Accuracy of Family History

In order to fully interrogate this question, evidence of accuracy had to be explored beyond the primary care setting. Although this encompassed broader clinical settings than the most comprehensive published review,¹⁰² the results were fairly similar. Most eligible studies examining accuracy of reporting of cancer family history focused on breast or colorectal cancer, with fewer examining accuracy for ovarian and prostate cancers. In contrast to a previous review, ¹⁰² we did not limit studies to those verifying the status of unaffected relatives. This strategy yielded a broader set of studies that evaluated aspects of reliability but there were no significant gains in the number or quality of studies evaluating the primary question of accuracy. Overall, the few rigorous studies which fully evaluated accuracy (i.e., accuracy of reported absence and accuracy of reported presence of cancer in relatives) appeared to suggest that informants are more accurate in identifying which relatives are free of cancer (specificity) than in identifying relatives who have been affected by cancer (sensitivity). Our results indicate that family history reporting may be more accurate for first degree relatives than second degree or beyond, although few studies examined accuracy in the latter. Our findings also suggest that accuracy may be different for different cancer types, and influenced by the method of ascertainment of family history.

Future efforts to improve accuracy of reporting would be improved by explicit consideration of whether sensitivity or specificity is the primary goal, which is dependent on the clinical context and purpose of a family history oriented strategy. For example, maximizing sensitivity prioritizes the goal of missing as few “at risk” family histories as possible, and is consistent with a policy in which the potential benefits from finding potential cases carry more weight than the potential costs and harms of investigating individuals or families with false positive histories. In contrast, maximizing specificity prioritizes avoiding the potential costs and harms of false positives, and is consistent with a policy which directs limited resources towards only identifying individuals or families with the greatest likelihood of being at significant disease risk, at the cost of missing some true positives.

The studies reviewed focused on accuracy as a binary concept (presence or absence of cancer); we do not have evidence relating to the accuracy of other information which is relevant in cancer risk assessment such as information on age of onset. We are unable to comment on which gold standard is “best” for judging accuracy, nor on the effect of clinical setting or tool format. The accuracy of reporting by patients or members of the population cannot be completely separated from the performance of tools to gather such data,⁵¹ but we had limited information on the latter and it was not always evident whether a structured Family History Tool (FHxT) was utilized in data collection.

We also have little insight into which informant characteristics are associated with more accurate reporting; future evaluations could consider formally examining factors such sex, age, and cultural background. It is possible that informants affected by cancer may seek out more complete information on their family history after their initial diagnosis, but we were unable to confirm this speculation.

Future research should also consider the issue of reliability of patient recall, including the issue of what is an “adequate” interval for studies of repeatability. We suggest that it would be helpful to try to separate the reliability of reporting as a psychometric property in an individual from the reliability of reporting as a function of extra knowledge sought by an individual from other family members in the period between first and second data collections.

In general, we might expect that the accuracy of family history reporting will improve in future, as current initiatives lead to more awareness on the part of the general public. It is not clear whether this will be countered by the effect that increased population mobility has on people's abilities to keep up to date with the health of more distant family members.

Family History Tools

The review identified a number of FHxTs developed for use in a primary care setting, most of which had not been evaluated against either best estimate gold standard or current primary care practice. Because of the limited number of studies, the evaluation of FHxTs was extended to relevant tools in non-primary care settings. Taken together, there was reasonable agreement between FHxTs and accepted best estimate gold standard, and, when compared to current primary care standard practice, FHxTs identified significantly more genetically relevant family history information. The clinical significance and added benefit of this added information still needs to be explored.

The tools identified in this review varied considerably, from those which took a comprehensive approach, emulating the geneticist's pedigree drawing interview to those which focused on identifying selected cancers in specific relatives. Many were designed to be used in the physician's office, in paper-based or electronic format. It has been suggested that other formats, such as web-based or mailed surveys, allow patients and consumers to (potentially) take “ownership” of their family history, offer them the opportunity to gather information from relatives,^{37, 43, 45, 49, 52} and may make for better use of primary care provider (PCP) time. Some electronic tools require patients to assemble family history information in advance of the office visit, which may also promote accuracy and ownership. Some studies have shown high response rates to mailed FHxTs from PCPs^{48, 54} and “consumer empowerment” was the basis of the previous U.S. Surgeon General's Thanksgiving “Family History Day.” ^{110, 111} Several organizations have set up similar web-based FHxTs for public use^{50, 112} (http://www.norwichunion.com/healthtree/index.htm ¹¹³; http://www.ama-assn.org/ama/pub/category/13333.html ¹¹⁴).

