- Email: [email protected]
Perception of the economic value of primary care services: A willingness to pay study
Health Policy 94 (2010) 266–272
Contents lists available at ScienceDirect
Health Policy journal homepage: www.elsevier.com/locate/healthpol
Perception of the economic value of primary care services: A willingness to pay study Jesús Martín-Fernández a,∗ , Tomás Gómez-Gascón b , Juan Oliva-Moreno c , María Isabel del Cura-González d , Julia Domínguez-Bidagor e , Milagros Beamud-Lagos f , Teresa Sanz-Cuesta g a
San Martín de Valdeiglesias Health Center, 1st La Bola St., 28680 San Martin de Valdeiglesias, Madrid Health Service, Madrid, Spain Puerta Bonita II Health Center, Madrid Health Service. Faculty of Medicine, Complutense University of Madrid, Spain c Economic Analysis Department, Castilla la Mancha University, CIBERESP-CIBER in Epidemiology and Public Health, Spain d Research Unit, 9th Primary Care Area, Madrid Health Service. Rey Juan Carlos I University, Madrid, Madrid, Spain e Health Education, Quality Unit, 11th Primary Care Area, Madrid Health Service, Madrid, Spain f Research Unit, 11th Primary Care Area, Madrid Health Service. University Nursing School, Complutense University of Madrid, Madrid, Spain g Preventive Medicine and Public Health, Research Unit, 9th Primary Care Area, Madrid Health Service, Spain b
a r t i c l e
i n f o
Keywords: Primary health care Economics Patient satisfaction Health services research Health services needs and demand
a b s t r a c t Objective: Identify the economic value the user attributes to the visit to the family physician, in a setting of a National Health System, by the Willingness to Pay (WTP) expressed. Methods: Economic evaluation study, by the contingent valuation method. Questions were asked about WTP using a payment card format. Interviews were conducted with 451 subjects, in areas with different socioeconomic characteristics. An ordered probit was used to evaluate model’s validity. Results: Median WTP expressed was D 18 (interquartile range D 8–28), not including “zeroanswers” of thirty-four subjects (7.5%). This value represents 2% of average adjusted family incomes. Patients with higher incomes or with chronic illnesses presented a probability of 5–14 percentage points of expressing a high WTP. For every point of increase of patient satisfaction, the probability of presenting a WTP in the lowest range decreases 7.0 percentage points. Subjects with a low education level and those older than 65 expressed a lower WTP. Accessibility, risk perception, nationality and having private insurance were not related to the WTP expressed. Conclusions: Users of primary care have a clear perception of the economic value of care received from the family physician, even in a framework of providing services financed by taxes and without cost at the moment of use. This value increases in subjects with higher incomes, with greater need for care, or more satisfied. © 2009 Elsevier Ireland Ltd. All rights reserved.
1. Introduction The permanent debate on the proper allocation of economic resources to provide effective health care policies is also present in Spain [1]. The Spanish National Health Sys-
∗ Corresponding author. Tel.: +34 918610273; fax: +34 918612320. E-mail address: [email protected] (J. Martín-Fernández). 0168-8510/$ – see front matter © 2009 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.healthpol.2009.11.001
tem is financed fundamentally by taxes, and health care services have no monetary cost at all. Only prescribed pharmaceuticals are paid for, to a maximum of 40% of their total value, except pensioners and some persons with chronic illnesses who receive them free. In this context a high consumption of resources is observed, also in primary care (PC). In 2005 there were thirteen million PC visits in the Community of Madrid (CM) corresponding to 3,757,000 different people, almost 63% of the population census [2].
