Background

Health inequalities, precisely inverse associations between socioeconomic status (SES) and morbidity and mortality, are well analysed and described [1]. Further, numerous studies prove evidence for vertical inequalities in utilization of healthcare according to education, income and occupation, which represent SES. In order to shed more light on the role of healthcare in explaining health inequalities, it is crucial to examine socio-economic inequalities in the utilization of treatment in a more differentiated way. It has been shown that socioeconomic inequalities in healthcare are present in both universal and non-universal healthcare system, and existence does not depend on the type and financing of health systems [2,3,4]. Distinctions were more likely to be found according to the different dimensions of healthcare for which inequalities are analysed. So far, international evidence on socioeconomic inequalities in treatment was mainly summarized on disease-specific, or country-specific basis and indicate that lower SES is associated with poor diabetes management, lower achievement of glycaemic control targets, and reduced visits of diabetes clinics and ambulatory care facilities for treatment of diabetes [3, 5]. In cancer patients, lower SES is associated with receiving less often (neo) adjuvant therapy for colorectal cancer, [6] and with receiving less often any treatment, surgery and chemotherapy for lung cancer [2]. In coronary heart disease patients SES was often associated with lower access to coronary procedures [4]. Nevertheless, SES was only partly associated with receiving radiotherapy and chemotherapy in colorectal cancer patients, [6] or with access to drug treatment and cardiac rehabilitation in coronary heart disease patients, [4] and not associated with radiotherapy for lung cancer [2]. In contrast, in diabetes patients, it was found that lower SES was associated with more visits to a diabetologist, and more often GP consultations [3]. For Germany it was summarized that higher status groups presented higher utilization in terms of specialist consultations and prevention services [7]. It should be noted, however, that systematic reviews often summarize studies that use different operationalisations of SES and healthcare utilization, and may therefore be difficult to compare.

In order to tackle inequalities in utilization of healthcare, we need to gain a better understanding of healthcare inequalities. More detailed evidence is needed in which domains of healthcare and indicators of utilisation and SES are specified [8]. Despite disease- and country-specific systematic reviews on socioeconomic inequalities, only two reviews summarized the international evidence of inequalities in utilization rates in the general population [8, 9]. However, the first limited their analyses to home health services in developed countries, and found that utilization of home health services in the general population was notably lower for persons with high compared to low SES [8]. The second limited the analyses to healthcare utilization rates in the elderly population and found that the association with SES varied strongly according to the type of healthcare analysed. While elderly patients with low SES were advantaged in home visits, they were disadvantaged in dental and medical appointments, and no association with SES was found for hospitalization rates and emergency use [9].

So far, it is evident that socioeconomic inequalities in utilization rates differ depending on the domain of healthcare analysed. Nevertheless, the evidence on socioeconomic differences in physician utilization in the general adult population has not been summarized so far. Moreover, individual studies suggest that inequalities in physician visits differ depending on whether utilization of primary-care or specialist physicians is analysed [10,11,12]. Therefore, enhanced knowledge is needed (1) if socioeconomic inequalities in visiting primary-care or specialist physicians do exist; and (2) if divergent results of inequalities in physician visits can be explored depending on methodical diversity, e.g., operationalization of SES or utilization.

Methods

This review aims to summarize the evidence on socioeconomic inequalities in consulting primary-care and specialist physicians in the general adult population in high-income countries. Studies from low- and middle income countries (defined by The World Bank 2019 [13]) were excluded as the nature of and issues related to healthcare utilization and health-care system differ significantly from health care systems in high-income countries. To perform this review, we searched the electronic databases Medline and Web of Science to identify relevant studies. In addition, we manually searched the reference lists of all included articles for further potentially relevant studies. The search was conducted in January 2019 and limited to articles published in either English or German within the last 15 years. Different combinations of keywords related to (a) primary-care or specialist physicians, (b) inequalities, and (c) SES were used for the search (see additional file 1).

Study selection and eligibility criteria

The identified records were independently screened by two researchers for eligibility criteria in three consecutive steps: titles, abstracts, full texts. SLL conducted the screening at any step, JH conducted the title- and abstract screening, and JR and JM each conducted half of the full text screening. After each step a joint decision was reached through discussions in cases of disagreement. The criteria used to identify articles of interest limited the search results to original quantitative studies. An article was included in the review if it met the following criteria: (a) analysing the general population aged 15 years or older in a high-income country; (b) analysing any SES indicator (income, education, occupation, social class, or any combination of these indicators) based on individual data; (c) analysing utilization of primary-care, or specialist physicians, or both independently from each other (d) presenting quantitative original data on differences in utilization between at least two different SES groups. The following exclusion criteria were applied: (a) specific populations, namely disease- or SES-specific; (b) differences in race, rurality, insurance status, financial barriers, or employment status; (c) utilization of medical interventions, dentists, inpatient treatment, healthcare in general, or of physicians without differentiating between primary and specialized care; (d) SES or utilization based on area data, or not linked to the individual; (e) conference abstracts and comments.

