Survivor Typologies Predict Medical Surveillance Participation: The Childhood Cancer Survivor Study

Abstract

Objective

Adult survivors of childhood cancer adhere poorly to recommended medical surveillance. We sought to identify modifiable factors that contribute to non-adherence.

Methods

Latent class analysis categorized survivors (ages 18–52 years) at risk of cardiac, breast, or bone late sequelae on the basis of their health-related concerns, fears, and motivation. These classifications were compared at two time points for self-reported adherence to recommended echocardiography, mammography, and bone densitometry screening.

Results

Three classes (worried, collaborative, self-controlling) characterized survivors in each of the 3 risk groups: cardiac (N=564; BIC=10,824.66; LRMLRT P=.002), breast (N=584; BIC=11,779.97, LRMLRT P<.001), and bone (N=613; BIC=11,773.56; LMRLRT P=.028). Only 9% of at-risk survivors in the self-controlling class reported undergoing bone density screening in 2005, compared to 17.2% in the collaborative class (P=.034). Thirteen percent of the self-controlling, 24% of the collaborative (P=.025), and 34% of the worried (P=.010) classes reported undergoing bone densitometry in 2009. While 73% of at-risk survivors in the worried class reported having had an echocardiogram in 2009, only 57% of the collaborative (P=.040) and 43% of the self-controlling (P<.001) classes did. In 2005 and 2009, respectively, fewer survivors in the self-controlling class (37% and 53%) than in the collaborative (51%, P=.038 and 70%, P=.01) and worried (58%, P=0.002 and 69%, P=0.025) classes reported undergoing mammograms.

Conclusions

Modifiable intrapersonal characteristics associated with these 3 classes predict self-reported participation in medical surveillance. Continued observation and validation of these survivor profiles may inform tailored interventions to enhance survivors’ screening participation.

INTRODUCTION

As childhood cancer survival rates rise,[1, 2] late treatment-related morbidity [3–6] is of increasing concern. Follow-up screening can modify the likelihood and severity of late effects [7–8], but many survivors do not adhere to these recommendations [9–15]. For example, despite female survivors’ increased risk of early breast cancer,[16–18]recent studies found that only 41%– 55% of those with a history of chest radiation underwent mammography;[12, 19] among survivors <40 years, 47.3% had never had a mammogram, and only 52.6% of those 40–50 years were regularly screened (2 mammograms within 4 years)[19].

Treatment of childhood cancer with anthracyclines and/or chest radiation incurs a risk of late cardiotoxicity [6, 20–24]. Thirty years after diagnosis, survivors of childhood cancer have a rate of cardiac death 7 times the age- and sex-matched national average [25]. However, only 28% of at-risk participants in the Childhood Cancer Survivor Study (CCSS) had received the recommended cardiac screening [12, 26]. Cranial radiotherapy, glucocorticoids, methotrexate, and prolonged corticosteroid therapy increase the risk of low bone density and osteonecrosis, [27] but only approximately 25% of survivors at greatest risk had recently undergone bone densitometry [28].

We recently identified 3 distinctive profiles that predicted adult survivors’ intent to undergo routine and/or cancer-related check-ups [29]. These profiles were defined by survivors’ health-related motivation, worry, and concern, -- established modifiable mediators and moderators of health outcomes [30–36]. In this study, we replicated the distinctive survivor profiles in different samples, identified specific profile covariates, and used the profiles to predict self-reported participation in echocardiography, mammography, and bone densitometry screening at two different time points.

METHODS

Data Source

The CCSS is an IRB-approved multi-institutional retrospective cohort study initiated in 1994 to examine late effects in survivors of pediatric cancers diagnosed and treated between 1970 and 1986. Eligible participants had survived 5 or more years after completion of treatment for a malignant disease diagnosed before age 21 years. Survivors completed a baseline questionnaire at study entry and they respond to follow-up questionnaires (available at http://ccss.stjude.org) at regular intervals [37–38]. Participants provided informed consent for study procedures and release of their medical records.

