Reliability and validity of the preparatory grief scale for use with hemodialysis patients

Abstract

BACKGROUND

During the gradual decline of physical and social functioning associated with end-stage renal disease, patients might experience a premonition of impending death, resulting in a series of pre-mourning grief responses called preparatory grief. The preparatory grief in advanced cancer patients (PGAC) scale is the most widely used preparatory grief scale for patients on hemodialysis in China.

AIM

To verify the reliability and validity of the PGAC scale in patients on hemodialysis.

METHODS

In total, 327 patients undergoing regular hemodialysis in the blood purification center of three grade-A tertiary hospitals in Guangdong and Guizhou provinces were selected by convenience sampling. The assessment was administered using the general information questionnaire and the Chinese version of PGAC. SPSS 25.0 and Amos 24.0 were used for item analysis, confirmatory factor analysis (CFA), convergent validity, and internal consistency reliability estimation.

RESULTS

In the modified Chinese version of PGAC, 7 dimensions covering 27 total items were retained. CFA revealed a good fit of the factor model (chi-square degree of freedom = 2.056, standardized root mean square residual = 0.0479, root mean square error of approximation = 0.0570, GFI = 0.872, AGFI = 0.841, IFI = 0.931, CFI = 0.930, TLI = 0.919). The factor loadings of the items ranged 0.503-0.884. The composite reliability ranged 0.664-0.914, and the average variance extracted ranged 0.366-0.747. Cronbach’s α of the scale was 0.945, and Cronbach’s α for various dimensions ranged 0.662-0.914.

CONCLUSION

The modified PGAC has good reliability and validity, and it can effectively measure preparatory grief in patients on hemodialysis.

Key Words: Preparatory grief; Hemodialysis; Scale; Reliability; Validity

Core Tip: Because of the high mortality rate of end-stage renal disease (ESRD), some patients might not achieve their expected survival times even if they receive hemodialysis. Similar to patients with advanced cancer, those with ESRD also exhibit preoperatory grief. The preparatory grief in advanced cancer patients (PGAC) scale is the most widely used preparatory grief scale for patients in China, and it can be used to measure a range of preparatory grief responses in patients with advanced cancer or terminally ill patients. However, the use of this scale to assess patients on hemodialysis has not been reported. Therefore, this study evaluated the reliability and validity of the PGAC scale in patients on hemodialysis the aim of providing a standardized measurement tool for assessing preparatory grief for this population.

INTRODUCTION

Epidemiological research has indicated that the prevalence of chronic kidney disease (CKD) is increasing annually, reach approximately 14.3% globally and 10.8% in China[1]. End-stage renal disease (ESRD) is the final stage of CKD, and it carries a low cure rate and requires hemodialysis as a life-sustaining measure[2]. According to statistics, the number of patients on hemodialysis is increasing at an annual rate of 11% in China[3]. In addition to receiving hemodialysis at the dialysis center 1-3 times a week, patients on hemodialysis also need to maintain a strict diet, water control, and self-care of vascular access, and they have no hope of cure, resulting in a huge psychological burden[4]. As their physical and social functioning gradually declines, patients might experience a premonition of impending death, resulting in a series of pre-loss grief responses termed preparatory grief. “Preparatory grief” originally described the preparation of a soldier’s wife for the possible death of her husband during World War II[5]. Later, the concept was expanded from a person preparing for the loss of a loved one to a person preparing for his or her own death. Preparatory grief is the grief experienced by terminally ill patients to prepare themselves for death[6]. In contrast to a standardized measurement tool, the preparatory grief in advanced cancer patients (PGAC) scale is the most widely used preparatory grief scale for patients on hemodialysis in China[7,8]. The scale was originally developed by the Greek scholar Mystakidou et al[9] in 2005 to measure a series of preparatory grief reactions in individuals with advanced cancer or patients on the verge of death. Research has indicated good internal consistency reliability, criterion validity, convergent validity, and discriminant validity for the instrument, which can effectively evaluate the preparatory grief response of subjects. Because of the high mortality rate of patients with ESRD, some patients might not experience their desired survival time even if they receive hemodialysis. Similar to end-stage cancer, patients with ESRD are also anticipating death. However, the use of this scale in assessing patients with hemodialysis has not yet been reported. Thus, this study assessed the reliability and validity of the PGAC scale were evaluated via item analysis, confirmatory factor analysis (CFA), convergent validity, and internal consistency reliability to provide a standardized measurement tool for evaluating preparatory grief in patients on hemodialysis.

