Development and preliminary validation of a health-related family quality of life instrument

Abstract

The researchers’ aims for this preliminary study was to develop an instrument, guided by the Theory of Health-related Family Quality of Life, to measure health-related family quality of life (HR-FQoL) as perceived by women with cancer. The researchers used a two phase design for instrument development: (1) establishing face validity of a 38-item instrument with an expert panel and patient review, and (2) focusing on the internal structure and construct validity of the instrument with responses from female patients (N = 236) with a cancer diagnosis (breast, gynecologic). The researchers identified a final 25-item HR-FQoL instrument with four sub-scales that each captured multiple concepts within the Theory of HR-FQoL. The resulting instrument maybe used by researchers and clinicians to assess various aspects of health-related family quality of life among female breast and gynecological cancer survivors.

Quality of life in the context of cancer diagnosis, treatment, and survival has long been the focus of psychosocial research in oncology. One area of this research that has not been clearly defined or explored is patients’ experiences of quality of life within family contexts. Researchers have explored health-related quality of life and family quality of life have been explored as their own distinct concepts. In order to understand where these concepts overlap, Radina (2013) developed the Theory of Health-related Family Quality of Life (HR-FQoL).

We aimed to develop a clinical instrument, guided by the Theory of Health-related Family Quality of Life, that is intended to measure HR-FQoL as perceived by women with cancer. In this paper, we report preliminary development of the measure in a two-phase study where we established content and construct validity of the instrument. While these data are based on US participants, we believe that the establishment of this instrument can serve as a foundation for additional iterations that reflect cultural differences relating the health beliefs and expectations around family functioning.

Background

Health-related quality of life includes physical, mental, emotional, and social functioning and emphasizes the influence of individual health status on quality of life. Family quality of life (Brown & Brown, 2014) includes the ability of family units to achieve their goals within larger social and situational contexts and how individuals perceive their own quality of life within the context of family (e.g., family dynamics, family cultures). Because illness occurs within family systems (Patterson & Garwick, 1994; Rolland, 1994), the concepts of patient care and family support should broadly include the multidimensional concepts of both health-related quality of life and family quality of life (HR-FQoL).

Despite recent explorations of HR-FQoL dimensions (e.g., Ginter & Radina, 2019; Radina et al., 2019), gaps remain in the availability of tools to measure HR-FQoL. For example, instruments exist to measure health-related quality of life (e.g., SF-36) and specific aspects of quality of life or specific health contexts (e.g., FACT-G; Sajid et al., 2008). However, these instruments (Sajid et al., 2008) do not sufficiently address the intersection between health-related quality of life and family quality of life. For example, some of these instruments measure concepts not traditionally considered aspects of family quality of life, including family functioning, perceived family well-being, and family members’ abilities to meet the needs of each other (Brown & Brown, 2014). Likewise, existing scales that measure family quality of life (e.g., FACES IV, Family APGAR, McMaster Family Assessment Device (FAD) General Functioning Subscale) are derived from family theory-based conceptual frameworks (e.g., Epstein et al., 1983; Hu et al., 2011; Olson et al., 2006; Smilkstein et al., 1982). Yet, none focus on the intersection between F-QoL and HR-QoL. The absence of a single patient reported instrument to measure HR-FQoL is problematic as lack of such a tool makes it difficult for researchers and clinicians to quickly and accurately assess patient perceived HR-FQoL concerns that may require intervention/support.

Researchers have explored how family functioning (e.g., adaptation, resilience) is modified during illness (Lennon et al., 2016), with practice implications (e.g., caregiver burden) for clinical and community providers (e.g., nurses, social workers). Breast and gynecological cancers are known disruptors of established family patterns and family’s quality of life (e.g., household roles, parenting responsibilities, sexual relations between patients and their partners; Akyüz et al., 2008; Laizner, 2018). Therefore, within the context of chronic conditions and/or long-term cancer survivorship failure to address HR-FQoL often becomes even more detrimental for women (Radina, 2013). For breast and gynecological cancer clinicians, this is problematic because their patients: (1) often experience changes in emotional satisfaction and feelings of psychological or affectional closeness with family members (Ginter & Radina, 2019; Radina, 2013, p. 44), and (2) play important roles such as mother, life partner, daughter, or caretaker for family members. Thus, an understanding how women experience HR-FQoL is needed to provide insights into overall family functioning and inform clinical interventions.

Methods

In designing this study, we were guided by Theory of Health-related Family Quality of Life, which addresses family functioning and sense of coherence within the health-related quality of life domains of physical, psycho-social, and emotional closeness (Radina, 2013). Of critical importance to us was the previous worked related to the theory (Radina, 2013) that centered on the unique roles that many women, specifically breast cancer survivors, play in family life (e.g., wives, caregivers, family managers; Radina, 2009; Radina & Armer, 2001, 2004; Radina et al., 2008). We used a two phase design for instrument development. In Phase One, we used the Theory of HR-FQoL (Radina, 2009; Radina & Armer, 2001, 2004; Radina et al., 2008) to develop items and refine the content and face validity of the instrument via expert panel and patient review. In Phase Two we used a cross-sectional survey design to examine several aspects of construct validity including the internal structure of the instrument and convergent and divergent validity.