Sample

Our study sample was drawn from CCSS participants who: a) had responded to two CCSS follow-up surveys (2005 and 2009) containing the screening outcome measures; b) participated in at least one of two CCSS ancillary studies containing the psychological latent class indicators (Health Care Barriers [HCB] in 2001, Mammogram Survey [MS]) in 2005; and c) were at high risk of treatment exposure–based late effects, including cardiac (anthracycline and/or chest or total body radiation), bone density (cranial radiotherapy, glucocorticoids, methotrexate, and/or prolonged corticosteroid use), and breast cancer (chest or spinal radiation) (Figure 1). The cardiac and bone risk groups had 387/808 survivors in common; the bone and breast-cancer risk groups shared 30/1226 survivors; the cardiac and breast-cancer risk groups shared 62/1141; and 28/1376 survivors belonged to all 3 groups. The overlap in the groups is related to multi-agent chemotherapy and/or multimodal therapy (e.g., chemotherapy, radiation, surgery). Each agent or modality confers excess risk for a spectrum of adverse treatment effects, some of which are overlapping. All survivors were age =18 years at the time of data collection.

Figure 1.

Description of study data sources

Measures

Class indicators (established by confirmatory factor analysis) were fears, health concerns, intrinsic motivation, and extrinsic motivation. These measures were identical and derived from either the Health Care Barriers (2001) or Mammogram (2005) ancillary CCSS studies and were antecedent to the surveys containing the screening outcomes. Three summed items defined survivors’ fears about future health, cancer recurrence, and discovery of a problem at a check-up visit (1=not at all; 5=extremely). Three items were summed to define survivors’ health concerns about general health, the chance of illness or health problems related to previous cancer, and the importance of a check-up visit (1=not at all; 5=extremely). Five summed items from the Multidimensional Health Locus of Control Scale [39] (e.g., “I am in control of my health”) defined survivors’ intrinsic motivation (1=moderately/strongly disagree; 4=moderately/strongly agree). Four summed items from the Multidimensional Health Locus of Control Scale [39] (e.g., “Health professionals control my health”) defined survivors’ extrinsic motivation (1=moderately/strongly disagree; 4=moderately/strongly agree).

Covariates included disease and treatment variables (diagnosis, age at follow-up, age at cancer diagnosis, years since cancer diagnosis), sex, education, race, total personal income, current health insurance status, perceived severity of late effects, and self-reported health status (1=excellent, 5=poor). Survivors indicated whether they had experienced chronic health problems lasting longer than 6 months and rated the severity of their main chronic health problem. These responses were categorized as moderate, severe, or life-threatening chronic problems vs. mild or no chronic problems.

Medical Surveillance

The Children’s Oncology Group has compiled risk-based, exposure-related guidelines for surveillance and management of late effects of treatment [7]. Echocardiography, mammography, and bone densitometry are recommended at specific intervals based on age at treatment, chemotherapy and/or radiation exposures, and clinical indications at the end of therapy. Recommended screening frequency ranges from as often as yearly to every 5 years based on the established criteria. Given the length of time between follow-up surveys, all patients would have been due for at least one of the three screenings. Survivors indicated the time of the most recent echocardiography (ultrasound or MUGA scan), bone densitometry (DEXA or CT scan), and mammography in both of the CCSS follow-up surveys (1=never, 5=5 or more years ago). Survivors who answered “don’t know” to any of these questions were excluded from analysis. We used a very conservative approach to code the medical surveillance outcome for the logistic regressions. For the echocardiogram and bone densitometry multinomial logistic regressions, “never” and “5 or more years ago” were re-coded as 0 (non-adherent); “more than 2 but less than 5 years ago”, “1–2 years ago” and “less than 1 year ago” were re-coded as 1 (adherent). In keeping with previous reports [19], for the mammogram logistic regressions, “more than 2 but less than 5 years ago” was considered non-adherent. Self-reports of medical screening have been established as valid study measures in the general population [40–44]. Recent unpublished ancillary CCSS studies report 90% agreement between self-reports of screening and medical record review.

