An Innovative Strategy for Collecting Family Health History: An Effectiveness-Implementation Trial in Primary Care Clinics

Abstract

PURPOSE

We aimed to evaluate an innovative strategy to collect family history (FH) and explore patients’ views of this strategy.

METHODS

We conducted a matched-pair effectiveness-implementation trial in family practices affiliated with the University of Toronto Practice-Based Research Network (UTOPIAN). The intervention group included family physicians (FPs) from randomly selected practices using electronic health records (EHRs) and an e-mailing platform, and randomly selected patients aged 30-69 years (4/FP/week) seen in clinic over a 6-month period. The matched control group included FPs (1:1) and patients (up to 5:1) from the UTOPIAN database. The intervention included patient and FP education, an e-mailed patient invitation to complete an FH questionnaire, automatic FH EHR upload, FP notification of completed FH questionnaire, and links to clinical support tools. Intervention patients were e-mailed a postvisit follow-up questionnaire. The assessed outcome was new documentation of FH in the EHR using mixed effects logistic regression and descriptive statistics for patient feedback.

RESULTS

Fifteen FPs and 576 patients were recruited from 3 multidisciplinary team practices to the intervention group, matched to 15 FPs and 2,203 patients in the control group. Within 30 days of visit, a new FH was documented in the EHR for 93/576 (16.1%) of intervention patients compared with 5/2,203 (0.2%) control patients (adjusted OR = 94.2; 95% CI, 36.8-240.8). New cancer FH documentation was greater in the intervention group compared with the control group (7.8% vs 0.1%; P < .01). Of patients who reported discussing FH (n = 296), 24.5% reported screening test recommended, 7.5% referral to a nongenetics specialist, and 2.4% referral to a genetics specialist. Most patients (60.5%) found this FH strategy helpful.

CONCLUSIONS

This study showed improved collection/documentation of FH. Contributors to success of the intervention included being patient completed and seamless EHR integration with a reminder. This FH strategy needs tailoring to different contexts.

INTRODUCTION

Family history (FH) has been described as the “most useful tool for risk assessment for common chronic diseases.”¹ Patients with a positive FH have a twice or greater chance of developing many diseases.¹ Family history can identify those who carry genetic variants that increase the risk of disease.^2,3 Many evidence-based guidelines incorporate FH into risk calculations and recommendations.^1,4,5 Despite advancing genomic technology, including genome sequencing and polygenic risk scores, FH remains the reference standard for risk assessment of a genetic condition and/or chronic disease and is used for interpretation of genomic tests.¹ Actions to mitigate increased risk, such as modifying screening frequency and/or modality, genetic counseling and/or testing, lifestyle changes, and risk-reducing surgery, can then be taken, potentially saving lives, especially for those with pathogenic variants.^1-3

Despite these potential benefits, FH is often not entered or updated in electronic health records (EHRs), and the adoption of FH tools for risk assessment and targeted risk reduction has been consistently low.^2,6-10 Challenges include clinician time, accuracy, difficulty accessing clinically integrated risk calculators during visits, and lack of EHR-integrated clinical guidance for personalized management.^1,2,10,11

Many FH tools exist, both patient and clinician facing, to meet these challenges.^1,3,12-22 Examples of computerized FH collection and risk-assessment tools include Family Healthware (US Centers for Disease Control and Prevention),^9,19,20 MeTree (MeTree&You Inc),²³ Health Heritage (NantHealth LLC),²⁴ and MyFamily (Family Care Path Inc).²⁵ A systematic review found a 46%-78% improvement in data recording, particularly in identifying high-risk individuals, with FH tools compared with standard practice.¹⁴ Many tools assessed only cancer and were confined to cancer clinics. Various strategies have been studied to increase clinical use of FH, with varying success.^9,26,27 Methods from implementation science could inform strategies to increase uptake of FH tools.² Ideal features of an FH strategy have been described as computerized, patient administered, user friendly, collects all data necessary for risk stratification, updateable, integrates risk algorithms with evidence-based clinical decision support, and EHR compatible.^1,28 Our objective was to evaluate an innovative FH strategy that incorporates many of these ideal features, using evidence-based implementation approaches, with the ultimate goal of improving personalized primary care.