The acceptability and ease of completion of FHxTs were assessed in only a few studies. These aspects of the tools' content and face validity should be an integral part of any evaluation of future primary care FHxTs.

While some authors³ have identified elements that could be included in an “appropriate” family history (see Figure 5), there is no explicit consensus on a minimum data set covering the extent and the nature of family history data appropriate to primary care practice. Until the evidence base is clear, it is suggested that a minimum adequate cancer family history should include information on siblings, parents and grandparents (and the paternal and maternal lineage of the latter), specific enquiry about whether other relatives had the cancers of interest, and the ethnicity of the respondent. When cancer is identified, the age of diagnosis should also be noted, and other relatives with similar or related conditions identified.

Figure

Figure 5. Typical information obtained in Three-Generation Pedigree.

In assessing individual tools, it is important to consider the notion of “appropriateness” in relation to individual patient factors (e.g., age) and in terms of patient population characteristics.⁶ For instance, for a 40-year old patient it may be appropriate to enquire about all siblings, parents and grandparents, but children's health may not be as relevant for eventually determining cancer risk. Where there is concern about risk of familial breast cancer, information on aunts and uncles may be more informative than that on grandparents. Also, while some authors have suggested that a minimum family history should cover three generations^{3, 115, 116} the reliability of information beyond first degree relatives and grandparents is unclear (see comments on accuracy, above). On the other hand, some genetic RATs require a count of the number of unaffected relatives, as well as those with a cancer of interest (e.g., Yang 1998⁵³). Accurate risk assessment generally requires information on the side of the family (maternal or paternal) to which relatives with cancer belong, and most FHxTs identified this. Finally, ethnicity (an indication of ancestry ¹¹⁷) may be associated with increased risk of particular disorders, including some cancers, but few tools were designed to capture such data on ethnicity.

We suggest that, in future FHxT development studies, it would be useful to distinguish between two different purposes for FHxTs - assembly and updating of “complete” family history information in a generic approach, and ascertainment of targeted information for specific disease risk assessment. For the latter, it may be logical to evaluate the performance of a FHxT as part of a disease-specific RAT, rather than as a stand-alone tool. For more generic tools, approaches to their rational development and evaluation would benefit from agreement on the “minimum family history dataset” for primary care purposes, bearing in mind that the goal in this setting is usually to stratify or triage risk rather than ascertain or diagnose a genetic condition. An evidence-based minimum dataset would take into account evidence on accuracy of patient reporting of family history under primary care office conditions and might not necessarily have to replicate the extent or type of data captured in a clinical genetics setting. Table 12 lists some of the elements which could be considered for inclusion in a minimum dataset. It is presented to foster discussion and evaluation only as it is not within the scope of this review to formally assess its utility or feasibility.

Table 12

Potential items for inclusion in minimum family history dataset.

Family histories are not static;^{45, 49} however, practical issues of updating family history have not been explored. On the one hand, PCPs may be able to assemble a patient's family history information over time, but on the other, necessary updates consume time and resources. Acheson¹ has reported that most family histories were completed on the first visit. It would be worth considering formally whether a staged approach over several visits leads to more accurate or extensive information, and clarifying the optimum interval for updates.

