Fax: (617) 632-6080
Measuring psychosocial distress and parenting concerns among adults with cancer
The Parenting Concerns Questionnaire
Article first published online: 19 APR 2012
Copyright © 2012 American Cancer Society
Volume 118, Issue 22, pages 5671–5678, 15 November 2012
How to Cite
Muriel, A. C., Moore, C. W., Baer, L., Park, E. R., Kornblith, A. B., Pirl, W., Prigerson, H., Ing, J. and Rauch, P. K. (2012), Measuring psychosocial distress and parenting concerns among adults with cancer. Cancer, 118: 5671–5678. doi: 10.1002/cncr.27572
- Issue published online: 30 OCT 2012
- Article first published online: 19 APR 2012
- Manuscript Accepted: 28 FEB 2012
- Manuscript Revised: 11 FEB 2012
- Manuscript Received: 10 AUG 2011
- psychosocial aspects;
- quality of life;
A 2-phase, mixed methods study was conducted to develop a Parenting Concerns Questionnaire (PCQ) for adults with cancer. Limited information about this area of psychosocial distress highlights the need for a measurement tool that can identify adult oncology patients with heightened parenting concerns who could benefit from additional intervention.
Telephone focus groups were conducted with 16 oncology patients who had children 18 years old and younger. Group interview transcripts were analyzed to generate qualitative themes and candidate items for the PCQ. A 38-item version of the questionnaire was completed by 173 oncology outpatients who had children 18 years old and under. Participants also completed the Distress Thermometer, HADS (Hospital Anxiety and Depression Scale), and FACT-G (Functional Assessment of Cancer Therapy–General). Exploratory factor analyses revealed the emergence of 3 subscales of 5 items each, yielding a 15-item questionnaire. Associations between total PCQ scores, standardized measures of distress, depression, anxiety, quality of life, and demographic and illness characteristics were examined.
The 15-item PCQ demonstrates good internal consistency (Cronbach's α = .83). PCQ scores were significantly associated (P < .01) with standardized measures of psychosocial distress (Distress Thermometer, HADS, and FACT-G) in the expected directions. Higher PCQ scores were associated with female sex, single parenthood, metastatic or recurrent cancer, subjective understanding of incurable disease, comorbid chronic health condition, and current mental health treatment.
The PCQ proved a reliable and valid measure of parenting distress among cancer patients, and thus merits further study. Cancer 2012. © 2012 American Cancer Society.
Approximately 22% of incident cases of cancer occur in adults between 20 and 55 years of age,1 during a time of intense social and vocational activity. For many of these adults, parenting is a primary role, and this essential activity and aspect of identity is often unattended to by multidisciplinary oncology clinical and research teams.2 Recent epidemiological data estimate that 18% of adults within 2 years of initial diagnosis, and 14% of a large sample of survivors (N = 13,385, mean years since diagnosis = 7.9 years) are parenting minor children, and an estimated 1.6 million adults and 2.85 million children are affected by parental cancer.3 Because parenting status is not routinely documented in cancer epidemiology, research, or clinical care, we have no actual prevalence data, and the above epidemiological data are acknowledged to be underestimates. Commonly used quality-of-life research instruments may ask about general family concerns and only rarely about children in particular.4 Instruments that do ask about children use a single general item, limiting capacity to assess psychosocial needs, the impact of having cancer on parenting, or the role of parenting status on oncology treatment decisions. There is no published measure that identifies concerns about the impact of parental cancer on children. The paucity of existing data highlights the need for a tool that could be used in research and clinical settings to identify patients who are struggling with parenting concerns.
Parenting status is known to affect medical decision-making in both early stage and advanced illness. Having children living at home predicts willingness to accept more aggressive treatment in patients with a range of disease stages.5 Women with early stage breast cancer and dependent children required smaller benefits in life expectancy or survival rate to make adjuvant chemotherapy worthwhile, compared with women without dependent children.6, 7 Among patients with advanced cancer, those with dependent children had higher rates of panic and worry than those without dependent children. These subjects were more likely to prefer more aggressive treatment and to be less likely to engage in advanced care planning. This same population of patients with dependent children was also judged by their caregivers to have worse quality of life in the last week of life.8
The impact of parental cancer on children is also significant; latency-age boys and adolescent girls are specifically at higher risk of psychosocial problems such as anxiety, depression, and aggression than norms.9 Consistent with other populations,10 risk for child internalizing disorders is higher among children whose parents not only have cancer, but are also depressed. Importantly, rates of depression among parents with cancer are high (up to 35%) and are associated with impaired family function as well as child externalizing disorders.11 The extent to which parental depression is attributable to parenting concerns is unknown.