Phase One (content and face validity)

We obtained Institutional Review Board and Scientific Review Committee approvals prior to study commencement. Phase One included: (1) item generation, (2) expert panel review, (3) instrument redesign, (4) patient pilot of items, (5) instrument redesign, and (6) patient review/piloting of the revised instrument.

Instrument construction

Authors AA, BB, and CC constructed an initial 38-item HR-FQoL instrument that captured the theoretical components of emotional closeness (13 items); family sense of coherence (10 items); and family functioning (13 items) from the theory. We derived items by examining other instruments for guidance on wording and response scales. Specific items from existing scales were not chosen for inclusion. Rather the content and wording of scales were used to inspire the development of new items. We determined that responses would be measured on a 4-point ordinal scale (1 = Strongly Disagree, 2 = Disagree, 3 = Agree, 4 = Strongly Agree). Individual perceptions of satisfaction with FHR-QoL are central to the theoretical framework. Thus, we developed one item capturing level of satisfaction with each of the three dimensions (i.e., satisfaction with practical support, satisfaction with the way the family faces challenges, and satisfaction with emotional support) was developed (Table 1). The same 4-point response options used for all of the other items in the measure are used for the three satisfaction items.

Table 1.

Instrument and satisfaction items by theory of HR-FQoL components.

HR-FQoL theoretical components	Item
Family sense of coherence	My family avoids dealing directly with problems. My family is easily overwhelmed by problems. My family uses resources creatively to deal with my illness. We believe we are a strong family that can deal with problems. My family works together to handle problems related to my illness. My family accepts stressful challenges as an expected part of life. Spiritual beliefs are important to my family. Our family shares a set of beliefs that gives us a sense of purpose in life. My family tends to look for meaning in upsetting circumstances. My family makes back-up plans for unpredictable health problems.
Emotional closeness	I readily share my feelings about my illness with my family. I avoid discussing concerns related to my illness with my family. My family members tend to withdraw when they are angry. Our family tends to calmly discuss problems. Angry communication is expressed frequently in my family. My family tends to take a negative view of life. Discussion of family problems tends to not resolved the problem. My concerns will be treated with respect by my family. My family tends to express warmth. My family is deeply concerned about any major threats to my health. My family tends to express encouragement. Family members really try to understand my concerns about my illness. Communication between my family members tends to be clear and direct.
Family functioning	Differing opinions tend to be respected in my family. My family members make assumptions about the help I need rather than asking me what I need. Some of my family members are critical of the way I manage my illness. Some of my family members try to control my choices about I take care of myself. My family members will take on my tasks when needed. My family will assist me when I ask for their help. I feel comfortable asking my family members for help. I cannot count on timely help from my family members. My family has rigid expectations about how men and women should behave. Some of my family members express strong emotional reactions to minor problems. My family has unspoken rules for behavior. My family expects me to be as self-sufficient as possible. My family members don’t share my concerns about completing tasks I used to do.
Satisfaction	I am satisfied with the way my family faces the challenges of my illness. I am satisfied with the emotional support I received from my family. I am satisfied with the practical support offered by my family for dealing with my symptoms.

Recruitment and data collection methods

Eligible experts from Vanderbilt University Medical Center (VUMC), Vanderbilt School of Nursing (VUSN), and Miami University were family/human service professionals and research or clinical management professionals who were currently active in the management of female cancer survivors. Experts without clinical or research experience with female cancer survivors were excluded. We recruited experts via convenience and snowball techniques as well as from existing IRB-approved patient registries. Inclusion criteria for patients were biological female, a history of breast or gynecological cancer diagnosis, receipt of treatment for breast or gynecological cancer within the last five years, 18 years of age or older, ability to read English and see print/computer screens, and valid email address. Patient exclusion criteria were biological males, biological females without a history of breast or gynecological cancer, no history of treatment of breast or gynecological cancer in the last five years, under 18 years of age, inability to read English or read print/computer screens, and current cognitive impairments that prevent one’s ability to provide informed consent. Participants were not compensated. Informed consent was obtained prior to study enrollment for all participants. Experts participated from their place of work or at VUSN. Patients participated at either VUMC, VUSN, or in a private location of the patient’s choice (e.g., home).

Both participant groups completed surveys using Research Electronic Data Capture (REDCap). REDCap is a secure, web-based application consisting of a software toolset and workflow methodology for electronic collection and management of research and clinical trial data (Harris et al., 2009). The study coordinator sent patients a link to the REDCap survey, which included a demographic form, a breast cancer-related information form, the 38-item revised HR-FQoL instrument, and the 3-item Satisfaction Scale. Experts were asked to rate each item for relevance to the experience of female breast cancer patients using a 5-point ordinal scale (ranging from 1 = not relevant at all, to 5= extremely relevant), complete a survey regarding their satisfaction with the instrument, make comments regarding the instrument itself, and provide suggestions for adding or removing/revising items. Researchers made instrument revisions based on experts’ feedback. This revised version was sent to the patient panel for review.

