Evaluation of a social determinants of health screening questionnaire and workflow pilot within an adult ambulatory clinic

In November 2018, the Epic EHR module for SDOH was incorporated into Sutter Health’s system-wide Epic EHR platform. The module included 24 questions to identify SDOH needs across 10 domains (financial resource needs, transportation needs, stress, depression, intimate partner violence, social connections, physical activity, alcohol consumption, educational attainment, and marital status), based on the recommendations from NAM and the Centers for Medicare and Medicaid Services (CMS) [30]. The EHR module was adapted to a paper questionnaire (using the verbatim wording of Epic questions) and data collection procedure (hereafter referred to as “workflow”) through 4 week-long iterative plan-do-study-act (PDSA) cycles [31] in Fall 2019 at the pilot ambulatory clinic. The preliminary 24-question SDOH questionnaire and standard workflow were informally introduced and intermittently used within the pilot clinic and between January and mid-February 2020 to gain additional feedback for further refinement prior to the pilot. A focus group discussion with physicians and clinic staff (N = 9) in February 2020 finalized the questionnaire and standard workflow for the pilot.

The SDOH questionnaire and workflow were piloted in a single ambulatory clinic from February 18th-March 25th, 2020 (pilot cut short due to COVID-19 pandemic). The questionnaire included 11 questions across eight SDOH domains: financial resource needs, transportation needs, stress, depression, intimate partner violence, social connections, physical activity, and alcohol consumption (Additional file 1). Physicians provided guidance on which responses for each question they felt identified a “need” and described the actions which identifying such a need would provoke from the following options: “No immediate action – document and monitor,” “Immediate physician intervention,” “Long-term physician management,” or “Referral to clinic case manager and/or social workers.” The final workflow describes the integrated data collection protocol from a patient checking into the clinic through the potential action steps following the identification of specific social needs (Fig. 1).

Table 2 compares the sociodemographic characteristics of 2019 and 2020 eligible patients. There were no statistically significant differences in sociodemographic characteristics (e.g., age, sex, race/ethnicity and payment type) between patients in the pilot and the comparison group in 2019. Fifty-seven percent (N = 164) of the 2020 eligible patients in the pilot were between 25 and 54 years of age, and 61% (N = 175) were female. Thirty-two percent of the 2020 patients (N = 92) self-identified as Hispanic, 31% (N = 89) as Non-Hispanic white, and 20% (N = 57) as Non-Hispanic Black. Nearly three-fourth of the 2020 patients (74%, N = 215) were privately insured, while 17% (N = 50) were Medicare patients.

Table 3 describes patient responsiveness to the questionnaire by specific SDOH domain. The majority of eligible patients (83%) responded to at least one SDOH question. Responsiveness for seven of the eight domains, by domain, ranged from 55% (social connection) to 67% (alcohol). The questions regarding depression had the lowest response rate, with only 7 % of eligible participants completing at least one of the questions in this domain.

We observed no statistically significant differences in responsiveness (response to at least one SDOH question) across sociodemographic groups and visit types (Additional file 2). The response rates for specific domains were uniform across domains and sociodemographic groups, with a few exceptions. Women were significantly more likely to respond to questions for the financial resource domain (54% vs. 46%; p < .05), the physical activity domain (66% vs. 34%, p < .05), and the intimate partner violence domain (66% vs. 34%, p < .05), compared to men. Though not statistically significant, women tended to provide more responses across the remaining domains. In addition, responsiveness was significantly different across age groups within the alcohol domain: 25- to 54-year-olds made up a significantly larger proportion of respondents (61%), and 65+ year-olds made up a significantly larger proportion of non-respondents (27%). Though significant differences were observed across age groups, insurance types, and visit types for responses to the depression domain, sample sizes were very small (only 20 people responded to the either of the domain questions).

Table 3 also describes whether a SDOH need was identified among all eligible patients who responded to at least one survey question (N = 240). The majority of participating patients (51%, N = 123) had at least one identified social need, and needs were identified within each domain. Among those who had any response to a given domain’s questions, the most commonly identified SDOH need was stress (33%, N = 78), followed by physical activity (22%, N = 52), alcohol (12%, N = 29), and social connections (6%, N = 15). Identification of at least one social need did not differ across sociodemographic groups and visit types (Additional file 2). Physical activity needs were identified for a significantly larger proportion of female patients compared to males (81% vs. 19%, p < .01) as well as at new patient/transfer visits compared to other visit types (48% at new patient/transfer vs. 13% at Medicare wellness and 38% at adult wellness visits, p < .05). Further, new/transfer patients made up a larger proportion of respondents without identified physical activity needs (68%) than other visit types. For three other domains, significant differences across sociodemographic groups were observed, but absolute numbers of patients with identified needs were small. A significant difference in reported intimate partner violence needs was observed by visit type: all reported intimate partner violence needs were among new/transfer patients (N = 13, 5.4%). All reported transportation needs (N = 6, 2.5%) were among Non-Hispanic Black individuals (N = 3) and Non-Hispanic Asian individuals (N = 3). All reported depression needs (N = 5, 2.1%) were among patients 55 and older, those with public insurance, and either Medicare Wellness or Adult Wellness visit types.

