Quality in Primary Care Open Access

Keywords

general practice, health service utilisation, multilevel analysis, patient preference, substitution

How this fits with quality in primary care

What do we know?

Substitution for medical specialist services by primary care is increasingly considered an instrument to improve quality and efficiency of care delivery. To effectively substitute medical specialist services in patientcentred health care systems it is important to take patient preferences into account alongside technical feasibility criteria, but insight into patients’ preferences is lacking.

What does this paper add?

This study provides strong indications that patients’ preferences for substitution are influenced by the type of medical intervention. According to patients, follow-up treatment and non-complex invasive treatment are more acceptable for substitution than complex invasive treatment, non-invasive treatment and diagnostic examinations.

Introduction

In the Netherlands and other countries with a strong primary care system, such as theUKand Scandinavian countries, general practitioners (GPs) provide first contact and ongoing care for most health care problems. They decide whether or not to refer patients to medical specialists in hospitals (‘gatekeeper’ function). This is regarded as cost-effective compared with health systems in which patients can access specialist services directly. To further improve the quality and efficiency of the health care system, health policies in many countries stimulate shifting medical specialist services delivered in hospitals to GPs.[1] Research suggests that this substitution offers cost benefits,[2–4] while quality of care and health outcomes seem to be unaffected if the transferred service does not demand competencies beyond those of the average GP.[1,5] Besides this, care by GPs is associated with improved access (proximity) and convenience for patients.[1,5]

To stimulate substitution, the Dutch government included several incentives for substitution in the reform of the Dutch health insurance system in 2006.[6] First, they introduced additional payments for specific services in primary care that substituted for medical specialist services.[6] Previous research has shown that additional payments for specific services in primary care indeed stimulate substitution.[7,8] Second, the Dutch government introduced a regulated market for health insurance, which emphasises cost-effective health care purchasing by insurers in order to attract patients.[6,9] For example, insurers can purchase care delivered by GPs instead of by medical specialists. To achieve this, they in turn have to spur on health care providers to provide high-quality health care which is produced as efficiently as possible. At the same time, insurers need to take into account their insurants’ preferences for the source of health care delivery, since insurants have a free choice of insurer and can change insurer every year.[9] However, to date it is unclear how shifting medical specialist services to GPs is valued by patients.

Previous research that assessed patients’ preferences for different provider types (e.g. medical specialist versus nurse practitioner) showed that patients base their provider choice mainly on providers’ expertise, [10–13] and less on convenience[10] and the location of care.[13] The importance of waiting time for patients’ preferences for provider type varies in different studies.[10,13] Furthermore, it was shown that patients’ preferences for different provider types are influenced by age,12 education[12] and previous experiences with the medical intervention.[12,14] Previous studies show slight differences between patients’ preferences for health care provider types by type of medical intervention. For example, while surgical patients mainly base their provider choice on care delivery characteristics such as provider experience and specialisation, [13] chronically ill patients base their provider choice on the continuity of care.[15] Accordingly, we expect that patients’ preferences for health care provider types, and thus for substitution, may be affected by medical intervention type (e.g. minor surgery, follow-up treatment and diagnostics). When this is true, health care policy makers, insurers and practitioners should take patients’ preferences into account next to quality, efficiency and technical feasibility as criteria to effectively shift medical specialist services to GPs.

The aim of this study was to assess whether patients’ preferences for service delivery by medical specialists or GPswere related to the type of medical intervention (e.g. diagnostics or minor surgery).

Methods

Data collection

We conducted an online survey among members of the Dutch Insurants Panel. The Insurants Panel consists of approximately 10,000 insurants of one of the biggest Dutch health insurers (market share = 26%).16 Compared with the Dutch population, older people (aged between 40 and 80 years), people with a bad to moderate self-reported health status and people who have been insured with the same health insurer for over 10 years are over-represented.[17] The aim of the panel is to gather information on people’s experiences with and expectations of health care in general and their health insurer in particular. Members for the panel were recruited through an announcement in the health insurer’s magazine. Compliance with privacy regulations was approved by the Dutch Data Protection Authority (number 1309664). According to Dutch legislation, neither obtaining informed consent nor approval by a medical ethics committee was obligatory for this study.

