess patients’ skills (knowledge, partnership, recognising and managing, coping) on symptoms and condition. This paper uses Battersby’s et.al definition of self-management as ‘the active involvement of the patient in the management of their chronic medical condition’ [14].

In this study, we aim to assess the psychometric properties (internal reliability and construct validity) of the Partners in Health (PIH) scale, a chronic condition self-management instrument to assess baseline self-management behaviours in a CHF cohort. For this syndrome, the literature supports the observation that validated CHF specific CDSM programs are few and, comorbid conditions are frequent in CHF [5,24]; thus, assessing both disease specific and generic CDSM tools could be important, and for the generic Flinders Program of CDSM, validating the PIH scale is an important first step in utilizing this tool.

2. Methods

The study data is from a subanalysis of the SELFMANagement in Heart Failure Study (SELFMAN-HF) study, as described in the protocol [24].

2.1. Design

A clinical audit was performed on 210 consecutive patients screened for the SELFMAN-HF study, assessing the efficacy of CDSM programs, in patients commencing Sodium Glucose Co-transport 2 inhibitor (SGLT-2i) for CHF. As a routine for chronic disease management, all patients completed the PIH tool at baseline. The cohort included patients with heart failure with reduced ejection fraction (HFrEF) and heart failure with preserved ejection fraction (HFpEF). The study team, and expert panel designed, oversaw the conduct and analysis of results. The analyses were conducted by 3 investigators (PI, FH, DS).

2.3. Participants

For this audit eligible patients were aged over 18 years who, were screened for use of SGLT-2i within May 2022 and Jan 2024, with a clinical diagnosis of CHF, were eligible for an SGLT-2i and had completed a PIH as routine clinical work-up and care. Patients screened were predominantly HFrEF cohort, aged over 18 years started SGLT-2i within 6 months - for systolic CHF, echocardiographic left ventricular ejection fraction (LVEF) <40%. Patients screened with LVEF > 40%, who completed with PIH scale at baseline were not enrolled into the prospective SELFMAN-HF study, however baseline data were used to form this psychometric audit analysis. Patients were excluded, should concerns be raised by any medical staff; has a life expectancy of ≤6 months whilst receiving palliative or nursing home care; have a significant cognitive impairment; or started SGLT2-i for more the >6 months [24].

2.4. Sample Size Calculations

The effect size for the introduction of SGLT-2i in a community cardiology population with treated CHF is not known. The study design chose >80 patients with LVEF <40% and New York Heart Association (NYHA) class 2 or 3. The documented event rates in similar settings for readmission and mortality at 1 month and 1 year were 25%. Thus, a follow up at 12 months would see >100 admissions and >10 mortalities. Interim analysis at 3-6 months, and the lower event rates the sample size was increased to the documented number.

2.5. Trial Instruments and Procedures

2.5.1. CFPI Program and PIH Tools

The CFPI is a generic and comprehensive CDSM program that utilises various tools (questionnaires) to obtain a patient centred, comprehensive and flexible program of care for patients. This tool is gold-standard and is on par with other programs in achieving its outcomes, i.e., assess self-management understanding and goals, and from this understanding can tailor education to achieve self-efficacy for managing chronic disease. The CFPI program extract patient reported outcomes (PRO) and health services data along 4 part [13] 1) The Partners in Health Scale (PIH) is a self-rated questionnaire for the patient to assess their self-management knowledge, attitudes, behaviours and impacts of their chronic condition. 2) The PIH, using 12 questions and 8-point Likert scale, to score patient on 4 self-management domains, knowledge, partnership in treatment, recognition and management of symptoms and coping; 3) the additional parts are the cue and response, problems and goals and chronic condition care plan, 4) CFPI are detailed in these references [1,13,14,15,24,25,26,27,28,29,30,31].