Statistical Analysis

Latent class analysis (LCA) enables the identification of a set of mutually exclusive typologies that account for the distribution of individuals in the population; typologies are created by cross-tabulation of observed discrete variables [45]. Unlike typical regression analysis, in which the population is assumed to be homogeneous and a single model holds for all cases, LCA can accommodate multiple populations (i.e., typologies) within the population. As recommended by Nylund and colleagues,[46] a combination of the Bayesian information criterion (BIC), Lo-Mendell-Rubin parametric likelihood ratio (LMRLRT), and bootstrap LMRLRT (BLMRT) tests were applied to determine the number of latent classes. Details regarding the interpretation of these parameters have been previously reported. [29] SAS 9.1 software (SAS Institute Inc, Cary, NC) was used to describe sample characteristics and MPlus Version 6.1 (Muthén and Muthén, Los Angeles, CA) was used to develop the latent class models.

Models that had the fewest substantively meaningful distinct classes and a BLMRT p ≤0.05 were accepted. Covariates that were significantly (p ≤.05) associated with the classes were included in the best-fitting latent class model. Binary outcomes (participation in echocardiography, mammography, and bone densitometry) were tested for equality of proportion to determine class-specific self-reported participation. All analyses were repeated for each risk group.

RESULTS

Latent Class Model Selection

Survivor classes were first identified within each risk group on the basis of our indicator variables, without controlling for covariates (Table 1). Five different models were tested for the cardiac and breast-cancer risk groups (1–5 classes) and 6 models (1–6 classes) for the bone risk group (Table 2). Both the 2- and 3-class models were tested for each risk group with covariates. Significant covariates included self-perceived health status and severity of late effects in all three models, health insurance status in the breast cancer risk group, race in the bone density risk group, and sex in the cardiac and bone density risk groups. Final fit statistics for the three models with covariates were: cardiac (N=564; BIC=10,824.66; LRMLRT P=0.002); bone (N=613; BIC=11,773.56; LMRLRT P=0.028); and breast (N=584; BIC=11, 779.97, LMRLRT P=<0.001). The final models with covariates demonstrated posterior probabilities (showing how well each participant fit the assigned class) of 85%-90%, supporting appropriate class assignment and model fit [47].

Table 1.

Descriptive summary of classes within risk groups before controlling for covariates