Patient demographic form

This form included age, gender, race, ethnicity, years of education completed, marital status, employment status, zip code, urban or rural locale, insurance status, household income, number of family members living in the home, number of children, length of marriage if married, and family members who provide the most support.

Patient breast cancer-related information form

This form included questions about time since cancer diagnosis, cancer stage at diagnosis, treatment type(s), and recurrence information. The presence of current treatment side effects (lymphedema; neuropathy; fatigue; memory and cardiac problems) was captured using a yes/no response option.

HR-FQoL instrument and feedback

Patients were asked to (1) indicate with a checkbox whether each individual item in the instrument should be retained or deleted, and whether any additional content or items of importance should be included and (2) comment on the length of the survey and the perceived usefulness of the survey.

Satisfaction scale

The scale included one item capturing level of satisfaction with each of the three proposed dimensions of HR-FQoL (practical support, the way the family faces challenges, and emotional support).

Data analysis

Experts

We collected and analyzed data to produce the most user-friendly instrument for subsequent patient review. We generated frequency distributions of the preliminary 38-item’s relevance responses using IBM SPSS Statistics software to evaluate potential for response variability. We undertook qualitative content analysis of the comments about each specific question and the overall instrument. After individually coding these data, authors (AA, BB, CC) discussed coding discrepancies and achieved consensus.

Patients

We generated frequency distributions of each of those item responses were generated. We again analyzed qualitative content using content analysis as described above.

Instrument revision

We made revisions to the 38-item FHR-QoL instrument based on experts’ and patients’ feedback all of whom suggested dropping two items. We then offered a 36-item version of the FHR-QoL to 10 patient participants for their initial review and feedback. In response to those patients’ feedback, no items were dropped however the research team revised seven items for clarity. Patients reported positive feedback about the usefulness of the survey and felt the length of the survey was appropriate. Researchers sent the revised 36-items to the same 10 patients for a second review, and feedback about this version was extremely positive. We made no additional revisions to the 36-item FHR-QOL.

Phase Two (content, construct, and convergent validity)

The goal of Phase Two was to assess the internal structure and convergent/divergent validity of the 36-item instrument. Participants were asked to answer each question based on whether their experience with their family matches or does not match the statement. Authors VS, ER, and SR reviewed literature of multiple published instruments to determine optimal comparators for convergent/divergent validity. The instruments we chose either reflected theoretical constructs expected to correlate with the 36-item tool, or in the case of Marlowe Crown, expected to not correlate with the tool. To reduced subject burden, we used the shortest available instruments that met these criteria.

Sample

A new sample of participants were breast and gynecological cancer patients who were recruited from: (1) an existing IRB approved patient registry; (2) the Vanderbilt-Ingram Cancer Center (VICC); (3) ResearchMatch.org, a national electronic, web-based recruitment tool (www.researchmatch.org/about/); and (4) advertisements posted through Facebook and the VUMC communications channel, fliers distributed to lymphedema therapists, oncologists, cancer centers, surgeons, YMCA’s, and other lymphedema and cancer support groups in the targeted geographical areas of Ohio and Tennessee. To enhance patient recruitment, study staff also conducted in-person recruitment and eligibility screening at the VICC clinic and provided eligible participants with a paper copy of the survey and a return stamped envelope. Eligibility criteria were the same as for Phase One except that we expanded the targeted population to include females with a history of gynecological cancer. This allowed for the recruitment of a broader range of women who have experienced female-specific cancers that may influence gender-related experiences within families. The goal was to recruit a total final, generalizable sample of five or more times the total number of items in the initial item pool (N = 180). With over-recruitment to account for potential incomplete responses, a total of 236 participants met the inclusion criteria and completed the 36-item initial HR-FQoL instrument. No patients from Phase One participated in Phase Two.

Data collection

Recruitment materials included a link to the REDCap IRB approved consent form and study survey. Participants who consented automatically advanced to the screening tool to determine eligibility and if eligible continued to the study measures. Some participants preferred to complete the survey (paper copy or electronic survey using a tablet) while study staff were onsite. Hard copies (n = 43) were double entered prior to analysis. Instruments included in the survey that were used to evaluate the validity of HR-FQoL instrument are listed in Table 2.

Table 2.

Instruments used to establish HR-FQoL instrument validity.

