Patient Safety in Primary Care: Conceptual Meanings to the Health Care Team and Patients

Alden Yuanhong Lai, PhD; Christina T. Yuan, PhD; Jill A. Marsteller, PhD; Susan M. Hannum, PhD; Elyse C. Lasser, MS; JaAlah-Ai Heughan, MS; Tyler Oberlander, BA; Zackary D. Berger, MD; Ayse P. Gurses, PhD; Hadi Kharrazi, MD, PhD; Samantha I. Pitts, MD; Sarah H. Scholle, PhD; Sydney M. Dy, MD


J Am Board Fam Med. 2020;33(5):754-764. 

In This Article


Study Design and Sample

We conducted a qualitative study that involved interviews with 101 frontline clinicians, administrators, and staff (hereafter, "personnel") and 12 focus groups with a total of 65 patients. We collected data in 10 patient-centered medical homes (PCMHs) that had achieved level 3 status in the National Committee for Quality Assurance (NCQA) PCMH recognition program.[26] At the time of study, level 3 PCMHs were practices that NCQA determined as performing at the highest level (there were 3 levels in total) based on 6 aspects of primary care delivery, including patient access, team-based care, care management and support, care coordination, quality improvement, and population health management.[27] The 10 PCMHs included in our sample were located in 4 US states of Colorado, Maryland, North Carolina, and Pennsylvania and had a median of 3.5 years as a level 3 PCMH before the point of data collection. As an initial step to understand patient safety in primary care, this study focused on level 3 PCMHs because they could potentially show what patient safety in high-performance primary care settings entails.[28] We first adopted a purposive sampling frame to maximize the types of PCMH practices (ie, ownership and geographical types) for our study because such variation could theoretically generate a dataset that was as rich as possible for the purpose of qualitative analysis. We then used volunteer sampling to recruit practices within each frame. This was because our study involved interviews with a majority of each PCMH's personnel and some of their patients, which necessitated certain changes to their daily operations and therefore the need for them to "volunteer" as a study participant. Patients were eligible to participate if they were over the age of 18, had a chronic medical condition, and visited the PCMH more than once a year. Each PCMH communicated to their eligible patients about the study through flyers, letters, and e-mails, and those who were interested to participate were requested to contact our study team directly. The Institutional Review Board at the Johns Hopkins Bloomberg School of Public Health (no. 7497) approved this study. All participants provided written informed consent.

Data Collection

We worked in teams of 3 to 4 members to collect data at the 10 PCMH practices and posed the broad question "What does safe medical care mean to you?" to all participants. We allowed the participants to elaborate as much as possible, including the use of related terms (eg, safe care, patient safety, and keeping patients safe) to elicit further descriptions.[29] We interviewed personnel individually and conducted focus groups for patients. Personnel did not share their perspectives in the presence of patients. Patients, on the other hand, only shared their perspectives in the presence of other patients in the focus group. All interviews and focus groups were audio-recorded and professionally transcribed. We visited each site once over a 2-day period between May 2017 and May 2018. This study was part of a parent study that sought to investigate the implementation of the PCMH model and how it related to patient safety. The data reported here were the result of a separate analysis of the responses to the aforementioned open question, which was used at the beginning of all interviews and focus groups in the parent study. We wanted to capture, in the broadest sense, how the participants conceptualized patient safety, and thus limited the current analysis to the responses to this question. The responses typically took the first 5 to 10 minutes. The interviews lasted approximately 30 to 90 minutes in their entirety and 60 minutes for the focus groups.

Data Analysis

We used a team-based coding process and conducted thematic analysis in 6 phases.[30,31] We used a 10% subsample from each data source to develop a team-based codebook. Because we interviewed 101 personnel, we first randomly selected 10 personnel interview transcripts, which were independently and inductively coded by 3 analysts to derive initial codes. The analysts then compared their coded segments for each transcript, refined the code definitions, and created a centralized codebook. During this process, the focus was on discussing disagreements in codes, ways of coding, and alternative explanations for codes as opposed to calculating intercoder agreement.[32,33] Afterward, another 3 analysts used this newly created centralized codebook to guide the coding of a 10% subsample of the focus group data, which translated to 1 patient focus group transcript. All 6 analysts then reconvened to compare their codes and coding, revised the centralized codebook, and used it to formally analyze all data. During this process, there was a third update to the centralized codebook as both teams conducted analysis iteratively. When all coding was completed, we began to search for themes by collating codes and reviewed major and minor themes for reporting.[30] See Figure 1 for a schematic of the analytic steps.

Figure 1.

Schematic representation of the team-based process used for thematic analysis.

To compare and contrast the perspectives of personnel, and those of patients, 1 analyst conducted a convergence assessment to ascertain areas of agreement, partial agreement, dissonance, and nonnarration.[34,35] Agreement refers to a direct overlap between the themes and descriptions given from both perspectives; partial agreement refers to a general overlap, but with deviating nuances to be accounted for; dissonance refers to a direct disagreement; and nonnarration refers to a lack of expression of a set of themes or descriptions by 1 stakeholder group compared with the other. We used MAXQDA 12 software for data analysis.[36] All researchers involved in data analysis were also involved in data collection. The research team comprised health care scholars and primary care physicians with a diversity of backgrounds, including patient safety, patient-centered outcomes research, health services research, implementation science, organizational behavior, gerontology, health information technology, and human factors engineering. To maintain reflexivity, we created and used team memos throughout the data analysis process[37] and held multiple rounds of team discussions that led to the iterative refinement of the codebook and coding process.