Effect of Preventive Messages Tailored to Family History on Health Behaviors: The Family Healthware Impact Trial

INTRODUCTION

Family history holds promise to target health promotion activities more effectively to individuals and families according to their genomic risk for disease.^1,2 Family history captures genetic and environmental components of disease, including shared cultures and behaviors.³ The efficiency of familial risk stratification was shown in a Utah-based study reporting that 72% of coronary heart disease clustered in 14% of families, and 86% of strokes occurred in 11% of families.⁴

The extent of family history collected varies widely.^5–7 Several issues contribute to the underutilization of family history, including time, complexity, and accuracy.^5,8,9 Even when family history is collected, physicians may not precisely identify familial risk.¹⁰ Furthermore, patients may have incomplete knowledge of their family history.¹¹ Interventions that efficiently collect family history, automate familial risk stratification, and allow patients to confirm and revise their family history with relatives are likely to increase the utility of family history¹¹ and the feasibility of family history research in primary care.

Family history allows recommendations for disease prevention and screening to be tailored to one’s level of familial predisposition.^12,13 Currently, however, there is insufficient evidence showing that family history assessment in primary care improves patient outcomes.^6,7 A recent systematic review of the usefulness of routinely collecting family history in primary care identified 2 uncontrolled before-after studies of adherence to breast cancer screening, with risk defined by family history of breast or ovarian cancer.⁷ In a recent British randomized trial, a computerized tool used by general practitioners for familial cancer risk increased the number and appropriateness of referrals to cancer genetics clinics.¹⁴ A National Institutes of Health State-of-the-Science Conference in 2009 concluded that research addressing this area is needed.⁶

To facilitate efficient collection and use of family history for disease prevention, the Centers for Disease Control and Prevention (CDC) created Family Health-ware, an interactive online tool that collects and records family history of 6 common diseases (coronary heart disease; stroke; diabetes; and colorectal, breast, and ovarian cancer) and prevention-relevant habits (physical activity, diet, smoking, alcohol use, aspirin use, and screening test use) for each disease. The software immediately analyzes the user input, generates a 3-tiered risk assessment based on family history for each disease, and provides tailored preventive health messages to the user focused on health behaviors, screening, and family risk category. The Family Healthware report includes a family tree, as well as information about the specific details of family history that increase the respondent’s risk. The development and features of Family Healthware are described elsewhere in detail.¹⁵

The algorithms to classify risk use individuals’ reports of their own health history and that of first- and second-degree relatives, including diseases in maternal or paternal lineages, and age at onset. The algorithm categorizes risks as weak, moderate, and strong according to rules derived from empirical data and validated in epidemiologic studies.¹⁶ The labels for each risk category were developed through focus group studies.¹⁵ Usually, a weak familial risk was assigned if there was no family history or if there was late-onset disease in only 1 second-degree relative. Moderate familial risk was generally assigned if there was only 1 first-degree relative with late-onset disease or 2 second-degree relatives from the same lineage with late-onset disease. Strong familial risk was generally assigned if there was a first-degree relative with early-onset disease, when multiple relatives were affected, or when a hereditary syndrome was present. Diabetes was also considered in stratifying risk for coronary heart disease. For most common chronic diseases, a moderate familial risk is associated with about a 2-fold increase in risk over a weak familial risk, and a strong familial risk is associated with about a 3-fold or greater increase.¹³

In 2003, the CDC selected 3 academic centers to evaluate this new tool: Evanston Northwestern Health-care (ENH) (now NorthShore University Health-System), the University of Michigan (UM), and Case Western Reserve University (CWRU) with the Ameri-can Academy of Family Physicians’ National Research Network (AAFP NRN). The goal of the Family Health-ware Impact Trial (FHITr) was to determine whether systematic family history collection, risk assessment, and messages tailored to familial risk influence the adoption of healthy behaviors, recommended health screenings, and family and physician communication related to the 6 diseases.

