EHR data were available from Renown via the Epic ² Clarity database, a large subset of the data in the Epic EHR application. EHR data were available from 2006 to 23 August 2021, although the EHR wasn’t fully implemented until 2011. The patient diagnosis data review was conducted in June 2021 and participants were included for EHR review if at least 3 months passed since the T1pos notification to ensure that participants had time to respond to their findings.

Diagnoses were retrieved using native application diagnosis codes (nDx) found in more than forty clinical and administrative/billing tables. Each nDx was associated with an entry date as well as a native diagnosis description and mapping (if available) to ICD-9-CM ³ and/or ICD-10-CM ⁴ codes. In general, nDxs map to one or more ICD codes and are often more specific than ICD codes. Because of their greater specificity, we used nDxs for our analysis rather than ICD codes.

All retrieved nDxs were initially reviewed based on their description and only diagnoses deemed relevant to the T1pos finding of an individual were retained (Supplementary Material S1). A detailed review of the remaining nDxs was conducted to determine relevance to each specific T1pos condition. All nDxs reviews were conducted by a physician (GE) with an Internal Medicine background. Prior knowledge of T1pos conditions was defined as a genetic diagnosis appearing in the EHR prior to the notification date.

For ancillary procedures, we focused on retrieving representative screenings for each condition: mammograms and other types of screening breast imaging procedures for HBOC (Winters et al., 2017; Lee et al., 2020), colonoscopies for LS (Jasperson et al., 2010; Peterse et al., 2020) and LDL tests for FH (Youngblom et al., 2014; Jellinger et al., 2017). LDL laboratory results were retrieved directly from the results table in Clarity based on native component codes, whereas imaging and other procedures were collected based on orders, results, and mentions in the clinical notes. We also retrieved indications that a mastectomy or oophorectomy was performed from diagnosis tables, clinical notes, and surgery log tables.

Clinical notes for all participants with medical records were retrieved based on a comprehensive keyword search using terms related to each individual condition (HBOC, LS, FH); to genetic testing, findings, or consultations; and to the HNP. Several iterations of the keyword search term collection were performed until no missed terms were found in two repetitive random samples of 100 notes from the entire collection of T1pos participants’ notes. All selected notes were then manually reviewed by a single reviewer (GE) for any references pertinent to T1pos findings.

To determine whether participants and their physicians possibly enacted changes to clinical care after notification, we visually examined patient timelines. Changes in care were suspected under the following conditions: if there was an increase in the frequency of mammographies or if prophylactic mastectomies or oophorectomies were performed (HBOC), if a new colonoscopy was ordered without prior history of screening colonoscopy or outside of the recommended timeframe of repeat colonoscopy (LS), for FH we used change in LDL levels as an overall indicator of lifestyle changes and outcome of possible prescribing of effective lipid lowering medications.

Explicit referrals for confirmatory genetic testing were not visible from the Renown EHR. However, we examined recorded referrals for LGCs within the Renown EHR as well as available data from the third-party vendor ⁵ that conducted genetic consultations on behalf of the HNP and was responsible for such recommendations for confirmatory testing.

Surveys were sent in January 2020 and October 2021 to 462 T1pos participants that had consented to further research participation (not all were included in our study due to a cutoff point of May 2021 for T1pos results). The survey (Supplementary Material S2) was electronic, and participants answered up to 24 questions, depending on their responses. Several reminders were sent within 2 weeks to participants who had not yet responded to the survey. Survey responses were then aggregated and analyzed, and the responses of participants who were also Renown patients were matched with their EHR.

Most results reported in this study were descriptive and did not require the use of statistical tests. However, Fisher’s Exact Tests were used to test whether the likelihood that a participant was T1pos, was notified, or had an EHR record differed due to sex or race, and Wilcoxon Rank Sum Tests were used to test whether there were differences due to age. Pearson’s Chi-squared Tests were used to test whether survey responses differed between T1pos conditions. A Bonferroni correction was used to adjust for multiple testing where appropriate.

On May 2021 there were 520 HNP participants (out of 41,835) that were T1pos for HBOC (268 participants), LS (102 participants) and/or FH (153 participants) (Figure 1). There were two participants with both HBOC and FH and one participant with both HBOC and LS. Participants in this study were notified between September 2018 and September 2020. Notification and counseling were completed for 293 (56.3%) of the 520 T1pos participants, and notification success was significantly higher for white participants (Table 1).