METHODS

We conducted a matched-pair study using a hybrid effectiveness-implementation design to test the effects of the intervention on relevant outcomes while gathering information on implementation.²⁹ We used the Consolidated Framework for Implementation Research³⁰ as well as ideal features of an FH strategy from the literature^1,2,28 to guide the design of our intervention strategy. This study followed the Consolidated Standards of Reporting Trials reporting guideline.³¹

Study Setting and Design

Seven eligible primary care multidisciplinary team³² practices affiliated with the University of Toronto Practice-Based Research Network (UTOPIAN)³³ were invited to participate. At the time, UTOPIAN comprised a network of more than 100 clinics in Ontario.³³ Eligible practices included those using the Ocean (OceanMD; WELL Health Technologies Corp)³⁴ e-mailing platform via PS Suite EHR (Telus Health).³⁵ Study investigators, trainees, and family physicians (FPs) who did not routinely contribute data to UTOPIAN were excluded. Six practices agreed to participate. We used a 2:1 allocation for intervention vs control practices using a random number generator to assign practices to each arm (4 intervention, 2 control practices). All FPs who contributed data to UTOPIAN were invited to participate. One intervention and 2 control practices dropped out, owing to research ethics board delays. We replaced the control arm with randomly selected matched control FPs using a 1:1 allocation of eligible FPs contributing data to UTOPIAN, whose practices had research ethics board approvals in place for routine collection of EHR data. The intervention was delivered for 6 months at each site from September 2021 to June 2022, during the COVID-19 pandemic. Ethics approval was obtained from the research ethics boards of the University of Toronto (RIS #40764), Mount Sinai Hospital (REB #20-0270-E), and North York General Hospital (REB #21-0038).

Intervention and Study Pathway

The intervention was multifaceted, consisting of patient and FP education, a family history screening questionnaire (FHSQ),¹⁶ clinician EHR alerts, seamless EHR integration, and clinical decision support via electronic links to point-of-care tools. Intervention group FPs were invited to a webinar regarding the study and importance of FH. Patients of participating clinics were informed via waiting room television presentations, clinic websites and e-mail (including a link to a YouTube [Google LLC] presentation). We recruited patients of FPs in the intervention group using clinic appointment schedules, with 4 patients randomly selected per physician per week over a period of 6 months by a research associate (S. Kukan) using a random number sequence generator to determine order of approach. Patients aged 30-69 years, not pregnant, with any type of appointment, were eligible. A study invitation, consent form, demographic questionnaire, and FHSQ were e-mailed to eligible patients 1 to 2 weeks ahead of their scheduled appointment. The FHSQ (Supplemental Appendix), adapted from a tool validated in PC,^16,36,37 is designed to identify people at increased risk of breast, ovarian, colorectal, and prostate cancer, melanoma, ischemic heart disease, and type 2 diabetes. Conditions on the FHSQ have clear FH criteria determining risk and effective interventions for risk management. We added questions regarding genetic testing and presence of a genetic disorder (as defined by the patient).

The FHSQ results were automatically uploaded to the EHR; FPs were notified by an EHR prompt on the day of the patient’s appointment that there was FH information in the EHR for review. Family physicians were directed to complete a multiple-choice FH action form in the EHR after the visit, indicating how FH information was used. Patients were asked to complete a postvisit feedback questionnaire 3 days after their appointment. Control patients and FPs received no intervention and could discuss and enter FH into the EHR per usual care.

UTOPIAN Data Collection

Consenting intervention patients were identified within the UTOPIAN EHR database and linked to their survey data. Intervention FPs were matched to control FPs using the UTOPIAN EHR database. We performed stratified sampling without replacement to match each intervention FP with a control FP via 1-to-1 matching with age (birth year ± 2 years) and exact match on sex and type of EHR. After control FPs were identified, we performed patient-level matching using each FP in the intervention group as the stratifying factor. Control patients who satisfied the same eligibility criteria were selected. Intervention patients were matched with a maximum of 5 control patients using exact match on sex and birth year ±5 years. This matching scheme was applied using sampling without replacement to increase statistical power and minimize statistical bias.³⁸ Deidentified UTOPIAN data were extracted biannually from the EHRs of FPs who had previously consented and their passively consenting patients and stored in a Data Safe Haven for analysis.³⁹ Standardized mean differences were used to estimate the difference in patient characteristics between intervention and control group patients.

Outcome Measures

The primary outcome was any new documentation of FH in the FH section of the EHR by a primary care clinician within 30 days of the clinic visit. This could be new or updated FH. Secondary outcomes included new documentation of FH of cancer, heart disease, or diabetes; incidence of positive FH on the FHSQ; process outcomes including FP actions after FH review; and patient and FP experiences with the intervention, its feasibility, and suggested improvements. Family physicians’ perspectives obtained via qualitative interviews are reported separately. To estimate the sample size, using UTOPIAN 2017 first-quarter data, we determined that 2% of patients aged 30-69 years had their FH updated after their most recent clinic visit, comparable to the literature rate.²⁷ We estimated that improvement to 20% would be clinically meaningful. We used an intracluster correlation coefficient of 0.1, with an average cluster size of 5 FPs per practice. With these assumptions, for the primary outcome, we estimated 80 patients per arm would give 80% power to detect the difference from 2% to 20% for an α value of .05.

