Equity in the use of publicly subsidized psychotherapy among elderly Danish cancer patients – a register-based cohort study

Abstract

Approximately 30% of cancer patients suffer from psychological distress, and psychotherapy has been shown to be effective in alleviating it. Based on the ‘Behavioral Model of Health Service Use’, we investigated equity in the use of publicly subsidized psychotherapy in a cohort of Danish cancer patients. We present descriptive data on patients’ use of psychotherapy and examine characteristics of those who used this service. Material and methods. The study population comprised 3646 participants in the prospective Diet, Cancer and Health cohort, diagnosed with a first cancer between 2003 and 2009, aged 56–80 years. Data on cancer diagnosis, psychotherapy use and comorbid conditions were obtained from registers, whereas data on demographics, social support and health status were obtained from questionnaires. Cox proportional hazards regression was used to identify factors related to use, which were subsequently evaluated with regard to equity. Results. Subsidized psychotherapy was used by 2.3% of the cancer patients. Longer education (> 10 years compared to < 8) was associated with greater likelihood of initiating psychotherapy [hazard ratio (HR), 1.97; 95% confidence interval (CI), 1.06–3.65], as was previous psychotherapy use compared to no previous use (2.86; 1.32–6.20). Patients with partners who did not reliably provide emotional support were significantly more likely to use psychotherapy than those without a partner (2.36; 1.05–5.30), a difference not found for those with partners who did provide support. Further, a higher SF-36 mental component score (0.96; 0.94–0.98, per point increase), and older age were associated with less use (65–69 years: 0.43; 0.21–0.89, 70–74 years: 0.17, 0.07–0.41; > 74 years: 0.07, 0.01–0.57, compared to < 60 years). Conclusions. The results allow several possible interpretations. We found that mental health plays a role for accessing subsidized psychotherapy, suggesting that use of psychotherapy is predicted by need and thus characterized by equity. However, education and previous psychotherapy use also play a role, suggesting elements of inequity.

Psychological distress is a significant problem, affecting about a third of cancer patients [1]; for some patients, psychotherapy may alleviate distress [2], and psychosocial care may be an integral part of cancer rehabilitation [3]. Since 1992, all Danish residents with cancer have been entitled to 12 subsidized psychotherapy sessions provided by licensed psychologists [4]. A wide range of therapies may be offered, including cognitive behavioral, psychodynamic and existential-humanistic psychotherapy. In this paper, therefore, ‘psychotherapy’ is used to refer broadly to all forms of psychological therapy offered under the subsidy. The subsidy covers 60% of a psychologist's fee, and patients must be referred by their general practitioner no later than 12 months after their diagnosis. Thus, the Danish healthcare system should provide equitable access to psychotherapy for cancer patients; however, in a recent cross-sectional survey of 1490 Danish cancer patients, more than one in five expressed a need to see a psychologist, but almost half reported no offer of this treatment [5]. This study is the first assessment of Danish cancer patients’ use of subsidized psychotherapy to date.

Based on searches of the Pubmed and Psycinfo databases (not systematic), we identified five previous studies that, among other outcomes, have assessed cancer patients’ use of outpatient services provided by psychologists in relation to a cancer diagnosis. Four studies reported utilization rates among 439–1199 patients ranging from 2.6% to 6.2% [6–9], and one study (n = 252) reported that patients’ wish to see a psychologist predicted use six months later [10]. The need for psychological counseling has previously been related to factors such as age, education, having a partner and disease-related factors [11].

The Behavioral Model of Health Service Use presents a theoretical framework for investigating factors related to use of services (Figure 1) [12]. At its core are three factor groups: predisposing factors, which form a dispositional basis for an individual's use of a given service; enabling factors, which mark an individual's ability to gain access to the service; and need factors, which represent the need for the service, i.e. the illness level. According to the model, equity in service use can be ascertained by determining the patient characteristics that are the strongest predictors of use. Equitable use is achieved when the dominant predictors are demographic and need factors [12].

Figure 1. The Behavioral Model of Health Service Use. Based on Andersen (1995) [12].