In an initial effort to fill the information gap regarding patient concerns about their children, this study used qualitative and quantitative methods to develop a parenting concerns measure that could be used for clinical and research purposes. We evaluated the psychometric properties of the resulting scale and examined the correlations between parenting concern scores, other psychometrically sound measures of psychological distress and quality of life, and demographic and illness variables. We hypothesized that higher parenting concern scale scores would be positively associated with more psychological distress and worse quality of life.
MATERIALS AND METHODS
After approval from institutional review boards at the Massachusetts General Hospital Cancer Center and the Dana-Farber Cancer Institute, both in Boston, Massachusetts, instrument development proceeded in 2 phases: 1) qualitative focus groups and 2) quantitative analyses to reveal emergent factors and items to retain. The resulting 15-item pilot instrument then underwent cognitive testing.
Qualitative Focus Groups
Adults with cancer who had children 18 years old or younger were recruited using flyers posted in outpatient oncology clinics and participated in telephone focus groups between January and April 2008. To engage parents in discussion of developmentally informed issues, participants were divided into groups of 3 to 5 participants by the age of their children: under 4 years, 4 to 8 years, 9 to 12 years, and 13 to 18 years old. Focus groups were recorded and transcribed. Interviewers used a semistructured script with questions about the impact of cancer on the family, concerns about children, and changes in interactions with children since diagnosis. Three independent raters manually coded transcripts using an iterative, multistep process of content analysis to develop major themes and subthemes. Coding disagreements were resolved by consensus and comparison to the raw data. Themes were then used to create scale items.
Introductory letters and survey packets were placed in outpatient clinic waiting rooms at 2 large urban cancer centers. Self-selected adult oncology outpatients with children 18 years old and younger anonymously completed the packet of self-report assessments described below and returned them during clinic visits between October 2008 and August 2009.
Measures for Quantitative Phase
Sociodemographic and Illness Characteristics
Participants were asked to provide information regarding age, sex, race/ethnicity, marital status, income, education, cancer diagnosis and stage, recurrence and metastases, time since diagnosis, treatment modalities and frequency, and mental health diagnoses and treatment. They were also asked to subjectively rate their current health status and goal of cancer therapy.
Participants completed the 38 candidate questionnaire items that were derived from focus group themes. Items were scored on a 5-point Likert scale where 1 = “not at all concerned” and 5 = “extremely concerned.”
Quality of Life
Participants completed the Functional Assessment of Cancer Therapy–General (FACT-G), a 27-item questionnaire with 4 subscales assessing physical, social/family, emotional, and functional well-being on a 5-point Likert scale where 1 = “not at all” and 5 = “very much.” The total FACT-G has a coefficient α = .89, with 51% of variance accounted for.12 The general adult population norm is a mean total score of 80.1 (standard deviation = 18.1).13
Depression and Anxiety
The Hospital and Anxiety Depression Scale (HADS) is a 14-item self-reported measure that has been well tested in cancer populations. It has two 7-item subscales assessing depression and anxiety.14 An overall cutoff score of 15 or greater results in 80% sensitivity, 76% specificity, and a positive predictive value of 41%.15
The Distress Thermometer uses a single-item rating of distress from 0 to 10 over the past week, and participants checked off current concerns on 32-item problem list that includes “dealing with children.”16 This measure compared favorably with the HADS and Brief Symptom Inventory 18, and was found to have optimal sensitivity and specificity at a cutoff score of 4.17
In this first phase of scale development, we had no a priori hypotheses about the nature of the measure's underlying factor structure. We conducted an exploratory factor analysis to assess the relationships between the items and identify underlying factors. Factor extraction assists in developing reliable factors from populations of questionnaire items. Alpha factor analysis with Varimax rotation was employed to facilitate item reduction.18 The 5:1 subject-to-item ratio falls within current practice standards in factor analysis.19 The SPSS Predictive Analytics SoftWare (PASW) Statistics 17.0 statistical software package was used for all statistical analyses.
Cognitive Testing of the 15-Item Questionnaire
Ten additional subjects in active cancer treatment were interviewed to explore their comprehension, interpretation, and level of comfort with each of the final 15 items.
A total of 16 adults (13 female) with a range of cancer diagnoses participated in qualitative focus groups. Qualitative analyses resulted in 5 major themes and 4 subthemes, which were reflected in the first version of the Parenting Concerns Questionnaire (PCQ) instrument.
Themes included 1) communication with children, 2) children's current and future coping or behavior, 3) parenting identity and evaluation of own parenting, 4) partner's ability to support patient and children, and 5) changes in parenting. The last theme included the relational impact of the ill parent's emotional, physical, and cognitive changes, as well as practical changes in social support and daily routines.
