Identifying subgroups of well-being among patients with cancer: Differences in attitudes and preferences around surveillance after curative-intent surgery

Abstract

Background

Patient perceptions and preferences related to postoperative surveillance are not yet well defined.

Methods

A cross-sectional analysis of the surveillance practice preferences and attitudes was undertaken based on subgroups derived from clustering participants for measures of well-being, including financial toxicity, emotional, family/social, and functional well-being.

Results

Among 212 participants, the average age was 58.1 years and most patients were female (57.1%) and white (90.2%). Common malignancies included melanoma/sarcoma (26.4%), thyroid (25.5%), breast (18.9%), gastrointestinal (18.4%), and lung (7.5%) cancer. Respondents within the highest well-being subgroup rated their perception of communication as being the highest more consistently compared with the other well-being subgroups (P = .005). Participants with the highest level of well-being felt more reassured by follow-up appointments (Subgroup 1, Med = 4.00, interquartile range (IQR) = 0.25 vs subgroup 4, Med = 3.75, IQR = 0.73, P = .023). In contrast, patients with the lowest sense of well-being had the highest level of nervousness related to surveillance (subgroup 1, Med = 1.60, IQR = 1.00 vs subgroup 4, Med = 2.20, IQR = 1.15, P < .001). There were no differences in surveillance frequency preferences among different well-being subgroups.

Conclusion

Attitudes towards postoperative surveillance varied with regard to perception of provider communication, nervous anticipation, and assuredness depending on overall patient well-being. Providers should attempt to assess patient well-being as part of a tailored approach to postcancer surgery surveillance.

1 |. INTRODUCTION

More than one-half of patients diagnosed with cancer will undergo some type of surgical procedure as a part of their treatment.¹ Advances in systemic therapy and surgical options have led to an improved prognosis for many patients diagnosed with cancer and, thus, a growing population of cancer survivors.^2,3 Current models of postoperative cancer surveillance among survivors may, however, be expensive, ineffective, and therefore not sustainable.^4–7 In addition, the volume of surveillance visits may be a burden on the health care system, as well as providers, survivors, and family members.^8–10 Despite widespread adoption of frequent routine clinic appointments, there is a dearth of empirical data on surveillance practice patterns or patient preferences regarding postsurgical cancer care.¹¹

Although patient-centered care (PCC) is widely recognized as the gold standard for cancer care, many providers often take a “one-size-fits-all” approach and treat all patients the same despite variations in patient needs and resources.^12–14 Patient preferences around the topic of surveillance have not been well-examined and whether current practices are patient-centered remains poorly characterized.¹¹ Providing optimal, tailored PCC should include assessment of patient-specific wishes related to both decisions about the surgical procedure itself, as well as how surveillance is performed in the years following operative intervention. In fact, patients can experience cancer-specific anxiety related to undergoing active surveillance.^4–7 In turn, cancer-related anxiety can be associated with depression, mental health issues, and adversely impact patient quality-of-life.^4–7 As such, data to understand patient perception, as well as preferences, related to postsurgical cancer treatment including postoperative surveillance and follow-up, may be important. Therefore, the objective of the current study was to evaluate potential differences in attitudes and preferences related to follow-up among patients who had undergone curative-intent surgery for a solid malignancy. Specifically, using cluster analysis methodology, we sought to define the potential association of patient well-being with surveillance practice preferences across various subgroups among patients who had a curative-intent resection for cancer.

2 |. METHODS

Participants were recruited during follow-up visits held with their providers at The Ohio State University Comprehensive Cancer Center—Arthur G. James Cancer Hospital and Richard J. Solove Research Institute (OSUCCC-James). Potential participants were prescreened for inclusion by research personnel who had identified the appointment as “return/surveillance” on the clinic schedule. Additional inclusion criteria for participants were aged ≥18 years old, underwent curative-intent resection for a solid tumor by an OSUCCC-James surgeon and had no evidence of active disease. Members of the research team approached patients in the waiting room before the appointment with the surgeon. If the patient was interested in participating, the research personnel reviewed the consent form and administered the survey to patients via Qualtrics on a tablet or hard-copy survey.¹⁵ All data were stored using Qualtrics until the data were exported for analyses. The study was approved by The Ohio State University Wexner Medical Center Institutional Review Board (protocol #2017C0190).