MATERIALS AND METHODS

Research participants

Patients undergoing regular hemodialysis in the blood purification centers of three grade-A tertiary hospitals in Guangdong and Guizhou provinces from March 2022 to October 2022 were selected by convenience sampling as the study subjects. The inclusion criteria were as follows: Age ≥ 18 years; diagnosis of ESRD according to the diagnostic criteria of Internal Medicine[10]; ongoing hemodialysis, normal cognitive function (Montreal Cognitive Assessment score ≥ 26 points) and communication ability; provision of informed consent; and willingness to participate in this survey. The exclusion criteria were as follows: Unstable conditions; concurrent malignancies; acute infections; serious diseases (dysfunction) of vital organs such as the heart, brain, and lungs; and dialysis duration ≤ 3 months. The sample size calculation followed the rule that the sample size was 10-fold larger than the number of items. At least 327 patients were required given the presence of 31 items in the scale and an expected effective recovery rate of 95%. This study was approved by the Ethics Committee of Guangzhou First People’s Hospital (K-2021-191-01).

Research tools

General information questionnaire: The demographic data and clinical characteristics of the patients, including sex, age, educational level, marital status, occupation, per capita monthly household income, type of medical expenses settlement, and religious beliefs, were investigated using a general information questionnaire designed by the research team.

Chinese version of the PGAC scale: In 2018, Chinese studies[11] tested the reliability and validity of the PGAC scale. The results showed illustrated that Cronbach’s α of the scale was 0.919, and Cronbach’s α ranged 0.533-0.926 for various factors, with a split-half reliability of 0.842 and content validity of 0.916, making the scale suitable for studying preparatory grief in patients with cancer and terminal diseases. The instrument contains 31 items and 7 dimensions, namely sadness and anger (12 items: 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 19, 20), disease adjustment (4 items: 13, 17, 23, 25), somatic symptoms (4 items: 24, 26, 27, 29), religious comfort (3 items: 14, 15, 31), perceived social support (3 items: 1, 2, 30), wish (2 items: 21, 22), and self-consciousness (3 items: 16, 18, 28). The scale is based on a four-point scale, with “0” indicating disagreement and “3” indicating agreement, and its total score ranges 0-93, with higher scores indicating higher levels of perceived sadness.

Pre-investigation: Convenience sampling was conducted to select 35 patients on hemodialysis from a grade-A tertiary hospital in Guangzhou for a preliminary survey. They were given the PGAC scale and asked about their understanding of the instructions for completing the questionnaire. The questionnaire was collected on site with an effective recovery rate of 100%. The age of the surveyed subjects was 47.91 ± 11.66 years, and this cohort included 19 men (54.3%) and 16 women (45.7%). Of these participants, 30 were married (94.3%), and five were unmarried (5.7%). Meanwhile, 19 subjects were unemployed (54.3%), and 16 were employed (45.7%). Twenty-two subjects (80.0%) had a per capita monthly household income of 1000-4999 yuan, whereas 13 subjects (20.0%) had a per capita monthly household income exceeding 4999 yuan. Thirty-four subjects (97.1%) paid for hemodialysis through medical insurance, and one subject (2.9%) received free medical care coverage. Twenty-eight participants (80%) had no religious beliefs, three (8.6%) had Buddhist beliefs, and four (11.4%) had Christian beliefs. In the pre-investigation, Cronbach’s α was 0.922 for the PGAC scale, and no changes were made to the items after this evaluation.

