This review demonstrated that there has been increasing attention to unmet needs of cancer patients in the literature over time, with greater attention given to descriptive studies than intervention studies. While the literature search describing trends in types of unmet need publications was not intended to be systematic, the focus on descriptive research is concerning. Descriptive research has been invaluable in highlighting the unique needs of cancer patients and the necessary urgency to ameliorate these issues [5, 7, 9, 13]. It is generally accepted that the process of conducting descriptive research is simpler than for intervention studies, in terms of conceptualization, feasibility, and publishing [38]. While this may explain the trends observed in unmet needs research, it raises several issues. If it is not possible to change unmet needs, then there may be limited value in continuing to describe these needs. Conversely, if it is possible to change unmet needs, it could be argued that a greater emphasis on evaluating the effectiveness of strategies is warranted. A more appropriate balance in research effort is necessary to capitalize on available research funding, to develop a best-practice evidence base, and most importantly, to improve the psychosocial outcomes of cancer patients.
A total of nine intervention studies were identified. Six of the nine trials included in this review failed to demonstrate an intervention effect on unmet needs. Therefore, the results of this review suggest that, while it may be possible to reduce some patients’ unmet needs through supportive discussions, therapies, or referral, these changes have not been consistently demonstrated. Results of the reviewed intervention trials do not provide strong evidence for any particular approach for reducing levels of unmet need. In particular, testing of multiple subscale scores at multiple time points and the use of post hoc subgroup analyses in two of the three positive trials suggest the possibility of spurious results arising from type II errors [30, 37]. The third trial was a pilot study [34] and, although it demonstrated an intervention effect, this was not replicated in the full-scale RCT [29]. Given increasing interest in the assessment of and use of unmet needs measures for screening and tailoring interventions, it is timely to consider which factors may contribute to the mixed and limited findings observed across trials.
Potential explanations for mixed findings among intervention trials which aim reduce unmet needs among cancer patients
Psychometric rigor of unmet needs measures
While measures such as the SCNS have been psychometrically tested for validity and reliability [17, 39], sensitivity to change over time has not been well explored [6, 40]. It is possible that current measures of unmet need are not sufficiently sensitive to consistently identify small or isolated changes in unmet need. This may be due to the fact that these tools were originally designed to capture a wide breadth of concerns across an entire population, rather than being sensitive to the particular needs of an individual. Such psychometric design attributes may, therefore, limit the use of these instruments as individual screening devices or intervention outcome measures.
Appropriateness of measures for the selected sample
It should also be noted that, while all studies included in this review used the term “cancer patient,” some of the samples included may have consisted of cancer survivors. While definitions of survivorship vary [41, 42], treatment completion is often used as a defining point [43]. In particular, for the three studies that reported recruitment via population-based cancer registries [29, 31, 34], participants were at least 4–6 months post diagnosis at the time of study entry, and current treatment status was not clearly reported. It is, therefore, likely that many of these participants may have been “survivors” rather than patients. Since the needs of cancer survivors are known to be different from cancer patients [44], this suggests that studies which include survivors should use needs measures which have been specifically developed for the survivor population [6]. The SCNS, used in all three of the latter studies, was developed for a patient population (undergoing treatment) rather than a survivor population (post-treatment) [17]. Therefore, this measure may not have been appropriate to adequately capture the needs of these groups. Future research with cancer survivors should use measures specifically developed for this population such as the Survivors Unmet Need Survey (SUNS) [19] or the CaSUN [45].
Study samples are insufficient to find an intervention effect
It may be that, when using wide-ranging unmet needs measures such as the SCNS or CNQ, or for potentially heterogeneous samples involving multiple cancer types, large samples are required to identify small changes in particular patient needs. These measures may be more suited to identifying small changes in the population prevalence of domains of unmet need, rather than identifying small patient-specific changes. However, this theory is not supported by the studies reviewed here, given that the study with the largest and potentially most homogenous sample failed to find an effect [29].
Analysis of unmet needs
Items in unmet need measures can potentially be analyzed in a number of ways, and the scoring and analysis of these measures have evolved over time. Most studies in this review used domain scores as their outcome measure. On both the NA-ACP and the SCNS, domain scores are calculated by summing item scores [18, 46]. Therefore, similar scores at two follow-up points may not necessarily reflect that participants are endorsing the same needs at both time points [23]. Similarly, the approach of using subscale scores does not allow determination of whether specific needs have been reduced over time as a consequence of the intervention. Similar problems arise when the number of needs endorsed as unmet is used as an outcome measure rather than a domain score. This may hide the fact that needs may change over time, with different needs contributing to the prevalence count for an individual at any given time point.
An alternative to using subscale scores may be to examine changes in specific items of unmet needs. Without evidence of test–retest reliability at the item level, however, examining changes in prevalence of need by item is also problematic. Item-level test–retest reliability has not been demonstrated for any unmet needs measures for adult cancer patients [40]. This means that it is impossible to tell whether any change in number of people endorsing an item is due to the intervention or lack of test–retest reliability in the item.