Table 4 presents the responses to the staff experience survey questions, across all responding staff and by staff category. With respect to the effectiveness of the pilot, 90% (N = 18) of respondents believed that social needs information could help improve patient care and health outcomes, and 90% (N = 18) felt that social needs information could improve therapeutic relationships with patients. Eighty percent (N = 16) of staff respondents reported being concerned that patients would feel uncomfortable answering the SDOH survey questions. Only one of the three responding physicians strongly agreed or agreed that having the social needs information would influence their medical decision making.

As shown in Table 4, 95% (N = 19) of staff strongly agreed or agreed with the statement that collecting social needs information was within the scope of clinical care. Eighty percent (N = 16) of staff believed that the clinic patients would in fact have unmet social needs, including only 57% (N = 4) of MAs. These two questions capture a perception of the relevance of the SDOH survey and standard work.

Regarding the utility of the SDOH survey itself in the context of the standard work, 80% (N = 16) of staff felt the survey asked all relevant questions for assessing unmet social needs, 90% (N = 18) strongly agreed or agreed that the survey improved the clinic’s ability to identify patients with unmet social needs, and 85% (N = 17) believed that it would increase the likelihood of patients actually being connected to case managers or social services to address identified social needs.

As described in Table 4, 85% (N = 17) of staff understood their role in relation to the SDOH survey and standard work. Only 50% (N = 10) were aware of available Sutter Health resources to support, including one physician, three MAs, and two PSRs, the primary actors in the SDOH standard work. Similarly, only 50% (N = 10) of staff were confident in their ability to help patients address social needs, with no physicians reporting confidence.

All proposed potential barriers to implementing the standard work were identified as major barriers or barriers by multiple staff participants. Patients lacking time to complete the survey was identified as a barrier by the largest percent of staff (75%, N = 15), followed by staff lack of time to respond to social needs (65%, N = 13), staff lack of training about or resources for responding to social needs (both 45%, N = 9), and staff lack of training about administering the survey (20%, N = 4).

Table 5 compares the 2019 and 2020 eligible patient visit length. During the 2020 pilot, average length of visit across all eligible visits was 39.8 min, which was 1.7 min longer than during the same time period in 2019; however, the difference was not statistically significant. In looking at the specific segments of the visit, there was a small but statistically significant difference in the length of rooming (the time spent with the MA), with average 2020 rooming lasting approximately 0.7 min longer than average length of rooming in 2019 (p < .05). In comparison with the same time period in 2019, average length of visit in 2020 pilot was longer across all visit types from 1.2 to 3.7 min in difference, but the difference was not statistically significant. We did observe significantly longer rooming time among new/transfer patients by 1.36 min (p < .01) and significantly longer exam time among Medicare wellness patients by nearly 5 min (p < .05) in 2020 compared to 2019.

Among the 2020 eligible patients, significant differences in visit length were observed across several sociodemographic groups and by visit type (Additional file 3). The average visit length was significantly longer among patients who are 65+ (43.36 min) compared to other age groups (p < .05) and on public insurance (43.64 min) compared to other insurance types (p < .05). Time for checking-in and rooming were also significantly different by insurance type: patients who were self-pay or having other insurance types had the longest checking-in time (10.4 min, p < .05), and patients with public insurance had the longest rooming time (12.2 min, p < .01)). Rooming is also significantly different across visit types, with the longest time among Medicare wellness patients (12.70 min, p < .01). Physician exam times were only significantly different by age group, with the 55-64 group had the longest exam time of 19.0 min (p < .01).