J. Martín-Fernández et al. / Health Policy 94 (2010) 266–272
Different uses of this public health system, and especially of PC, can be related to several factors in addition to health needs [3]. Therefore other elements, in addition to the need stated must be considered in health planning. Among these variables we found the user’s perception of value of the service to be fundamental. In the absence of a market, whereby society’s valuation of a health care service program could be expressed, the problem is: how do we measure perception of value in monetary terms? A possible applicable measure is the willingness to pay (WTP) for these services, calculated by the Contingent Valuation (CV) Method. Willingness to pay is a ‘theoretically correct’ approach, to the service valuation because of its foundation in welfare economics. In addition, WTP imposes no restrictions to which attributes of a program people are allowed to value, and allows comparability across other public services [4]. This methodology has been employed in different situations in health care that include the evaluation of the approach to various pathologies [5], the valuation of health technologies and preventive or curative treatments [4,6]. CV has also been used to study the “need” for certain services in prevalent illnesses [7], or preferences for preventive programs [8,9]. In the primary care setting, CV has been used to estimate the willingness to defray the costs of improvements in health care in systems in development [10], to evaluate health promotion programs [11], mental health care [12], the acceptability of payment for the continuation of care by the family physician [13], or the extension of certain kinds of health coverage [14]. Primary care services provided by the Public Health System in the CM are satisfactorily perceived [15], but the valuation that people make of them in scales comparable with other services considered necessary in a “welfare” society are unknown. Establishing the economic value attributed to this service could be valuable in the design of future health care policies [16], as well as to comparatively estimate the utility attributed to the service. The purpose of this paper is to study the economic value attributed by the user of the public health care system in the CM (Spain) to the service received at a visit to the family physician, by the WTP expressed. 2. Material and methods A cross-sectional study was designed that used the Contingent Valuation Method, in a framework of compensatory variation of the classification by O’Brien and Gafni [17]. Interviews were conducted with 451 subjects, over 35 years, who had just visited a family physician. These were randomly selected from the lists of visits at two rural and four urban centers of areas with different socioeconomic characteristics in the Community of Madrid (Spain). 2.1. Scenario and payment format Field work was conducted by a single trained interviewer, between December 2007 and March 2008. The interviewer explained that the study was an attempt to attribute value to a service whose price is not established and it was not at any time pursuing the estab-
267
lishment of direct payments for it. The following question was then asked: “Imagine for a moment that there is no public health care system. Suppose you have a health problem similar to what brought you here today to the visit and you have to be attended by the physician who received you today, but you must pay for this care directly. How much money would you be willing to pay for this visit?” The response was provided by a “double payment card” method. The first payment card only contained three values: less than D 20; between D 20 and D 40; more than D 40. These cutoff points were chosen following the valuation of the distribution of responses obtained in a pilot with twenty patients. The second card was fitted to the WTP expressed and contained the values: 0, 1–5, 6–10, 11–15, 16–20, 21–25, 26–30, 31–40, 41–50, 51–60, 61–70, 71–80, 81–90 and more than D 90. 2.2. Characteristics of the subjects and the service evaluated The subject’s characteristics collected included age, sex and place of birth, state of health, accessibility to the service, other types of insurance, risk perception, relation with the physician, socioeconomic situation and family income. The evaluation of health need was made noting the existence of chronic pathologies, hospital admissions in the last year, and number of visits to the family physician in the last year. The visual–analogical scale of EuroQol-5D was used to evaluate the subjective perception of the state of health. Accessibility to the service was studied by the time necessary to obtain an appointment and by the waiting time from time of appointment to the visit. To evaluate risk perception questions were asked about the existence of the following risky behavior: smoking habit, driving vehicles without obeying safety measures, risky sexual behavior, non-observation of job-safety measures and excessive alcohol consumption (AUDIT-C questionnaire) [18]. The relation with the physician was evaluated with the PDRQ questionnaire, adapted to Spanish [19]. To evaluate the socioeconomic situation questions were asked about the highest level of education completed (illiterate, no education, primary, secondary, superior), by job (class I for the most skilled job to V for the least skilled) [20], and by current employment situation. Adjusted family income was calculated adding up all the income of the family unit and weighting by family size N (average income = family income/N0.4 ). The exponent 0.4 reflects that the family economy is an economy of scale, in which each new member needs an increasingly smaller marginal contribution of resources to maintain the buying power of the family group. 2.3. Analysis The WTP declared was transformed into a continuous variable, applying to each interval its “middle value”, for the descriptive study of the distribution of the variable.