From these themes, coders created a pool of 38 candidate items, scored on a 5-point Likert scale with the prompt “In the past month, I have been concerned that…”.
A total of 194 subjects responded to the survey, and 175 completed at least 34 of the 38 candidate parenting concerns items (<10% missing items); there was a 90% completion rate. A total of 173 subjects completed these parenting items and all of the standardized instruments, and were included in the analyses of external validity. Median age was 46 years (range, 26-65 years). Sample characteristics are described in Table 1. Illness characteristics and subjective assessments of health status are listed in Table 2.
|Sex (female)||141 (73)|
|American Indian||2 (1)|
|Native Hawaiian/Pacific Islander||1 (1)|
|Black/African American||6 (3)|
|Living with spouse or partner||172 (89)|
|Less than high school||1 (0.5)|
|High school graduate/GED||25 (13)|
|2-4 y college||111 (57)|
|Graduate degree||56 (29)|
|Treatment in past month|
|Intravenous chemotherapy||100 (52)|
|Oral chemotherapy||28 (14)|
|Hormonal therapy||13 (7)|
|Relatively healthy and not likely to die within 6 months||166 (86)|
|Relatively healthy and likely to die within 6 months||2 (1)|
|Seriously ill but not likely to die within 6 months||16 (8)|
|Seriously ill and likely to die within 6 months||1 (0.5)|
|Goal of treatment:|
|Cure so no cancer in body||107 (55)|
|Treatment to live as long as possible with illness||72 (37)|
|Palliative treatment||3 (2)|
|Cure and treatment||7 (4)|
|Has another chronic health problem||88 (45)|
|Ever treated for mental health problem||100 (52)|
|Currently receiving treatment for mental health problem||65 (34)|
Factor Analysis and Item Reduction
Following recommendations for exploratory factor analysis best practices,18 the scree test guided the decision about how many factors to retain. The scree plot suggested a 3-factor solution; eigenvalues for the first 5 factors were 9.29, 3.49, 2.90, 1.88, and 1.72. Examination of the item loading tables for 2-, 3-, and 4-factor solutions suggested that the 3-factor solution had the cleanest factor structure. These factors were: 1) practical impact of the illness on children, 2) emotional impact of the illness on children, and 3) concerns about the coparent. The 3-factor solution accounted for 41.3% of the item variance and was used for all subsequent analyses of individual items.
Three items did not have factor loadings ≥0.32 on any of the 3 factors and were dropped, to ensure adequate overlapping variance among items on each factor.18 Also dropped were 12 items with variance not well explained by the 3 factors (communality <0.40). Six items with borderline communalities were retained for content. Five items cross-loaded on 2 factors (factor loadings ≥0.32); 4 were dropped, and 1 was retained for content.
Inter-item correlations among the remaining items were examined. Three items had high correlations with other retained items (r ≥ 0.60). Two were dropped to minimize redundancy, and 1 was retained for content. Two other items were dropped for redundant content. The final PCQ scale has 15 items loading on 3 factors, with 5 items on each factor. These items represented all of the themes yielded by the qualitative analyses of focus group transcripts supporting face validity. Item means and standard deviations are shown in Table 3.
|Factor 1: Practical impact of illness on child|
|My own mood, worries or emotions are affecting my children||2.92||1.08||192||0.64||0.08||0.03|
|My physical limits or low energy are affecting my children||2.88||1.19||193||0.53||0.00||0.29|
|I am not able to spend as much time with my children as I'd like||2.86||1.41||190||0.72||0.16||−0.09|
|My illness is changing my children's routine||2.34||1.21||190||0.65||0.17||0.19|
|Changes in my memory or attention are affecting my children||2.16||1.13||188||0.63||0.27||0.14|
|Factor 2: Emotional impact of illness on child|
|My children are emotionally upset by my illness||3.12||1.18||190||0.33||0.65||−0.09|
|My children are worried that I am going to die||2.63||1.33||185||0.20||0.55||0.07|
|My children get upset when we talk about the illness||2.12||1.07||187||0.12||0.55||0.17|
|My children might be in need of professional mental health care||2.10||1.11||188||0.05||0.78||0.11|
|My children get confused or upset by what others say about my illness||1.86||0.93||183||0.13||0.47||0.15|
|Factor 3: Concerns about coparent|
|My children's other parent would not be able to meet their emotional needs if I die||2.45||1.40||183||0.15||0.15||0.70|
|There is no one to take good care of my children if I die||1.89||1.21||190||0.01||0.15||0.60|
|My partner is not providing me with enough practical support||1.88||1.23||178||−0.03||0.19||0.53|
|My partner is not providing me with enough emotional support||1.83||1.24||173||−0.14||0.19||0.62|
|My children's other parent would not be a responsible caregiver if I die||1.72||1.25||185||0.14||0.01||0.67|
Individually interviewed subjects demonstrated good comprehension and accurate interpretation of each of the 15 questions. They felt that items represented important areas of concern. Some subjects expressed mild discomfort with the question about the possibility of death, but all recognized the importance of asking about this concern and did not endorse deleting it.