Instrument	Source	Description	Cronbach’s alpha
Convergent validity
Family APGAR	Smilkstein et al. (1982)	5-item tool captured “family functioning” concept across five aspects: Adaptability, Partnership, Growth, Affection, and Resolve. Response options were on a 3-point scale ranging from 0 (hardly ever) to 2 (almost always).	0.92
Medical Outcome Study Social Support Survey short version	Sherbourne and Stewart (1991)	8-item version of the 19-item Medical Outcomes Study Social Support Survey (). The MOS-SS consists of two sub-scales, emotional support (4-items) and instrumental support (4 items), with response options of 1 (none of the time) to 5 (all of the time). Higher scores indicate greater perceived support.	Total: 0.93 Emotional subscale: 0.89 Instrumental subscale: 0.94
McMaster Family Assessment Device (FAD) General Functioning Subscale	Epstein et al. (1983)	Measured the structural, organizational, and transactional characteristics of families in the “family function’ concept (Byles et al., 1988; Mansfield et al., 2015). The general functioning scale asked participants to rate how well each statement describes their own family on a four-point scale ranging from strongly agree to strongly disagree. The FAD was scored by averaging the responses to the 12 component items. Higher scores indicate worse levels of family functioning.	0.94
Couples Illness Communication Scale (CICS)	Arden-Close et al. (2010)	4-item scale assessed communication when one member of a couple has a chronic illness and captured the “emotional closeness” concept. The measure uses a 5-point Likert scale ranging from strongly disagree to strongly agree.	0.85
Family Sense of Coherence—Short form	Antonovsky and Sourani (1988)	12-item scale assesses a family’s general belief that their family circumstances are comprehensible, manageable, and meaningful and captured our “family sense of coherence’ concept. The differential items are scored from 1 to 7 with extreme anchor phrases.	0.90
Differentiation in Family Systems Scale	Anderson and Sabatelli (1992)	11-item scale measured family differentiation and captured the “emotional closeness” concept. A 6-point Likert scale (0 = never to 5 = always) was used.	0.92
Divergent validity
Marlowe-Crowne Social Desirability short form 13 (MC-C)	Crowne and Marlowe (1960)	instrument was used to access participants’ likelihood to respond to items in order to present themselves positively (Andrews & Meyer, 2003; Ballard, 1992; Reynolds, 1982). Items are true/false with a score of 0–13. A higher score indicates a more unrealistic favorable impression.	0.71

Data analysis

We determined the final sample to include participates who completed all of study measures. We generated item response frequency distributions to evaluate variability in responses prior to inclusion in subsequent evaluation of the proposed structural and convergent validity of the measure. We investigated and ultimately confirmed the structure of the HR-FQoL instrument with a combination of Principal Components and Principal Axis exploratory factor analyses (EFA). Resulting factors with a minimum eigenvalue of 1.0 were interpreted using varimax rotation to ease that interpretation. We used confirmatory factor analyses (CFA) to evaluate the fit of our proposed internal structure of the HR-FQoL and subsequently to confirm the fit of our revised measure to our proposed factor structure generated with the assistance of EFA. A combination of the comparative fit index (CFI), Tucker-Lewis Index (TLI), and root mean squared error of approximation (RMSEA) (Hu & Bentler, 1999) were used for that evaluation. CFI and TLI values equal to or above .90 and SRMR values below .08 indicated a good model fit. Further evaluation of the internal consistency of the item responses within each proposed factor was conducted using Cronbach’s alpha coefficient. Finally, we used Spearman correlations of the proposed HR-FQoL total and factor scores, as well as the satisfaction scores with the scores from the included established measures of family structure and satisfaction, as well as social desirability, to evaluate the convergent and divergent validity of the FHR-QoL. STATA 15 (StataCorp, 2017) was used for CFA; IBM SPSS Statistics 27 (IBM Corp, 2020). was used for all the other data analyses.

Results

Phase One results

Experts (n = 8)

Based on frequency distributions of each of the items we determined that all but two items had ≥ 50% of relevance ratings in the moderate relevance/extremely relevant categories. Qualitative findings were: (1) the “tone” was more negative than desired for two items and it was suggested that these be dropped prior to patient review, (2) all other items were thought to be a “good fit with the theory concepts,” “easy to understand,” and believed by all experts to be “relevant,” and (3) the instrument was deemed to be “clear with acceptable formatting,” and “readable,” and “relevant.” A comment about length indicated that the instrument would be “acceptable if under 40 items.”

Patients (n = 10)

We determined that there was satisfactory variability of each of the items based on our examination of frequency distributions. Qualitative findings indicated: (1) “poor wording” for seven items that required revisions for clarity, (2) “keep all the items” suggesting that patients did not want any individual item dropped from the survey, (3) positive feedback about the usefulness of the instrument, and (4) a theme of “we need better instructions,” which included the need for a definition of family. These same participants also review a revised instrument based on their feedback. Responses were extremely positive with “no further revisions” suggested.

Phase Two results

Sample description

Tables 3 and 4 provide demographic, clinical, and treatment characteristics of the participants for Phase Two. All participants were from the US.

Table 3.

Demographic characteristics of phase two participants (N = 236).