	BONE RISK					CARDIAC					BREAST RISK
	Total n=624	Self-Control n=163	Collab n=353	Worried n=108	P value	Total n=571	Self-Control n=136	Collab n=326	Worried n=109	P value	Total n=632	Self-Control n=110	Collab n=330	Worried N=192	P value
		26%	56%	17%			24%	57%	19%			17%	52%	30%
Age at survey completion, 2003					.918					.219					.149
Mean	30.36	30.19	30.38	30.56		31.65	30.93	32.12	31.13		38.92	38.36	38.70	39.62
SD	7.2	6.8	7.3	7.3		7.5	7.1	7.5	7.95		6.18	6.09	6.35	5.89
Gender(%)					<.001					<.001					NA
Female	52	33	57	67		51	38	51	66		100	100	100	100
Male	48	67	43	33		49	62	49	34
Race/Ethnicity (%)					.004					.015					.859
White	74	81	74	65		75	79	76	68		89	90	88	89
Black	7	4	8	10		8	5	7	12		1	0	2	1
Hispanic	12	4	12	20		12	6	13	16		4	3	5	4
Other	7	10	6	5		6	10	4	5		6	7	6	6
Annual Personal Income (%)					.005					.092					.834
< $40,000	67	58	68	80		64	63	61	72		68	65	68	67
>=$40,000	25	32	24	17		31	32	33	22		23	21	25	22
Not reported	8	10	8	4		5	4	6	6		9	14	7	11
Education (%)					.005					.320					.084
Less than HS	7	9	7	5		4	4	4	5		2	3	2	2
Completed HS	49	38	49	62		46	40	45	53		38	41	33	44
Completed College	42	50	42	31		49	54	50	40		53	48	58	47
Not reported	2	3	1	2		1	2	0	2		7	8	6	8
Chemotherapy (%)					.520					.571					.312
Yes	93	91	93	94		89	92	89	89		75	81	73	76
No	7	9	7	6		11	8	11	11		24	19	26	24
Not reported											1		1
Radiation%					.478					.499					NA
Yes	70	66	71	73		76	71	77	76		100	100	100	100
No	30	33	29	27		24	28	23	24
Not reported	0	1	0	0		0	1	0	0
Chemotherapy and radiation (%)					.218					.924					.312
Yes	63	57	64	67		65	63	66	65		75	81	73	76
No	37	42	36	33		35	36	34	35		24	19	26	24
Not reported	0	1	0	0		0	1	0	0		1		1
Age at cancer diagnosis					.723					.371					.213
Mean	9.02	9.31	8.95	8.81		10.11	9.65	10.41	9.77		11.54	10.78	11.56	11.95
SD	5.5	5.1	5.7	5.7		5.94	5.7	6.0	6.1		5.57	5.34	5.60	5.63
Years since cancer diagnosis					.238					.577					.373
Mean	21.34	20.89	21.43	21.75		21.54	21.27	21.71	21.37		27.38	27.58	27.14	27.67
SD	4.4	4.2	4.4	4.3		4.5	4.3	4.7	4.3		4.46	4.53	4.42	4.50
Cancer type (%)					.210					.786					.060
Leukemia	50	48	53	44		29	31	29	26		9	9	11	5
Central nervous system	9	12	7	9		1	1	1	1		3	4	4	2
Hodgkin lymphoma	11	10	10	17		23	18	23	30		54	46	52	63
Non-Hodgkin lymphoma	14	15	15	11		14	17	13	13		7	6	8	7
Wilms tumor	0	0	0	0		8	8	9	7		12	18	11	10
Neuroblastoma	0	0	0	2		5	5	5	3		5	2	6	6
Soft tissue sarcoma	7	7	7	7		8	10	6	7		4	6	5	3
Bone cancer	8	7	8	10		13	10	14	13		5	8	5	4
Chronic late effects (%)					<.001					<.001					.003
Moderate, severe or life- threatening	20	7	20	38		22	10	21	41		57	47	55	67
Mild or none	79	93	78	58		77	90	78	56		43	53	45	53
Not reported	1	0	1	4		1	1	1	3
Self-rated health (%)					<.001					<.001					<.001
Excellent/very good/good	84	96	88	55		83	96	86	59		78	83	83	68
Fair/poor	15	4	12	43		16	2	13	39		15	9	11	24
Not reported	1	1	1	3		1	1	1	2		7	8	6	8
Limited ability to participate in vigorous activities (%)					.002					.014					.041
Yes	29	18	33	37		30	21	33	34		38	32	43	34
No	69	82	67	58		69	79	66	64		54	60	50	58
Not reported	1	0	1	5		1	0	1	2		7	8	6	8
Current health insurance (%)					.381					.181					.003
Yes	87	83	88	89		90	90	90	87		87	79	89	88
No	12	17	11	10		10	10	10	11		6	13	4	5
Not reported	1	1	1	1		0	0	0	2		7	8	7	8

Abbreviations: SD, standard deviation; HS, high school; self-control, self-controlling; collab, collaborative

Table 2.