Cronbach's alpha for the 15-item PCQ scale = .83, n = 141. Cronbach's alpha for each subscale are as follows: 1) practical impact of the illness on children = .79, n = 185; 2) emotional impact of illness on children = .79, n = 171; 3) concerns about coparent = .85, n = 157.
The Distress Thermometer, FACT-G, and HADS data were tabulated and scored according to instrument directions.
Table 4 lists mean total scores for the 15-item PCQ, the 3 subscales, and the HADS, FACT-G, and Distress Thermometer. Mean scores fall in the mild to moderate range of clinical severity. The mean Distress Thermometer score of 4.31 just crosses the threshold for clinical intervention (= 4), and HADS Anxiety scale mean of 7.6 was just over the clinically significant cutoff score (= 7).
|PCQ Total score||2.33||.68||182||1.00||4.57|
|PCQ Subscale 1: Practical impact||2.64||.89||189||1.00||4.80|
|PCQ Subscale 2: Emotional impact||2.38||.82||185||1.00||4.40|
|PCQ Subscale 3: Coparent||1.95||1.03||179||1.00||5.00|
|FACT-G total score||70.29||18.94||176||0||108.00|
There were moderate correlations20 in expected directions between PCQ total scores and these standardized measures (r values = 0.30-0.59, P values < .05) (Table 5). This suggests that being concerned about parenting issues is related to anxiety, depression, functioning, and overall distress. However, the correlations also indicate that the PCQ captured a unique aspect of participants' concerns, and support the discriminant validity of this new measure. The PCQ is expected to measure specific aspects of patient distress related to parenting, and thus scores may be similar to other areas of emotional distress such as anxiety or depression. However, these scores are not identical and therefore the PCQ may capture previously unmeasured forms of distress among adult cancer patients.
|Variable||PCQ total score||Subscale 1: Practical||Subscale 2: Emotional||Subscale 3: Coparent|
|FACT-G total score||−0.59**||−0.64**||−0.44**||−0.24**|
Total PCQ scores were normally distributed and t tests were used to compare scores between different groups of patients. Patients who were female, single, had metastatic or recurrent disease, had another chronic health condition, and were in current mental health treatment had significantly higher mean PCQ scores. Patients who described the goal of their treatment as living as long as possible with the disease, versus being cured, also had significantly higher levels of parenting concern (Table 6).
|Metastatic disease||−3.83 (166)***|
|Cancer recurred||−2.69 (162)**|
|Receiving mental health treatment||−2.38 (166)*|
|Other chronic health problems||2.66 (180)**|
|Single parent||−3.52 (179)***|
|Active treatment in past month||−1.96 (180) T|
|Goal of treatment||−3.85 (168)***|
|Cure (no cancer left in body)||2.16||0.56||101|
|To live as long as possible with illness||2.54||0.71||69|
A 1-way, between-subjects analysis of variance examined the relationship between reported income level as a categorical variable, and parenting distress. Total PCQ scores had homogeneous variance across income levels as demonstrated by Levene's test. There was a significant overall effect [F(4,173) = 3.38, P = .011], with lower income associated with higher levels of parenting concerns, especially at incomes below $20,000 (median = 2.95, standard deviation = 0.86).
To our knowledge, this study is a first comprehensive approach to measuring parenting concerns among adults with cancer, and is an initial step in identifying and addressing the specific needs of millions of parents with cancer and their families. The PCQ demonstrates good internal consistency and face validity, and PCQ scores had significant but moderate correlations with standardized measures of distress, suggesting that parenting issues may be a unique contributor to psychological distress associated with cancer. The utility of a new parenting concerns instrument may be to capture the psychological distress specifically associated with parenting that is not fully identified by standardized tools for general distress, anxiety, or depression.
This study is limited by the sample demographics, because the subject population was predominantly female, caucasian, well-educated, and married. Research with more diverse subject populations might reveal different patterns of overall distress and parenting concern, additional psychometric data, as well as more specific associations with standardized measures of psychosocial distress or parenting stress.