In February 2008, all panel members filled in a questionnaire on several background variables (age, gender and educational level). In June 2008, questionnaires were sent to 500 men and 500 women of the Insurants Panel who had agreed to complete Internet questionnaires. We used stratified random sampling to select these panel members and to create a subgroup that showed the same distribution regarding age as the whole Insurants Panel. After oneweek, a reminderwas sent to 512 panel members who had not responded, and again after one week a second reminder was sent to 339 panel members.

Conceptual model

To assess the effect of medical intervention type on patients’ preferences for type of health care provider, we constructed a conceptual model. In this model, the patients’ preference for health care provider type is the dependent variable. The independent variable is represented by the medical intervention type. Andersen’s Behavioural Model of Health Services Utilisation[18] was used to determine which covariates had to be included in our model to correct for confounding. Originally, the purpose of Andersen’s framework was to discover inequities in health care utilisation,[18] but it has also been used to analyse choices of type of health care utilisation,[19] and choices for using conventional or complementary (alternative) medical services.[20] The framework portrays the process of choosing health care as a complex of three interrelated sets of predisposing (sociocultural characteristics of individuals), enabling (logistical aspects of obtaining care) and need factors.[21]

The survey included questions on all variables of our conceptual model:

• Preferences for health care provider type regarding 11 different medical interventions (Table 1). There were three options: (1) treatment by a GP, (2) treatment by a medical specialist and (3) no preference. According to the purchasing department of the insurer, these 11 medical interventions were eligible for substitution.

• Predisposing factors: age (continuous variable), gender and educational level (highest level of education was categorised into three levels, lower, intermediate and higher, according to the classification of Statistics Netherlands).

flenabling factors: travel distance—four-level categorical variable (< 5, 5–10, 10–20 or > 20 km, perceived effort needed to visit a GP or medical specialist—five-level categorical variable (none, little, intermediate, great or very great). Health insurance was not included as an enabling factor because the study population consists of insured patients of one health insurer, and all medical interventions included in this study were covered.

• Need factor: perceived health status—five-level categorical variable (poor, fair, good, very good or excellent)]. Perceived health status was used as a proxy for need because of its relation to use of health services.16

In addition, we included ‘treatment experience’—use of health care services for one of the medical interventions in the past 12 months (dichotomous variable indicating whether or not a respondent had one of the medical interventions in the past 12 months)—as a covariate in our analysis, assuming that previous experiences could influence patients’ preferences.

In addition, all panel members were asked why they preferred being treated or examined by a GP or medical specialist. The respondents could choose from 12 reasons. These reasons were similar for both health care providers.

The complete questionnaire used in this study is shown in Supplementary file 1.

Data analysis

We used descriptive statistics to describe the survey results. To analyse the effect of medical intervention type on preferences for health care provider type, we first categorised the medical interventions into five groups using dummy variables: complex invasive treatment (invasive I), non-complex invasive treatment (invasive II), follow-up treatment (follow-up), non-invasive treatment (non-invasive) and diagnostic examinations (Table 1). Categorisation was based on substantive considerations and a first analysis of the medical interventions. Next, we conducted multilevel multinomial regression analyses comparing three responses: (1) preference for GP, (2) preference for medical specialist and (3) no preference. The ‘GP preference’ was regarded as a reference group. In the multilevel analyses, two levels were distinguished: 11 medical interventions (level 1) within individual respondents (level 2). By analysing the data with multilevel models, we took into account this hierarchical data structure; the preferences of a given individual for each of the medical interventions are supposed to be correlated.[22,23]

Two models were tested using MLwiN 2.0 (see Supplementary file 2, available online only, for model specifications). In the first model, the influence of the type of medical intervention on preferences for health care provider type was assessed. In the second model, the covariates were added to the model to correct for confounding. The level 2 variance and the covariance between the dependent variables were inspected.

Results

The questionnaire was returned by 694 members of the Insurants Panel (response rate 69.4%). All 694 respondents filled in all questions about their preferences for health care provider type regarding the 11 medical interventions. Respondents were similar to non-respondents in terms of age, gender and educational level (Table 2).

Table 3 shows the unadjusted preferences for health care provider type. For complex invasive treatments, non-invasive treatment and diagnostic examinations, respondents more often preferred treatment or examination by a medical specialist than by a GP. Respondents more often preferred to visit a GP for follow-up treatment and non-complex invasive treatments.

Table 4 shows the five most common reasons to prefer utilisation of GP or medical specialist services. GP services were preferred because of better accessibility (52% of respondents), shorter access times (50%) and more comfort (28%) in comparison with medical specialist services. Reasons for preferences for medical specialist services included better skills of a medical specialist (82% of respondents), a lower perceived risk of treatment by a medical specialist (43%) and more confidence in medical specialists (31%) compared with GP services.