As inequality may also be an issue in rehabilitation [13], and knowledge of utilization patterns may help to determine whether psychotherapy is received by those who need it and whether use is equitable, we examined the extent to which a large population-based sample of more than 3500 Danish cancer patients used the public subsidy for psychotherapy. In addition, we investigated which characteristics of cancer patients determined their use of psychotherapy.

Material and methods

Participants

The study sample was drawn from the prospective Danish Diet, Cancer and Health cohort, established in 1993–1997 to investigate the relation between diet, lifestyle and risk for cancer [14]. The cohort consists of 57 053 participants, aged 50–64 at inclusion. Cohort participants who responded to a follow-up questionnaire and in whom a first primary cancer was diagnosed after 31 December 2002 (n = 3646) formed the sample for the present study.

Clinical and demographic information

Date of birth, gender and vital status were obtained from the Civil Registration System, which contains this information for all individuals residing in Denmark since 1968. Data on education were obtained from the cohort baseline questionnaire [14] and were grouped according to years of education completed (< 8, 8–10 and > 10 years). From the Danish National Cancer Register [15], we obtained dates of diagnosis and cancer types (ICD-10 codes), which were grouped according to site. Data on hospital admissions and outpatient contacts (registered since 1977 and 1995, respectively) were obtained from the Danish National Patient Register [16]. The Charlson comorbidity index was calculated from the ICD-7 and ICD-10 codes registered for each person up to the date of cancer diagnosis [17] and was dichotomized for the analyses, as a score of 0 vs. ≥ 1. The unique personal identification number assigned to all Danish residents was used to link the data from questionnaires with the registers.

Data on psychologist use

The National Health Service Register, an administrative register of the Danish National Board of Health, contains data on payments to psychologists from the National Health Insurance [18]. Data on use of the subsidy are available from its adoption in 1992, and all services billed to the National Health Insurance for individuals in the Diet, Cancer and Health cohort were obtained for this study through 2010. Payment data were also obtained from the Danish regions, including date, service type and referral category from 2003 until November 2011. Both sources were used to ascertain the use of psychologists before the date of cancer diagnosis. Any patient for whom a ‘first-session’ service was registered within 18 months of his or her cancer diagnosis was considered to have seen a psychologist in relation to the diagnosis. All subsequent sessions until another ‘first session’ were considered part of that treatment course, and treatment courses were considered to be completed after 12 months without a new session, if the patient was alive at that time. Only sessions for which patients were referred due to ‘severely debilitating illness’ were included, and an 18-month time frame was chosen to allow for delay after referral. As clients ordinarily receive the subsidy for only one session a day, all duplicate sessions were excluded.

Health-related quality of life and social support

The cohort follow-up questionnaire, completed in 1999–2002, included the SF-36, from which the ‘mental and physical component summaries’ were used [19]. Scales were scored based on the ‘Full Missing Data Estimation’ procedure [19]. A social support survey, which has previously shown good face-validity and fair-to-excellent eight-day test–retest agreement [20], was also part of the questionnaire. Single items from this survey were analyzed, as it yields no global indices [20]. Perceived emotional support was assessed by the item “Can you have a good talk with any of the following persons if you need their support?” The original response categories (‘always’, ‘often’, ‘now and then’, ‘rarely’, ‘never’ and ‘have none’) were used to create a binary variable indicating the reliable availability of emotional support by combining the two first (‘always’, ‘often’) and the remaining categories (‘now and then’, ‘rarely’, ‘never’ and ‘have none’). Participants’ intimate relationship status was obtained from the item “Do you have a partner (spouse, significant other)?” (Being a widow or widower was considered not having a partner.) The existence of children and other family members was inferred from two items measuring contact frequency, combined with a cohabitation-status item for children. The existence of a relationship and the availability of emotional support were combined to yield measures with three categories: no relation (partner, children, other family members), relation not providing reliable emotional support and relation providing reliable emotional support.