This analysis examines the effects of Family Health-ware–tailored messages compared with generic health messages among participants reporting suboptimal behaviors with respect to smoking, aspirin use, diet (fruit and vegetable intake), physical activity, blood pressure screening, cholesterol screening, and diabetes screening. We hypothesized that patients who completed Family Healthware and received a risk assessment and messages tailored to family health history of the 6 diseases would be more likely to change these lifestyle behaviors or obtain indicated screening when compared with patients receiving a generic preventive health message (Figure 1⇓).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Conceptual model of how tailored messages on family risk status and recommended preventive strategies would result in behavioral changes.

METHODS

The study methods are described in detail elsewhere.¹⁷ Briefly, FHITr used a 2-arm cluster-randomized design, with primary care practices as the unit of randomization. Eligible patients were recruited from 41 primary care practices; they were aged 35 to 65 years and did not have coronary heart disease, stroke, diabetes, or any cancer except skin cancer other than melanoma. Patients were excluded if pregnant or unable to speak or read English. At ENH and CWRU/AAFP NRN, potentially eligible patients with upcoming appointments were sent invitation letters signed by their primary care physicians. At the UM site, invitation letters were sent to the entire panel of eligible patients of participating clinics. Figure 2⇓ displays recruitment and accrual of practices and patients.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Consort diagram of practice and participant recruitment.

Individual protocols were approved in 2004 by institutional review boards (IRBs) at all 3 centers, and a combined protocol was approved by the CDC’s IRB. Recruitment took place from November 2005 to March 2007. The allocation to control or intervention was designed to be about 1 to 2.

In the intervention arm, participants first completed an online baseline questionnaire followed by Family Health-ware, which provided personalized familial risk assessment and prevention messages (see Supplemental Appendix 1, available at http://annfammed.org/cgi/content/full/9/1/3/DC1, for a summary of messages). The control group completed the same online baseline questionnaire and received standard prevention messages for the same 6 diseases (Supplemental Appendix 2, available at http://annfammed.org/cgi/content/full/9/1/3/DC1, for the standard health message). Approximately 6 months later, both the intervention and control groups completed a follow-up questionnaire. Then the control group also completed Family Healthware tool.

The baseline survey collected demographics, health status, and use of medical services; screening behaviors and intention to screen; and lifestyle choices, health beliefs, and family history communication patterns (with family members and physicians). The 6-month follow-up survey assessed behavior and health belief changes over time.

Participants viewed their prevention messages on a computer screen after completing the baseline questionnaire. In 2 of the study sites (ENH and CWRU/AAFP NRN), accounting for 67% of the study population, printed prevention messages were provided to intervention and control groups at a scheduled primary care appointment, with a copy provided to the physician at the patient’s request. At the third site (UM), the prevention messages were mailed or e-mailed to the patient—who was asked to bring the messages to the next doctor visit—and copies were given to their physician and placed in the electronic health record.¹⁷ Physicians at CWRU sites were to complete a questionnaire at the time of each participant’s visit. Responses were available for 75% of intervention group participants, showing that the physician looked at the Family Healthware report during 74% of these visits (range among physicians, 41%-100%).

RESULTS

Of the 4,248 enrolled participants, 3,786 completed the baseline survey. FHITr had an 18% recruitment rate and an 88% retention rate (from baseline to follow-up, as summarized in Figure 2⇑). An unknown proportion of the invited patients (estimated by chart review at ENH to be 13%)¹⁹ were ineligible because of chronic diseases. As seen in Figure 1⇑, sites that enrolled patients with an upcoming appointment had approximately double the participation rate compared with patients invited without an appointment. Overall, the mean age of participants was 50.6 years, and most were white (91%), female (70%), married (76%), insured (97%), and at a relatively high socioeconomic level. The distribution of participants by primary care specialty was family medicine 48%, internal medicine 39%, and obstetrics/gynecology 12%. Table 1⇑ displays the demographic characteristics of 2,364 intervention and 1,422 control group participants. After controlling for study site and practice-level clustering, the only statistically significant difference (P = .003) was proportion of participants beginning the study during different seasons (defined by 4-month periods) of the year. The distribution of familial risk of diabetes differed by study arm, with the control group exhibiting a slightly higher proportion of weak family risk (P = .03). No other differences in familial risk by study arm were observed (Table 2⇓). Among the participants using Family Healthware, 91% viewed their risk report and messages online.