Data collection methods

Data collection was initiated after contacting the nursing departments of various hospitals in advance and obtaining the informed consent of the patients. To ensure the scientific validity of the research results, the research subjects were selected rigorously according to the inclusion and exclusion criteria, and corresponding data were collected. The investigators included three graduate students. Before the questionnaire survey, the three investigators were trained in a unified manner to clarify the standards for completing the questionnaire survey. An electronic questionnaire was compiled using the online survey tool SoJump (http://www.wjx.cn/vm/texLrFS.aspx#). When patients were waiting for dialysis or during dialysis sessions, the investigators explained the purpose and significance of the survey to patients and informed them of the completion requirements with unified instructions after obtaining their informed consent. Patients were asked to click on the questionnaire link or scan the QR code to complete and submit the questionnaire. Patients who could read and complete the questionnaire were asked to complete the survey independently. For patients who could not read and write but could correctly understand and answer the questions, they were allowed to complete the questionnaire with the help of the investigators. In total, 341 questionnaires were distributed. After the completion of data collection, the questionnaire data were exported to Excel (Microsoft, Redmond, WA, United States) for one-by-one checking by the researchers. Fourteen invalid questionnaires that were completed in fewer than 3 minutes and those with irregular or inconsistent responses were excluded. Finally, 327 valid questionnaires were recovered, with an effective rate of 95.89%.

Statistical analysis

We used the “Analyze Download” function of SoJump to import the raw data of the electronic version of the questionnaire into the SPSS database (IBM, Armonk, NY, United States). Cronbach’s α, the critical ratio, and Pearson’s correlation coefficient were calculated using SPSS 25.0, and the internal consistency evaluation and item analysis of the scale were performed (test level α = 0.05). A theoretical model was constructed using Amos 24.0 (IBM), and CFA was used to verify the fit of the model and evaluate its structural validity. Based on the factor loading of each item in the CFA, the average variance extracted (AVE) and composite reliability (CR) of each factor were calculated using the Exps application to evaluate the convergent validity[10].

RESULTS

General information of the research subjects

Among the 327 patients surveyed, 60.6% were male, and the average age of the patients was 57.35 ± 12.78 years (range, 18-84). In total, 46.8% of the patients had an education level of junior high school. Concerning the marital status, 87.5% of the patients were married. Meanwhile, 48.6%, 29.4%, and 22% of the patients were unemployed, emeritus/retired, and employed, respectively. The per capita monthly household income ranged 1000-4999 yuan for 59.3% of the patients. In addition, 76.1% of the patients received medical care through social medical insurance, and the remaining 23.9% received free medical care expenses. In total, 262 patients (80.1%) had no religious beliefs, 28 patients (8.6%) had Buddhist beliefs, and 37 patients (11.3%) had Christian beliefs (Table 1).

Table 1 General information of the research subjects, n (%).

Characteristic	n = 327
Gender
Male	198 (60.6)
Female	129 (39.4)
Age (years)	57.35 ± 12.78
Education level
Junior high	153 (46.8)
Senior high school	106 (32.4)
College	68 (20.8)
Working status
Unemployed	159 (48.6)
Emeritus/retired	96 (29.4)
Employed	72 (22.0)
Monthly household income (yuan)
1000-4999	194 (59.3)
> 4999	133 (40.7)
Medical care payment approach
Social medical insurance	249 (76.1)
Free medical care expenses	78 (23.9)
Religious beliefs
Buddhist beliefs	28 (8.6)
Christian beliefs	37 (11.3)
No beliefs	262 (80.1)

Item analysis

According to the total score, the top and bottom 27% were set as the high-skill and low-skill groups, respectively. The independent-samples t-test revealed that the decision value of each item ranged 2.780-24.819 (P < 0.001). According to Pearson’s correlation analysis, the correlation coefficients of items 11, 25, and 28 concerning their relationships with the total score were 0.317, 0.386, and 0.188, respectively (P < 0.001), and the correlation coefficients of the other 28 items regarding their relationships with the total score ranged 0.483-0.770 (P < 0.001).

Validity analysis

CFA: Correlation analysis revealed a low-to-moderate correlation among the seven dimensions. Therefore, the validation model hypothesis proposed in this study was a first-order, five-factor oblique model. The maximum likelihood method was used to estimate the initial model, and estimation of path coefficient analysis illustrated that the standardized coefficients of the 31 items spanned 0.314-0.882 (P < 0.001), with the path coefficients being lower than 50 for 11, 13, 25, and 28, indicating that they should be deleted. The results indicated that the model fitted well after deleting items 11, 13, 25, and 28, as presented in Table 2.

Table 2 Comparison of the fit indices of the not modified and modified models.

Fit indices	χ2/df	SRMR	RMSEA	GFI	AGFI	IFI	CFI	TLI
Not modified	2.181	0.0592	0.0600	0.835	0.801	0.902	0.901	0.888
Modified	2.056	0.0479	0.0570	0.872	0.841	0.931	0.930	0.919
Reference value	< 3	< 0.05	< 0.08	> 0.9	> 0.9	> 0.9	> 0.9	> 0.9

SRMR: Standardized root mean square residual; RMSEA: Root mean square error of approximation.

