Social disparities in patient safety in primary care: a systematic review

A literature search was conducted using two databases: PubMed and Web of Science. As the topic of patient safety is susceptible to changes over time, the search was limited to publications published between January 1st 2006 and January 31st 2017.

The lack of funds for translation of publications made it necessary to restrict this systematic review to publications published in languages mastered by the researchers namely English, French, Dutch and Italian. The search terms were based on the patient-safety domains according to the WHO-framework [5] and consequently discussed with patient safety experts in order to increase the quality of the search strategy. The search strategy is presented in Table 1. During the screening stage, no distinction was made in terms of research design and of population’s age, including studies on adults, adolescents and children. Only studies on primary care, ethnicity, gender, income, education and that were carried out in high-income countries (World Bank Classification) were included.

Results of the search strategy were uploaded in COVIDENCE, a software developed by Cochrane Library to facilitate and improve the collaboration among reviewers.

Figure 1 provides an overview of the study selection. A total of 2050 studies were retrieved. At the first screening round, all titles and abstracts were screened independently by two reviewers (JD and CP) and 2024 articles were excluded. In the second round, the full-texts of the 26 remaining studies were reviewed independently by the two researchers (JD and CP). During this round, 11 articles were excluded. Disagreement was automatically recorded in COVIDENCE at each screening stage and discussed until consensus was reached. Eventually, a total of 15 studies were included for analysis.

For each included study, a data-extraction form and a quality assessment were completed by two independent researchers (JD and CP). The standardised Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies was used [7]. For the included systematic review, the Critical Appraisal Skills Programme (CASP) was used. Using these instruments, the included publications were rated good, fair or poor.

To ensure that the two reviewers were collecting the same information from each study, the use of the Quality Assessment Tool was first piloted and tested. Hereto the researchers independently used the tool in the assessment of five papers and then discussed the discrepancies in their analysis. They analyzed whether these discrepancies occurred due to a different interpretation of the items in the Quality Assessment Tool and came to consensus on the points where their interpretation was different. Also a method for dispute resolution was discussed a priori.

Given the heterogeneity of study population, study design, interventions and outcomes, it is not possible to conduct a meta-analysis. Information extracted from the included studies can be consulted in Table 2 . This table provides information regarding first author, publication date, country, study design, study population, outcome measure and relevant study outcomes.

The results of the quality assessment can be consulted in Table 2. Seven studies [8‐14] were rated “good” while the other eight [15‐22] were rated as “fair”.

Five of the fifteen studies are carried out in Europe [11, 12, 14‐16] while ten are carried out in the United States [8‐10, 13, 17‐22]. Ten [8, 10, 12‐16, 19, 20, 22] of the fifteen studies explicitly looked at social disparities in their research questions whilst the other five studies look at general factors associated with the occurrence of patient safety events.

The following results are clustered into the domains of the WHO-framework on patient safety: administrative errors, diagnostic errors, medication errors, and transition of care errors.

Two of the fifteen studies [17, 21] report on administrative errors in primary care. The first study [17] finds that ethnic minorities have higher odds of experiencing harm and adverse consequences due to errors in the testing process (ordering, implementing, and performing the test, reporting results to the clinician, notifying the patient of the results and following up) compared to white patients. The second study [21] evaluates chronically ill patients’ perception about the coordination of care, and it describes no significant disparities regarding patients’ ethnicity and income. Gender and education differences are not described in neither of the two studies.