2.1 |. Assessment instrument

A cross-sectional descriptive survey methodology was utilized. The survey was part of a larger study that focused on patient preferences for postoperative surveillance practices.¹⁶ For the purposes of the current study, questions from established, reliable, and valid questionnaires related to patient well-being, demographic factors, and attitudes/preferences towards follow-up were used. Demographic variables were collected including age, race, income, and current relationship status. Cancer demographics were obtained from the patient medical record including date of cancer diagnosis, operation type, date of cancer operation, and current surveillance schedule. Birthdate and gender were included in the survey to cross-reference and validate the accuracy of the information obtained from the medical record. Determinants of well-being included financial toxicity, social/family well-being, emotional, and functional well-being. Levels of patient financial toxicity were assessed using the COmprehensive Score for Financial Toxicity (COST), from the Functional Assessment of Chronic Illness Therapy (FACIT) measurement system. The COST scale has 11 items with strong internal validity (Cronbach’s alpha = .90).¹⁷ Psychosocial well-being was assessed with the emotional, family/social, and functional subscales from the General Functional Assessment of Cancer Therapy Questionnaire (FACT-G). FACT-G was designed for use among patients with any cancer type. Responses to questions were on a five-point scale and ranged from 0 (not at all) to 4 (very much); Cronbach’s alpha statistic for the scale (α = .89) and subscales (α = .69–.82) was strong. Reliability of the measure has also been established; test-retest correlation coefficients for the total scale (r = 0.92) and subscales (r = 0.82-0.88) were strong for the 3 to 7 day readministration period.^18,19

The reassurance, nervous anticipation and communication subscales of the Attitudes Towards Follow-up Scale (ATF). The ATF questions were measured with Likert scales ranging from 1 to 4 and had demonstrated internal reliability (α = .77).²⁰ Investigator-created questions were used to assess preferences for follow-up, including frequency of visits, comfort level seeing another provider, and communication of abnormal/normal blood and imaging results.

2.2 |. Statistical analysis

Descriptive statistics for continuous and categorical variables were presented as median (interquartile range [IQR]) and frequency (%), respectively. Differences among patient subgroups, as well as demographic and patient characteristics were examined using Kruskal-Wallis one-way analysis of variance and χ² tests for continuous and categorical variables, respectively. To identify unique patient subgroups based on all four measures of patient well-being, patients were clustered on the averages of each measure. Ward’s minimum-variance hierarchical method was used to identify and classify patient subgroups. This method of unsupervised learning did not require specifying a priori the number, types, or characteristics of the different patient subgroups. Instead, an unsupervised learning approach grouped patients in a way that maximized differences between subgroups and minimized variability within each subgroup.^21,22 The cluster analysis was performed using SAS v9.4 and all other analyses were performed using SPSS v.24.²³ Statistical significance was assessed at α = .05.

3 |. RESULTS

Among 300 potential participants who were screened for study inclusion, 271 individuals completed the survey (response rate, 90.3%). Following review of survey responses, 14 participants were excluded due to index operation performed not at OSUCCC-James (n = 6), surveillance was being performed for a premalignant condition (n = 3), or patient had not undergone curative resection (n = 3); 47 patients did not complete one of more of the well-being measures and were therefore also excluded from the analysis. The final analytic cohort consisted of 212 participants.

3.1 |. Baseline participant characteristics

The average age of study participants was 58.1 years (SD = 13.5; range, 23.3-88.5) and most patients were female (57.1%) and white (90.2%). Less than one-half of respondents had a college degree or higher (55.7%) and the majority of individuals had a combined household income less than $100 000 (<$50 K, 34.9%, $50-$99 K, 24.0%, $100-$150 K, 17.5%, >$150 K, 18.4%) (Table 1). The most common diagnosis was melanoma/sarcoma (26.4%); other cancer diagnoses included thyroid (25.5%), breast (18.9%), gastrointestinal (18.4%), and lung (7.5%). Of note, the majority of respondents (81.6%) reported currently not receiving any cancer treatment, while a minority reported receiving chemotherapy (6.6%) and/or radiation therapy (1.9%); few patients (6.6%) reported receiving “other” treatments. On average the survey was completed 4.95 years (SD = 6.08; range, 0-30.47) after curative-intent surgery.

TABLE 1.