268
J. Martín-Fernández et al. / Health Policy 94 (2010) 266–272
“Zero-answers” were analyzed separately. We believe that a zero valuation of the service does not correspond to the real perception of the person interviewed. In fact, there are other options approaching zero that enable expressing more realistically a low-value perception. We therefore assume that these responses represented a rejection to the question posed (“zero protests”), or the inability to pay, and were therefore treated separately. The theoretical validity of the WTP was studied analyzing its relation to need, satisfaction, socioeconomic position and average income based on the conceptual framework of the “welfare theory” [17,21]. We therefore expected a higher WTP in people with more health needs, with greater economic resources or better social situation (level of superior studies, higher skilled work) and in those more satisfied with the service received. Given that the dependent variable is not observable a model was constructed in which the latent variable y* is not observed (it is the “real valuation” of the individual), and depends upon a linear combination of explicative variables. An ordered probit was utilized, a model that enables grasping the ordered nature and intensity of the dependent variable [22]. Thus, the higher the value of our latent variable, the higher the probability will be that the individual will state a higher category in the valuation scale. The dependent variable acquires a value of 1–4 depending on the intensity of the order. Thus y = 1 corresponds to the lowest valuation, WTP less than D 10, y = 2 represents a mid-low WTP, between D 11 and 20, y = 3 a mid-high WTP, between D 21 and 40, and y = 4 the highest WTP, more than de D 40. We can express our model as follows: y∗ = ˇ0 + ˇ1 x1 + · · · + ˇn xn + ε where x1 , xn are the explicative variables, ˇ0 is a constant, ˇ1 . . .ˇn are the coefficients that adjust the model and ε is a coefficient that represents a random error with normal distribution. The explicative variables were chosen depending on the requirements of the model, eliminating those that presented colinearity (they could be partially explained by the linear combination of other selected variables). The model was initially constructed without accounting for “zero-answers.” Subsequently, there is a sensitivity analysis with all the subjects, substituting the “zeros” by their “estimated WTP.” These values were predicted based on the responses observed in individuals who revealed their WTP. Thus, the “zero values” were substituted for each individual by the average values of the WTP observed in the similar individuals whose WTP > 0. To reduce possible biases, a matching was established between each individual with WTP = 0 and more similar individuals with WTP > 0. Age, gender, level of studies, and level of income as control variables were taken into consideration. 3. Results An invitation to participate was extended to 487 subjects; thirty-six (7.4%) did not accept, twenty-seven for not having time to be interviewed, two for lack of inter-
est in the subject, two who gave no reason and in five cases for other reasons. Of those who did not accept participating ten were men and twenty-six women, with an average age of 51.1 years (57.3 years for the group included p = 0.013) and no differences were found with respect to the rest of the group as far as area of residence, number of chronic illnesses or frequency of utilization of the service. Table 1 shows the characteristics of the 451 subjects that finally consented to participate in the study. The mean value of the WTP was D 19.6 (CI 95%, D 18.0–21.2), for those subjects who expressed values greater than zero. Median distribution was D 18 (interquartile range D 8–28). Looking at the average number of visits in a year, this value represents approximately 2% of average adjusted family incomes. Thirty-four subjects (7.5%) expressed a WTP equal to zero. The profile of these subjects is a person of middle age somewhat older than the rest of the group (63.4 vs 56.8 years; p = 0.011), with a better perception of his/her quality of life (EVA-EuroQol-5D 74.5 vs 63.6; p = 0.004), who makes the same use of the system (same number of admissions, visits, and waiting time), with higher aggregation of illiterate or uneducated subjects (44.1% vs 19.4%, p < 0.001) and with lower adjusted family income (D 1011.6 vs 1311.4, p = 0.080). Table 2 shows the results of the empirical model without “zero-answers.” The reference category is a male, between 35 and 49 years of age, Spanish nationality, without chronic illnesses, not hospitalized in the prior year, with less than ten annual visits, with a wait at the last visit of less than 15 min, without private health insurance, without recognized risky behavior, with secondary studies and a monthly adjusted family income level ≤D 600. Patient satisfaction with the relationship with the physician was introduced as a continuous variable. Table 3 shows the results incorporating the estimates effected on the WTP by “zero-answers.” These results are expressed in marginal terms, that is, in comparison with the category of reference. In the case of satisfaction, the result refers to a change of one unit on its measurement scale. Age is a variable associated with WTP. While there are no differences in intermediate