Out of the 520 T1pos participants, 417 had reviewable EHRs. After filtering out diagnoses clearly unrelated to HBOC, LS or FH, 14,584 nDXs were collected for those 417 participants (corresponding to an average of 35 unique native diagnoses per participant). 250 (60%) of the 417 participants with Renown EHR were successfully notified. Their mean age was 47.5, they were 33.2% male, and they had a total of 9,034 nDXs, or an average of 36 unique nDXs per individual. All notified participants with EHR record met the minimum required 3 months time span between notification and EHR review (mean 2.2 years, minimum 0.9 years, maximum 2.9 years). 41 of these participants had EHR records with 20 or less nDXs, while mean time span for nDXs was 8.6 years (standard deviation: 6.0 years). Therefore, none of these participants were excluded due to lack of follow-up.

Among T1pos participants with reviewable EHRs, there were 72 out of 417 individuals with malignancies typically associated with HBOC or LS. Fifty such malignancies occurred prior to the initiation of the HNP in 2018 and only five individuals were referred to genetic consultation. Three of the five had meaningful related family history documented in the EHR prior or around the time of the diagnosis of malignancy. Sixteen participants were diagnosed with HBOC/LS typical malignancies after 2017 and prior to notification by the HNP of their T1pos findings. Five of them were referred to a LGC, three of them with strong family history documented at the time of or prior to the cancer diagnosis.

Based on review of genetic diagnoses among the 250 notified participants with EHR, 38 (15%) had EHR evidence that knowledge of their condition preceded notification, while 212 had no evidence of prior knowledge in their EHR (Figure 1). 47 (19%) could have benefited from earlier notification due to prior presentation of disease (27 HBOC, 9 LS, and 11 CVD before the age of 50 years).

Many of the genetic nDxs were non-specific even though specific nDxs, including some mentioning specific variants, exist in the system (Table 2). The four most frequent nDxs were non-specific and map directly to correspondingly non-specific ICD-10-CM codes. For HBOC, the nDx often indicated only breast or ovarian cancer susceptibility rather than susceptibility to all cancers associated with HBOC. Of the 212 participants who were notified, had an EHR, and did not have prior knowledge, 63 (30%) had at least a non-specific genetic diagnosis in their EHR, 55 (26%) had a specific genetic diagnosis in their EHR, and 25 (12%) had a specific genetic diagnosis listed in their problem list (PL) (Figure 1). We also noted that among more than 35,000 HNP participants with EHR records, 354 have a diagnosis of “Familial Hypercholesterolemia”. However, 316 of these participants were not T1pos. Also, only 11 (19%) of the 59 FH-notified participants without prior knowledge had a specific FH diagnosis, and only one of these diagnoses appeared in the PL. Review of the clinical notes of the 354 HNP participants with an FH diagnosis found only a single case where the clinical FH diagnosis was supported by a documented Dutch Lipid Clinic Network Criteria (DLCN). No evidence of use of the Simon Broome or the Making Early Diagnosis Prevents Early Death (MEDPED) clinical criteria was found.

Visually examining patient timelines for the 85 female HBOC patients (Figure 2A), we found 10 patients (12%) who appeared to have a change in care. Seventy-five female HBOC patients (88%) exhibited no change in care, of which 40 either had prior cancer or prior knowledge of their HBOC status. Among the 49 LS patients (Figure 2B), four (8%) did not have prior colon cancer and appeared to have received a colonoscopy related to their notification. Forty-five LS patients (92%) exhibited no change in care, of which 10 had prior cancer or prior knowledge of their LS status. Among the 66 FH patients (Figure 2C), LDL levels improved (at least temporarily) for six (9%) of the patients. Sixty (91%) of the patients positive for FH variants exhibited no change in care, of which 10 had CVD prior to both the notification of their FH status and the age of 50 years.