Differences in participant demographic variables between subgroups

	Frequency (% within subgroup)
	Total n = 212	Subgroup 1 n = 83	Subgroup 2 n = 72	Subgroup 3 n = 21	Subgroup 4 n = 36	P value
Sex						.730
Male	84 (39.6)	33 (41.3)	30 (42.9)	9 (47.4)	12 (33.3)
Female	121 (57.1)	47 (58.8)	40 (57.1)	10 (52.6)	24 (66.7)
Concurrent malignancy						.903
No	172 (83.9)	66 (82.5)	59 (84.3)	17 (89.5)	30 (83.3)
Yes	33(16.1)	14 (17.5)	11 (15.7)	2 (10.5)	6 (16.7)
Insurance						.348
Private	138 (65.1)	54 (69.2)	44 (62.9)	16 (84.2)	24 (70.6)
Medicare	67 (29.7)	24 (30.8)	26 (37.1)	3 (15.8)	10 (29.4)
Cancer Dx						.009
Breast	40 (18.9)	16 (20.0)	12 (17.1)	1 (5.3)	11 (30.6)
Sarcoma	56 (26.4)	16 (20.0)	23 (32.9)	6 (31.6)	11 (30.6)
Lung	16 (7.5)	7 (8.8)	4 (5.7)	2 (10.5)	3 (8.3)
GI	39 (18.4)	12 (15.0)	21 (30.0)	1 (5.3)	5 (13.9)
Thyroid	54 (25.5)	29 (36.3)	10 (14.3)	9 (47.4)	6 (16.7)
Income						.001
<60 K	77 (36.3)	18 (23.7)	29 (43.9)	8 (44.4)	22 (62.9)
>60 K	118 (55.7)	58 (76.3)	37 (56.1)	10 (55.6)	13 (37.1)
Education						.558
<BA/BS	118 (55.7)	41 (52.6)	44 (63.8)	11 (61.1)	22 (61.1)
≥BA/BS	83 (39.2)	37 (47.4)	25 (36.2)	7 (38.9)	14 (38.9)

Abbreviations: BA/BS, Bachelor of Arts/Bachelor of Science; GI, gastrointestinal.

3.2 |. Patient subgroups relative to well-being factors

Cluster analysis identified a four-factor solution with differences among groups based on three of the four well-being subscales, including social/family well-being, emotional well-being, and functional well-being (all P < .001; Figure 1). In addition, there was a trend toward a difference in the subgroups relative to the financial toxicity scale (P = .08; Table 2). On the basis of an unsupervised learning approach to the data, patients were stratified into four subgroups based on overall self-reported well-being (ie, subgroup 1, highest composite well-being vs subgroup 4, lowest composite well-being, with subgroups 2 and 3 being mixed).

FIGURE 1.

Median scores on scales of well-being by cluster subgroup

TABLE 2.

Summary scores of participant well-being by cluster subgroup

	Median (IQR)					P value
	Total n = 212	Subgroup 1 n = 83	Subgroup 2 n = 72	Subgroup 3 n = 21	Subgroup 4 n = 36
Financial toxicity	3.18 (0.36)	3.18 (0.36)	3.18 (0.45)	3.09 (0.27)	3.18 (0.27)	P=.08
Social/family well-being	4.67 (1.0)	5.00 (0.17)	4.67 (0.57)	4.00 (0.43)	3.59 (0.86)	P<.001
Emotional well-being	4.33 (0.83)	4.83 (0.50)	4.17 (0.67)	4.67 (0.33)	3.58 (1.17)	P<.001
Functional well-being	4.29 (1.15)	4.86 (0.29)	4.00 (0.50)	4.43 (0.54)	2.93 (1.0)	P<.001

Abbreviation: IQR, interquartile range.