Fit statistics of the risk LCA models, not controlled for covariates

Cardiac(N=571)
Classes	Log-Likelihood	BIC	LMR Adjusted LRT P Value
1	−5597.15	11,245.07
2	−5489.75	11,062.02	P<.001
3	−5445.57	11,005.38	P=.01
4	−5431.99	11,009.97	P=.526
5	−5426.30	11,030.33	P=.332
Breast (N=632)
Classes	Log-Likelihood	BIC	LMR Adjusted LRT P Value
1	−6482.38	13,016.36
2	−6372.91	12,829.65	P<.001
3	−6321.50	12,759.08	P<.001
4	−6301.62	12,751.56	P=.269
5	−6288.56	12,757.69	P=.234
Bone (N=624)
Classes	Log-Likelihood	BIC	LMR Adjusted LRT P Value
1	−6157.20	12,365.88
2	−6027.78	12,139.23	P<.001
3	−5978.91	12,073.66	P =.166
4	−5962.22	12,072.46	P=.228
5	−5930.69	12,042.59	P=.352
6	−5915.41	12,043.22	P=.275

Abbreviations: BIC, Bayesian information criterion; LMRLRT, Lo-Mendell-Rubin parametric likelihood ratio test; BLMRT, bootstrap LMRLRT

Interpreting Class Profiles

Figure 2 and Table 3 show the class-specific means of the indicator variables. This estimate divided by the standard error (EST/SE) indicates the strength of the relationship between the indicator and the latent class variable. In each risk group, one class reported poor perceived health status and moderate to life-threatening chronic illness more frequently than did the other two classes; these survivors also reported the greatest worries and health concerns, the lowest level of intrinsic motivation, and the highest level of extrinsic motivation (Table 3). We labeled this group “worried.” This class was much larger in the breast-cancer risk group (30%) than in the bone (17%) or cardiac (19%) risk groups.

Figure 2.

Estimated mean values of the latent class indicators by risk group.

Table 3.

Latent class indicators within risk groups, controlled for significant covariates

	Worried				Self-Controlling				Collaborative
	Mean Estimate (95% CI)	S.E.	EST/SE	P	Mean Estimate (95% CI)	S.E.	EST/SE	P	Mean Estimate (95% CI)	S.E.	EST/SE	P
Health Fears
Cardiac(n=564)	11.66 (10.50–12.83)	0.594	19.63	<.001	5.12 (4.67–5.58)	0.233	22.00	<.001	7.77 (7.23–8.30)	0.273	28.43	<.001
Breast(n=584)	11.68 (10.96–12.40)	0.366	31.92	<.001	5.37 (4.84–5.89)	0.267	20.12	<.001	7.01 (6.57–7.45)	0.227	30.93	<.001
Bone (n=613)	11.36 (10.32–12.41)	0.535	21.25	<.001	5.12 (4.69–5.55)	0.219	23.31	<.001	7.66 (7.10–8.22)	0.287	26.73	<.001
Health Concerns
Cardiac(n=564)	13.35 (12.88–13.83)	0.242	55.15	<.001	6.69 (5.92–7.45)	0.388	17.22	<.001	10.26 (9.70–10.82)	0.285	36.06	<.001
Breast(n=584)	13.65 (13.35–13.95)	0.152	89.97	<.001	7.58 (6.88–8.28)	0.358	21.16	<.001	11.64 (11.12–12.15)	0.261	44.54	<.001
Bone(n=613)	13.26 (12.66–13.87)	0.309	42.91	<.001	6.73 (5.99–7.47)	0.378	17.81	<.001	10.25 (9.65–10.85)	0.307	33.40	<.001
Extrinsic Motivation
Cardiac(n=564)	9.26 (8.36–10.17)	0.462	20.05	<.001	6.76 (6.25–7.27)	0.259	26.08	<.001	8.48 (8.04–8.93)	0.228	37.25	<.001
Breast(n=584)	12.92 (12.30–13.54)	0.317	40.73	<.001	10.68 (9.69–11.66)	0.503	21.22	<.001	12.64 (12.11–13.17)	0.269	46.99	<.001
Bone(n=613)	9.43 (8.29–10.57)	0.582	16.21	<.001	6.65 (6.21–7.09)	0.224	29.74	<.001	9.03 (8.41–9.66)	0.317	28.50	<.001
Intrinsic Motivation
Cardiac(n=564)	15.17 (14.05–16.29)	0.571	26.55	<.001	17.23 (16.67–17.78)	0.285	60.49	<.001	16.73 (16.29–17.17)	0.225	74.49	<.001
Breast(n=584)	22.36 (21.52–23.21)	0.433	51.70	<.001	23.42 (22.43–24.42)	0.509	46.02	<.001	23.99 (23.43–24.56)	0.288	83.29	<.001
Bone(n=613)	14.57 (13.13–15.78)	0.735	19.82	<.001	17.40 (16.98–17.83)	0.218	79.85	<.001	16.69 (16.22–17.16)	0.238	70.14	<.001