The Behavioral Model of Health Service Use

The patient characteristics were entered into the factor groups of the Behavioral Model of Health Service Use as follows. Age, gender, previous psychotherapy use and education were considered predisposing factors. Education might also have been considered an enabling factor, providing means for navigating the healthcare system, or a proxy for income, rather than a proxy for social position and culturally patterned attitudes and health beliefs. The presence of and emotional support from partners, children or other family members were considered enabling factors, as social relations may serve as a lay referral network or stress buffer, decreasing need [21]. All clinical variables were regarded as need factors.

As social support and quality of life measures were assessed before diagnosis, they express baseline functioning, not the status after diagnosis. It is assumed that baseline functioning will affect adjustment to cancer and thus be related to using psychotherapy after a diagnosis.

Statistical analyses

The determinants of psychotherapy use were studied by means of Cox proportional hazards regression. Effect estimates were presented as hazard ratios (HRs) and 95% confidence intervals (CIs). We performed crude analyses and a model with all predisposing factors (age at diagnosis, gender, education, previous psychotherapy use), a model combining the predisposing and enabling factors (adding emotional support from partner, children and other family members), and a full model with all factor groups (including need factors: SF-36 component summary scores, and Charlson index score). To examine a possible effect of cancer site, we attempted to perform subanalyses for men and women separately, including cancer diagnosis groups. However, due to few cases, statistical power was insufficient to complete such analyses. To estimate separate HRs for the SF-36 mental component summary subscales (vitality, social functioning, role emotional and mental health) [19], these were entered into the full model separately in place of the mental component summary. The assumption of proportional hazards was checked by inspecting Schoenfeld residual plots. For the continuous variables (SF-36 scales and component scores), the assumption of linear association with the log relative hazard was evaluated by means of exposure–response curves, which were estimated and visualized using restricted cubic splines [22].

In a subanalysis, the association of study variables with the number of psychotherapy sessions attended was studied by univariate linear regression. This analysis was performed for 69 patients who were alive 12 months after their last psychotherapy session, for whom the therapy course could thus be considered completed.

Differences between respondents and non- respondents to the follow-up questionnaire among eligible participants in the Diet, Cancer and Health cohort were analyzed using χ²-tests. All analyses were performed with the R software, version 2.15.0 [23].

Results

The response rate to the follow-up survey among patients eligible for this study was 82%. Due to missing data on social support, education or quality of life, 599 individuals were excluded, leaving 3646 patients as the study sample. Of these, 83 patients (2.3%) initiated psychotherapy within 18 months of their cancer diagnosis. The mean time between diagnosis and the first session was 180.9 days (six months), and a mean of 4.6 sessions were attended (Table I).

Table I. Descriptive characteristics of study population and of psychotherapy users.

	All cancer patients* (n = 3646)		Users of psychotherapy (n = 83)
	n	%	n	%
Age at diagnosis (years)
< 60	151	4.1	10	12
60–64	917	25.2	31	37.3
65–69	1293	35.5	32	38.6
70–74	1003	27.5	9	10.8
> 74	282	7.7	1	1.2
Gender
Female	1708	46.8	52	62.7
Male	1938	53.2	31	37.3
Length of education (years)
< 8	1168	32	18	21.7
8–10	1696	46.5	39	47
> 10	782	21.4	26	31.3
Cancer site
Prostate	709	19.4	9	10.8
Breast	658	18	24	28.9
Colorectal	422	11.6	7	8.4
Lung	381	10.4	11	13.3
Urinary tract	297	8.1	3	3.6
Other digestive organs	281	7.7	6	7.2
Leukemias and lymphomas	249	6.8	6	7.2
Gynecological	167	4.6	2	2.4
Other	160	4.4	8	9.6
Malignant melanoma	151	4.1	1	1.2
Head and neck	94	2.6	1	1.2
Central nervous system	77	2.1	5	6
Charlson comorbidity index
0	3379	92.7	79	95.2
≥ 1	267	7.3	4	4.8
SF-36 mental component summary mean (range)	54.34 (10.61–72.02)		49.11 (13.16–64.56)
SF-36 physical component summary mean (range)	51.40 (10.28–66.16)		51.72 (27.42–64.10)
Reliable emotional support from partner
No partner	694	19	17	20.5
Partner, no reliable support	169	4.6	11	13.3
Partner, reliable support	2783	76.3	55	66.3
Reliable emotional support from children
No children	350	9.6	10	12
Children, no reliable support	426	11.7	9	10.8
Children, reliable support	2870	78.7	64	77.1
Reliable emotional support from family members other than partner and children
No other family members	78	2.1	2	2.4
Other family members, no reliable support	1218	33.4	34	41
Other family members, reliable support	2350	64.5	47	56.6
Previous psychotherapy use
Yes	75	2.1	8	9.6
No	3571	97.9	75	90.4
Days from cancer diagnosis to first psychotherapy session mean (range)
			180.9 (9–545) (median: 139)
Number of psychotherapy sessions attended mean (range)			4.6 (1–12)† (median: 4)