The survey sent to T1Pos-notified (Supplementary Material S2) had an overall response rate of 39.6% and a 42.8% response rate among individuals with an EHR, for a total of 107 respondents with an EHR (Figure 3). Among these 107 respondents, 18 (17%) indicated that they did not recall being notified, and 76 (71%) indicated that they reported their findings to their healthcare providers. Of the 76 who reported their findings, 42 indicated that an action plan was formulated for them with 40 of those indicating that they were following their plan. Nine indicated they were not sure whether a plan was formulated for them; however, seven of those indicated they were following their plan. Altogether, 62% indicated that they were following their plan. 26 (34%) indicated that they had prior knowledge of their T1pos status (but only 11 of these had prior EHR documentation), and at least 45 (59%) indicated that they reported their findings to a Renown-affiliated provider. Of these 45, 18 (40%) had a diagnosis in the PL. Of the 59 participants that indicated no prior knowledge of their T1pos status, three (5%) had EHR documentation of their finding that preceded notification and 50 (85%) indicated that they reported their findings; none of the nine participants who did not report their findings (15%) had documentation in their EHRs (p = 0.02, Fisher exact, 2-tail). 22 (44%) of the 50 without prior knowledge and who reported their findings had a relevant nDx in their EHR, but only 13 (26%) had a relevant nDx in their PL. 81 (91%) of 89 respondents who recalled being notified indicated that they had shared or planned to share their T1pos results with their family members. Differences in responses to the survey between the three T1pos conditions were not statistically significant.

The review of the clinical notes indicated that one participant notified their provider of their finding but specifically requested for it not to be documented. Their finding subsequently does not appear in their EHR.

According to data from HNP’s third-party vendor for 94 of the participants, only 18 participants (19%) sought confirmatory testing while 48 (51.1%) declined confirmatory testing during the initial counseling session. One confirmatory test resulted in no finding. However, among all notified participants with EHRs, 49 (19.6%) had a referral for a LGC and 32 (15.1% of those without prior knowledge) had a referral after notification.

Since our study period coincided with the COVID-19 pandemic, we also reviewed the frequency of encounters and procedures for the 250 notified participants with EHR to ensure that our findings were not affected by a persistent decline in healthcare services. Supplementary Figure S1 shows that other than a temporary decline in procedures and encounters at the beginning of the pandemic, healthcare utilization levels for participants in this study rebounded after several months to pre-pandemic levels.

Although most participants shared their pathogenic genetic screening T1pos results with their healthcare providers, we observed a much lower rate of documentation of those results in their EHR (Figure 1). Survey results indicate that participants’ sharing their previously unknown genetic finding with their provider increases the likelihood of its documentation in the medical record. However, even when a participant says they have shared their results, less than 11% of such participants had a sufficiently specific diagnosis in their PLs. Since the PL is the primary method for indicating and sharing a patient’s active health problems between providers, these low documentation rates in the PL are especially worrisome. The discordance between sharing the results with providers and recording the finding in the EHR was not due to participants’ reluctance to have the finding documented and thus argues that the “hands-off” approach is not necessarily problematic but would benefit from overcoming some of the gaps in knowledge providers have with genetic testing and clinical decision support of genetic testing positive findings. The low EHR documentation rate does not appear to be due to participants’ reluctance to have the finding documented. It occurs despite significant promotional efforts within Renown in support of the HNP.

Even when findings were recorded, quite often diagnoses were not as specific as they could have been, considering the available nDxs in the EHR system. A diagnosis of “Genetic susceptibility to other malignant neoplasm” (Z15.09, 54 instances, Table 2) is too vague to inform clinical action. Similarly, recording “Genetic susceptibility to malignant neoplasm of ovary” (Z15.01) as a single code to document a finding of BRCA1 or BRCA2 does not convey the scope of the risk (as BRCA1 and BRCA2 also increase the risk of cancer of the breasts and other organs). Such non-specific coding may prevent appropriate risk-reduction interventions from being implemented. However, codes documenting specific variants were occasionally used (Table 2), indicating that more specific nDxs are available to providers.

We also observe cases where specific codes were used for documenting FH without the support of required clinical criteria. “Familial hypercholesterolemia” diagnosis (ICD-10-CM E78.01) is mostly used for patients without documented genetic findings of FH or evidence that a clinical criteria such as the DLCN was applied, thus reducing its significance, and necessitating the recording of a genetic variant for a provider to be certain that a patient was FH-T1pos. However, we could only find two such records for T1pos participants with FH.

The frequent use of non-specific diagnoses may simply reflect the widespread use of ICD-10-CM codes for clinical documentation and their relative inappropriateness for documenting genetic findings (Topaz et al., 2013; DeAlmeida et al., 2014; Fung et al., 2014). In contrast to ICD-10-CM, SNOMED CT ⁶ has specific codes for BRCA1 or BRCA2 variants (SNOMED CD IDs 412734009/412738007 respectively). The use of non-specific diagnoses may also reflect documented issues in current EHRs with effective integration of genetic data with patient medical records (Kho et al., 2013) as well as issues with template designs, such as having to select codes from exhaustive lists.