It seems logical that FHxTs are likely to produce most benefit if they are accompanied by management plans for patients at familial cancer risk; otherwise “proactive” family history collection by PCPs and/or consumers may be wasteful of time, energy, health care resources, and may even be harmful. While some guidelines¹¹⁸ recommend that family history information should only be collected in response to patient enquiry about familial breast cancer risk or if the provider suspects increased cancer risk, others argue that family history collection is an integral part of good clinical practice in primary care and that failure to do so should be considered negligence.^{51, 119}

Risk Assessment Tools

An inclusive definition of RAT was used to capture the widest range of interventions potentially applicable to primary care. Their formats varied from fairly simple tools designed solely to stratify risk to those in which the capture of family history data was closely linked with management recommendations within a format designed to promote implementation in practice. We chose to focus on only those guidelines that had been formally evaluated in their own right, or embedded in some form of tool designed to promote use in practice. This decision recognized the very large number of familial cancer stratification guidelines which had been published over the time period of the review. We judged that an exhaustive approach to describing such guidelines would have provided little insight into the review questions and would likely be quickly out of date. However, for information, we listed the guidelines developed by national agencies or professional organizations in an Appendix B.^*

Similarly, we focused only on those RATs which produced as output a risk of cancer, and excluded those for which the only output was risk of a given mutation. Our rationale was that family history reflects an integration of risk generated by genetic factors (including gene variants which may confer only modest increase in risk), shared environments, and common behaviors² and is an important predictor, in its own right, of disease risk. We suggest that this approach is consistent with the overall primary care perspective of the review, and increases the likelihood that the tools included would be accepted as relevant and usable by the target professional groups, outside the specialist genetics setting. In addition, clinically valid RATs which generate disease risk strata should, by definition, allocate families with high risk of mutation into the highest risk category, therefore alerting practitioners to their need for specialist assessment.

A large number of studies reported outcomes in terms of the distribution of patients across risk strata compared with an independent standard (e.g., an accepted guideline or an assessment by a specialist geneticist). This is an approach to assessing clinical validity (i.e., predictive value) and is of course dependent on the validity of the gold standard comparator. This review was not designed to assess this component of clinical validity, which ultimately requires studies that rigorously evaluate how well risk categorization predicts eventual disease outcome. We found that very few studies examined effectiveness in terms relevant to the questions posed in this review—either professional practice outcomes (e.g., improved confidence in clinical decision making) or patient outcomes (e.g., more accurate risk perception). Taken together, the evidence is not sufficient to make definitive recommendations, but it does tentatively indicate that RATs may improve the appropriateness of referral of patients for genetic counseling. Whether this is clinically or administratively worthwhile depends on the local clinical context. The extra benefit from a RAT must be set against the costs of implementation, particularly if there is already high compliance with referral guidelines. There is insufficient evidence to determine whether RATs, by themselves, are likely to improve physician confidence or skills in broader aspects of patient care related to familial cancer.

Just as with FHxTs, the potential effectiveness of RATs may be confounded by the strategy used to implement them in practice. Decision tools are complex interventions, and thus present challenges in their development, application, and evaluation.^{36, 120} Recent analyses have begun to identify the characteristics of decision tools that appear most likely to promote effectiveness in practice but few studies have evaluated patient outcomes. One of the most significant predictors of decision tool effectiveness appears to be the automatic provision of decision support as part of a practitioner's workflow.¹²¹ This should become increasingly straightforward to achieve as electronic medical records become more widely implemented and linked with computer-based RATs. Other predictors of tool effectiveness include the provision of actionable recommendations (rather than just assessments); the provision of decision support at the time and location of decision making; the periodic feedback on performance to users; built-in features that promote the sharing of recommendations with patients; and systems that request documentation of reasons for not following recommended actions.¹²¹ It is plausible that this emerging evidence on desirable characteristics of decision tools, while still preliminary, is applicable to family history based RATs. It should be noted that many tools have been evaluated by the same investigators who developed them, and that such studies seem to report higher levels of practitioner performance than studies where tools are evaluated by independent observers.

The barriers to the use of FHxTs and RATS tools in practice include lack of time,¹²² lack of PCPs' confidence in their knowledge and skills in genetics,^{80, 123, 124} and reimbursement policies.³ Finally, even though a typical PCP may provide care to a significant number of patients with a history of familial cancer,⁶⁴ they may make up only a very small part of his or her daily practice. Hyland et al.¹²⁵ suggested that the rate of physician contact with women with a family history of breast cancer was about 0.6 consultations per month per family physician. Systems to implement apparently efficacious tools therefore need to take account of these barriers, and broader consideration could be given to the cost-effectiveness of developing tools which assess familial risk across a range of common chronic disorders.