Our 3-factor solution explained 41.3% of the total variance in the original pool of 38 items. There often remains much unexplained variance in the original correlation matrix following exploratory factor analysis. A weakness of this study is that the ratio of subjects to items was 4.6:1, which is toward the lower end of recommended ratios,20 and may result in errors in eigenvalues and factor loadings. Replication of this exploratory factor analysis in other populations is needed to determine the stability of the factors identified here, followed by testing with confirmatory factor analysis. Exploration of the PCQ relative to other measures of parenting distress, such as the Parenting Stress Index,21 might provide specific criterion validity, although this measure asks more about inherent child issues, rather than specific stressors such as parental illness.
This study assesses parenting concerns, an unaddressed area of distress among adults with cancer that may be related to parental depression and anxiety, as well as medical decision-making in this population. This focus on subjective parental concern is distinct from examination of the children's distress and yet is pertinent to the need for intervention around the experience of the children living with adults with cancer.22 Interactions between parental distress and child distress have been noted and bear further exploration.23 Challenges in communication between parents and children about the illness are also cited as a source of distress and misconceptions for children.24–26 Several of the original 38 candidate PCQ items about communication were in the borderline range of factor loadings. Factor loadings for the current 3 factors demonstrates simple structure, yet additional items in the areas of communication, children's misconceptions, and parental attributions about child distress bear further exploration. In research populations, a more comprehensive PCQ may eventually be useful to specifically assess the role of parenting in quality of life and medical decision-making.
This largely well-educated, married, self-selected sample indicated mild to moderate concerns about their children, and slightly increased levels of depression and anxiety, and quality of life concerns. However, there is a subset of patients with elevated PCQ scores, potentially signaling a higher risk population of women with advanced illness, or comorbid health and mental health diagnoses, who are most concerned about their children. Parents with a subjective understanding of their disease being incurable also had higher levels of concern. Single parenthood and low income may be additional risk factors and may confound one another in these analyses. These populations with increased levels of parenting concern may benefit from further assessment and intervention.
Interventions for families coping with medical illness have used cognitive behavioral strategies,27 psycho-education, and parent guidance to support child coping,26, 28 with the premise that communication with children about parental illness may reduce child anxiety.24 The parental end-of-life period is known to be particularly stressful for children,29, 30 and literature on bereaved children31, 32 emphasizes the important role of family functioning and the surviving parent.
Clinical experience,26 and data in early-stage treatment decisions6, 7 and end-of-life decision-making8 indicate that having dependent children is a primary mediator for the cancer experience. Studies showing high rates of parental depression during cancer and its impact on family and child function11, 23, 33 further support the need for specific intervention for these patients. It is possible that directly asking about parenting concerns may facilitate discussions about medical decision-making, end-of-life care, and quality of life for patients and their families.
In quality-of-life research, the refined PCQ may eventually identify parenting issues as mediators of distress and decision-making in oncology populations. As a subjective measure of parental concern, such an instrument may also become helpful in clinical screening to help oncology clinicians initiate discussions about family with patients, and guide intervention around parenting, or help with evaluation before and after such interventions. Evidence-based clinical interventions could ultimately improve outcomes for adult oncology patients and their families by mediating parent and child distress, and clarifying medical decision-making in this patient population.
We acknowledge Nancy Borstelman, LICSW, and Rebecca Babcock, BA, for their help in subject recruitment.
Research supported by National Institutes of Health grant R03 CA126394-01.
CONFLICT OF INTEREST DISCLOSURE
The authors made no disclosure.
- 1Howlader N, Noone AM, Krapcho M, Neyman N, Aminou R, Waldron W, Altekruse SF, Kosary CL, Ruhl J, Tatalovich Z, Cho H, Mariotto A, Eisner MP, Lewis DR, Chen HS, Feuer EJ, Cronin KA, Edwards BK, eds. SEER Cancer Statistics Review, 1975–2008. Bethesda, MD: National Cancer Institute. Available at: http://seer.cancer.gov/csr/1975_2008/, based on November 2010 SEER data submission, posted to the SEER web site, 2011.
- 18Using Multivariate Statistics. 4th edition. Boston, MA: Allyn & Bacon; 2001., , editors.
- 19Best practices in exploratory factor analysis: four recommendations for getting the most from your analysis. Pract Assess Res Eval. 2005; 10: 173-178., .
- 20Statistical Power Analysis for the Behavioral Sciences. 2nd edition. Hillsdale, NJ; Lawrence Erlbaum Associates; 1988..
- 28Massachusetts General Hospital Cancer Center. PACT–Parenting At a Challenging Time. http://www.mghpact.org.