We found no observable differences in sociodemographic characteristics between the participating patients in this pilot and the reference group of patients in 2019, indicating that the eligible patients in this pilot are likely similar to the patient population historically seen in the clinic. While 83% of the patients answered questions within at least one SDOH domain, only 3.5% of the patients completed all domains. Reasons for non-responsiveness may include discomfort with answering sensitive questions or the skipping of questions that patients did not feel were relevant to their lives. Much of current literature shows that patients perceive social risk screening as appropriate; however, research also emphasizes different factors which influence acceptability, such as trust in clinicians, clinical settings, and patients’ concern on privacy of social health data within their EHR [20, 38‐41]. For example, a study by Cunningham & Sobell showed that adults feel that it is appropriate to be asked questions on alcohol use but may feel uncomfortable or underreport [42]. Stigma or shame was cited as barriers to disclosure of social needs like food insecurity [43]. The overall non-responsiveness to the depression domain questions (with only 7% of eligible patients responding) may also be indicative of stigma against mental illness, though evidence from other clinical sites suggests that depression screening is increasingly perceived as positive [44]. Further investigation is needed to understand why depression screening in the context of the overall SDOH questionnaire was unsuccessful. Additionally, our finding that females were more likely than males to respond to survey questions may be indicative of females’ greater recognition of the importance of screening. A study of patient perspectives on SDOH screening by Rogers and colleagues found that females were significantly more likely than males to agree that social needs screenings were necessary and should be a part of healthcare settings [45]. Overall, these findings emphasize the importance of patient-centered implementation of social risk screening. Due to the scope of this evaluation, we were not able to understand patients’ perspectives regarding answering the SDOH questionnaire. Future iterations of this intervention should incorporate a qualitative component to understand patients’ experiences of and feelings about answering the SDOH questions in the context of this standard workflow [45, 46].

Half of the participating patients had at least one identified social need, with the most commonly identified social needs being stress, physical activities, alcohol, and social connections. These findings are similar to those reported in other studies that used different SDOH screening tools and workflows. A study by Page-Reeves and colleagues of patients in family medicine clinics also found identified 46% of patients having social needs [19]. A recent study by Tong and colleagues on a target population with a higher risk of having social needs reported a higher proportion (71–86%) of patients who screened positive for social needs, with the most common social needs included physical activities, dental, and alcohol use [23]. Despite the SDOH questions being available at Sutter Health in Epic since 2019, these questions were not used by physicians before the pilot. The fact that SDOH needs were identified in this pilot indicates the importance of having a standard workflow to gather SDOH information in a systematic way [15, 18]. Although we found certain differences in identified social needs across sociodemographic characteristics, small sample sizes limit the interpretability of these findings. Given the dearth of literature on whether identified social needs from screenings in clinical settings differ across patient populations, future research with a larger sample size and an incorporated qualitative component would provide the opportunity of examining potential disparities in the identification of social needs.

Overall, there was a strong support from staff in using the SDOH questionnaire and workflow to collect SDOH information as they were perceived to be important and effective in identifying social needs among patients. Although evidence was mixed, recent studies also show that social risk screening is perceived to be important by clinicians and healthcare team [23, 47, 48]. In this pilot, staff identified one possible challenge to effectively being able to collect SDOH information: that patients could feel uncomfortable answering the questions. Previous studies also reported concerns from clinicians related to ensuring that screening was done empathetically, without negative judgement, and with attention to privacy protections [19]. A qualitative study by Byhoff and colleagues found that patients actually felt more “cared for” or “listened to” when asked about social needs within a clinical setting, while emphasizing the need for “empathy” and “compassion” from staff conducting the screenings [46]. Accordingly, future implementations of this intervention must prioritize understanding patients’ experiences with answering the SDOH questions.

In the pilot’s standard workflow, physicians are a key gatekeeper of effectiveness, as they directly impact the translation of identified needs to action (whether within the exam space or through referral to case managers or social workers). Despite feeling that social needs information could help improve therapeutic relationship with patients, only one of the three responding physicians reported that having the information would influence their medical decision-making. Other studies found that knowing patients’ social needs not only improved patient-provider communications but also changed what clinicians do [23], such as providing more exercise and dietary counseling, being mindful of medication costs when prescribing, and helping with transportation to access to clinics. As only three physicians completed the staff survey, it was challenging to interpret physicians’ perceptions on changes of medical decisions without in-depth conversation, suggesting the utility of qualitative research with physicians in future pilots.

The overwhelming majority of surveyed staff indicated that SDOH screening was relevant for their specific clinical population and within the scope of clinical care more generally. These findings align with the overarching recognition in health care delivery that understanding and engaging with patients’ SDOH needs should be incorporated into primary care settings [5, 15]. The majority of staff also felt the survey itself asked relevant questions and would be useful to support patients in connecting with resources to address social needs. In their study of 258 clinicians (physicians, social workers, nurses, and pharmacists) at Kaiser Permanente Southern California, Schickendanz et al., similarly found that the majority of those surveyed agreed that social need screening should be incorporated into clinical care and that knowing such information could be beneficial to patients [21]. These findings from our pilot site suggest the staff’s willingness to initiate SDOH screening. They also suggest that implementing plan-to-study-act (PDSA) cycles with a small group of staff before launching the pilot could be a potential strategy to engage staff in the pilot and their willingness to initiate SDOH screening. As the SDOH questionnaire and workflow is adapted and implemented at other Sutter Health sites, attention should be paid to any differences in adoption across locations so that factors which best facilitate adoption in the Sutter Health network can be identified.