ages, those older than 65 have a higher probability of 18.2 percentage points of indicating a low WTP. Women tend to show a lower WTP than men (p < 0.10). Having been diagnosed with chronic illnesses is associated with greater WTP. The probability of presenting a high WTP in these subjects is 5.0 percentage points greater, and 13.9 percentage points less of expressing a low WTP. For every unit of increase on the Likert scale in the PDRQ-9 questionnaire, which indicates higher satisfaction, the probability of presenting a WTP in the lowest range decreases 7.0 percentage points. Those persons who have not completed any level of education present a higher probability of expressing a lower WTP than those who have secondary studies (22.2 percentage points higher). There is no difference between superior and secondary studies. Family income level is also associated positively with a higher WTP. This effect is observed at the threshold of
J. Martín-Fernández et al. / Health Policy 94 (2010) 266–272
269
Table 1 Characteristics of the subjects participating in the study. Mean (CI 95%) Age Sex (male/female) Nationality (Spanish/other) Other insurance (yes/no) Chronic illness (yes/no) VAS-EuroQol-5D Visits to the doctor per year Hospital admissions (yes/no)
Median (IQ range)
57.3 (56.0–58.7)
Percentages over total
57.0 (45.0–70.0) 36.6%/63.4% 89.6%/10.4% 23.5%/76.5% 71.2%/28.8%
64.4 (62.5–66.4) 15.3 (14.3–16.4)
50.0 (60.0–80.0) 13 (7–20) 20.0%/80.0%
Time in making appointment Same day One day Two days Three days More than three days
26.6% 33.0% 19.1% 5.1% 16.2%
Waiting time at visit Less than 15 min Between 16 and 30 min Between 31 and 60 min More than 1 h
70.5% 23.1% 5.5% 0.9%
Relation with family physician (1 worst, 5 best possible) Number of risky behaviors
4.4 (4.3–4.5) 0.5 (0.4–0.6)
4.8 (4.0–5.0) 0.0 (0.0–1.0)
Education level Illiterate No education Primary education Secondary education Superior education
0.9% 20.4% 37.9% 25.5% 15.3%
Social group Managers, directors Intermediate positions Skilled non-manual worker Skilled manual worker Partially skilled manual worker Unskilled manual worker
12.9% 16.0% 13.5% 40.8% 2.7% 14.3%
Adjusted family income (D )
1288.8 (1200.0–1377.6)
966.6 (682.1–1591.5)
CI 95%: Confidence interval 95%; IQ range: interquartíle range (percentile 25 to percentile 75); VAS-EuroQol-5D: visual analog scale of EuroQol-5D questionnaire. D 1000, where there is a higher inclination to refer to a high or mid-high WTP and a lower inclination to refer to a low WTP. As far as the model’s non-explicative variables, these were nationality, having private insurance or not, number of visits, waiting time before the visit and not communicating risky behavior. The incorporation of the abovementioned values for the “zero-answers” signifies passing from a sample of 417 to 451 individuals. The joint significance of the model improves, and the estimated coefficients remain relatively constant. Only gender and hospital admission lose significance, while the value of the estimated coefficients for the significant variables is very similar in both models.
4. Discussion Despite monetary cost being non-existent for the user of visits to the family physician in the Spanish National Health System, the latter has a clear perception of the economic value of the service received, which is expressed in a WTP of around D 18, approximately 2% of average adjusted family disposable income. The value of D 18 for each visit
represents approximately 60% of the value of the rates the public health system charges other insurers for repeated visits to a primary care physician, which range between D 30 and D 39 [23,24]. The difference does not appear too great taking into account that the second figure refers to the price and the first to the value attributed and considering the complete lack of experience of users with respect to the real cost of the service, given that co-payment for the care of the family physician does not exist in the Spanish Health System. The WTP declared in this paper is in the range of the WTP for dependency insurance in our country (about D 18/month of 1999, was an acceptable price for 40% of those who accepted paying for the concept) [25]; or the WTP for a prostate cancer screening test, about $15 of 2006 [9]; or the WTP for all the care required by a common illness like a cold, D 29 of 2005 [5]. It is important to note that this study is conducted in a context of high consumption of this health resource, as in 2005 almost 63% of the population census in the Community of Madrid utilized primary care services [2]. These circumstances signal that the value of the WTP offered is quite realistic, despite there being no direct payment and the existence of a situation of relative overuse,
270
J. Martín-Fernández et al. / Health Policy 94 (2010) 266–272
Table 2 Ordered probit model not including answers “zero”. Variables
Age 50–65 years Age > 65 years Female Foreigner Chronic illness Hospital admissions 10–19 visits/year >20 visits/year Waiting time 15–30 min Waiting time > 30 min Private insurance Satisfaction One risky behavior More risky behaviors No education Primary education Superior education AFI D 600–1000 AFI D 1000–1500 AFI D 1500–3000 AFI > D 3000 Probability (Y)
WTP > D 40
WTP: D 20–40
WTP: D 11–19
WTP: D 0–10
dy/dx
Std. err.
dy/dx
Std. err.
dy/dx
Std. err.
dy/dx
Std. err.