Data Analysis

We used SAS version 9.4 software (SAS Institute Inc) for data analysis. We analyzed questionnaire data descriptively using frequency distributions. For the primary outcome, adjusted odds ratios (ORs) were calculated using mixed effects logistic regression modeling, accounting for clustering and physician/patient matching using random effects. Differences in incidence of new recording of cancer, heart disease, or diabetes were evaluated using Rao-Scott adjustment to χ² test with false discovery rate criteria applied.^40,41 Patient responses to short-answer questions on the postvisit questionnaire were independently reviewed by 4 researchers (J.C.C., E.B., S.W., S. Kukan) who met to agree on descriptive coding and themes.^42,43

RESULTS

Three primary care practices (2 community, 1 academic) participated, with 15/20 (75%) eligible FPs agreeing to participate. We e-mailed 1,857 study invitations to eligible intervention group patients, of whom 676 (36.4%) consented to participate (Figure 1). Of these, 646 (95.6%) completed the FHSQ. Of these consented patients, a final sample of 576 (89.2%) could be linked to the UTOPIAN database. The mean intervention patient age was 52.2 years, two-thirds were female, and 94% lived in urban areas (Table 1).

Figure 1.

Intervention Patient Flowchart

FHSQ = family history screening questionnaire; UTOPIAN = University of Toronto Practice-Based Research Network.

Table 1.

Patient and Family Physician Characteristics

Characteristicsa	Intervention	Control	Standardized mean difference
Patient
No.	576	2,203
Age, y, mean (range)	52.2 (30-69)	51.5 (30-69)	0.03
Sex, female, No. (%)	387 (67.2)	1,460 (66.3)	−0.02
Region, urban, No. (%)	543 (94.3)	2,017 (91.6)	0.41
Marital status, married, No. (%)a	396 (68.9)	nab	nab
Education, completed university and/or post-graduate/professional training, No. (%)a	485 (84.3)	nab	nab
Household income, >$100,000, No. (%)a	339 (58.9)	nab	nab
Ethnicity, White/North American/European, No. (%)a	419 (72.9)	nab	nab
Physician
Number	15	15
Age, y, mean (range)	48.5 (35-68)	48.2 (33-68)
Gender, female, No. (%)	11 (73.3)	11 (73.3)
>10 years in practice, No. (%)	8 (53.3)	11 (73.3)

na = not applicable.

^aAnalyses conducted on 575 participants, owing to missing demographic data.

^bData on patient marital status, education, household income, and ethnicity only available for intervention patients.

The documentation of positive FH of heart disease, diabetes, and 5 types of cancer, self-reported by patients on the FHSQ, is summarized in Table 2. In addition, 68/576 (11.8%) reported themselves or a family member as diagnosed with a genetic disorder.

Table 2.

Positive Family History, as Completed by Patients (n = 576)

Questionnaire item	Total No. (%)
Close relative with diagnosed heart disease before 60 years of agea	189 (32.8)
Close relative with diagnosed diabetes	264 (45.8)
Close relative with diagnosed melanoma	103 (17.9)
Close relative with diagnosed bowel/colon cancer before 60 years of age	55 (9.5)
>1 relative on same side of the family with diagnosed bowel/colon cancer at any age	67 (11.6)
Close male relative with diagnosed prostate cancer before 60 years of age	34 (5.9)
Close female relative with diagnosed ovarian cancer	65 (11.3)
Close relative with diagnosed breast cancer before 50 years of age	78 (13.5)
>1 relative on same side of the family with diagnosed breast cancer at any age	113 (19.6)
Completed genetic testing	93 (16.1)
Family member with diagnosed genetic disorder	68 (11.8)

^aClose relative defined as parent, child, brother, or sister, living or dead.

Recording of Family History

We matched 576 intervention patients with 2,203 control patients. Before the intervention, 470 (81.6%) of intervention patients and 1,904 (86.4%) of control patients had some FH recorded on their charts, indicating comparability (Table 3). Within 30 days after the clinic visit, new FH was documented in the FH section of the EHR by their primary care clinician for 93 (16.1%) intervention patients and 5 (0.2%) control patients (adjusted OR = 94.2; 95% CI, 36.8-240.8, P < .001).

Table 3.