Demographic and patient characteristics were largely comparable among the four subgroups (Table 1). Specifically, there were no differences among patients in the four different self-reported well-being subgroups with regard to age, sex, insurance type, or education level (all P > .05). There were, however, differences in other factors such as cancer diagnosis and income level, as well as whether a patient brought an additional person regularly to surveillance appointments (Table 1). Specifically, patients in ongoing surveillance for a history of thyroid cancer were more likely to report the greatest sense of well-being (subgroup 1, 53.7% vs subgroup 4, 11.1%); in contrast, patients who were in surveillance for a history of melanoma/sarcoma or a gastrointestinal cancer were more frequent in subgroup 2 (41.1% and 53.8%, respectively vs subgroup 1, 28.6% and 30.8%, respectively; subgroup 3, 10.7% and 2.6%, respectively; subgroup 4, 19.6% and 12.8%, respectively, both P < .05; Figure 2). Whereas approximately half of the respondents who reported the greatest sense of well-being (subgroup 1) had an income level over $60,000 (49.2%), patients with a lower income were more likely to report lower well-being (subgroup 2, 3, and 4 were 31.4%, 8.5%, and 11.0%, respectively; P = .001). In addition, almost three-quarters of patients in the lowest self-reported well-being group (subgroup 4, 72.2%) reported that someone regularly attend follow-up appointments with them; in contrast, only roughly one-half of participants who reported the greatest sense of well-being noted that someone routinely joined them at their surveillance visits (subgroup 1, 57.5% vs subgroup 2, 65.3%, vs subgroup 3, 27.8%; P = .012).

FIGURE 2.

Frequency of cancer diagnosis by cluster subgroups

3.3 |. Differences between patient subgroups regarding attitudes and preferences towards follow-up

While most patients reported being satisfied with their surveillance care, attitudes toward cancer surveillance did differ across cluster subgroups (Table 3). In general, most patients reported high satisfaction around communication with their surgical oncologist pertaining to surveillance (Med = 4.0, IQR = 0.25). However, respondents with the highest levels of reported well-being (subgroup 1) were the most likely to rate their perception of communication as being the highest, while there was more variability in communication scores across the other three well-being subgroups (P = .005). Similarly, while participants reported feeling reassured by their follow-up appointments overall (Med = 3.75, IQR = 0.50), there was variation among patients based on the reported well-being. Specifically, participants who had the highest sense of well-being reported the highest feelings of reassurance (subgroup 1, Med = 4.00, IQR = 0.25); in contrast, patients who had a lower sense of well-being were less likely to be reassured by surveillance visits (subgroup 2, Med = 3.75, IQR = 0.50; subgroup 3, Med = 3.75, IQR = 0.50, subgroup 4, Med = 3.75, IQR = 0.73; P = .023). Self-reported nervous anticipation also varied by overall well-being. Specifically, patients with the lowest sense of well-being had the highest level of nervousness related to surveillance related to their follow-up appointment (subgroup 1, Med = 1.60, IQR = 1.00 vs subgroup 4, Med = 2.20, IQR = 1.15; P < .001).

TABLE 3.

Summary scores of attitudes towards follow-up by cluster subgroup

	Median (IQR)					P value
	Total n = 212	Subgroup 1 n = 83	Subgroup 2 n = 72	Subgroup 3 n = 21	Subgroup 4 n = 36
Reassurance	3.75 (0.50)	4.00 (0.25)	3.75 (0.50)	3.75 (0.50)	3.75 (0.73)	P=.023
Communication	4.00 (0.25)	4.00 (0.00)	4.00 (0.25)	4.00 (0.50)	4.00 (0.25)	P=.005
Nervous anticipation	1.80 (1.00)	1.60 (1.00)	1.90 (1.00)	1.80 (0.80)	2.20 (1.15)	P<.001

Abbreviation: IQR, interquartile range.

There were no differences related to preferences around the frequency of surveillance follow-up visits among patients relative to well-being regardless of the time from surgery (range 1 to over 10 years postsurgery, P values range, 0.543-0.975). Overall patient well-being was also not associated with patient acceptance/comfort level to see a surgeon versus nonsurgeon provider during the surveillance visit (P = .715). In addition, patient self-reported well-being did not impact whether patients preferred to receive follow-up surveillance information on normal/abnormal blood and imaging in person or via technology (eg, secure email, MyChart, etc, P value range, 0.319-0.940).