Some percentages may not add up to 100 because of rounding.

*Patients in the Diet, Cancer and Health cohort with a first cancer diagnosed after 2002; †14 patients died within 12 months of their last session, and their treatment course might hence not have been completed had they not died. After exclusion of these patients, the mean number of sessions attended is slightly higher (4.9), while the median remains 4.

Univariate Cox proportional hazards analyses

Older patients (aged 65–69, 70–74, or > 74 years) were significantly less likely to use psychotherapy than the youngest patients (< 60 years), and men were half as likely to use psychotherapy as women (Table II). Previous psychotherapy use (compared to no previous use) and longer education (> 10 years compared to < 8) increased the likelihood of psychotherapy use almost six-fold and more than two-fold, respectively. Among the enabling factors, only having a partner who was not a reliable source of emotional support was significantly associated with psychotherapy use (HR, 2.60; 95% CI 1.22–5.55) when compared to patients without a partner. Of the need factors, the SF-36 mental component summary was significantly associated with using psychotherapy (HR, 0.95; 95% CI 0.93–0.97 for one point increase).

Table II. Hazard ratios (HRs) with corresponding 95% confidence intervals (CIs) for psychotherapy use by 3646 patients within 18 months of first cancer diagnosis after 2002.

	Crude analyses HR [95% CI]	Predisposing factors^1 HR [95% CI]	Predisposing & enabling factors^2 HR [95% CI]	Predisposing, enabling & need factors^3 HR [95% CI]
Predisposing factors
Age at diagnosis (years)
< 60	Reference	Reference	Reference	Reference
60–64	0.50 [0.25, 1.02] ˙	0.51 [0.25, 1.04] ˙	0.50 [0.25, 1.03] ˙	0.54 [0.26, 1.11] ˙
65–69	0.37 [0.18, 0.75] **	0.41 [0.20, 0.83] *	0.39 [0.19, 0.80] **	0.43 [0.21, 0.89] *
70–74	0.14 [0.06, 0.34] ***	0.16 [0.07, 0.40] ***	0.16 [0.06, 0.39] ***	0.17 [0.07, 0.41] ***
> 74	0.06 [0.01, 0.44] **	0.07 [0.01, 0.51] **	0.06 [0.01, 0.50] **	0.07 [0.01, 0.57] *
Gender
Female	Reference	Reference	Reference	Reference
Male	0.53 [0.34, 0.83] **	0.58 [0.37, 0.92] **	0.64 [0.40, 1.02] *	0.68 [0.42, 1.10] *
Length of education (years)
< 8	Reference	Reference	Reference	Reference
8–10	1.44 [0.82, 2.52]	1.19 [0.68, 2.09]	1.19 [0.67, 2.09]	1.23 [0.69, 2.17]
> 10	2.06 [1.13, 3.76] **	1.74 [0.95, 3.18] *	1.84 [1.00, 3.37] *	1.97 [1.06, 3.65] *
Previous psychotherapy use
No	Reference	Reference	Reference	Reference
Yes	5.65 [2.72, 11.70] ***	4.41 [2.11, 9.20] ***	3.90 [1.85, 8.25] ***	2.86 [1.32, 6.20] **
Enabling factors
Reliable emotional support from partner
No partner	Reference		Reference	Reference
Partner, no reliable support	2.60 [1.22, 5.55] *		2.52 [1.11, 5.74] *	2.36 [1.05, 5.30] ˙
Partner, reliable support	0.78 [0.46, 1.35]		0.91 [0.51, 1.63]	1.11 [0.61, 2.01]
Reliable emotional support from children
No children	Reference		Reference	Reference
Children, no reliable support	0.74 [0.30, 1.83]		0.52 [0.20, 1.36]	0.59 [0.23, 1.53]
Children, reliable support	0.76 [0.39, 1.48]		0.85 [0.43, 1.70]	0.98 [0.49, 1.97]
Emotional support from other family members
No other family members	Reference		Reference	Reference
Family, no reliable support	1.02 [0.25, 4.25]		0.93 [0.22, 3.94]	1.03 [0.24, 4.38]
Family, reliable support	0.73 [0.18, 3.02]		0.64 [0.15, 2.65]	0.78 [0.18, 3.28]
Need factors
SF-36 mental component summary	0.95 [0.93, 0.97] ***			0.96 [0.94, 0.98] ***
SF-36 physical component summary	1.00 [0.97, 1.03]			1.00 [0.98, 1.03]
Charlson comorbidity index
0	Reference			Reference
≥ 1	0.69 [0.25, 1.87]			0.88 [0.31, 2.44]