0.0016 −0.0639 −0.0358 −0.0300 0.0496 −0.0311 0.0092 0.0160 0.0235 0.0474 0.0255 0.0290 −0.0063 −0.0197 −0.0666 −0.0489 −0.0168 0.0440 0.1265 0.0686 0.1408
0.0223 0.0213 0.0200 0.0212 0.0178 0.0168 0.0198 0.0229 0.0220 0.0444 0.0231 0.0101 0.0181 0.0246 0.0191 0.0209 0.0224 0.0279 0.0482 0.0360 0.0759
0.0026 −0.1212 −0.0556 −0.0583 0.0933 −0.0583 0.0150 0.0254 0.0360 0.0619 0.0389 0.0483 −0.0108 −0.0364 −0.1443 −0.0859 −0.0302 0.0656 0.1292 0.0924 0.1226
0.0367 0.0428 0.0289 0.0478 0.0361 0.0348 0.0317 0.0349 0.0308 0.0453 0.0324 0.0168 0.0310 0.0501 0.0485 0.0378 0.0431 0.0370 0.0306 0.0389 0.0312
−0.0004 0.0033 0.0105 0.0016 −0.0040 0.0032 −0.0024 −0.0045 −0.0073 −0.0192 −0.0079 −0.0070 0.0014 0.0025 −0.0116 0.0080 0.0027 −0.0141 −0.0598 −0.0259 −0.0721
0.0055 0.0093 0.0078 0.0059 0.0069 0.0042 0.0055 0.0074 0.0084 0.0232 0.0089 0.0042 0.0038 0.0026 0.0175 0.0063 0.0027 0.0116 0.0285 0.0178 0.0452
−0.0038 0.1818 0.0809 0.0867 −0.1390 0.0862 −0.0218 −0.0369 −0.0522 −0.0901 −0.0565 −0.0704 0.0157 0.0536 0.2224 0.1268 0.0443 −0.0955 −0.1959 −0.1351 −0.1912
0.0535 0.0657 0.0416 0.0727 0.0544 0.0519 0.0460 0.0505 0.0445 0.0665 0.0467 0.0236 0.0453 0.0749 0.0788 0.0558 0.0638 0.0536 0.0484 0.0570 0.0593
0.0763
0.3080
0.3163
0.2994
Characteristics of the model: N 417, LR Chi2 84.11, p < 0.0001, pseudo-R2 0.0768. In bold p < 0.05. In italics p < 0.10; AFI: adjusted family income.
without being able to contribute to the debate on the need or not of establishing systems of co-payment to finance health systems [26,27]. A clear perception exists of the value of the service, without valuing the preferences with respect to the mode of financing (by taxes alone or accompanied by co-payments). These preferences, moreover, could not be deduced from exclusive view of users of the system, given that the value of use versus non-use would be over-represented [16].
A discussion of the potential problems of the method chosen is called for, which relate fundamentally to the scenario presented and the formulation of risk, payment vehicle chosen, and administration of the survey [28]. Given that the description of the scenario may have implications for the WTP expressed [29], we chose as realistic a description as possible, making reference to a good immediately obtained. The payment card format causes
Table 3 Ordered probit model assigning value to answers “zero”. Variables
WTP > 40D
WTP: 20–40D
WTP: 11–19D
WTP: 0–10D
dy/dx
Std. err.
dy/dx
Std. err.
dy/dx
Std. err.
dy/dx
Std. err.