4 |. DISCUSSION

High-quality PCC is necessary across the continuum of cancer treatment, including postoperative surveillance and follow-up. Data on patient preferences around surveillance and the association of patient well-being and surveillance are lacking. The current study helped address this gap by assessing patient surveillance preferences among patients across a range of self-reported well-being. The current study was novel in that we prospectively collected survey data using validated subscales of patient well-being that incorporated aggregate data on financial toxicity, chronic illness, as well as emotional and family/social wellness. Using an unsupervised learning approach, we were able to identify empiric patient subgroups who had marked differences in composite baseline well-being measures (ie, subgroup 1 vs subgroup 4). Of note, patient well-being was strongly correlated with several important factors such as perception of physician communication, as well as how reassured or nervous a patient was relative to cancer surveillance. Of note, patients who had a great sense of well-being were the most likely to report the highest levels of satisfaction with provider communication and feel the most reassured with their surveillance visit. In contrast, well-being was not associated with other surveillance-related factors such as preferences around the frequency of surveillance or whether the surgeon vs an allied health professional was present for the follow-up visit.

Despite national guidelines concerning the frequency of surveillance visits, several studies have reported marked heterogeneity in the approach and timing of surveillance among patients in the postoperative period.^24,25 In addition, several studies conducted among patients with prostate or breast cancer have noted that surveillance may be associated with increased anxiety and adversely impact patient quality-of-life.²⁶ In a study that investigated anxiety and cancer-related worry among patients with cancer at routine follow-up visits, Lampic et al²⁷ noted that one in five patients reported moderate or strong anxiety. In a separate study, Collins et al²⁸ reported that most patients were anxious about their follow-up appointments, yet some found the visits to be reassuring. In the current study, we expanded on this previous work. Rather than simply asking patients if they were reassured or anxious, we used validated assessment tools such as FACIT, FACT-G, and ATF to assess more holistically the various aspects of the patient experience related to surveillance. In turn, we were able to correlate composite measures of patient self-reported well-being with attitudes toward surveillance to identify which subsets of patients were more reassured than anxious about follow-up cancer appointments. Of note, patients who had a higher sense of well-being were more likely to be reassured by surveillance visits. Perhaps as expected, patients who had a lower general baseline sense of well-being were considerably more likely to report nervousness, anxiety, and a lack of reassurance (Table 3). These data highlight how caregivers need to be aware of the broad range of experiences patients have relative to surveillance. Also, the data serve to emphasize that understanding a patient’s baseline sense of well-being might inform how he/she perceives surveillance (eg, anxiety provoking, reassuring, etc). In turn, physicians should tailor their interpersonal approach and communication style to suit best the individual patient need.

Another interesting finding in the current study was that patient well-being also correlated with patient satisfaction around communication with their surgical oncologist about surveillance. Specifically, patients who had the greatest sense of well-being also had the highest satisfaction with patient-provider communication. Communication is a crucial element of the patient-provider relationship and is the cornerstone of shared decision-making. Poor shared decision-making has been associated with worse patient-reported health outcomes, worse established quality indicators, and higher health care utilization.²⁹ The finding that worse overall well-being adversely affected patient perception of provider communication, as well as the general health care experience (ie, increased anxiety, lower reassurance), demonstrated that “wellness” could impact the postsurgical patient experience. It was interesting to note, however, that overall well-being was not associated with preferences around the frequency of surveillance follow-up visits or the manner (eg, in person, secure email, MyChart, etc). While data from the current study could not establish causality or the direction of the association, the findings did demonstrate the strong interwoven nature of patient well-being with communication, anxiety, and assuredness around postoperative cancer surveillance.

Several limitations should be considered when interpreting the data. The majority of patient participants were recruited from surgical oncology offices at a high-volume comprehensive cancer center, which may have resulted in volunteer bias and affect generalizability of the results.³⁰ Recall bias may also have been a limitation given that the study was a cross-sectional survey. To reduce recall bias, rather than ask the patient to recall the specifics of their care, we extracted the cancer diagnosis, treatment, and management data from the electronic health record. In addition, the use of cluster analysis to create the varied subgroups was empiric by nature as this process used an unsupervised learning technique based on the variables of interest.³¹ As such, a similar methodological approach applied to different patient populations who had varied characteristics might yield different subgroups and attitudes/preferences.

In conclusion, data from the current study demonstrated that attitudes towards postoperative surveillance varied with regard to perception of provider communication, nervous anticipation, as well as assuredness depending on overall patient well-being. As such, providers should attempt to assess patient well-being as part of a tailored approach to postcancer surgery surveillance. Patients with a low general sense of well-being were at increased risk of perceived poor communication and anxiety. Finding innovative ways to foster a more patient-centered approach to the needs and preferences of cancer patients during the postoperative surveillance period are needed.