Data extraction and quality assessment

Data extraction was conducted by SLL, and checked by JR or JM. The following information was extracted from texts, tables, and figures of the included studies: author, year, countries, database, number of participants, participant’s age, physician (primary care or specialist), measurement of utilization, measurement of SES, confounder variables, and the result if an association of SES with physician visits has been found. As the included studies analysed different aspects, it was difficult to compare them in a common scheme that would account for all differences. Consequently, assumptions and simplifications had to be made in order to compare the studies. The results in the tables were abstracted to the most relevant finding analysing if a relationship (and the direction) between SES and utilization of primary-care or specialist physicians was found with the following simplifications:

  1. 1.

    the results comparing the highest SES with the lowest SES (when more than two SES-groups were compared);

  2. 2.

    significant differences at a p ≤ 0.05 or lower (when several p values were designated);

  3. 3.

    the most recent findings (from studies analysing trends of socioeconomic inequalities);

  4. 4.

    the results from the best fitting final model (if an analysis was conducted using different types of adjustments);

  5. 5.

    the results including the broadest variety of the population (if subgroups, e.g. private and public healthcare, were analysed)

Further, to simplify the description of the extracted information and the comparison:

  1. 1.

    we only report the results from high-income countries and report only the most frequent result (if several analyses have been conducted for more than four countries);

  2. 2.

    we dichotomized adjustment variables to “↑” if adjustments were made for at least gender, age and any general health variable; and “↓” if the required need adjustments were not made, including only adjusting for age, gender, and mental health.

Risk of bias was assessed in accordance with RoBANS, [14] and assessed independently by (1) SLL and (2) either JR or JM. The assessments were subsequently discussed to achieve a consensus regarding the rating of each domain in each included article. In a joint decision it was defined that register and national survey data are defined to present “low risk of bias” for the selection of participants, but “high risk” when only sub-populations were analysed without rationale. Second, confounding variables presenting a “low risk of bias” are age, gender, and a minimum of one need-variable of chronic diseases or self-rated health. Further, register data and standardized questionnaires measuring self-reported values are defined to be a “low risk for bias” for measurement of exposure. Fourth, register and national survey data are defined to present “low risk of bias” for the blinding of outcome measure. Fifths, the risk of bias for incomplete outcome data was defined unclear, when missing values were not mentioned or imputed, but high risk when missing values were evident but not tested, and defined low risk when missing values were mentioned and tested. Lastly, for secondary data analyses and analyses of register, panel or national survey data without a study protocol, selective outcome reporting was rated “low risk of bias” when descriptions in the methods section match with the results section.

Results

We found 1229 unique abstracts published between January 2004 and December 2018. Among these, 57 examined socioeconomic differences in physician visits and met all inclusion criteria. The flowchart of the study selection procedure is presented in Fig. 1. Most studies were based on register-data or secondary data from population surveys (see Table 1). In total, the studies comprised data from 32 high-income countries, of which seven were non-European countries, namely, Australia, Canada, Chile, Hong Kong, Israel, New Zealand, and the USA. Whereas three studies analysed pooled data from several European countries, the majority analysed data from one country (n=44), or several countries separately (n=10). These 54 studies most often reported data from Spain (n=15), Germany (n=14), and Belgium (n=10).

Fig. 1
figure 1

Flowchart of the systematic literature research

Full size image
Table 1 Characteristics of the 57 studies included in the systematic review
Full size table

Overall, 70% (n=40) of the studies analysed both primary-care and specialist physician visits, another 25% (n=14) of the studies only primary-care physician visits, and 5% (n=3) only specialist physician visits (Table 2). The definition of primary- and specialist care differed between the studies and health-care systems. Primary-care implied family physicians, and/or general practitioners, but in some cases after excluding prevention services, child or maternity care, physicians at healthcare centres, or internal medicine physicians. Specialist care was defined as medical outpatient specialists, any specialist except while being hospitalized, or generally physician visits at the hospital without being hospitalized. Utilization of physicians was measured according to probability (having visited a physician or not) in 72% (n=41) of the studies, according to frequency (number of visits) in 37% (n=21) of the studies, or according to conditional frequency (number of visits conditional to having visited a physician at least once) in 30% (n=17) of the studies. 95% (n=54) of the studies adjusted the analysis for need according to at least gender, age, and either self-rated health or chronic conditions. SES was measured by income (58%; n=33) and/or education (54%; n=31) in most of the studies. 4% (n=2) of the studies measured SES by income, education, and occupation; 9% (n=5) of the studies only by occupation; and 7% (n=4) of the studies by an SES-index. The period for which physician utilization was reported by the participants, ranged from two weeks to two years. 68% (n=39) of the studies analysed utilization rates within the last 12 months, 14% (n=8) of the studies within the last four weeks or one month, 11% (n=6) of the studies within the last three months, 5% of the studies within the last two months, each 4% (n=2) of the studies within the last two weeks, and last two years, and 2% (n=1) of the studies within the last six months. Most of the 57 studies have carried out several calculations (for different countries, age groups, utilization or SES measures; see Table 2). Therefore, and through rounding the percentages are more than 100%. The following results are based on a total of 548 different analyses.