Interventions tested are ineffective
It is possible that limited intervention effects observed across trials are a result of ineffective interventions. Descriptive studies have indicated that a range of sociodemographic, disease, physical, and psychological factors are associated with unmet needs [16]. Lack of effect may, therefore, indicate that the intervention is not powerful enough to address the many factors that may influence unmet needs. Similarly, the “dose” of the intervention may also have been inadequate to achieve an effect. A recent meta-analysis has found that longer-term interventions (minimum 12 weeks) had a greater impact on QoL of adults with cancer than short-term interventions (d = 1.19, d = 0.47) [47]. Intervention intensity and frequency were quite varied across the studies reviewed, ranging from single assessment and printed feedback to multiple sessions and multiple referrals. It is, however, difficult to assess the true intensity of any particular intervention, given that referral or feedback were a key part of the majority of both the effective [30, 34] and ineffective interventions [31–33, 35, 36]. Limited data were provided on the consequences of the referrals. It is highly likely that this was variable both within and between studies.
Interventions are not delivered as intended
Lack of intervention effect may also reflect lack of adherence to key intervention components by patients or providers. For example, in McLachlan’s study, recommended services from the tailored management plan were declined by patients in 38% of instances [24]. Reasons reported for refusal of services included inappropriate timing of the referral and preferences for other forms of support including other formal services and informal support of self-management [24]. Similarly, in Boyes’ pilot study, only two of the four doctors involved in the study reported that they had discussed the feedback on needs with their patients during the consultation [36]. Given this low level of adherence to the intervention, the lack of effect may not be surprising.
Floor effects preclude demonstration of an intervention effect
Post hoc analysis in Aranda’s study indicated that an intervention effect may be possible if those who have high unmet needs at baseline are selectively targeted [30]. Similarly, Girgis and colleagues noted that participants in their trial reported higher levels of QoL and psychological well-being than expected [31]. The authors suggested that this may have contributed to the lack of intervention effect observed. This echoes a common criticism of psychological interventions to reduce distress, anxiety, and depression among cancer patients, whereby interventions to improve these outcomes are targeted at all patients rather than at those with demonstrated need at baseline [48]. This may reflect an assumption that all people with cancer have high levels of unmet supportive care needs.
However, selective targeting of high needs individuals poses its own challenges. These difficulties may include logistic and cost difficulties in screening large numbers of patients to identify a subset requiring intervention prior to commencing a trial. It may also reflect that there is no clear threshold for needs measure, which indicates clinical significance [23]. Therefore, by arbitrarily excluding those below a certain score, it is possible that some people who may benefit from an intervention may be excluded. Research focusing on the establishment of clinical significance of need measures may aid in the interpretation of scores and the application of such measures to assessing intervention effectiveness.
Needs reflect a desire for certainty and reassurance which cannot be met by the health care system
Longitudinal studies indicate that needs may change over time [25, 49, 50]. However, it is not clear from intervention studies whether needs can be reduced more rapidly or by a greater magnitude by intervention. Potentially life-threatening illnesses such as cancer are associated with high levels of uncertainty [51]. It is plausible that this uncertainty and desire for reassurance may be expressed as unmet psychological or information needs. If this is the case, it is possible that no amount of information, support, or service provision will be able to address this need. Therefore, during some phases of the illness trajectory, certain unmet needs may be endemic to the cancer experience [23]. In these circumstances, it may be important for providers to explore patient concerns, acknowledge uncertainties, and provide appropriate reassurance.
Lack of clarity regarding the nature of unmet needs
In addition to considering the possibility that unmet needs may reflect a desire which cannot be met, it is timely to also consider the nature of the concept of unmet need. The concept of “unmet needs” is a relatively new one, and there is little literature regarding the nature of the construct [23]. The concept of unmet needs appears to have arisen in the context of identifying the range of patient experiences which, if addressed, might ameliorate disease-related psychosocial impacts such as depression, anxiety, and poor QoL. While associations between these outcomes are supported by descriptive research [15, 16], there is no evidence of a causal relationship between unmet needs and patient psychosocial outcomes. It is possible that unmet need surveys, while being helpful in identifying particular patient concerns, are not appropriate as a focus for intervention development or outcome measurement.
Future directions
Given that there is increasing attention directed at describing the unmet needs of people with cancer, it is important to establish whether, and if so, how unmet needs can be reduced. There are several areas where future work could inform the development and testing of unmet need interventions. Firstly, to progress work in this field, it is necessary to develop clear guidelines about the scoring of unmet needs scales. Such guidelines should describe how needs should be scored for intervention trials so that it is possible to attribute change in needs to the intervention. This may necessitate examining test–retest reliability at the item level for existing scales. Further, to determine the magnitude of change needed to establish an intervention effect, the issues of sensitivity and clinical significance need to be considered. Unlike measures of anxiety or depression, criterion validity against a gold standard clinical interview cannot be used to determine clinical significance [23]. Greater clarity about the construct of unmet needs and its mechanism of operation would also assist in the development and design of appropriate interventions. One possible way forward may be to examine what level of needs predict future adverse outcomes such as greater health care utilization, poorer QoL, or greater risk of developing depression.
Limitations
It is possible that some relevant articles were missed by the current review. Articles published in languages other than English and unpublished articles, for example, may have been missed. The use of a funnel plot to assess publication bias was considered. However, given the relatively small number of studies included, it is likely that these results would have been unreliable [52].