Survey strategies to increase participant response rates in primary care research studies

Understanding and measuring primary care provider and patient knowledge, attitudes, behaviours, practices and views on health care issues are critical to improving health and health care outcomes. Surveys are one of the most useful and frequently employed methods in primary care research and help improve the delivery of health services and to guide development of policies. While relatively easy and cost effective, validity of findings from survey research is highly dependent on having an adequately large sample that is also representative of the target population of interest. A response rate is most commonly defined in two ways: the number of subjects who respond to a survey divided by (i) the number of subjects in the selected sample and/or (ii) the number of eligible subjects in the sample (1,2). It is important that research studies accurately report the denominator for calculation of response rates so that readers can assess potential bias, particularly related to reasons for non-response. There is no established threshold for defining a high response rate but a rate of 80% or higher is considered excellent. High response rates are an essential attribute of a quality research study for generating valid (3), reliable (3) and generalizable (4) results for survey and prospective observational studies. With statistically powerful response rates, researchers can ensure that their study sample represent the target population of interest. However, patient and provider response rates to surveys have been declining (5,6). Low response rates can result in a number of methodological biases (i.e., non-response bias, delays in project timelines, budgetary problems and studies that are underpowered) that have important implications for how researchers identify appropriate targets for future interventions or programs or assess if an intervention achieved the desired effect (7,8). Primary care research involves the collection of survey data from a variety of sources (patients, clinicians and staff) to assess a variety of characteristics and attributes, such as patient-reported outcomes, attitudes and beliefs around care delivery, burnout, practice culture and practice organizational characteristics. The growing demand and interest in primary care research has the potential to increase survey burden resulting in low response rates. Response rate challenges in primary care research studies can occur for primary care clinicians and staff particularly in patient surveys. Low response rates on clinician and staff surveys may be due to demanding work schedules, disruptions in routine practice, lack of motivation to participate in research, confidentiality and increasing demand to participate in studies (9). Low response rates in patient surveys can occur related to the type of patients researchers wish to survey (10,11), sensitivity of questions asked, e.g. intimate partner violence, mental health issues, etc. (12), mode or survey administration (10,13,14) and mistrust in science, among other factors. Given the increasing demand to conduct primary care research, this methods brief highlights strategies to increase patient, staff and provider survey response rates with illustrative examples from primary care research studies. Comprehensive reviews examining survey response rates within primary care literature have reported response rates varying from 10.3% to 61% (15–17). A number of factors can impact the response rate including the survey length, content, mode of administration, incentives and number of follow-up surveys. These factors are inter-related and addressing them using multiple strategies may result in better response rates than just addressing one of them (18). For instance, survey studies with primary care practice clinicians and staff found that response rates increases from 50% in one study to closer to 80% (19,20) when all factors contributing to response rates were addressed; below we describe strategies to address these most common factors. Achieving optimal survey response rates when conducting research among ‘hard-to-reach’ patient populations susceptible to disparities in primary care delivery may be difficult (21). Hard-to-reach patients include transient patients with inconsistent access to primary care, those suffering from sensitive and vulnerable health conditions such as serious and persistent mental illness, those with social and economic vulnerability and other social determinants of poor health. Engaging hard-to-reach populations in primary care research presents a key challenge and is especially problematic for longitudinal studies requiring multiple survey time points. A few strategies have shown to help increase response rates (and participation) in research for hard-to-reach populations. Partnering with community primary care clinics and using community members to identify, invite and engage vulnerable patients in research can be effective (22). This strategy, however, requires meaningful relationship building with community partners, sometimes over many years. Working with practices from networks such as practice-based research networks (PBRNs) in the USA can also help increase response rates in surveys. PBRN staff cultivate long-term relationships with practices to answer community-based health care questions and translate research findings into practice. PBRNs engage clinicians in quality improvement activities and an evidence-based culture in primary care practice to improve the health of all Americans (22). Engaging PBRNs is especially effective in reaching clinicians for survey research. For example, Galliher et al. found that surveys assessing clinician-reported behaviours and beliefs related to hepatitis C, hyperlipidaemia and pharyngitis was higher among PBRN members than the general membership of the American Academy of Family Practitioners (23). Both the content and the length of a planned survey can enhance or diminish final response rates. To the extent possible, using previously validated survey instruments is highly recommended (23). When developing new measures, it is important to engage survey research experts and use rigorous methods (2). Creating new instruments requires intricate detail and sensitivity to measuring construct(s) without leaving any room for vagueness, creating bias or uninterpretable data (24). Further, new instruments need to be tested extensively to establish validity and reliability. For some populations, it is also important to consider cultural and language relevance of the selected instrument. If the survey is not culturally relevant or not translated in commonly used languages among the target population of interest, it could significantly reduce response rates. Thus, it