The distribution of lifestyle and screening behaviors at baseline are summarized in Table 3⇓. Adjusting for site, age, sex, and practice clustering, there was a significantly higher percentage of participants not at goal in the intervention group with respect to aspirin use (P <.001). With the exception of aspirin use, there was no statistically significant difference across the 2 study arms. Ninety percent or more of participants were already at goal for blood pressure, cholesterol, or blood glucose testing.

Table 4⇓ summarizes the changes in preventive behaviors from baseline to follow-up for each study arm. The highlighted categories for each behavior represent the comparison groups used to determine the outcomes. The intervention group was significantly more likely than the control group to move from not at goal status to at goal for both fruit and vegetable consumption (absolute difference in proportion = 3%; odds ratio [OR] = 1.29; 95% confidence interval [CI], 1.05–1.58) and physical activity (absolute difference in proportion = 4%; OR = 1.47; 95% CI, 1.08–1.98), even after adjustment (Table 5⇓). Contrary to expectation, the study participants exposed to the message tailored to their family history were significantly less likely to move from not having a cholesterol test in the last 5 years to having it done within 5 years than were the participants exposed to the standard health message (absolute difference in proportion = 15%; OR = 0.34; 95% CI, 0.17–0.67). There was no significant intervention effect on smoking cessation or blood pressure and blood glucose testing. Controlling for the number of high-risk messages, the total number of increased-risk messages received by each participant, or baseline perceived risk of each disease did not affect the results. We found no difference in outcomes between sites that enrolled patients with appointments compared with the site that invited eligible patients regardless of upcoming appointments.

Overall, older participants moved to at goal more frequently for fruit and vegetable consumption (P <.001), blood glucose testing (P = .002), and cholesterol testing (P = .003), whereas younger participants showed a higher propensity (P <.001) to move to at goal for aspirin use. Women were significantly more likely than men to move from not at goal to at goal for fruit and vegetable consumption (P = .01) and aspirin use (P = .001), whereas the reverse was true with respect to blood glucose testing (P <.001) and smoking cessation (P = .048). Nonsmokers were more likely than current smokers to move to at goal for fruit and vegetable consumption (P = .002).

DISCUSSION

FHITr is the first randomized controlled trial to assess whether systematically screening for family history and tailoring prevention messages to familial risk improves any health behaviors.⁷ We study hypothesized that tailoring preventive recommendations for common diseases based on an individual’s family history risk of each disease could increase the percentage of patients completing the recommendation. Our results support this concept for some behaviors, showing statistically significant increases in self-reported physical activity and fruit and vegetable intake for primary care patients using Family Healthware compared with those receiving a standard prevention message. Use of Family Healthware and delivery of the tailored message did not significantly increase the percentage of adults who were at goal for other health behaviors or screening compared with a standard message. Future analysis will determine whether the effects reported here were mediated by changes in perceptions and communications as summarized in Figure 1⇑.

Improved adherence to recommended levels of physical activity and fruit and vegetable intake after using Family Healthware was modest, but potentially influential if the changes are maintained over time, especially given the onetime intervention used to achieve this initial change. The magnitude of changes observed in daily fruit and vegetable intake²⁰ and physical activity²¹ are similar to effects observed with other minimal interventions.

The family risk level was categorized as weak for most participants for every disease except coronary heart disease, as noted in Table 2⇑. This group of participants could have interpreted the message of no familiar risk as approval to engage in unhealthy behaviors. There was no evidence of this behavior change in Table 3⇑, however.