Seven of the fifteen studies [9, 10, 12, 13, 16, 20, 22] describe social disparities in diagnostic procedures. Four of them [9, 12, 16, 22] describe gender disparities in diagnosis. Henning et al. [16] and Maserejian et al. [22] describe that women have a lower likelihood of receiving proper and timely diagnosis respectively of cancer and coronary heart disease, compared to men. Henning et al. [16] describes differences in the interpretation of clinical symptoms and referral patterns in patients with Urothelial carcinoma of the bladder (UCB) visiting the General Practitioner (GP) for primary consultation and demonstrates that, despite the fact that women have worse prognosis and there are no gender-related differences in clinical symptoms of UCB, they are more likely to be treated for alleged urinary tract infections without further referral to an urologist compared to men. Maserejian et al. [22] describes disparities in physicians’ diagnosis of coronary heart disease (CHD) using a factorial experiment presenting videotaped CHD symptoms, systematically altering patient gender, age, socioeconomic status (SES) and race, reporting that physicians are less confident about CHD diagnosis in middle-aged female patients, indicating that their gender and age combination misleads physicians. Contrarily, Hansen et al. [12] reports that men experience longer doctor delays, that is the timeframe from the first contact with the GP presenting with symptoms up to time of investigation. One study [9] reports no association between gender and diagnostic errors. Five [9, 10, 13, 20, 22] of the seven studies describe ethnic disparities in diagnosis. One study [20] describes that although black patients experience lower levels of depression than white patients, they are more likely to be underdiagnosed with depression during the first primary care visit compared to whites. Two studies [9, 22] find no association between ethnicity and diagnostic errors whilst Fleming-Dutra et al. [10] find the opposite. The latter study reports that black children are less likely to be diagnosed with otitis media, compared to their white counterparts despite presenting the same symptoms. One study [13] reports that diagnosis of diabetes mellitus type 2 by physicians is associated with race resulting in underdiagnoses for white patients despite the same symptoms as black and Hispanic patients, whilst the prevalence of undiagnosed signs and symptoms of diabetes in the community is patterned more strongly by income and education (SES) and not by ethnicity. Similarly, Eva et al. [9] reports that patients’ SES is not associated with the physicians’ change of opinion regarding their diagnosis.

Five of the fifteen studies [8, 10, 11, 14, 15] report social disparities in medication procedures. Three of the five studies describe gender disparities in medication procedures. A systematic review [14] describing gender differences in antibiotic prescription demonstrates that n the amount of antibiotics prescribed to women is 36% higher than that prescribed to men in the 16–34 years age group and 40% higher in the 35–54 years age group. In particular, the amount of cephalosporins and macrolides prescribed to women is 44 and 32% higher, respectively, than those prescribed to men. Additionally, Cooper et al. [15] reports that women have greater odds of receiving potentially inappropriate prescriptions while Green et al. [11] describes that women have lower likelihood of receiving opioid prescriptions for pain treatment. Two of the five studies [8, 10] describe ethnic disparities for black patients in medication procedures. Becker et al. [8] finds that blacks are less likely to receive opioids for pain treatment compared to whites. Lastly, Fleming-Dutra et al. [10] reports that among children with otitis media, white children are more likely to receive broad-spectrum antibiotics than their black counterparts. One [15] of the five studies describes income disparities, reporting a higher likelihood to receive potentially inappropriate prescriptions for low-income patients compared to their wealthier counterparts. Disparities based on the patients’ educational attainment are not described in neither of the five studies.

Two of the fifteen studies [12, 19] describe social disparities in the transition of care procedures. Both studies report an unequal referral pattern with regard to the patients’ gender, describing that women are more likely to be referred or undergo cardiac stress test compared to men [19] while Hansens et al. [12] reports that men are more likely to experience longer doctor-system delays, namely referral to the hospital, first in-hospital visit, referral to treatment and its initiation, compared to women. Only one [19] of the two studies describes ethnic disparities, reporting no association between ethnicity and transition of care procedures. One [12] of the two studies describes income disparities in regard to transition of care procedures, reporting that high-income women experience shorter system delays, but they longer doctor delays, compared to their less wealthy counterpart.

The use of the framework based on the most recent WHO guidelines and the comprehensive approach of this study are certainly major strengths. This review of the literature has also some limitations. We used Pubmed/MEDLINE and Web of Science to search for relevant papers considering the topic of this review i.e. patient safety in primary care. However, an additional search in social science databases might have resulted in additional papers looking at the issue with a different theory base e.g. discrimination.

A considerable number of studies on patient safety refers to “adverse drug events” in randomised clinical trials in medication research. However, those have been excluded from the search and inclusion because these incidents may be attributable genetic, physiological factors of the individuals [38] rather than a dosage or prescription error. Furthermore, data on ethnic minorities other than Black/African- Americans such as Hispanics or Asians were not documented as often as in the hospital setting.

Data regarding social disparities in patient safety in primary care are somehow fragmented which do not allow to grasp a thorough overview of the problem. Future research should focus on matching data on patients’ gender, ethnicity and socioeconomic status, in countries where this type of data are collected, with the data of critical incident registers in order to observe whether these events differ according to patients’ characteristics and in which patient safety domain.