2.5.2. Data Collection

The PIH scale was completed by all patients screened at the heart failure clinic. Forms with done in the waiting room, the doctors’, or heart failure nurse office. Patients were offered guidance if they had difficulty interpreting the study and asked for assistance. The PIH use Likert-type scales (an 8-point rating scale), which allow change and progress to be measured and recorded during reviews. The PIH is provided to the patient in the waiting room or the consulting room. Time is provided to fill in the scale. Patients are provided assistance if they are not able to understand or if they have carer or partner to help.

2.5.3. Ethical Considerations

This study has been approved by the St Vincent’s ethics committee (approval no. LRR 177/21). All participants will complete a written informed consent prior to enrolment in the study.

2.5.4. Statistical Aspects and Data Analysis

We investigated the convergent and discriminant validity of PIH using Bayesian Confirmatory Factor Analysis (BCFA). This approach is a compromise between maximum likelihood (ML) CFA and exploratory factor analysis (EFA). In a single step analysis, BCFA enables the specification of the prior hypothesized major factor patterns as well as informative small-variance priors for cross-loadings and residual covariances. The cross-loadings model the relationships between PIH indicators and nontarget factors, and residual covariances model shared sources of influence on the PIH indicators that are unrelated to the factors. A strategic use of such priors facilitates a more substantive interpretation of model parameters in context of the current study population. This approach also minimizes the capitalization on chance that otherwise may occur through a sequence of model modifications. Statistical inferences about parameter estimates and model fit are made from a posterior distribution. A BCFA is like a standard CFA but does not rely on large-sample theory and performs better with small samples compared to ML algorithms. In other words, the Bayesian framework does not require traditional model identification to estimate model parameters. Previous studies have shown that using prior information (e.g., small variance priors for cross loadings and residual covariances) in a Bayesian estimation framework provide reliable estimates for small study samples (e.g., N=20; N =40).

Bayesian confirmatory factor analysis was first used to estimate the PIH-CFA model with zero cross-loadings. We then repeated the BCFA analysis, firstly with the addition of small, informative priors for cross-loadings and then with additional non-zero priors for the residual correlations. The observed variables were standardized in accordance with the standardized priors. To identify a suitable cross-loading prior variance a range of increasing values (0.001, 0.005, 0.01, 0.15, 0.20, and 0.30) were initially tested. The aim was to find a value that made a substantial reduction in the 95% confidence interval for the difference between the observed and the replicated chi-square values relative to the original CFA model without cross-loadings whilst not sacrificing speed of convergence [32,33,34,35,36,37,38,39,40,41,42].

The Inverse-Wishart (IW) distribution was then used for the addition of residual correlations to the model. This is the conventional prior distribution for covariance matrices in Bayesian analysis [32]. Priors were set for IW(dD, d) where d is the degrees of freedom of the distribution and D is a diagonal matrix comprised of residual variances from the BCFA model [32,33]. The starting value for the degrees of freedom parameter was set at d =50 and then varied to find the largest value that generated a posterior predictive p-value (PP p) greater than 0.05 to show acceptable BCFA model fit and at the same time being close to the CFA model. This was to enable the identification of the smallest possible modifications to address CFA model misfit [32]. BSEM does not require the assumptions of normality and is independent of large sample theory [33].

Bayesian estimation was comprised of 8 independent Markov chain Monte Carlo (MCMC) and two processors. The minimum number of iterations was set at 15,000 and maximum of 100,000 and monitored using the potential scale reduction (PSR) criterion where values less than 1.1 provide evidence for convergence [34]. To evaluate BCFA models, Bayesian versions of approximate fit indices root mean square error of approximation (RMSEA), confirmatory fit index (CFI) and Tucker-Lewis index (TLI) were used. The suggested cut off values for reasonably well-fitting models are: RMSEA<0:06, TLI > 0:95, CFI > 0:95. The Discrepancy Information Criterion (DIC) was also used to compare models as it takes into account model complexity via the estimated number of parameters or effective number of parameters [32,35]. Lower values are indicative of a better fitting model. To assess the internal reliability of PIH raw scores, McDonald’s omega coefficients and 95 % Bayesian credible intervals were calculated using estimates from the final model [36]. For health status questionnaires, coefficient values between 0.70 and 0.95 are indicative of good internal consistency [37]. All analyses were conducted using MPlus software (Version 8.9).