The results of the multilevel multinomial regression analysis are presented in Table 5.

The preference for health care provider type was significantly related to medical intervention type. The regression coefficients confirm the pattern in Table 3. In order of preference for GP services compared with GP preference in cases of complex invasive treatment, medical specialist services are preferred slightly less often for the examples of diagnostic examination (Model I:
= –0.45, p < 0.01; Model II:
= –0.48, p < 0.01), less so for the examples of non-invasive treatment (Model I:
= –1.49, p < 0.01; Model II:
= –1.51, p < 0.01), even less so for the examples of less complex invasive care (Model I:
= –2.98, p < 0.01; Model II:
= –3.01, p < 0.01), and least for follow-up care (Model I:
= –3.11, p < 0.01; Model II:
= –3.17, p < 0.01). The preference for GP services compared with no preference shows the same pattern.

Comparison of model 1 and model 2 showed that covariates did not confound the effect of type of medical intervention on respondents’ preferences.

Instead, they provided additional information about the respondents’ preferences. It appeared that older respondents (
= 0.01, p = 0.02), respondents for whom the effort to visit a GP is relatively great (
= 0.26, p < 0.01) and respondents with treatment experience (
= 0.47, p < 0.01) more often preferred medical specialist services. Respondents with a better perceived health status (
= –0.13, p = 0.04) preferred GP services more frequently.

The bottomrows of Table 5 show the variance in the responses at the level of the respondents and the covariance between the responses. Multilevel analysis allowed us to analyse variance and covariance. In a traditional analysis we would have analysed the responses for each of the 11 interventions separately.

This is inefficient and does not take into account that a respondent’s preference in the case of one of the interventions might be related to their preference in the case of the other interventions. This relationship between preferences, elicited for the 11 different interventions, is expressed by the variance at the level of the respondents (level 2 variance in Table 5).The variance between respondents who have no preference is much higher (2 = 2.84, p < 0.01) than that between respondents who prefer medical specialist services (2 = 0.88, p < 0.01). This means that that stating no preference is more of a general pattern of reacting to the questions about their preferences rather than being influenced by the specific medical interventions. Conversely, the specific preferences for either GPs or medical specialists depended more on the specific medical interventions. The use of multinomial multilevel analysis allowed us to analyse the three options for preferences, no preference, preference for GP or preference for medical specialist, simultaneously. Again, this is a more efficient analysis and it takes into account the fact that these preferences might be mutually related. This is expressed by the covariance. The covariance shows that respondents who more frequently prefer medical specialist services also frequently have no preference rather than preference for a GP.

This study provides strong indications that patients’ preferences for substitution are influenced by medical intervention type. Additional findings suggest that patients’ age, the effort needed to visit a GP, perceived health status and previous treatment experiences also influence preferences for substitution but do not confound the effects of medical intervention type on patients’ preferences. These additional findings are in line with previous studies[12] and utilisation patterns of health services. Rodriguez et al[24] found, for example, that having fair or poor health increases the chance of visiting a medical specialist, and that longer travel times to hospitals (i.e. a greater effort) had a significant negative effect on the probability of seeing a medical specialist. Being older increases the chance of having fair or poor health due to having an increased chance of (multiple) illness. This is probably not completely captured by self-rated health. Therefore it seems plausible that older people more often prefer treatment by a medical specialist.

Our study has some limitations. First, due to the limited number of treatment examples per medical intervention type, the results cannot be generalised to all treatments. Second, the medical interventions included in this survey were not all applicable to male and female respondents (e.g. male sterilisations). However, the statistical model adjusted for gender. Although these limitations prevent us from applying our results to every type of diagnostic examination, com-

plex invasive, non-complex invasive, non-invasive and follow-up treatment, our study does give clear indications that patients do not always prefer substitution. A third limitation of our study is that, compared with the Dutch population overall, older people (aged 40– 80 years) and people with a bad to moderate selfreported health status were over-represented in the panel used. However, these groups use health care services more frequently, which makes their preferences more relevant. We therefore do not think that the overrepresentation of these groups has led to distortions in the results, and it does not influence the relevance of the results. Another limitation is that we performed our study among a sample of panel members instead of patients. These panel members are not typical of all patients. This may have led to selection bias.