	Mean (SD)
Age (years)	53.4 (10.7)
Race	n (%)
Black/African American	13 (5.5)
White/Caucasian	213 (90.3)
Multiple/Other	10 (4.2)
Employment status
Full-time	123 (52.1)
Part-time	20 (8.5)
Homemaker	16 (6.8)
Retired	47 (19.9)
Unemployed	8 (3.4)
On disability	19 (8.1)
Other (self-employed, student)	3 (1.3)
Insurancea
Any government insurance	58 (24.6)
Any non-government insurance	186 (78.8)
No insurance	3 (1.3)
Annual household income (n = 208)
≤ $30,000	25 (12.0)
> $30,000	183 (88.0)
Married/partnered	169 (71.6)
Family member providing most support (n = 235)
No one	9 (3.8)
Spouse/partner	152 (64.7)
Child	30 (12.8)
Parent	15 (6.4)
Grandparent	1 (0.4)
Sibling	13 (5.5)
Friend	11 (4.7)
Other	4 (1.7)
	Median (IQR)
Education (n = 135, years, range: 5–20)	16.0 (15, 18)
If married, number of years (n = 158, range: 1–57)	25.0 (14, 32)
Number of family members in home (range: 0–10)	2.0 (2, 3)
Number of children (range: 0–8)	2.0 (1, 3)

^aCategories are not mutually exclusive.

Table 4.

Clinical and treatment characteristics of phase two participants (N = 236).

	N	n (%)
Type of cancer	229
Breast		210 (91.7)
Gynecologic		19 (8.3)
Time since cancer diagnosis (years)	236
< 1		46 (19.5)
1–2		56 (23.7)
2–3		52 (22.0)
3–4		32 (13.6)
4–5		30 (12.7)
6+		20 (8.5)
Stage at diagnosis (breast cancer)	200
0		19 (9.5)
I		69 (34.5)
II		75 (37.5)
III		25 (12.5)
IV		12 (6.0)
Stage at diagnosis (gynecologic cancer)	14
I		1 (7.1)
IA		6 (42.9)
IIB		1 (7.1)
IIIB		2 (14.3)
IIIC		3 (21.4)
IVA		1 (7.1)
Surgical treatment	236	217 (91.9)
Chemotherapy treatment	236	144 (61.0)
Radiation treatment	236	135 (57.2)
Recurrence of cancer	236	26 (11.0)
Time since recurrence (years)	26
< 1		15 (57.7)
1–2		8 (30.8)
3–5		3 (11.5)
Treatment side effectsa	207
Lymphedema/swelling		66 (31.9)
Neuropathy/tingling		103 (49.8)
Fatigue		178 (86.0)
Memory issues		115 (55.6)
Heart-related problems		12 (5.8)

^aCategories are not mutually exclusive.

Structural/construct validity

All 36 items had sufficient response distribution variability for inclusion in the measurement structural analyses. Criteria included distributions where all possible response categories were represented for each item and the overall distribution had no less than 20% on either side of the distribution (i.e., at least 20% in the “Strongly Disagree” and Disagree categories or vice versa). Most distributions were considerably more balanced across all response options than the minimal criteria. The Kaiser-Meyer-Olkin (KMO) Test for sampling adequacy was 0.954 indicating more than adequate sampling for conducting the factor analyses. The Bartlett Test of Sphericity had a null probability of < .001 indicating our ability to detect underlying factor structure to the correlation matrix.

Our initial evaluations of the inter-item correlation matrix and confirmatory factor analysis did not demonstrate good convergence or fit for the proposed three factor model of family “Coherence,” “Closeness,” and “Function.” While the root mean squared error of approximation was within the good-adequate range (RMSEA = .07, 95% CI: .065–.076), the fit indices were considerably lower than the goal of at least 0.90 (CFI = 0.84, TLI = 0.33). Therefore, we conducted Principal Components and Principal Axis factor analyses. We determined that both consistently suggested four factors within a core set of 25 items. Eleven of the 36 items either loaded extensively on multiple factors suggesting multiple interpretations of the item intent or did not load on any of the four core factors. Furthermore, confirmatory analysis of the four-factor new structure demonstrated a considerably improved and parsimonious fit (CFI = 0.92, TLI = 0.91) with an RMSEA still within the good-adequate range (RMSEA = .07, 95% CI: .059–.075). Table 5 shows the standardized coefficients for the four-factor solution resulting from exploratory and confirmatory analyses. The resulting factors produced scores with generally good internal consistency values (factor Cronbach’s alphas ranged from 0.69 to 0.94; Table 5). The overall 25-item scale scores had a Cronbach’s alpha of 0.95 and a Flesch Kincaid grade level of 6.4.

Table 5.

Standardized coefficients resulting from exploratory and confirmatory analyses final solution (N = 236).