There are several potential reasons why the FHITr study failed to show benefits for other health behaviors and screening. First, an unexpectedly high proportion of the study participants at baseline were already at goal for some of the recommended lifestyle and screening behaviors. As a result, study participants had very little room for improvement, thus limiting the power to detect a difference for smoking cessation, aspirin use, and blood pressure and blood glucose testing. Although participants were active patients in primary care practices, it is unclear why the recruitment process resulted in such a high proportion of at goal participants. Recruitment materials entitled, Evaluation of Tools for Health Promotion and Disease Prevention, described the study goal to help people make decisions about health services. Patients with a commitment to disease prevention may have been more likely to participate.

Second, the study population overrepresented white, female, married, and insured patients with a relatively high socioeconomic level compared with the populations served by the clinics. Efforts to recruit a more diverse study population failed. A barrier to recruiting a more diverse sample may have been the need for computer literacy and access. Participants, however, could also participate in the study by telephone or in person, as was done by 9% of participants.¹⁷ Although computer and Internet use is expanding rapidly in the United States, large disparities exist between ethnic and racial groups.^22,23 Thus, comfort with using computers may have contributed to a biased study sample. Related studies^24,25 have identified lack of access to a computer, absence of family history, and concerns about Internet security and privacy as some of the main barriers to using an Internet family history risk assessment tool.

The Family Healthware–tailored messages ranged from 8 to 15 pages. Furthermore, the reading grade level of the messages has been assessed at 11.8.²⁶ It is thus possible that the messages were not fully comprehended by participants. A simple 1-page generic message may have been easier to access and appeared to be more effective in convincing participants to have a blood cholesterol test. The messages delivered by Family Healthware could be shortened, highlighting key components, reducing the reading grade level, and providing more directive action steps. The messages used were standard health messages that had been used in public health settings for years. It may be time to reexamine these messages in terms of reading level, communication theory, and behavioral change theories.

We do not have data on whether the physicians reviewed the messages the study participants received or made any recommendations concerning the behaviors. Our survey of a limited physician subsample suggests they reviewed the messages for three-fourths of the participants. The degree to which physicians reinforced the messages, however, is unknown.

The study outcomes are also limited by a single (6-month) follow-up observation time. The observational interval was determined by the funding agency and resources available to conduct the study. Conclusions are also limited by the self-reported nature of the data, in that more intervention participants may have decided to endorse the socially acceptable responses once prompted by the Family Healthware reports.

This study was carried out in primary care practices because they are the locus for delivery of many preventive services.²⁷ We did not examine the impact of Family Healthware on physicians’ recognition of their patients’ familial disease risks or on physician-patient interactions regarding preventive care. Medical record reviews at ENH practices did show that Family Healthware disclosed increased familial disease risks not otherwise found in a substantial proportion of participants’ medical records.²⁸ Effective ways to engage practices and physicians in recognizing familial disease risk and enacting risk-appropriate prevention strategies require further exploration. For example, electronic family history tools like Family Healthware could reduce some of the time spent collecting and processing information in the office. A key consideration will be how to integrate the information into the patient flow of the primary care practice. Simply placing a note in the electronic health record along with an electronic message to physicians outside the context of a patient visit will be insufficient to induce physicians to make familial risk-based changes to their patients’ medical care.²⁹ Importing the information and risk classification into an electronic health record’s problem list with decision support built around the risk level might be more effective than the intervention evaluated here.

Data are still needed to show that familial risk data change preventive behaviors and health outcomes. Future research needs to recruit a more diverse study population, including participants not up to date on recommended screening. In addition, future research should examine how physicians use the output from tools like Family Healthware and how it affects communication between patients and physicians. Does Family Healthware provide new information to the patient or physician? Further, in the emerging context of the patient-centered medical home and models of team-based care, might such a tool help nonphysician staff to engage patients around familial risk and health behavior? Finally, it remains to be determined whether the development, deployment, and integration of tools such as Family Healthware into practice information systems will be cost effective.

Physicians have long used information about a patient’s family history, lifestyle, and exposures. The modest positive impact of the FHITr intervention on dietary intake and exercise behaviors suggests that familial risk may be an important motivator of health behavior change. This study, however did not address how to improve the delivery and use of this information. FHITr has shown that technology to systematize family history assessment is feasible, but we need more insights into how it will be used and its place in preventive care.