Significance level: *** < 0.001, ** < 0.01, * < 0.05, ˙ < 0.1. ¹Age at diagnosis, gender, education, previous psychotherapy use; ²As the predisposing model, with emotional support from partner, children, and other family members added; ³As the predisposing and enabling model, with SF-36 mental and physical component summaries, and Charlson index scores added.

Multivariate models

The multivariate models were based on the succession of factor groups in the Behavioral Model of Health Service Use (Figure 1). The first model contained the predisposing factors, the second further contained the enabling factors, and the third model included all three factor groups (Table II). Gender was significantly associated with psychotherapy use only in the predisposing model. Length of education reached borderline significance in the two first models and was significantly associated with psychotherapy use in the full model with an HR of 1.97 (95% CI 1.06–3.65) for longer (> 10 years) compared to shorter (< 8 years) education. Previous psychotherapy use was significantly associated with psychotherapy use in all models, with an HR in the full model of 2.86 (95% CI 1.32–6.20) for previous psychotherapy use as opposed to no previous use. Having a partner who did not reliably provide emotional support compared to not having a partner was significantly associated with psychotherapy use in the last two models, with an HR of 2.36 (95% CI, 1.05–5.30) in the full model. The effect of age on psychotherapy use remained significant and almost unchanged throughout, as did the estimate for the mental component summary score (HR, 0.96; 95% CI 0.94–0.98 for one point increase in the full model). When the SF-36 mental component summary subscales where entered into the model separately instead of the summary score, subscale HRs were very similar to the HR of the full component summary (Table III).

Table III. Hazard ratios with corresponding 95% confidence intervals (CIs) for psychotherapy use by 3646 patients within 18 months of first cancer diagnosis after 2002, with SF-36 mental component summary or subscales in fully adjusted model.†

SF-36 scales	Hazard ratio [95% CI]
Mental component summary	0.96 [0.94, 0.98] ***
Vitality	0.96 [0.94, 0.98] ***
Social functioning	0.97 [0.95, 1.0] *
Role emotional	0.96 [0.94, 0.99] **
Mental health	0.96 [0.94, 0.98] ***

Significance level: ***< 0.001, **< 0.01, *< 0.05. †Model containing predisposing, enabling and need factors, with SF-36 scales entered separately.

Factors associated with number of psychotherapy sessions and differences between respondents and non-respondents

Having reliable emotional support from a partner was the only factor that was significantly associated with the number of sessions attended (β, –2.82; 95% CI –4.6, –1.04), compared to patients with no partner. In χ²-tests, non-respondents (n = 943) were found to have significantly shorter education and be more likely to be male than respondents.