Because our study assesses a pilot program, we cannot speak to fidelity to the intervention with respect to the implementation of the SDOH questionnaire and workflow in other sites. Rather, our results speak to potential barriers that could hinder implementation at future sites – confusion regarding components of the standard workflow and the impact of the workflow on time [21, 22, 26]. While most staff felt they understood their role in the standard work, staff were less knowledgeable about and lacked training on available system resources to support patients and were also less confident in the ability to act on identified needs. Other studies have identified clinicians’ concerns that SDOH screening may not ultimately be helpful due to lack of availability or knowledge of, or access to, resources to address patients’ social needs as a potential barrier to implementation [23, 49]. The uncertainty expressed by staff speaks to the importance of incorporating information on this dimension of the workflow into staff trainings, including for staff who will not be directly supporting patients with next steps.

Time – specifically the additional time needed by patients and staff to complete the SDOH workflow – was a key concern for staff in the pilot. These concerns have also been identified in other studies of clinician perspectives [21, 26]. Our study did find that the average length of visits among eligible patients in 2020 was slightly longer than those in 2019, but the difference was not statistically significant. The longest added time period was the statistically significant difference of 5 min in exam time for Medicare wellness patients. One possible explanation for the exam-time difference for Medicare wellness patients is that Medicare wellness patients may have greater social needs, and so the results of the SDOH questionnaire may ultimately require more time with the physician to identify next steps. However, we did not observe that Medicare wellness patients had significantly more needs identified as compared with patients of other visit types. Medicare wellness patients may also have more complex health needs, and the evaluation for these patients may take more time due to Medicare requirements. Another possible explanation is the potential for particularly long visits to have a greater effect when comparing average values for a small number of visits (6% of our 2020 pilot population). Medicare wellness visits did have the most variability in lengths of the whole visit, rooming, and exam. Similarly, we observed small but significant differences in average visit lengths and lengths of specific sections of the visit across different sociodemographic groups (by age, insurance type, and visit type). However, without in-depth conversations with patients and providers, we cannot understand the impact these average length increases had on the experience of care, nor can we determine what may have been gained or lost by incorporated the SDOH workflow into the limited available time. In a qualitative study of social risk screening among patients and caregivers, participants expressed concern for the addition of the screening to already overworked clinician’s schedules [46]. In future iterations of the intervention, close attention should be paid to the quantitative and qualitative impact of the SDOH workflow on time so that adjustments may be made, whether to visit lengths or to the SDOH workflow, to minimize the time-pressure felt by clinic staff and patients and maximize the impact of the intervention.

This pilot evaluation adds to the limited literature of SDOH assessments in clinical settings that examine whether there are disparities in screening and identification of social needs across different patient populations. Though observed in a small sample size, we identified some differences in reach, effectiveness, and implementation across patient sociodemographic groups. Understanding potential reasons for those differences could improve equitable implementation of SDOH assessment among patient groups. As addressing social needs is a strategy for reducing health inequities, SDOH assessment should be disseminated in clinical settings with a mindful approach that minimizes the potential disparities across patient sociodemographic groups.

As the first study in the Sutter Health network to assess the incorporation of an SDOH questionnaire and workflow into a primary care setting, this evaluation not only provides feedback for further development of the intervention within the pilot clinic but also lays the groundwork for system-wide scale-up. Rather than assessing a single dimension of the intervention, this study synthesizes data from different sources to evaluate multiple elements of the intervention simultaneously. Through the use of the RE-AIM framework, this study presents a systematic approach to assessing social risk screening interventions that can be replicated by other clinics. Our study’s focus on identifying observable differences across sociodemographic groups also sets an important precedent for future studies to center considerations of equity throughout evaluations of SDOH screening interventions.

There are also important limitations of this pilot evaluation. Due to insufficient resources, we were unable to assess the effectiveness of the standard workflow in addressing identified patient social needs, a gap in the evaluation which must be prioritized in future intervention studies. A growing body of literature is examining the impact of screening for social risks in clinical settings on patient access to resources, healthcare experiences, and health, with evidence suggesting that such screenings are beneficial for patients [19, 49‐51]. As we could not examine how patient’s social needs were addressed in this pilot, future pilots should prioritize assessment of referrals, receipt of social services, and overall impact on patients [28]. In addition, our study timeline was impacted by the onset of the COVID-19 pandemic, which in turn limited our sample size with respect to comparing metrics across sociodemographic groups. Relatedly, as the pilot only focused on a single clinic, small sample sizes precluded inferential statistics across staff groups. Also, as previously discussed, future studies should examine the reasons for non-responsiveness to SDOH questionnaires in order to suggest effective ways for collecting SDOH information. Finally, due to the COVID-19 pandemic, we were unable to complete the intended patient experience component of the evaluation. Gaining a deeper understanding of patient perspectives remains a priority for future implementation.