is recommended that previously validated surveys be cognitively tested and evaluated in a small sample of target participants prior to full scale administration (25,26). In addition to selecting instruments that are relevant, valid and reliable, it is important to pay attention to time burden to participants for completing the survey. Surveys that take too long to complete can not only make it harder to engage participants in the study but can also result in missing responses on survey items. We recommend shorter survey length (i.e., 15–20 minutes) and prior cognitive testing of survey content and time to complete it is important to achieve high response rates (27,28). Surveys come in a plethora of formats; the most commonly employed methods are face-to-face, telephone and mail. More recently, there is a significant increase in use of electronic surveys (5,28,29) (e.g. email and online surveys) (29). Mode of survey administration is important both for recruiting study participants as well as for ensuring high response rates. Table 1 describes the advantages and challenges of different survey modalities pertaining to response rates. While traditional modes of administration such as in-person pen and paper, telephone and mail have the advantage of familiarity and ease of providing incentives for survey completion, newer methods such as web-based surveys are convenient and easy to complete on smart devices. Web-based surveys also allow for a more comfortable environment for participants to complete surveys on their own time. Combining methods and modes helps reduce barriers associated with each data collection format to optimize response rates. Comparison of survey modalitiesa aAs defined by SAGE Research Methods—‘Designing and Doing Surveys: Survey Formats’ (2012) (22). Comparison of survey modalitiesa aAs defined by SAGE Research Methods—‘Designing and Doing Surveys: Survey Formats’ (2012) (22). Use of mixed modes of survey administration is especially advantageous for research studies conducted in diverse primary care settings. Each primary care practice may have its own challenges reaching their patient populations and these challenges must be balanced with participant preference of survey mode and logistical challenges of survey administration (5,30). Research that a priori allows flexibility in selecting mode of survey administration suited to the unique context of the practice may result in higher response rates. For instance, in a large US initiative to improve cardiovascular preventive care involving over 1500 primary care practices, national evaluators recommended that each local research team use survey modes most suited to achieving high response rates. Local research teams selected mixed modes including web-based surveys, in-person and telephone. This initiative achieved survey response rates upwards of 70% (30,31). In the My Own Health Report Study (32), practices chose the method most conducive to assess patients’ health risks using a survey and achieved an average response rate of 49.1%, with higher response rates when surveys were completed in person in the office rather than by phone. This brief describes three key methodological strategies to enhance survey response rates in primary care research: (i) target population, (ii) survey content and (iii) mode of administration (Fig. 1). Over the past two decades, survey response rates have been steadily decreasing in primary care and health care research more broadly (10,11,33,34). It is important that the quality of the survey data be considered to assess the relative contribution to the literature. With these strategies, primary care researchers have experienced increases in response rates across patients and providers and by race/ethnicity and health condition (i.e., diabetes and chronic pain) (34). Prior studies show that deploying multiple strategies concurrently, such as using mixed modes, financial incentives and frequent follow-up and reminders can help increase response rates (33). However, the use of multiple strategies can increase methodological expense, although this may be offset by the reduced need for further sampling (35). Summary of key evidence-based strategies to improve survey response rates in a primary care setting. Summary of key evidence-based strategies to improve survey response rates in a primary care setting. It is critical that primary care research employ strategies to increase representation of hard-to-reach populations. This brief points to the importance of using community-based approaches, such as engaging with community organizations to improve participant retention and ensuring that survey content aligns with participant preferences, culture and circumstances to improve response rates in harder to reach and underserved populations (36). Community-based approaches have been adapted to primary care settings through PBRNs, where primary care practices meet to investigate questions related to community-based practices and to improve the quality of care (37). Despite their usefulness in increasing response rates, researchers and funding agencies should acknowledge that these community-based approaches require extensive time, staff resources and come with considerable costs. Response rates continue to be the primary method for documenting survey quality but even a high response rate can be problematic (38). Studies over the past decade have concluded that the response rate may not be as strongly associated with the quality or representativeness of the survey (39,40). There is increasing recognition that the degree to which sampled respondents differ from the survey population, or nonresponse bias, is central to evaluating the representativeness of a survey (39). Although nonresponse bias is more useful for understanding survey limitations, its assessment can be more difficult than estimating response rates and often requires advanced planning. Researchers often aggressively pursue high response rates to compensate for nonresponse bias, but even a survey with a high response rate may be more biased than a survey with a low response rate from a random and representative sample (41). Monitoring response rates from key subgroups (i.e., race/ethnicity, sociodemographics and number of chronic conditions) throughout the data collection process may help to reduce nonresponse bias while also providing information that researchers can use to adapt methods and increase reach. To this end, it is important that primary care researchers consider both response rates and nonresponse bias before and after survey administration (42). Funding: funding was issued by the affiliated departments. Ethical approval: no data were used for the completing of this brief. Conflict of interest: none. No data were used for the completing of this brief.