Item	Positive CFA (EFA)	(Dis)Connected CFA (EFA)	Distressing CFA (EFA)	Oppressive CFA (EFA)
Take on tasks when needed	.74 (.79)
Assist me when I ask for their help	.72 (.77)
Uses resources creatively	.73 (.69)
My concerns treated with respect	.83 (.67)
Believe we are a strong family	.76 (.65)
Try to understand my concerns	.84 (.64)
Works together to handle problems	.78 (.72)
Tend to express warmth	.80 (.61)
Cannot count on timely help*	.72 (.55)
Express encouragement	.82 (.67)
I feel comfortable asking for help.	.81 (.60)
Concerned about major threats to my health	.61 (.60)
Readily share my feelings about my illness		.62 (.68)
Avoid discussing concerns*		.74 (.71)
Avoids dealing directly with problems*		.78 (.40)
Withdraw when they are angry*		.59 (.57)
Tends to calmly discuss problems			.73 (.59)
Anger expressed frequently*			.62 (.72)
Takes a negative view of life*			.73 (.65)
Easily overwhelmed by problems*			.76 (.67)
Differing opinions respected			.77 (.49)
Discussions do not resolve problems*				.58 (.40)
Assume about help I need rather than asking*				.70 (.66)
Critical of way I manage my illness*				.62 (.76)
Try to control my care choices*				.50 (.70)

CFA = Confirmatory Factor Analysis; EFA = Exploratory Factor Analysis.

^*Item responses are reverse coded to enable combining into an overall score.

Table 6 shows how each item loaded on each subscale and the original planned HR-FQoL theoretical component it was designed to capture. To maintain the original item response scale as detailed above (range 1–4), each factor subscale was generated by averaging the response values for the items comprising the subscale. Four factors were identified.

Table 6.

HR-FQoL subscale items with planned theoretical components.

Item
	Planned theoretical component
Positive support
Take on tasks when needed	Family Functioning
Assist me when I ask for their help	Family Functioning
Uses resources creatively	Family Sense of Coherence
My concerns treated with respect	Emotional Closeness
Believe we are a strong family	Family Sense of Coherence
Try to understand my concerns	Emotional Closeness
Works together to handle problems	Family Sense of Coherence
Tend to express warmth	Emotional Closeness
Cannot count on timely help*	Family Functioning
Express encouragement	Emotional Closeness
I feel comfortable asking for help.	Family Functioning
Concerned about major threats to my health	Emotional Closeness
(Dis)Connected
Readily share my feelings about my illness	Emotional Closeness
Avoid discussing concerns*	Emotional Closeness
Avoids dealing directly with problems*	Family Sense of Coherence
Withdraw when they are angry*	Emotional Closeness
(Low) problematic support
Tends to calmly discuss problems	Emotional Closeness
Anger expressed frequently*	Emotional Closeness
Takes a negative view of life*	Emotional Closeness
Easily overwhelmed by problems*	Family Sense of Coherence
Differing opinions respected	Family Functioning
(Low) oppressive support
Discussions do not resolve problems*	Emotional Closeness
Assume about help I need rather than asking*	Family Functioning
Critical of way I manage my illness*	Family Functioning
Try to control my care choices*	Family Functioning

Positive family health-related quality of life (positive HR-FQoL)

This factor includes 12 items that reflect a mood of warmth and concern, clear and direct communication, and attempts to understand and respect patient concerns (Tables 4-6). One item is reverse coded prior to generating a positive factor score. Scores derived from these 12 items had a Cronbach’s alpha of .94 (Table 6). Higher scores could be interpreted as having more positive experiences of health-related family quality of life. A minimum of 10 item responses for the set of 12 items is required to generate the average score for this construct.

Disconnected family health-related quality of life (Disconnected HR-FQoL)

The four items in this factor reflect a sense of disconnected communication among family members. While family members may not communicate in a hostile manner, patients may feel uncomfortable talking about their concerns as direct communication about their concerns is typically thwarted by avoidance (Table 6). To enable creation of a continuum with higher scores reflecting more disconnected communication, one of the four items was reverse coded. The resulting scores had a Cronbach’s alpha of .78 (Table 7). A minimum of three item responses for the set of four items is required to generate the average score for this construct.

Table 7.

HR-FQoL subscale item-total correlations and reliability (N = 236).

Item
	Item-total correlation
Positive support (Cronbach’s alpha = .944)
Take on tasks when needed	.73
Assist me when I ask for their help	.71
Uses resources creatively	.70
My concerns treated with respect	.79
Believe we are a strong family	.74
Try to understand my concerns	.80
Works together to handle problems	.76
Tend to express warmth	.77
Cannot count on timely help*	.69
Express encouragement	.79
I feel comfortable asking for help.	.78
Concerned about major threats to my health	.59
(Dis)Connected (Cronbach’s alpha = .78)
Readily share my feelings about my illness	.53
Avoid discussing concerns*	.69
Avoids dealing directly with problems*	.59
Withdraw when they are angry*	.52
(Low) Problematic support (Cronbach’s alpha = .84)
Tends to calmly discuss problems	.65
Anger expressed frequently*	.59
Takes a negative view of life*	.65
Easily overwhelmed by problems*	.68
Differing opinions respected	.68
(Low) Oppressive support (Cronbach’s alpha = .69)
Discussions do not resolve problems*	.39
Assume about help I need rather than asking*	.54
Critical of way I manage my illness*	.54
Try to control my care choices*	.42

^*Item responses are reverse coded to enable combining into an overall score.