Our study is, to our knowledge, the first to quantitatively investigate patients’ preferences for health care providers regarding several types of medical interventions. Previous studies have focused on patients’ preferences for one type of medical intervention (e.g. follow-up care for breast cancer patients12) or for one patient group (e.g. dermatology services10). Despite the limitations, our study provides useful findings for health care policies aiming to shift services from medical specialists to GPs in patient-centred health care systems. The results of this study show health policy makers, purchasers and practitioners strong indications that shifting follow-up treatment and noncomplex invasive treatment from medical specialists toGPs has the largest support among(potential)patients. However, health care policy makers, purchasers and practitioners have to be aware that patients may be resistant to changes in the organisation of health care. Patients tend to prefer what they know best or have experienced previously.[12,25] However, acceptance by the public is not enough to effectively implement substitution of care. To ensure that patients get the opportunity to act according to their preferences, they must be adequately informed and the health care system must allow it. Patients have to learn, in turn, to express their preferences and act as autonomous health care users. Furthermore, health care policy makers and purchasers have to be aware of the risk that the quality of services may change when medical specialist services are substituted with primary care services. A change in quality could go either way, improving or decreasing quality. A risk to quality is, for example, that GPs will not be sufficiently skilled to undertake the work previously done by medical specialists.[1] It is therefore important to first assess whether GPs are equipped and skilled to provide these treatments, or otherwise start educational interventions to address this before shifting services. Another concern in the literature is that transfer will lead to increases in total health care expenditure at a macro level.[26] Although previous research showed that reductions in costs were achieved through lower salaries or reimbursements in primary care2–4 and reduced time and travel costs for patients,[1,5] these can be offset by rises in costs generated through increases in the volume of care and loss of economies of scale.1 More important probably is that the capacity that is freed in medical specialist care by shifting services to primary care might be quickly filled by higher demand for other services. The extent to which this happens depends on the incentives in the system. Specifically, for the interventions that were used as examples in our survey there is—as far as we know—no literature about improved quality when comparing medical specialists and GPs. In general, research about disease management [27] and bundled payments for the treatment of chronic disease[28,29] shows some evidence of improved care. However, this is related to programmes, rather than separate interventions, and the evidence is still not very strong.

Therefore, first of all, medical specialist services that can be substituted by GPs should be chosen carefully. Second, purchasing should be part of an integral policy that focuses on broader care programmes into which separate interventions fit, and that as far as possible takes cross-over effects of substitution into account.

Conclusion

This study demonstrated that patients’ preferences for substitution are influenced by medical intervention type. These findings provide strong indications that health policy makers, purchasers and practitioners in patient-centred health care systems should take patients’ preferences into account next to quality, efficiency and technical feasibility as criteria for effectively substituting medical specialist services with GP services.

Funding

This work was supported by the Dutch health care insurer VGZ.

Ethical Approval

According to Dutch legislation, neither obtaining informed consent nor approval by a medical ethics committee was obligatory for this study.

Peer Review

Not commissioned; externally peer reviewed.

Conflicts of Interest

Leti van Bodegom-Vos, Judith de Jong, Peter Spreeuwenberg and Peter Groenewegen declare that they have no conflicts of interest. Emile Curfs is an employee of VGZ (not involved in decisions about policy and health care purchasing of VGZ), the funder of this research.

Supplementary files

Supplementary file 1: Questionnaire

Note: This questionnaire was originally written in Dutch and was only translated for this publication.

Survey questions: ‘Preferences for type of health care provider’