A second class of survivors demonstrated class indicators and health status perceptions markedly opposite to those of the ‘worried’ class. Across all 3 risk groups, survivors in this class reported good/excellent health and little concern about health history or future health problems, and they placed little value on medical check-ups. While strongly intrinsically motivated for self-care, this group was minimally extrinsically motivated to involve health professionals in their long-term care management. We labeled this group “self-controlling.” In the cardiac and bone risk groups, this class was predominantly male; it had the lowest percentages of black and Hispanic survivors, the highest percentage of college graduates, and the highest reported incomes.

A third class, the largest class in each risk group, showed intermediate scores on the class indicators and on self-reported health status, was balanced in sex distribution, and endorsed moderate fears and health concerns. Members were highly intrinsically motivated to manage their health but also willing to work with health providers, as evidenced by their greater extrinsic motivation than the self-controlling group. We labeled this group “collaborative.”

Survivor Participation in Recommended Screening

Self-reported participation in echocardiography, bone densitometry, and mammography in two different follow-up surveys was compared within risk groups by class, controlling for significant covariates in each group (Table 4).

Table 4.

Comparison of participation in screening (2005 and 2009) according to class after adjustment for covariates

	2005			2009
Class	Worried vs. Self-Controlling	Collaborative vs. Self-Controlling	Worried vs. Collaborative	Worried vs. Self-Controlling	Collaborative vs. Self-Controlling	Worried vs. Collaborative
Type of screening	% (p value)	% (p value)	% (p value)	% (p value)	% (p value)	% (p value)
Echocardiography (N=564)	54% vs. 40% (P=0.082)	50% vs. 40% (P=0.092)	54% vs. 50% (P=0.550)	73% vs. 43% (P=<0.001)	57% vs. 43% (P=0.036)	73% vs. 57% (P=0.040)
Mammography (N=584)	58% vs. 37% (P=0.002)	51% vs. 37% (P=0.038)	58% vs. 51% (P=0.243)	69% vs. 53% (P=0.025)	70% vs. 53% (P=0.013)	69% vs. 70% (P=0.861)
Bone densitometry(N=613)	17.4% vs. 9% (P=0.155)	17.2% vs. 9% (P=0.034)	17.4% vs.17.2% (P=0.976)	34% vs. 13% (P=0.010)	24% vs. 13% (P=0.025)	34% vs. 24% (P=0.232)

Comparison used chi-square equality tests of means across classes with posterior probability-based multiple imputations (2 degrees of freedom for the overall test and 1 for the pair wise tests).

(2005)

While the classes did not differ in echocardiography, a significantly larger proportion of survivors in the collaborative than in the self-controlling group underwent bone densitometry. Similarly, significantly larger proportions of the collaborative and worried classes than of the self-controlling class underwent recommended mammography (Table 4).

(2009)

A greater proportion of survivors reported echocardiography in 2009 than in 2005 across all 3 classes, but the proportions within each class differed significantly (Table 4). Similarly, the proportion of survivors participating in bone densitometry increased in all classes in 2009. A significantly larger proportion of survivors in the worried and collaborative classes than in the self-controlling class participated in both bone density and mammography.