Table 2 Results on relationships between socioeconomic status and utilization of primary-care and specialist physicians
Full size table

Socioeconomic differences in primary-care and specialist physician visits

Overall, 52% of the analyses on utilization of primary-care physicians found no inequalities, and 35% found higher utilization for the lowest SES group (Fig. 2; primary care ‘all’). Contrary, 71% of the analyses on utilization of specialist physicians found higher utilization for the highest SES group, and 28% found no inequalities (Fig. 2; specialist care ‘all’). While taking a closer look at the various measures of utilization (Fig. 2; probability, frequency and conditional frequency), we found that 62% of the analyses on the probabilities of utilizing a primary-care physician found no socioeconomic inequalities, while 55% of the analyses on frequencies, and 54% of the analyses on conditional frequencies of primary-care physician visits found higher utilization in the most deprived. The results on specialist physicians also differed according to the operationalization of utilization in the way that 78% of the analyses on the probability and 75% of the analyses on the frequency of specialist visits found higher utilization for the highest SES-group. Higher utilization for the highest SES-group was found in only 50% of the analyses on the conditional frequency of specialist visits, whereas another 47% of the latter found no inequalities.

Fig. 2
figure 2

Socioeconomic differences in primary-care and specialist physician visits in all analyses, and subdivided for the different operationalisations of utilization (probability, frequency, and conditional frequency)

Full size image

Various measures of socioeconomic differences in physician utilization

In a second step, we took a closer look at further variations of measures, in order to examine whether they might cause distinct results. Therefore, we contrasted the study’s results (additional file 2) according to differences in time periods for which physician utilization was reported, and SES indicators.

Regarding time periods of utilization, it was found that in shorter time periods of 6 months or less, higher probabilities for primary-care physician visits in the lowest SES-group emerged, whereas studies analysing longer time periods found more often no inequalities. Contrariwise, for specialist visits higher probabilities in the highest SES-group were found less often in short compared to long time periods. Because only a very limited number of studies analysed the frequency or conditional frequency of utilization in a short time period, we renounced the comparison.

Regarding different SES indicators, higher probabilities and frequencies of primary-care physician visits were found for those with low education compared to those with low income, but higher conditional frequencies of primary-care physician visits were found more often for those with low income compared to those with low education. Results for socioeconomic inequalities in specialist physician visits seemed to hardly differ according to SES measurement. As only few studies measured SES by occupation or by an index, we renounced the comparison.

Quality of the studies

The quality of the included studies was fairly high, as the majority of the studies was rated to have a low risk of bias in at least five of the six domains of RoBANS. Only three studies were ranked to have a high risk of bias in more than one domain (additional file 3). The risk of bias of confounding variables, measurement of exposure, blinding of outcome measure and selective outcome reporting was rated low in 54 studies, whereas the risk of bias of incomplete outcome data was rated high in 23 studies (additional file 3).

Discussion

Principal findings

In general, socioeconomic inequalities in utilization of physicians were more prevalent among specialists than among primary-care physicians. The probability of utilizing primary care was often not influenced by SES in the general population, but the disadvantaged visited their primary-care physician more frequently. Moreover, the highest-SES groups often had higher probabilities for specialist visits, but studies often found no associations of SES with (conditional) frequencies of specialist visits.

Interpretation

This systematic review confirms that the existence of socioeconomic differences in healthcare utilization heavily depends on the health services analysed [9]. The existing review on socioeconomic inequalities in physician visits in the elderly population, which did not differentiate between primary-care and specialist physicians, found more medical appointments for the highest-SES group [9]. Accordingly, we found that a distinction of medical appointments between primary and specialized care is necessary when analysing socioeconomic inequalities in physician visits, because the results differed greatly according to the type of doctor and the type of service. We found that not all medical appointments, but mainly specialist were visited with higher probabilities and frequencies by the highest-SES groups. In contrast, most studies indicated that the probability of visiting primary-care physicians was not determined by SES, comparable to the evidence for hospitalization and emergency use, which rather presents access to need- and emergency-oriented healthcare [9]. Lastly, the frequency of primary-care physician visits often was higher in the lowest-SES groups, and is therefore comparable to the evidence on inequalities in utilization of home health services und visits [8, 9].