3. Results

Patient Demographic and Characteristics

From May 2022 till January 2024, 210 CHF patients (117 HFrEF and 88 HFpEF) were audited from a cohort eligible for SGLT-2i who had completed the PIH and had (Figure 1) available baseline information.

Baseline Characteristics (HFrEF)

The HFrEF study population comprising 117 patients, the baseline study characteristics are summarised in Table 1. The mean age was 66.8 years old (SD 13.5) and the gender, 88 (75 %) were male and 29 females. The majority of patients ethnicity were Caucasian

90 (77%)], the others South Asian, Asian, African or Indigenous. Most patients were married, 75 (64%)]. At least 71 (65%) described spouse of family supports for their ailment. A majority in 90 (76.9%) of patients had education up to high school level. Renal impairment defined as eGFR <60 ml/min were recorded in 48 (39%) patients. Thirty (25,6%) of patients did not record any associated comorbidity at baseline, 26 (22.3%) had one and the remainder had 3 or more. Hypertension and hypercholesterolemia were the most common comorbidity in 79 (68%) and 73 (62.4%) patients 79 and 73% respectively. Coronary artery disease (CAD) and diabetes was recorded in 51 (44%) and 42 (36%), obstructive sleep apnoea was also common, recorded in 31 (26.5%) of the cohort. Smoking history was recorded in 53 (45%) CHF was not a new diagnosis, i.e. a chronic condition, having been diagnosed prior to 12 months in 34 (29%) of the cohort. At baseline all patients had commenced Sodium Glucose co-Transporter-2 Inhibitors (SGLT2-i), within 6 months of identification for study enrolment. At this point at least 3 of 4 CHF [beta-blocker, renin-angiotensin aldosterone and/or neprilysin inhibitor (RAAS/ ARNI) or mineralocorticoid receptor antagonist (MRA)] pillars were already started in patients diagnosed prior to 12 months. Most patients with diagnosis of CHF within the 12 months period were also optimised on CHF pillar therapy. As was the requirement for eligibility of SGLT-2i, all patients were NYHA class 2 or more, with left ventricular ejection fraction (LVEF) Class III (<40%) and greater. CHF therapy was good all patients on antiplatelet or anticoagulant at baseline. CHF pillars were 112 (96%) for BB, 111 (95.1%) for RAAS/ARNI, and 46 (39%) for MRA. Polypharmacy was common

101 (68%) had at least 5 prescribed medications classes. Cardiac rehabilitation was only recorded among 16 (14%) of patients.

Baseline Characteristics (HFpEF)

The study population HFpEF comprising 88 patients, baseline study characteristics are summarised in table 1. The population had was 71.3 years old (SD 9.76); 46 (52%) were male and 42 (48%) females, highly very distinct demographics to HFrEF. The majority of patients ethnicity were Caucasian [90 (77%)], the others South Asian, Asian, African or Indigenous. Most patients were married [75 (85%)]. At least 71 (82%) described spouse of family supports for their ailment. A majority in 63 (71%) of patients had education up to high school level. Renal impairment defined as eGFR <60 ml/min were recorded in 23 (26%) patients. Only 3 (3.5%) of patients did not record any associated comorbidity at baseline, 32 (36.3%) had one and the remainder had 3 or more. Hypertension and hypercholesterolemia were the most common comorbidity in 76 (87%) and 67 (76%) patients respectively. Coronary artery disease (CAD) and diabetes was recorded in 30 (41%) and 29 (33%), obstructive sleep apnoea was also common, recorded in 21 (24%) of the cohort. Smoking history was recorded in 29 (33%) Heart failure was not a new diagnosis, i.e. a chronic condition, having been diagnosed prior to 12 months in only 7 (8%) of the cohort. At baseline 71 (80.7%) patients had commenced Sodium Glucose Co-Transporter-2 Inhibitors (SGLT2-i), within 6 months of identification for study enrolment. At this point with HFpEF the heart failure pillar was not relevant and only 2 (2.3%) were on ARNI, 46 (52.3%0 on BB, and 15 (17%) on MRA starkly different to HFrEF cohort. Most patients with diagnosis of CHF within the 12 months period were also optimised on CHF pillar therapy (BB, RAAS/ ARNI or MRA). As was the requirement for eligibility of SGLT-2i, all patients were NYHA class 2 or more, however all patients had left ventricular ejection fraction (LVEF) Class 1 (>50%) a clear separation with HFrEF patients. CHF therapy was not as good where diuretic, MRA and calcium channel blocker (CCB) use were lower. Antiplatelet or anticoagulant at baseline varied on comorbidity. CCF pillars for HFpEF is only the SGLT2i and were 71 (80.7) reflecting difficulty in diagnosis, ongoing understanding of HFpEF treatments and overlap with other indications such as DM and proteinuric chronic renal impairment. Polypharmacy was common 27 (31%) had at least 5 prescribed medications classes, although lower than HFrEF. Cardiac rehabilitation was not recorded in any participant for HF.