However, another possibility may be that healthcare personnel are uncomfortable dealing with genetic testing and the resulting findings. Numerous studies have shown that healthcare personnel, especially in the primary care setting, do not feel adequately equipped to order genetic tests or interpret, communicate, and follow up on such results (Overby et al., 2014; George et al., 2015; Hamilton et al., 2017; Hann et al., 2017; Briggs et al., 2018; Hauser et al., 2018; Laforest et al., 2019; Menko et al., 2019; Demeshko et al., 2020). Reservations regarding insurance discrimination and the social impact the findings might have for the patient play a role as well, although we note that only one person in the results herein asked to have no mention of the finding in the medical record. Additionally, physicians may not pay attention to unsolicited genetic results within EHRs (eMERGE (Gottesman et al., 2013; Williams et al., 2019; Nestor et al., 2021)) and it may be unclear to healthcare personnel who is responsible for positive genetic testing results (Pet et al., 2019). Ours was not a usability study and we cannot attribute the relative weight of the factors that may contribute to the observed poor documentation. Nevertheless, it is likely that if integrated clinical decision support tools were available for the PCPs seeing patients with CDCT1 findings, better documentation rates would follow. Such tools might suggest the appropriate diagnostic codes for the condition, the risk and the genetic variant detected, as well as recommended follow up steps and intervals.

Although we could not demonstrate improved practice patterns following T1Pos-finding notification for most participants (Figures 2A–C), many of the participants failed to benefit due to their old age, prior knowledge of their condition, prior presentation of outcomes, and prior interventions related to their findings. It is also possible that, because of the voluntary nature of the HNP, participants tend to be more health conscious than the general population and that this paradoxically contributed to our inability to detect improved practice patterns. Nevertheless, our results suggests that the timing of the genetic testing was a key factor. Had genetic screening been conducted earlier in life, many more participants would have benefited from T1Pos notification. Other studies (including our previous HNP publication) have reported similar findings (Grzymski et al., 2020; Guzauskas et al., 2020; Patel et al., 2020).

Since genetic testing after the presentation of a disease is clearly suboptimal, mandated testing in younger adult populations should be considered as a possible solution. In Nevada, a recently signed bill (SB251 (Nevada Legislature, 2021)), based on 42 U.S.C. 300gg-13 (GOVINFO, 2010), requires PCPs to obtain genetic counseling in compliance with the USPSTF recommendation (US Preventive Services Task Force, Owens et al., 2019) for risk assessment and possible genetic counseling and testing for all women with “a personal or family history of breast, ovarian, tubal, or peritoneal cancer or who have an ancestry associated with breast cancer susceptibility”. Even though the USPSTF recommendations were published in 2019, our review of the EHR indicates that widespread genetic screening under those circumstances is not yet common practice, especially if there was no evidence of relevant family history. Others have reported similarly low rates (Cham et al., 2022). Mandates such as Nevada’s may help identify many individuals at a younger age, prevent additional malignancies, and expand the scope of prevention by cascade testing. However, without sustained educational efforts within the general and medical communities, these types of efforts are more likely to increase screening after the presentation of symptoms rather than improve the ascertainment of family history in the medical record that will yield much earlier detection and risk reduction.

We are not aware of directly comparable studies attempting to measure the clinical outcomes of a “hands-off” return of results approach. The most similar study is probably Buchanan et al. (2020), which reported on the clinical outcomes of Geisinger’s genomic medicine experience (Williams et al., 2018), and their clinical data extraction and evaluation methods were similar. They provide similar information concerning diagnostic documentation and risk management, but in a different clinical setting and initiative design. In their report, post-disclosure diagnoses were evident in the EHRs of 13.4% of participants without prior knowledge, a rate comparable to the rate we observed in the PL. However, Buchanan et al. reported a much higher rate of post-disclosure risk management activities (70.2%) than we observed in our study. This may be because the definition of risk management activities used by Buchanan et al. for the T1pos conditions was significantly more encompassing (especially for FH) than our definition of behavior change. Nevertheless, the most likely cause of the difference in outcomes between our study and Buchanan et al. is the more integrative and proactive design of the Geisinger initiative.