Discussion

In this cohort of elderly cancer patients, only some 2% had used publicly subsidized psychotherapy after their diagnosis. We observed that younger cancer patients with longer education and prior experience with psychotherapy, who had a low score on the SF-36 mental health measure and partners who were not reliable sources of emotional support were more likely to use subsidized psychotherapy than other cancer patients.

Utilization rate and factors associated with use

The 2.3% rate of psychotherapy use among cancer patients found in this study is comparable to, but slightly lower than, the lowest rates found in previous studies, which have reported use of psychologists by 2.6–6.2% of patients [6–9]. Limited use thus appears to be consistent across studies. In our study, the low rate may have been influenced by several factors: the restricted duration of eligibility for subsidized psychotherapy (12 months after diagnosis), the copayment (about 40€ per session), participants’ high age and the availability of other counseling offers to Danish cancer patients, which are not systematically available (such as through private insurance, the Danish Cancer Society or services paid for privately). It was not possible to establish a rate of use of subsidized psychotherapy related specifically to severe illness among the general Danish population. However, between 0.5% and 0.6% of the population aged 56–80 years used subsidized psychotherapy for any reason during the years 2006–2009 [24]. This illustrates an infrequent use in this age group overall and suggests that the low utilization rate observed in our sample is plausible.

Only one of the five previous studies investigated predictors of psychologist use. Among 252 cancer patients, this study showed that only the wish to see a psychologist was associated with using this service six months later [10]. The other four studies examined predictors for a broad range of services together, thus not allowing for service-specific conclusions to be drawn [6–9]. The effect of length of education found in our study corresponds to an earlier finding by Ernst and colleagues [10]; however, those authors did not find that age was associated with use of psychologists, while we found a significant association.

With the exception of the study by Ernst and colleagues, previous studies were cross-sectional [6–9] and all but one study [6] used outcome data based on self-report, potentially biasing the results [7–10]. In three studies, the samples were limited to patients within a precisely defined time from diagnosis (between a maximum of 3 and 36 months) [6,8,10], whereas the remaining studies had no inclusion criteria for time since diagnosis and were thus based on much less specific samples [7,9]. Two studies included only patients with breast or colorectal cancer [6,8], limiting the generalizability of their results, while the remaining studies included patients with various diagnoses [7,9,10].

Tendencies of equity and inequity

Our finding that predisposing, enabling and need factors were associated with psychotherapy use could indicate tendencies of both equity and inequity of use. Within the framework of the Behavioral Model of Health Service Use, the influence of age (demographic predisposing factor), and of mental health (need factor) would indicate that use of psychotherapy was equitable [12]. In contrast, the effects of length of education, previous psychotherapy use (predisposing factors related to social structure and health beliefs, respectively) and having a partner who does not provide reliable emotional support (enabling factor) would be seen as indicators of inequity [12]. For some of these factors, however, the conclusions may be equivocal.

The effect of age may be based on several mechanisms. Older cancer patients may be less distressed than younger patients [1] and hence less in need of psychotherapy; however, the finding that age remained significant and the trend of decreasing HRs remained almost unchanged when need factors were entered into the model (Table II) suggests that age has an impact beyond that of need. It has previously been reported that older individuals may prefer to receive mental healthcare in the context of a visit to a primary care physician for a somatic illness [25]; their general physical condition may present a practical barrier to using specialty mental healthcare, which often requires travel; and they may prefer to deal with distress privately [26]. Further, primary care physicians may be less likely to identify, treat or refer older patients for psychological morbidity, perhaps due to the belief that psychotherapy is not effective for older individuals [27]. If such factors, related to health beliefs and enabling resources, underlie the effect of age, it may no longer exclusively be considered a sign of equitable use.

