Family Healthware, a self-administered, internet-based tool, was used by primary care adult patients to record their family history outside of the clinician-patient encounter and to receive tailored familial disease risk and prevention information. Elevated familial cancer risk was highly prevalent: 34% of participants were at strong or moderate risk for at least one of the three cancers. Before enrollment, participants at increased familial risk for colon or breast cancer were more likely than those with low familial risk to be adherent if eligible for cancer screening recommended for the general population of their age and sex. Family Healthware also identified that 4.4% of all participants were potentially eligible for colon cancer screening beyond that recommended for the general population, and 9% of female participants aged 35–39 years seemed eligible for earlier breast cancer screening. Furthermore, a sizeable percentage of all participants—approximately 2.5% for colon cancer, 10% for breast cancer, and 4% for ovarian cancer—were potential candidates for genetic risk assessment. However, fewer than half of the additional participants, for whom colon cancer screening was indicated based on family history, were adherent with risk-based colon cancer screening, and a mere 17% of eligible women aged 35–39 years had obtained early mammography. Consultation for familial cancer risk was rare, reported in only 3% of participants at baseline. Therefore, Family Healthware identified a group at increased cancer risk in need of targeted preventive measures.
Although small in absolute numbers, those at strong risk represent a group in whom screening and referral are of the utmost importance. Referral of these individuals for genetic counseling and cascade genetic testing within their families is a powerful way to focus screening recommendations for many of the highest risk individuals within a population.
34,35We hypothesized that tailoring preventive recommendations on an individual’s family history of breast, ovarian, and colon cancers should increase the percentage of patients completing these recommendations, especially from a population of adults aged 35– 65 years with an established primary care provider relationship. In this setting, both the Family Healthware intervention and the standard prevention messages were associated with an increase in the percentage of adults adherent with screening for breast or colon cancer 6 months later. However, there was no significant difference in screening rates between the control group and the group using Family Healthware, either as a whole or within levels of familial risk. Thus, we also found no evidence that tailored messages conveying information about weak familial risk resulted in complacency about cancer screening. Conversely, there was no evidence that use of Family Healthware resulted in overscreening for those at weak risk.
The FHITr demonstrated the feasibility of implementing self-administered family history risk assessment in primary care, but conducting this study with unselected, healthy patients limited the enrollment of people in need of screening and the power to detect an effect. The population recruited to the FHITr study had far higher screening adherence rates than a national sample, as measured by the Behavioral Risk Factor Surveillance System.
36 FHITr compared with national mammography rates for women aged 40 years or older were 76% vs. 61%, and FHITr compared with national combined fecal occult blood testing/endoscopic colon cancer screening rates for adults age 50 years or older were 79% vs. 60%.
36 This reflects the sociodemographic features of our study population such as having private insurance, being non-Hispanic white, a higher education level, and having a physician,
36–38 factors associated with higher cancer screening rates. Furthermore, as others have also observed,
38 participants with increased familial risk of a specific type of cancer were more likely to be already up-to-date on screening for that type of cancer, before study enrollment. The resulting ceiling effect left only a small proportion of participants in whom effects of the intervention could be observed. Smaller still was the proportion of subjects who had strong or moderate familial cancer risks. The posthoc power to detect a difference between the overall intervention versus control group ranged from 0.19 for mammogram screening to 0.06 for colon cancer screening.
The nature of the study as an evaluation of a computerized health tool may have created a perceived need for computer literacy and access and thereby limited the diversity of the study population. We attempted to mitigate this effect by facilitating participation for those without internet access at home. Participants could complete the study by phone by interview or in the clinic. Telephone assistance was used in lieu of direct computer access by 9% of the study subjects, and there were no significant demographic differences among groups using these two modes of study participation.
27The greatest potential to measure the effect of an intervention (and to serve that group) is in a population where baseline adherence is low. Paradoxically, however, barriers to healthcare will be larger in a group with lower levels of insurance, household income, and education. Recruitment may also be more challenging. Therefore, we believe that in designing future studies of this kind, strategies will be needed to reach eligible subjects who have not been screened. In addition to the medically underserved, it will be important to include younger participants (aged 20 – 40 years) with a strong family history of cancer, as this study demonstrated that the majority of these had not had cancer screening that may have been indicated based on their familial cancer risk.
In addition to the screening behaviors, we measured self-reported visits to a medical professional other than the participant’s regular doctor because of concerns about cancer and visits to a genetic specialist or genetic counselor. Either event was extremely rare and was not influenced by the Family Healthware messages. We are not able to determine whether physicians made the recommendation to see such a provider.
Although it may seem surprising that the study participants at strong risk did not avail themselves of cancer genetics services, low rates of genetic assessment have been a recurrent theme, even in academic settings. For example, evidence of physician referral was found for only 7% of high-risk patients who were seen at a National Cancer Institute-funded comprehensive cancer center ambulatory clinic and used a patient-administered computerized cancer risk assessment program.