The factor loadings of the measured variables ranged from 0.503 to 0.884 (P < 0.05; Figure 1 and Table 3).

Figure 1  Path analysis diagram of confirmatory factor analysis of the preparatory grief scale.

Table 3 Parameter estimation results of confirmatory factor analysis of the preparatory grief in advanced cancer patient scale.

Sub-scales	Item	Parameter significance estimation				Factor loading
		Unstd.	SE	t	P value	Std.
Sadness and anger	Item 3	1				0.772
	Item 4	1.002	0.068	14.778	a	0.764
	Item 5	1.032	0.069	14.943	a	0.773
	Item 6	0.812	0.072	11.351	a	0.614
	Item 7	0.916	0.068	13.413	a	0.710
	Item 8	1.015	0.067	15.199	a	0.783
	Item 9	0.764	0.063	12.092	a	0.647
	Item 10	0.871	0.073	12.004	a	0.644
	Item 12	0.714	0.062	11.513	a	0.620
	Item 19	0.86	0.069	12.412	a	0.663
	Item 20	0.846	0.064	13.272	a	0.701
Disease adjustment	Item 17	1				0.693
	Item 23	1.142	0.104	11.025	a	0.717
Somatic symptoms	Item 29	1				0.503
	Item 27	1.336	0.17	7.841	a	0.627
	Item 26	1.412	0.167	8.455	a	0.719
	Item 24	1.104	0.152	7.273	a	0.548
Religious comfort	Item 14	1				0.709
	Item 15	0.8	0.07	11.357	a	0.630
	Item 31	0.847	0.075	11.305	a	0.630
Perceived social support	Item 1	1				0.536
	Item 2	1.815	0.194	9.362	a	0.805
	Item 30	1.866	0.199	9.368	a	0.823
Wish	Item 21	1				0.844
	Item 22	0.995	0.054	18.345	a	0.884
Self-consciousness	Item 16	1				0.811
	Item 18	0.847	0.057	14.804	a	0.776

^aP < 0.001.

Convergent validity: In this study, CR and AVE were used as indicators of convergent validity. The results demonstrated that after deleting items 11, 13, 25, and 28, CR exceeded 0.6 for all seven factors (range, 0.664-0.914), whereas AVE was greater than 0.36 for all factors (range, 0.366-0.747; Table 4).

Table 4 Convergent validity of the preparatory grief in advanced cancer patient scale.

Categories	CR	AVE	Categories	CR	AVE
Sadness and anger	0.914	0.493	Perceived social support	0.771	0.538
Disease adjustment	0.664	0.497	Wish	0.855	0.747
Somatic symptoms	0.694	0.366	Self-consciousness	0.773	0.63
Religious comfort	0.695	0.432

AVE: Average variance extracted; CR: Composite reliability.

Reliability analysis: Cronbach’s α for the entire scale was 0.945. In addition, the value was 0.914 for “sadness and anger”, 0.662 for “disease adjustment”, 0.689 for “somatic symptoms”, 0.689 for “religious comfort”, 0.760 for “perceived social support”, 0.854 for “wish”, and 0.769 for “self-consciousness”.

DISCUSSION

Hemodialysis is the most common renal replacement therapy globally, accounting for approximately 69% of all renal replacement therapies and 89% of all dialysis treatments[12]. Over the past 60 years since the inception of hemodialysis, significant advances have been made in dialysis technology and patient access to treatment, particularly in high-income countries. However, the availability, accessibility, cost, and outcomes of Hemodialysis vary widely around the world, and overall, the rates of quality of life impairment, morbidity, and mortality are high. In addition, patients on hemodialysis have high symptom loads, and they often experience considerable financial stress[13]. Despite the many advances in technology and delivery systems since hemodialysis introduction, the poor prognosis of patients receiving hemodialysis remains a major public health concern. According to several previous studies, psychological disorders in patients with ESRD arise because of decreases in physiologic status and other effects of long-term hemodialysis[14]. Qualitative data have helped clarify the general lived experience of patients on hemodialysis, such as grief and loss[15].