Note: Each of the subscale Cronbach’s alpha decreased if any of the single items. Comprising the respective scale was removed from that scale.

Distressing family health-related quality of life (Distressing HR-FQoL)

This factor contains negative attributes reflecting pessimistic tone, angry reactivity, avoidance of open discussion of concerns and feelings, and lack of respect for differing opinions (Tables 5 and 6). To enable a summary score that reflects higher levels of distress than lower scores, two of the five items comprising the distressing factor were reverse coded and the resulting factor scores had a Cronbach’s alpha of .84 (Table 7). A minimum of four item responses for the set of five items is required to generate the average score for this construct.

Oppressive family health-related quality of life (Oppressive HR-FQoL)

This factor is comprised of four items that reflect hostile, unhelpful, and critical qualities (Table 3). Higher scores on this factor reflect more “oppressive” qualities. The resulting factor scores had a Cronbach’s alpha of .69 (Tables 5, 6, 7). A minimum of 3 item responses for the set of 4 items is required to generate the average score for this construct.

Overall family health-related quality of life (overall HR-FQoL)

If desired, it is possible to generate an overall HR-FQoL score by reversing the direction of the average Disconnected, Distressing, and Oppressive subscale scores and then averaging those reversed scores along with the Positive scores. A higher score reflects a higher self-assessment of FHR-QoL. The overall score had a Cronbach’s alpha of 0.87. All four component subscale scores are required for computation of this overall HR-FQoL score.

Summaries of the subscales (factors) and overall HR-FQoL scores, as well as of the scores from each of the established family support and functioning measures are shown in Table 8. Median scores were generally in the middle of the possible range of 1–4 with considerable variability in each of the score distributions. As expected, inverse correlations of the Disconnected, Distressing, and Oppressive scores with the Positive HR-FQoL scores were observed (r_s: −0.74, −0.70, and −0.58, respectively). Correlations among those three subscales ranged from 0.50 to 0.64.

Table 8.

Descriptive summaries of scores from HR-FQoL and established measures.

	N	Median (IQR)
HR-QoL (Range 1–4)
Overall	236	3.1 (2.7, 3.5)
Positive HRF-QoL	236	3.3 (2.9, 3.7)
Connected HRF-QoL	236	2.8 (2.2, 3.3)
Problematic HRF-QoL	236	3.0 (2.8, 3.4)
Oppressive HRF-QoL	236	3.0 (2.7, 3.5)
Satisfaction with HR-FQoL (Range 1–4)	236	3.0 (3.0, 4.0)
Family APGAR (Range 0–10)	226	9.0 (7.0, 10.0)
CICS (Range: 4–20)	202	15.0 (12.0, 18.0)
McMaster FAD General Functioning (Range 1–4)	225	1.7 (1.2, 2.2)
MOS Emotional Social Support (Range 1–5)	225	4.5 (3.7, 4.8)
MOS Instrumental Support (Range 1–5)	226	4.6 (3.7, 5.0)
MOS Overall (Range 1–5)	225	4.4 (3.8, 4.9)
Family Sense of Coherence (Range 12–84)	221	66.0 (54.0, 72.0)
DIFS (Range 1–5)	226	4.2 (3.6, 4.6)
Marlowe Crowne Social Desirability (Range 0–13)	228	8.0 (6.0, 10.0)

Abbreviations: CICS = Couples Illness Communication Scale, MOS = Medical Outcome Study, DIFS = Differentiation in Family Systems.

Convergent validity

Associations of the HR-FQoL scores with the scores from established measures are summarized in Table 8. Given the sample size of 236, any correlation coefficient > ±0.18 would have been statistically significant (p < .05) and thus meaningless for informing construct validity. Therefore, the relative magnitude of the coefficients was used to evaluate convergent validity. Lower scores on the Family Assessment Device measure indicate better functioning. Thus, all of the correlations of the HR-FQoL subscale and overall scores were in the expected direction. The strongest of those correlations (in terms of absolute values) were with the Differentiation in Family Systems Scale and the Family Sense of Coherence (several r_s > ± 0.70). With the exception of the Marlowe Crowne measure of social desirability, the weakest correlations were observed for the Oppressive and Disconnected HR-FQoL scores with the MOS Social Support scores (r_s= 0.36 and 0.40, respectively). Finally, all of the HR-FQoL score correlations with the Marlowe Crowne were less than .30 demonstrating that social desirability did not play a large part in the participants’ responses (Table 9).

Table 9.

Correlations of HR-FQoL scores with established measures.