A limitation of our study was the structure of healthcare in northern Nevada, where some subspecialties are predominantly private practice groups that have not provided us with access to their medical records. However, our review of the clinical notes indicated that procedures outside the reach of Renown’s Epic EHR are often documented in clinical notes during subsequent visits at Renown. Thus, even if a participant’s PCP was not an affiliated Renown physician and user of Renown’s Epic EHR system, it is reasonable that a significant genetic finding would eventually appear in the EHR record, given the typical rate of encounters at Renown and follow up time. We believe that the partial availability of clinical data due to the gradual implementation of Renown’s EHR from 2006 to 2011 had a minimal effect with regards to the recorded date of the finding but no effect on its actual documentation. When possible, additional specific dates were incorporated upon review of clinical notes.

Our review of EHR data was conducted at least 10 months after T1pos notification by the HNP. This was deemed sufficient time to allow T1pos participants to share their results with their physicians and for the findings to appear in the medical record. The existence of private practice groups was also the reason that for procedures such as colonoscopies, we considered orders as well as completed procedures. Although Renown’s coverage of primary care is roughly 50% in northern Nevada, at least 59% of survey respondents who shared their results shared with a Renown provider, suggesting a higher capture rate in our population. Although this was a single center study, the training and practice of medicine are comparable to other integrated networks and medical centers and our results should be considered in that broader context.

While our survey was not designed to evaluate how likely participants were to share their results with different types of family members, more than 90% of respondents indicated that they shared their finding with family. This level of uptake is comparable to the highest levels reported by others (Menko et al., 2019).

From the limited data set obtained from the third-party vendor that provides the genetic counseling, it is worth noting that more than half declined confirmatory testing and only 19% completed confirmatory testing. Thus, it seems that there is little value in recommending confirmatory testing. Financial or insurance considerations did not appear to be a significant contributing factor to the low rate of confirmatory testing. It may have been that HNP assurances regarding the robustness of the genetic testing results negated the importance of seeking confirmation for some participants.

The COVID-19 pandemic overlapped with our study period. We examined the possibility that this might have reduced participant utilization of healthcare, and thus affected our ability to detect responses to notification in the EHR. However, after a 2–4-month period of decreased utilization at the beginning of the pandemic, utilization rebounded to pre-pandemic levels (Supplementary Figure S1). Given that the minimum observation time was at least 10 months, we believe the pandemic had a minimal effect on our ability to detect responses to T1pos notification.

Only 60% of T1pos consenting participants with EHR were successfully notified and counselled, but the HNP has observed that the notification success rate was significantly higher when participants were contacted by Renown physicians than when they were contacted by the third-party vendor. This is likely due to Renown’s name recognition by participants. However, the third-party vendor success rate appears to be comparable to the rest of the industry. This highlights the need to find much more effective ways of reaching out to T1pos participants. Lack of notification was also associated with being non-white, who are underrepresented in the HNP (Table 1). This is likely a reflection on the socioeconomic disparities of certain non-white ethnic groups in northern Nevada ⁷ , negatively affecting their communication means and access to healthcare. Modifications to the HNP protocol including integration of the study into the EHR and improvements to the clinical decision support available to Renown Health providers will help address these disparity gaps moving forward.

As a result of these findings and in conjunction with the new state law, SB251, Renown and the HNP have made significant changes including obtaining informed consent to report positive findings directly into the medical record of the consented patient. We have expanded physician and other provider education, created order sets within the EHR specific to the CDC Tier 1 conditions, as well as study- and CDC Tier 1-specific tip sheets for providers.

Altogether, our findings indicate significant missed opportunity to maximize the benefit of the HNP voluntary population-based genetic screening and suggests that a change of design is required when it comes to the integration of the results into the participants’ medical record. Relying on participants to share their T1pos status with their healthcare providers appears to be inefficient, suggesting that a much more proactive approach should be taken. To improve results, we propose that participants’ consent be obtained at the time of recruitment for the study to automatically integrate T1pos findings with their EHR and to directly contact the participants’ healthcare providers. Persistent training of medical staff regarding CDC Tier 1 conditions is also needed to maintain a high level of awareness of the significance of such results and ensure appropriate documentation. Medical staff should use the most specific available codes and should document the findings in the PL. Failing to document findings in the PL could result in a loss of knowledge regarding the patients’ at-risk status for years to come. However, as Nestor et al. (2021) showed, even documented findings can often be ignored. This highlights the need for continued outreach to T1pos participants and especially their healthcare providers on follow-up steps and documentation that needs to be taken to effectively manage disease risk and to ensure optimal outcomes of PbGS.