39 A study of computerized risk assessment among patients with breast cancer in an academic medical oncology setting suggested that referral uptake was reflected by stages of readiness that were not traversed for all patients within a 6-month period,
40 the time frame of our study. Recognizing that opportunities for colonoscopy or mammography screening might be limited during the 6-month follow-up period, FHITr used a modified Stages of Change model as an intermediate outcome measurement of intention to screen.
41,42 Intention to undergo any cancer screening, as assessed by the Stages of Change measures, did not differ significantly between study arms (data not shown).
The full set of Family Healthware tailored messages for all six diseases ranged from 8 to 15 pages depending on the familial risks. Clearly, there is a chance that the messages were not fully read by the study participants. Furthermore, the added exhortation to action in the tailored prevention recommendation when compared with the standard messages () may have been too subtle and may not have been fully appreciated by the participants. One interpretation of our results is that a written message is generally not sufficient as a stand-alone intervention for prompting cancer-related health behavior changes, even when provided in the context of a primary care visit as was the case for most subjects. However, our study may lack sufficient power to state this definitively.
The messages delivered by Family Healthware need to be refined. Because too much information might dilute the message and paradoxically lead to inaction, the importance of various elements of the message in prompting behavior change should be better understood and used as parsimoniously as possible. Family Healthware is programmed to inform individuals about which family history features (affected relatives and early age of cancer diagnosis) led to their strong or moderate risk categorization, because there is some evidence that tailored risk communication based on age, family history, or risk category influences risk perception and screening uptake.
43,44 However, it is unclear which specific elements of the messages are key. Coupling optimized Family Healthware messages with additional interventions warrants further investigation.
Although screening did not differentially improve in the intervention versus control group, the percentage of all eligible subjects adherent to risk-based screening increased at 6-month follow-up by 7% for colon cancer screening (change from 77 to 84% adherent) and 8% for mammography (75% to 83% adherent). Among those not up-to-date at baseline, we observed a 60% increase in the proportion current on mammography and a 36% increase in the proportion current on colon cancer screening (). The increases observed in FHITr are commensurate with effects observed for other practice-based interventions. For example, in a systematic review of strategies for improving colorectal cancer screening,
38 the percentage point increase owing to various types of interventions ranged from patient reminders (0 –15%; highest figure with a nonsignificant
P value), education videos or brochures (not effective), decision aids (mixed results; highest 14 –23%), and group education (not effective). One-on-one interventions and those focused on eliminating barriers such as access to care and language were associated with a 15– 42% increase in screening rates. System-level interventions (e.g., clinician reminders and patient navigators) showed a 7–28% improvement in screening.
Thus, it seems likely that in FHITr both the control activities (baseline questionnaire and generic prevention messages) and the intervention (baseline questionnaire, Family Healthware questionnaire, and family history-tailored prevention messages) served as prompts for patients and clinicians to accomplish cancer screening. All evaluable FHITr subjects completed a baseline survey assessing health behaviors, lifestyle choices, risk perceptions, and communication of family health history. The lengthy survey may have had unintended consequences of leading participants in both groups to ponder over their own health and/or may have been inferred as health recommendations. Two thirds of participants had an upcoming visit with the primary care clinician, offering an opportunity for advising screening, which may have been enhanced by either generic or family history-based prompts. With this study design, effects attributable to Family Healthware itself may have been masked.
In the setting of significant family history for some diseases, there is a risk that healthcare providers may order or patients may request screening tests with no evidence to support a benefit. Indeed, some of the physicians who were asked to participate in the study voiced objections to the use of messages about ovarian cancer screening and even the recommendation that patients speak to their doctor about the issue. Although such conversations may have occurred, our study shows that a low proportion of women in this situation report getting serum CA-125 testing or transvaginal ultrasounds.
The FHITr study provides limited insight into the use that clinicians made, or might make, of the automatically generated family history report and prevention prompts. Noninclusion by Family Healthware of known risk factors such as colorectal polyps, breast biopsy, and gestational diabetes may have limited its utility for clinicians. Despite its computerized format, Family Healthware was not integrated with the electronic medical record during this study; in the future, family history assessment has the opportunity to link risk stratification with clinical decision support.
45 Factors associated with improved clinical outcomes using clinical decision support systems include automatic provision of recommendations (not only assessments) as part of clinician workflow and using computer-based methods at the time and location of decision making.
46 Design of future studies and clinical implementation should incorporate these factors to enhance utility of familial risk stratification.
In summary, the FHITr study did not find an effect of family history-based prevention messages on cancer screening or consultation behaviors in a largely white, well-educated, and affluent population whose screening rates were already high at baseline. Unexpectedly high baseline screening rates resulted in low power to detect an intervention effect. Nonetheless, Family Healthware identified a substantial proportion of unscreened participants for whom cancer screening and consultation for risk assessment were recommended based on family history but would not have been recommended for the general population at average risk. It will be challenging but particularly important to couple risk assessment to a more active intervention than written messages and ultimately to achieve behavior change in a less adherent population. Engagement of clinicians, incorporation of nonfamilial risk factors for comprehensive assessment, and implementation of clinical decision support systems seem to be some of the key factors necessary to achieve the full potential of familial risk assessment.
The publisher's final edited version of this article is available at 