Documentation of FH in the EHR at Baseline and Within 30 Days of Clinic Visit

	Control (n = 2,203)	Intervention (n = 576)	P a	Unadjusted OR (95% CI)	Adjusted OR (95% CI)a	P b
Patients with any FH recorded in EHR at baseline, No. (%)
No	299 (13.6)	106 (18.4)		nac	nac	nac
Yes	1,904 (86.4)	470 (81.6)	.37
New documentation of FH within 30 days of clinic visit, No. (%)
No	2,198 (99.8)	483 (83.9)		84.6 (34.2-209.3)	94.2 (36.8-240.8)	<.001
Yes	5 (0.2)	93 (16.1)	<.01

EHR = electronic health record; FH = family history; na = not applicable; OR = odds ratio.

^aRao-Scott χ² adjustment with false discovery test.

^bAdjusted OR calculated using mixed effects modeling with nested specifications, adjusting for income quintiles and rurality.

^cUnadjusted/adjusted ORs not calculated for baseline data because there was no statistical difference between study arms.

Within 30 days of the clinic visit, 45 (7.8%) intervention patients and 2 (0.1%) control patients had documentation of FH of ≥1 of the 5 cancers (P < .01) (Supplemental Table). Nineteen (3.3%) intervention patients and 1 (<0.01%) control patient had heart disease FH documented (P < .01).

We assessed a possible spillover effect for patients not involved in the study but seen by intervention FPs during the study period. In this group, new FH was recorded in the EHR for 10/4,459 (0.2%), compared with 8/7,481 (0.1%) for patients of control FPs.

Patient and Family Physician Feedback

Intervention patient feedback on the FH strategy is summarized in Table 4. The majority of those who attended their appointment (296/409, 72.4%) indicated discussing FH with their FP at the visit; of those, 251 (84.8%) said their FP initiated the discussion. Approximately two-thirds considered this a helpful way of informing FPs of their FH and recommended continuing to collect and update FH in this way.

Table 4.

Patient Feedback on Family History Intervention (n = 478)

Feedback	Patient response No. (%)
Comfort filling out the FHSQ (very/somewhat)	414 (86.6)
Time to fill out the FHSQ 10 min	339 (70.9)
Discussed family health history with FPa	296 (72.4)
FP initiated FH discussionb	251 (84.8)
Length of conversation regarding FH 5 minb	229 (77.4)
Helpful/very helpful as way of informing FP of FH	290 (60.7)
Agree/strongly agree FH should continue to be collected and updated in this way	319 (66.7)

FH = family history; FHSQ = family history screening questionnaire; FP = family physician.

^aLimited to those who attended their clinic visit (n = 409).

^bLimited to those who attended and discussed their FH (n = 296).

Patient and FP reports on what actions were taken at the visit based on FH are summarized in Figure 2. Patients reported recommendations for screening tests (24.5%) and/or lifestyle changes (7.8%). Family physicians reported a change in screening or management at 3.4% of visits. Referral to a genetics specialist was reported by approximately 2% of patients and FPs, whereas nongenetics specialist referrals were reported by 7.5% of patients compared with 0.8% of FPs.

Figure 2.

Patients’ (n = 296) and FPs’ (n = 267) Reports of Actions Taken at Clinic Visit

FH = family history; FP = family physician.

Note: >1 response option allowed; percentages do not sum to 100%.

Themes from the patient short-answer questions indicated that more clarity was needed on the FHSQ, including definition of family, age cutoffs, and conditions to report (n = 25). Patients wanted the opportunity to disclose additional conditions in the family, particularly stroke, dementia, mental health, bone disease, colitis, and conditions for more distant relatives (n = 20).

DISCUSSION

This multifaceted intervention resulted in significant improvements in EHR documentation of FH and its discussion with patients. Consistent with the literature, the effects of this intervention were likely due to time efficiency (patient completion at home, EHR integration); addressing knowledge gaps (easily accessible education, evidence-based clinical decision support); and system integration (incorporation into usual care flow with physician reminders). The ascertainment of cancer, heart disease, and diabetes FH in first-degree relatives replicated or exceeded reports from other studies, suggesting clinical utility of the FHSQ strategy.^44-46

Murray and colleagues compared patient-collected FH before an appointment using 3 strategies (telephone voice response system, FH tool on a secure website at home, or waiting room laptop computer) to usual care.²⁷ Similar to our findings, there was documentation of new FH in 16.7% of intervention patients compared with 1.7% control patients. Participation was 9.8% to 14.6% across intervention arms, compared with 36% in our study. In that study, although the majority of patients reported discussing FH, 50% of patient-reported FH information was never reviewed by their provider, possibly due to EHR challenges, and if not reviewed, did not become part of the EHR.^10,27 In contrast, in the present study, although the FHSQ was automatically uploaded into the EHR, the FP had to enter new FH information into appropriate EHR fields for it to be counted as new FH documentation. Family history discussion occurred during most visits (72.4%), usually initiated by the FP (84.8%), and both patients and FPs reported actions taken as a result of FH.