Consequently, socioeconomic inequalities disadvantaging the deprived are a matter of concern especially in specialist visits. Based on this review, we are not able to infer whether these inequalities are a matter of need, a matter of access barriers to specialist physicians, a matter of different information, or a matter of different preferences and patient choice. Nevertheless, nearly all studies adjusted for patient’s need according to gender, age, and any physical health condition. Either self-rated health or the number of self-reported chronic conditions was applied as an indicator for the latter. Although this indicates good quality, these indicators remain only approximate to real need of receiving healthcare. Accordingly, we cannot conclude that probability of primary-care physician visits is needs-based even though most studies did not find significant associations with SES. In order to avoid underestimating or disregarding differences, when analysing only probabilities of visits using register-based data, Agerholm et al. concluded that health status should be considered in analyses on socioeconomic differences in healthcare utilization [16]. Nevertheless, self-rated health remains a subjective rating of people’s perception of their health. Although studies found that self-rated health is a good proxy for objective health in the general population, [69] one study found that the evaluation of self-rated health is biased by SES, because the more educated rated their subjective health worse with the same level of objective health, [70] which implies socioeconomic differences might be underestimated.

Given the results that low-SES populations often visit specialist physicians less often, but primary-care physicians more frequently at concurrently equal probabilities compared to high SES populations, an intuitive explanation is that barriers in access to specialists are important in explaining healthcare inequalities. One possible reason for access barriers to specialists might be rurality of low-SES populations [71]. Thus, waiting time and distance might carry more weight in visiting specialists, because those are often distributed regionally more widely than primary-care physicians. Furthermore, the results suggest that different information, preferences, and patient choices are relevant reasons for socioeconomic inequalities in physician visits, because the relationship with primary-care physicians is more trusting and familiar than with specialist. As a consequence, the lower educated might feel less exposed to existing communication problems (language barriers, terminology, information gap) [72] with their longtime, well-known primary-care physicians, and they might prefer visits to them compared to specialist physicians [10]. The perceived role in healthcare varies between SES groups, as those with low SES tend to delegate responsibility to healthcare professionals [73]. Given the trustful and longtime relationship with primary care physicians compared to specialists, and given that primary-care physicians have the task of gatekeeping in some countries, might emphasize the importance of primary-care physicians from the perspective of the low-SES population when delegating responsibility for their healthcare, and might therefore explain the more frequent visits from the most deprived.

This review found that income inequalities advantaging high-SES groups in primary-care physician visits are more pronounced than educational differences. This may be an indication that financial barriers are a relevant additional factor explaining socioeconomic inequalities in utilization of primary-care physicians [74]. The finding that detrimental inequalities were found less often in shorter time periods is consistent with the finding that detrimental inequalities were found less often when utilization was operationalized with frequency versus probability. Accordingly, a higher frequency of physician visits among the most deprived means that they are more likely to have visited a physician at least once in a short period. A possible explanation could be that frequencies are more likely to be influenced by preferences and patient choice, whereas probabilities are more likely to be influenced by access barriers.

Limitations

Although we have screened 1229 references we might have missed relevant publications, especially those not differentiating between primary and specialized healthcare in the abstract, but only in the main text. Second, the selection criteria might bias the results, which are not generalizable to children, disease-specific populations, low- and middle-income-countries, or inequalities induced by other (horizontal) disadvantages. Because very few studies based SES on area data, we excluded them even when area SES was linked to the individual on postal codes. Third, we made various simplifications in order to compare the studies, which influenced the reported results, which must be interpreted carefully. We described the results only by comparing the highest with the lowest SES group, and we did not include effect sizes in our descriptions. The health systems of the countries are very different, e.g., primary und specialized healthcare was defined differently in different studies. Primary health and its connection with specialist care is organised differently between the countries. For these reasons and because some studies analysed the same data basis, frequencies must be interpreted with caution, and comparisons are rather explorative hints than robust results. Finally, data on utilization, SES, and health were often self-rated, and even though instruments are valid, the accuracy is affected by different factors, [75] which limits expressiveness.

Conclusions

In order to tackle socioeconomic inequalities in healthcare to the detriment of the deprived population, utilization of and access to specialist physicians is essential. The fact that predominantly no inequalities in probabilities of visiting primary-care physicians were found is generally a good result. Not visiting a primary-care physician can be interpreted as more fatal in maintaining good health than visiting specialists less frequently. This emphasizes the fact that the general population in high-income countries might have access to physicians largely independent of their SES, but the deprived might experience more barriers in accessing specialized healthcare. We assume that higher frequencies of primary-care physician visits from the low-SES groups with the same level of need might be subject to patient preferences in order to compensate for different levels of health literacy, information and communication, and therefore improve equal opportunities in receiving health maintenance.