Based on the evaluation data of the PGAC scale applied to patients on hemodialysis, this study evaluated the structural validity of the scale through CFA. Scholars[16] previously suggested the following evaluation indicators for CFA: The factor loading of each item in the specified dimension is suggested to be greater than 0.50 but lower than 0.95; χ² degrees of freedom (χ²/df) of 1–3; standardized root mean square residual (SRMR) lower than 0.05; and root mean square error of approximation (RMSEA) lower than 0.08. GFI, AGFI, IFI, CFI, and TLI lower than 0.90 suggest that the model fits well, and values closer to 1 indicate a better model fit. Based on item analysis and parameter estimation results for the initial model, this study deleted four items (items 11, 13, 25, 28) with correlation coefficients lower than 0.40 and factor loadings lower than 0.50, and model estimation was conducted to obtain a model with a good fitting effect. A scale with seven dimensions (27 total items) was generated. The analysis of the modified scale illustrated that the indices of fit such as χ²/df, SRMR, RMSEA, GFI, AGFI, IFI, CFI, and TLI were all at acceptable levels. The factor loadings of various items in the specified dimensions all exceeded 0.50 (range, 0.503-0.884; P < 0.001). These findings indicated that the latent variables have a high degree of explanation for the measured variables, and the model fit well with good structural validity.

In the study, the following four items were removed: Item 11 (“I am very afraid of the unknown”) belongs to the dimension of “sadness and anger”; items 13 [“I am very afraid of being abandoned by my family (dislike, alienation and indifference)”] and 25 (“I believe that there will be an afterlife after death”) belong to the dimension of “disease adjustment;” and item 28 (“I want to live a few more years or a dozen years, the longer the better”) belongs to the dimension of “self-consciousness.” With the initial establishment and improvement of China’s universal medical insurance system in recent years, medical insurance support for major treatments such as hemodialysis has continued to increase. Many of the patients’ treatment and drug expenses were included in the reimbursement scope of “outpatient special diseases”, and the reimbursement ratio is gradually increasing, leading to lower medical expenditure. Of the 327 patients treated in the blood purification center of three grade-A tertiary hospitals in Guangdong and Guizhou provinces, treatment was covered by social medical insurance for 76.1% of patients, and the remaining 23.9% enjoyed free medical care. It can be inferred that, thanks to China’s medical security system, patients have less concerns and fear about being abandoned by their families (disgust, alienation, indifference), as they believe that the direct economic impact of their illness on their families is relatively small. In addition, compared with the findings for patients with advanced cancer, patients on hemodialysis have less intense emotions of sadness, anger, and attitudes and perceptions of death. Moreover, the survival of patients with ESRD is gradually increasing as hemodialysis technology continuously advances, with an increasing survival rate annually[17], whereas the mortality rate of cancer remains high[18]. Therefore, it is speculated that the preparatory grief of patients on hemodialysis is primarily attributable to the loss of kidney function and the loss of normal social and family role functioning caused by the maintenance of hemodialysis, whereas the preparatory grief of patients with advanced cancer is mostly caused by the perceived imminent loss of life. It can be inferred that if the PGAC scale is to be applied to the hemodialysis population, the aforementioned four items should be deleted because they do not reflect the actual situation of such patients.

This study used AVE and CR to measure the convergent validity of the modified scale. AVE refers to the degree of similarity in the results of different measurement methods when measuring the same feature. Higher AVE indicates a more convergent dimension and a greater average explanatory ability of the latent variable to the measurement index. It is suggested that AVE should be greater than 0.5, although 0.36-0.5 is an acceptable range[19]. CR represents the consistency of a set of indicator variables that constitute the factor, with higher values suggesting higher internal consistency of the model. CR is suggested to exceed 0.60[16]. In this study, AVE and CR exceeded 0.36 and 0.6, respectively, for the seven dimensions of the modified scale, indicating good convergent validity.

CONCLUSION

This study used Cronbach’s α to measure the internal consistency of the modified scale. Cronbach’s α for the modified scale was 0.945, and coefficient ranged 0.662-0.914 for the seven dimensions, indicating good internal consistency reliability[20]. Our team believes that grief might change with changes in the patient’s physical and mental condition as well as external circumstances, and thus, Cronbach’s α is an unstable variable that is unsuitable for measuring retest reliability.