We note the discordant reporting of actions taken as a result of FH discussions. Patients seemed more likely than their FPs to report that new preventive actions were initiated because of FH information (eg, recommendation of screening tests, lifestyle changes, referrals). Perhaps FPs thought they were suggesting screening and management based on FH they were already aware of, whereas patients perceived it as due to FH discussions. Family physicians might not think of lifestyle changes as a change in management; hence, lower numbers for the latter. Patients’ perceived change in management might have contributed to two-thirds finding this FH strategy helpful and agreeing to continue to collect and update FH in this way. Although the 2% reported rate of genetics referral might appear to be low compared with the positive FH incidence, it is compatible with the estimated 1.33% population risk for pathogenic/likely pathogenic variants for hereditary breast/ovarian cancer, Lynch syndrome (hereditary nonpolyposis colorectal cancer), and familial hypercholesterolemia, 4 conditions queried in the FHSQ.⁴⁷ It is likely that FPs took further FH details before initiating referral.

Patient-completed FH questionnaires have been shown to be fairly accurate compared with the reference standard of a 3-generation pedigree.¹⁵ Ginsburg and colleagues assessed the analytic validity of FH tools and found that patient-facing FH platforms performed better in data completeness and accuracy compared with routine clinical care.¹

A remaining challenge is to increase integration of FH tools into the clinical workflow. Consolidated Framework for Implementation Research–defined characteristics of interventions that enhance uptake include credibility of the source, evidence for desired outcomes, superiority to alternatives, adaptability to local needs, ability to test the intervention on a small scale, low complexity, ease of introduction, and low cost.^2,30 Our intervention was developed by a multidisciplinary team including FPs from different practice environments. Evidence for clinical utility of FH tools was presented at the FP webinar, FPs were given autonomy regarding how to manage FH information, and the FH strategy was integrated into patient flow with automatic EHR upload and a reminder on the day of the clinic visit, with no cost to participating physicians. Education is important for successful implementation. Allen and colleagues reported that brief educational modules that discuss the utility of FH can improve collection of information and increase the likelihood that people engage in FH conversations and tool use.² In our study, educational information for patients was available on the waiting room television, clinic website, and by e-mail. Patients indicated satisfaction with the FH strategy, suggesting successful future spread of the intervention.

We conducted this study during the COVID-19 pandemic. At that time, we perceived that FPs had lower receptivity to an intervention of this type and diminished capacity to introduce change to their practice. The apparent increase in FH documentation might therefore be an underestimate of its true potential effect.

Ideally, patients who have been educated regarding the value of FH would know how to gather the information, confer with relatives, and fill out a patient-facing FH tool. Family history would be automatically uploaded into the EHR, prompting an automatic risk algorithm providing clinical decision support for both patient and clinician to facilitate management decisions.¹ MeTree has provided proof of concept of this approach.^48,49 Further work is needed with EHR vendors to incorporate multiple languages and risk/management algorithms. An advantage of our strategy is that the FH tool is freely available, can be distributed using any secure e-mail platform, and is downloadable for clinical use.

Limitations

The baseline completion of FH was high for both intervention and control FPs compared with other studies, perhaps providing a ceiling effect. The one-third of patients invited to the intervention group who actually consented likely selected for those with a positive FH, perhaps skewing toward positive results. In addition, participating patient interviews could have enhanced understandings gained from the study.

Equity is also a concern, given that most intervention patients were White and well educated. Patients needed to be able to read English, have internet access, be computer proficient, and lack significant privacy concerns. Ensuring broader acceptability and usability of the FHSQ will require attention to these issues.

We plan to revise the FHSQ as suggested by patients and FPs, taking into account its usability by diverse groups, and implement this as a quality improvement project, maintaining features that led to success. Future strategies might include offering completion of the FHSQ in the clinic and assessing the management/screening tests ordered, whether they matched patients’ risk level, and whether patients followed recommendations.

CONCLUSION

This strategy showed significant improvement in collection and documentation of FH. Patients welcomed the opportunity to provide FH information before appointments. Factors contributing to the intervention’s success included being completed by the patient and seamless EHR integration with a reminder. Further research is needed to determine if these findings can be replicated with more diverse populations and to tailor the intervention to different practice contexts to improve collection and use of FH.