	Health-related family QOL score
Established measure score	Positive FHR-QoL(12 Items)	Connected FHR-QoL(4 Items)	Problematic FHR-QoL(5 Items)	Oppressive FHR-QoL(4 Items)	Satisfaction w/ FHR- QoL(3 Items)	Overall FHR-QoL(25 Items)
Family APGAR	0.64	0.61	0.57	0.46	0.61	0.68
CICS	0.47	0.52	0.43	0.41	0.47	0.54
McMaster FAD General Functioning Subscale	−0.71	−0.68	−0.74	−0.49	−0.68	−0.77
MOS Social Support— Emotional Social Support Scale	0.62	0.54	0.51	0.47	0.58	0.64
MOS Social Support— Instrumental Social Support Scale	0.52	0.40	0.41	0.36	0.46	0.52
MOS Social Support Short Form	0.62	0.50	0.49	0.45	0.56	0.62
Family Sense of Coherence	0.71	0.60	0.71	0.53	0.65	0.76
DIFS	0.78	0.65	0.74	0.61	0.75	0.83
Marlowe Crowne Social Desirability Scale SF	0.25	0.23	0.28	0.19	0.23	0.28

Note: Values in cells are Spearman coefficients.
	> = 0.70		> = −0.70
	0.60–0.69		−0.60–−0.69
	0.50–0.59

Abbreviations: CICS = Couples Illness Communication Scale, MOS = Medical Outcome Study, DIFS = Differentiation in Family Systems.

Discussion

The cluster analysis of responses from 236 cancer patients who participated in Phase Two along with empirically-driven decisions suggested to us that four patterns of family health-related quality of life reflected in four subscales with adequate reliability. These four subscales make up a final 25-item HR-FQoL instrument. The psychometric properties of this HR-FQoL instrument and each of its sub-scales demonstrated good internal reliability and construct validity with other scales that measured major concepts from the theory. As expected, the final 25-item HR-FQOL instrument (Table 4) and its four sub-scales directly or indirectly reflect the theoretical components of affectional closeness, family communication, family members’ roles/responsibilities, social support, perceived family abilities to cope, and family strengths (Figure 1). In addition, a three-item scale assessing the level of satisfaction with dimensions of HR-FQoL was identified.

Figure 1.

Theory of health related family quality of life.

The HR-FQoL instrument assesses the multi-dimensional nature of FHR-QoL. In reviewing the literature we found that there are consistent findings of positive associations between psychosocial support and quality of life in diverse patient populations (e.g., Anderson et al., 2019; Lalithambigai et al., 2019). We also found that family sense of coherence (Hsiao et al., 2018), problem-oriented coping (Fuochi & Foà, 2018; Kim & Han, 2020), close relationships (Merluzzi et al., 2019), and positive communication patterns (Lim & Shon, 2018; Psihogios et al., 2019) have also been positively correlated with psychological well-being and HR-QoL. The HR-FQoL instrument captures these aspects of family quality of life, along with perceived effectiveness of social support as an aspect of family functioning (Psihogios et al., 2019).

Limitations

As Connidis (2020) points out, who is included as “family” can be considered as (1) a tally (who is in and who is out), (2) based on the subjective meaning of those relations with important others that we consider to be “family,” (3) or including those with whom we have family-like relationships. Also, there are diverse social and cultural influences that shape who counts as family. In this study, we used a broad and inclusive definition of family to avoid limiting our sampling to specific family types and allowing participants to determine internally, for themselves, who they considered family. By not defining who is included or excluded as family members makes it difficult to determine how the instrument might be used to examine dyadic or group relationships within the family as they related to the patient.

The Phase Two sample was fairly homogenous (White, married) and reflected higher socio-economic status (i.e., highly educated, employed full-time). Therefore, the perspectives of individuals with different demographics are not represented, and the utility of the instrument in diverse samples is unclear.

Future directions

Preliminarily, our results suggest the need for future research to further refine and test the 25-item HR-FQoL instrument. Researchers should explore how responses on the HR-FQoL instrument may differ based on how family is defined (externally provided definition vs. internally). Given the homogeneous nature of participants, researchers should explore the use of this instrument across and among more diverse populations to determine its wider utility. Cultural differences in expectations for family life, as well as the delineation between roles and responsibilities of family members, may influence the experiences of HR-FQoL. The correlations within each sub-scale are high relation to what is acceptable. Further work is needed to explicate if these are different or measuring the same construct. Finally, we can only draw conclusions about the psychometric properties of the instruments in the current sample. Additional research is needed to determine the usefulness and efficacy of this tool for screening HR-FQol concerns in clinical settings

The resulting instrument may enable researchers and clinicians to assess various aspects of health-related family quality of life among female breast and gynecological cancer survivors. In clinical setting, use of the instrument may inform clinicians’ knowledge regarding the impact of illness on the family and shed light on the importance of support and connectedness in families and the role of these factors in how women with cancer adapt to and manage their cancer illness. This would inform their approach in caring for women with cancer and possibly lead to the adoption of interventions targeting both improvement in family functioning and clinical outcomes related to their illness. Further research is needed once the instrument is perfected to determine clinicians’ likelihood of utilizing this instrument as an assessment.