All of these factors taken together suggest that effective RATS require a coherent, evidence-informed approach to their design, consideration of their integration with other clinical and office systems, and attention to contextual factors which might moderate their effect, and their marginal benefit in practice.

Limitations

The studies reviewed in this report were limited to those published in English; however, the impact of any language bias is offset by the optimal applicability to English speaking countries for which this report was prepared. Our peer review process allowed content experts in this area to identify any additional studies (both published and unpublished) of relevance for this review thereby minimizing the likelihood of publication bias. In addition to using several web-based search engines, our search of relevant grey literature was limited to sites specified by the investigators, our technical expert panel (TEP), and peer reviewers. We contacted the authors of eligible studies to request copies of the tools or methods used to ascertain eligibility of family history method for this review. The majority of authors contacted did respond, but some did not. Language bias also limited the ability to interpret non-English FHxT, however this had a minimal impact on the studies described and evaluated. The budget and timelines available, however, were limiting factors in pursuing complete retrieval of all the instruments used to collect family history in the eligible studies.

Our criteria for defining a systematic FHxT or RAT resulted in the exclusion of guidelines, recommendations or mutation risk calculators (see above). These are all “decision tools” and, even though a rationale was provided, their exclusion was arbitrary. The result may be that the review has underplayed the value of guidelines (however published) in promoting effective clinical practice, and overlooked “specialist” tools which might actually be useful in primary care without further modification. Similarly, the definition used for applicability to family practice was based on criteria developed within our investigative team and has not been subject to external scrutiny. In the context of accuracy of family history reporting, eligible studies did not use the same method to ascertain family history or verify status within all relatives. As such, interpretation of the metrics of accuracy was limited to the methods of family history ascertainment and verification used in these studies.

Conclusion

The accuracy of self reported family history has implications for the correct risk assessment and management of patients. Accuracy of cancer family history reporting appears to be dependent on cancer type and method of collection, and accurate reporting of absence of cancer (specificity) appears to be greater than accurate reporting of presence of cancer (sensitivity). Accuracy of recall and reporting may be influenced by both patient factors and by the method used to capture the data (the tool). No studies appear to have examined both of these together, so it is impossible to comment definitively on their relative contributions to any lack of accuracy.

Family history is a fundamental element of health information, and the ability to take an adequate and accurate family history should be recognized as a core skill for all PCPs, irrespective of the availability of tools. Very few FHxTs have been developed for, and evaluated in, primary care settings. Further, few tools have been compared with either “best practice” (genetic interview) or current primary care practice (family history as recorded in charts). Although the evidence is very limited, and depends on extrapolation of studies of tools in settings other than primary care, it suggests that systematic FHxTs may add significant genetic family history information compared to current primary care practice.

A number of RATs, of varying format and complexity, have been developed for primary care settings, and a few of these have been evaluated in controlled trials. These studies provide tentative evidence for the effectiveness of such tools, but their utility in routine practice has not been established.

Recommendations

1.

Consensus should be reached on the extent of family history enquiry necessary for different clinical purposes and circumstances, taking into account the likelihood of accuracy of self reported information for different relatives, and the use to which the information will be put (e.g., overall or specific risk assessment).

2.

The benefits, costs and harms of using patient-completed tools for systematic family history collection and risk assessment, as a substitute for, or complement to, professional tools should be further examined. As well as assessing technical outcomes such as accuracy and completeness of data captured, evaluations should consider outcomes which relate to patient “empowerment” and the use of practitioner and health care resources.

3.

Further research is required to identify the specific strategies (e.g., sending tools home with patients) and tool features which promote the most accurate reporting of family history information.

4.

The optimum interval for updating a patient's family history information in primary care should be formally evaluated.

5.

Further evaluation of FHxTs and RATs in routine clinical settings and practice is required. Studies should: adopt appropriate comparators (generally current practice); ensure that tools are optimized (in terms of, for example, face and content validity) before evaluation; measure outcomes that relate to utility in routine practice; measure outcomes that provide information on potential costs or harms as well as benefits; and address or explore contextual factors which may modify utility in practice (e.g., practice infrastructure, time available).

Footnotes

*

Appendixes cited in this report are provided electronically at http://ahrq​.gov/clinic/tp/famhisttp.htm