The effect of having a partner who did not reliably provide emotional support was here considered an indication of inequity in the use of psychotherapy. We observed that patients without a partner were significantly less likely to attend psychotherapy than those with a partner who did not provide reliable emotional support. However, we found no difference in psychotherapy use between patients without partners and patients with partners who did provide reliable support. It could be speculated that this might reflect a role of partners either in the process of referral to health care services, in encouraging and helping patients to seek care, or in decreasing patients’ need to see a psychologist for those partners who give emotional support. Still, we cannot draw a firm conclusion with regard to equity, as the underlying mechanisms are speculative.

Another factor for which the conclusions based on the Behavioral Model of Health Service Use could be challenged is previous psychotherapy use. Previous use of psychotherapy is seen as an expression of an individual's underlying beliefs about and attitudes toward psychotherapy and therefore related to his or her disposition to use psychotherapy again. The effect of health beliefs and attitudes is seen as an indication of inequity in the Behavioral Model; however, this view is based on a judgement of whether patients’ attitudes are to be regarded as preferences that should be left unquestioned or as barriers to use based on potentially erroneous beliefs. Thus, previous psychotherapy use may not be an unequivocal marker of inequity.

Although some of the conclusions drawn for individual factors based on the Behavioral Model of Health Service Use may be questioned, the overall conclusion remains that clear tendencies of both equity and inequity are apparent in the use of psychotherapy by the sample studied.

Study strengths and weaknesses

A major strength of this study is its reliance on register-based data on psychotherapy use from insurance claims in the administrative National Health Service Register, effectively excluding selection and recall bias as well as loss to follow-up. Clinical data and vital status were from validated, complete registers, almost entirely excluding misclassification of the exposures. Moreover, this study of cancer patients’ use of psychologists is based on the largest sample to date. It is the first study to address use of outpatient psychologist services and associated factors as the primary outcome and to examine equity within a theoretical framework. Further, the time frame covered was precisely determined by the time during which patients were eligible for subsidized psychotherapy; we thus included only newly diagnosed patients and assessed their use within 18 months of diagnosis. This resulted in a study focused on the time just after diagnosis, in contrast to the less specific temporal foci in some previous studies [7,9].

A weakness of this study is the small number of patients who used psychotherapy, which limits its statistical power, which may also limit the strength of the inferences regarding equity of use. Further, as referral records were not obtainable, only the referral category ‘severely debilitating illness’ and proximity in time were available to link patients’ use with their cancer diagnosis. The fact that only 7.3% of the total sample and 4.8% of psychotherapy users had a score of 1 or more on the Charlson index may indicate that potential confounding by other somatic illnesses is a minor problem. Some factors of potential importance were not included in the study, such as income, cancer site and prognostic clinical factors, which might have independent effects on use of psychotherapy. The addition of these factors could have further nuanced the findings although the sample size would limit their usefulness. As the perceived availability of emotional support was assessed by only one item, the validity of this measure may be questioned. However, the validation of this questionnaire did not find the item to be problematic [20].

Only 36% of the individuals originally invited were included in the Diet, Cancer and Health cohort, with respondents being more likely to be women, younger, married, have longer education and higher income [14]. The response rate to the follow-up questionnaire among patients diagnosed with a first cancer after 2002, however, was 82%. Comparisons between respondents and non-respondents to the follow-up showed that non-respondents were more likely to be men with shorter education. Two of the previous studies of cancer patients’ use of psychologists did not report exact response rates [9,10], and the rates in the other three studies were 62%, 65% and 100% [6–8]. Differences between study populations and the general population arising from low response rates in the cohort might limit the generalizability of the findings.

This study is a valuable addition to existing knowledge about cancer patients’ use of psychotherapy, as it is a prospective investigation of a very large sample, using highly reliable register-based clinical and outcome data with minimal bias, and draws on a theoretical framework for assessing equity.

Conclusion

Only some 2% of the cancer patients studied used psychotherapy. Although psychotherapy use was highly influenced by need, measured as mental health, the results reveal elements of inequity in psychotherapy use as length of education and previous use of psychotherapy also affected psychotherapy use after a cancer diagnosis. These results suggest that increased attention should be paid to patients who are not usual psychotherapy users in the referral process.

Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.