DISCUSSION

We found 3 distinctive groups of survivors—worried, self-controlling, and collaborative—defined by indicators of health-related concerns, motivation, and affect, as in our previous work. This multiple nominal classification, rather than a single linear descriptive model, is supported by the fact that each survivor has a most likely class, the classes make up different proportions of the sample, and the classes have no ordered interrelation [48]. With the exception of the worried class in the breast-cancer risk group, the distribution of survivors across the 3 classes in each risk group was similar. The exclusively female sex of the breast-cancer risk group and the tendency toward female predominance in the worried class is likely to explain this difference.

Because of their self-reported good health and low levels of health concerns and worries, survivors in the self-controlling group may see no need for cancer-related follow-up and screening [49]. Their high levels of intrinsic motivation and low levels of extrinsic motivation indicate that they are unlikely to initiate medical follow-up. This typology may reflect survivors’ unawareness of their long-term risks [50]. Based on our collective and cumulative clinical experience and on-going clinical trials, an optimal first intervention for this group might be distance-based strategies (e.g., web and/or print media detailing treatment-related risks and surveillance recommendations). Because of their strong intrinsic motivation, they may be more likely to initiate cancer-related follow-up and screening after being informed of their risks. Because these survivors report the least worry about their future health, the information provided should be detailed and graphic in explaining the probability and nature of their treatment-related risks [10, 50].

The worried group’s demographic profile suggests the most difficulty in obtaining risk-based health care (Table 1). Survivors who are African-American, older at interview, or uninsured were reported as less likely to receive risk-based, survivor-focused care [12]. The worried group was also distinguished by greater cancer-related fears. Fear and worry exert both positive and negative influences on health-related behaviors. For example worried or health-anxious survivors may avoid screening to moderate their fear of bad news [51, 52]. However, some excess worry can make the individual more sensitive to potential threats or more vigilant about lifestyle and surveillance behaviors [51–56]. Misconceptions and lack of specific risk information can exacerbate fear or contribute to denial of the possibility of significant health problems [57–62]. While our analysis does not specifically consider which particular worries were prominent in each risk group, it is highly likely that worries will differ from one diagnostic group to another, as well as across survivors with different co-morbidities and late effects experiences. Face-to-face encounters, where the clinician could identify specific worry targets, would likely provide the best care for this group; however, individualized print summaries together with supportive telephone interactions detailing long-term risks and ways to reduce those risks would potentially be useful. The worried group should be offered information in a manner that avoids exacerbating fears and concerns.

Survivors in the collaborative group were both receptive to professional care (extrinsically motivated) and highly intrinsically motivated to maintain their health. However, they may not fully understand their risks or the need for periodic screening, as reported among childhood cancer survivors in general [50, 57–63]. Collaborators may be highly motivated to follow screening recommendations if they are provided risk-based information and specific recommendations. Print summaries of their treatment, including exposure risks and recommendations for medical follow-up and screening, would likely be sufficient to motivate collaborators to initiate care and work with their providers to minimize their risks.

In a previous report [29], the worried and collaborative classes differed markedly in that survivors in the worried class were more likely to obtain routine and cancer-related medical care. In this study, the two groups differed significantly only on echocardiography in the 2009 survey. While a larger proportion of the worried class than of the collaborative class participated in screening, statistical power may have been lacking in this small sample.

Limitations

While the CCSS population is a large and heterogeneous cohort of 5-year survivors, our results may not be generalizable to all childhood cancer survivors. Although observation of the same class structure in the risk groups could reflect non-independence of groups, the overlap across all 3 risk groups was only 2%, and the same class structure was previously determined in different samples [29]. Finally, there may be variables that differ across classes but were unavailable in our data set.

Conclusions

Our findings suggest that survivors’ participation in medical screening varies predictably across survivor typologies derived from personal endorsement of indicators of health-related concerns, motivation, and affect. Our current clinical trials are assessing the tailoring of interventions to these profiles to support physical activity among patients undergoing treatment and medical surveillance among adult survivors of childhood cancer. Future studies will examine to what extent these profiles predict other behavior-related health outcomes in other survivor samples.