Supplementary file 2: The multilevel multinomial model

References

1. Sibbald B, McDonald R, Roland M. Shifting care from hospitals to the community: a review of the evidence on quality and efficiency. Journal of Health Services Research & Policy 2007;12(2):110–17.
2. World Health Organization. Advisory Support to Primary Health Care Evaluation Model: Estonia PHC Evaluation Project. World Health Organization: Geneva, 2004.
3. Health Council of the Netherlands. European primary care, publication no. 2004/20E. 2004.
4. Ros CC, Groenewegen PP, Delnoij DM. All rights reserved, or can we just copy? Cost sharing arrangements and characteristics of health care systems. Health Policy 2000;52(1):1–13.
5. Roland M, Mead N, Sibbald B et al. Outpatient Services and Primary Care. A scoping review of research intostrategies for improving outpatient effectiveness and efficiency. National Primary Care Research and Development Centre, Manchester: 2006.
6. Minister of Health, Welfare and Sport. Substitution of Care [in Dutch]. Document number CZ/FBI 2861735. Minister of Health, Welfare and Sport: The Hague, 12 September 2008.
7. Groenewegen PP. Subsitution of primary care and specialist care: a regional analysis in Denmark. Social Science & Medicine 1991;33(4):471–6.
8. Krasnik A, Groenewegen PP, Pedersen PA et al. Changing remuneration systems: effects on activity in general practice. British Medical Journal 1990;300:1998–2001.
9. Custers T, Onyebuchi AA, Klazinga NS. Is there a business case for quality in the Netherlands? A critical analysis of the recent reforms of the health care system. Health Policy 2007;82:226–39.
10. Coast J, Salisbury C, de Berker D et al. Preferences for aspects of a dermatology consultation. Epidemiology and Health Services Research 2006;155:387–92.
11. Gerard K, Lattimer V. Preferences of patients for emergency services available during GP surgery hours: a discrete choice experiment. Family Practice 2005;22: 28–36.
12. Kimman ML, Dellaert BGC, Boersman LJ, Lambin P, Dirksen CD. Follow-up after treatment for breast cancer: one strategy fits all? An investigation of patient preferences using a discrete choice experiment. ActaOncologica 2010;49:328–37.
13. Schwappach DL, Strasmann TJ. Does location matter? A study of the public0s preferences for surgical care provision. Journal of Evaluation in Clinical Practice 2007; 13:259–64.
14. Scott A. Identifying and analysing dominant preferences in discrete choice experiments: an application in health care. Journal of Economic Psychology 2002;23:383–93.
15. Berendsen AJ, De Jong GM, Meyboom-de Jong B, Dekker JH, Schuling J. Transition of care: experiences and preferences of patients across the primary/secondary interface—a qualitative study. BMC Health Services Research 2009;9:62.
16. Hoeymans N, Melse JM, Schoemaker CG. Health and Determinants: part of the report of public health status and forecasts 2010 [in Dutch]. National Institute for Public Health and the Environment: Bilthoven, 2010.
17. Hendriks M, Delnoij DMJ, Groenewegen PP. VGZ Insurance Panel: Background characteristics 2006 [in Dutch]. Netherlands Institute for Health Services Research: Utrecht, 2006.
18. Andersen RM. Revisiting the behavioral model and access to medical care: does it matter? Journal of Health and Social Behavior 1995;36:1–10.
19. Thind A, Mohani A, Banerjee K, Hagigi F. Where to deliver? Analysis of choice of delivery location from a national survey in India. BMC Public Health 2008;8(29).
20. Sirois FM, Gick ML. An investigation of the health beliefs and motivations of complementary medicine clients. Social Science & Medicine 2002;55:1025–37.
21. Kelner M, Wellman B. Health care and consumer choice: medical and alternative therapies. Social Science&Medicine 1997;45:203–12.
22. Leyland AH, Groenewegen PP. Multilevel modelling and public health policy. Scandinavian Journal of Public Health 2003;31:267–74.
23. Snijders TAB, Bosker RJ. Multilevel Analysis. An introduction to basic and advanced multilevel modeling. Sage: London, 1999.
24. Rodriquez M, Stoyanova A. The effect of private insurance access on the choice of GP/specialist and public/ private provider in Spain. Health Economics 2004;13: 689–703.
25. Salked G, Ryan M, Short L. The veil of experience: do consumers prefer what they know best? Health Economics 2000;9:267–70.
26. Sibbald B, Pickard S, McLeod H et al. Moving specialist care into the community: an initial evaluation. Journal of Health Services Research & Policy 2008;13:233–9.
27. Gress S, Baan CA, Calnan M et al. Co-ordination and management of chronic conditions in Europe: the role of primary care—position paper of the European Forum for Primary Care. Quality in Primary Care 2009;17:75– 86.
28. de Bakker DH, Struijs JN, Baan CB et al. Early results from adoption of bundled payment for diabetes care in the Netherlands show improvement in care coordination. Health Affairs (Millwood) 2012;31:426–33.
29. Struijs JN, de Jong-van Til JT, Lemmens LC, Drewes HW, de Bruin SR, Baan CA. Three Years of Bundled Payment for Diabetes Care in the Netherlands: Impact on health care delivery process and the quality of care. RIVM: Bilthoven, 2012.