BCFA Model

The BCFA with exact zero cross-loadings and residual correlations corresponding to the pre-specified regular CFA model provided a posterior predictive p-value of almost zero (Table 1). The specification of small (close to zero) informative variances for cross-loadings ranging from 0.001 to 0.03 did not lead to an acceptable model fit based on PP p < 0.001. The addition of residual correlations produced almost perfect model fit for a larger prior (d=50) as shown by PP p values being close to 0.5 and confidence intervals symmetric about zero. Table 2 results show acceptable model fit for small informative cross-loadings of 0.005 and residual correlations d = 200 as indicated by goodness of fit indices whilst being close to the BCFA-based CFA model (PP p = 0.144). Table 2 shows factor loadings and factor correlation estimates for the BCFA and CFA model. All cross-loadings for the BCFA model were less than or equal to 0.1 and statistically non-significant in that 95% Bayesian credibility intervals covered zero. All hypothesised factor loadings were substantial and significant at p < 0.001 and factor correlation estimates were in the moderate range (0.409 – 0.662). For estimates of residual correlations, 95% (63/66) were statistically non-significant in that 95% Bayesian credibility intervals covered zero. Three residual correlations were statistically significant though ignorable in terms of statistical magnitude (<0.14) and practical implications. In conclusion, posterior predictive assessments of model fitness showed mostly ignorable discrepancies between BCFA and CFA models from a statistical and practical perspective. This provided additional support for a PIH four-factor solution using a standard CFA approach.

To assess the reliability of subscale scores for the BCFA 4-factor model with small variance priors, McDonald’s omega coefficients were calculated using unstandardised factor loadings and item residual variances. Coefficients (95 % Bayesian credible intervals) for the PIH subscales knowledge, partnership, management, and coping were 0.84 (0.79–0.88), 0.79 (0.73–0.84), 0.89 (0.85–0.91) and 0.84 (0.79–0.88), respectively. These values indicated that the subscales internal reliability in producing raw scores was in the acceptable range [32,33,34,35,36,40].

5. Conclusions

In this study we describe the first validation of the PIH scale, on the spectrum of heart failure syndromes (HRrEF and HFpEF). The PIH scale, a component of the Flinders Program, is validated across a range of chronic diseases and now CHF. This study has confirmed the reliability, validity and dimensionality of the PIH as a patient reported tool for measuring self-management capabilities for key domains, including collective self-efficacy for the CHF syndrome. For the next step, in other chronic diseases, studies have shown that the translation of patient scores, to tailored self-management support with the CFPI, improves patients’ knowledge, skills to disease and case management. This study lays a foundation to assess this concept for CHF. Importantly it also raises the questions on the steps required to introduce CHF (disease specific) aspects within the generic program. The traditional CHF approach has been predominately disease specific. This is an area that requires further exploration in the future.

6. Patents

Not relevant.