Age 50–65 years Age > 65 years Female Foreigner Chronic illness Hospital admissions 10–19 visits/year >20 visits/year Waiting time 15–30 min Waiting time > 30 min Private insurance Satisfaction One risky behavior More risky behaviors No education Primary education Superior education AFI D 600–1000 AFI D 1000–1500 AFI D 1500–3000 AFI > D 3000
−0.0001 −0.0664 −0.0238 −0.0253 0.0456 −0.0254 0.0036 0.0117 0.0193 0.0362 0.0278 0.0254 −0.0002 −0.0105 −0.0664 −0.0388 −0.0180 0.0405 0.1121 0.0661 0.1373
0.0192 0.0194 0.0168 0.0189 0.0154 0.0150 0.0170 0.0200 0.0188 0.0381 0.0210 0.0091 0.0163 0.0231 0.0169 0.0184 0.0194 0.0245 0.0431 0.0334 0.0718
−0.0001 −0.1373 −0.0421 −0.0537 0.0948 −0.0519 0.0066 0.0209 0.0332 0.0551 0.0466 0.0470 −0.0003 −0.0206 −0.1572 −0.0755 −0.0362 0.0677 0.1329 0.0982 0.1339
0.0354 0.0402 0.0283 0.0457 0.0339 0.0335 0.0310 0.0345 0.0301 0.0477 0.0318 0.0287 0.0301 0.0476 0.0442 0.0369 0.0419 0.0366 0.0330 0.0393 0.0339
<0.0001 −0.0073 0.0041 −0.0024 0.0036 −0.0008 −0.0005 −0.0019 −0.0039 −0.0113 −0.0063 −0.0032 < 0.0001 0.0005 −0.0270 0.0014 <0.0001 −0.0091 −0.0481 −0.0208 −0.0682
0.0024 0.0118 0.0047 0.0080 0.0082 0.0048 0.0025 0.0043 0.0055 0.0174 0.0073 0.0232 0.0020 0.0018 0.0210 0.0054 0.0037 0.0088 0.0256 0.0160 0.0442
0.0002 0.2110 0.0617 0.0813 −0.1440 0.0781 −0.0097 −0.0307 −0.0486 −0.0800 −0.0681 −0.0693 0.0005 0.0307 0.2506 0.1130 0.0541 −0.0992 −0.1969 −0.1435 −0.2030
0.0522 0.0638 0.0411 0.0713 0.0525 0.0512 0.0456 0.0504 0.0437 0.0684 0.0458 0.0235 0.0444 0.0717 0.0755 0.0553 0.0637 0.0530 0.0492 0.0569 0.0593
Probability (Y)
0.0676
0.2868 2
0.3296 2
0.3160
Characteristics of the model: N 451, LR Chi 100.18, p < 0.0001, pseudo-R 0.0852. In bold p < 0.05. In italics p < 0.10; AFI: adjusted family income.
J. Martín-Fernández et al. / Health Policy 94 (2010) 266–272
the user to behave as if he or she would be in a setting in which the same product was being sold at different prices [30]. Its suitability, and the advantages and inconveniences compared with other formats and other methods of estimating the WTP have been widely discussed [28,31,32], but it is a commonly accepted tool. To minimize any possible bias that the range presented in the card might introduce [28,31,33], we asked a question in two phases, the first with a wide range and the second attempting to determine the WTP. As recommended in the principal guides, the survey was administered by a trained interviewer [34], outside the health care process, immediately after the visit, which may favor the response rate. To evaluate the validity of the response we then studied the explicative variables in the framework of an election process that situates the user as a rational consumer. WTP was expected to be higher in those subjects with higher economic capacity, with greater need and with a higher perception of quality of the care received. As expected, rise in family income is associated with a higher probability of increasing the WTP, especially above the threshold of D 1000, and a positive elasticity can be intuited with respect to income depending on demand for the service studied. Education level was more explicative than social group in the model chosen, although there is a high correlation between social class, level of education and economic capacity [35]. Need is expressed several ways. The existence of chronic illnesses, as in other studies [5,9,10], is related to higher WTP. However, no association with the number of visits to the physician in the last year is revealed. The relation between this variable and health need may not be immediate in a services-provider system in which the direct cost is zero. That the number of visits depends on other factors, besides objective need, in our health system, has already described [36]. The relation between WTP and any indicator of quality perceived in the service evaluated is crucial. The personal style of the physician and the capacity to involve the patient in decision-making have been found to be characteristics for which the patient expressed a higher WTP in primary care [37]. Here, these qualities in the perception of a better-quality relation with the family physician are synthesized. Age and gender should not have a predetermined influence in the theoretical framework. We observed that women and those older than 65 showed a lower WTP. Women are known to visit the family physician more frequently and the same occurs with the older years in life, independently of morbidity [36]. In both cases there is lower disposable income, which is congruent with previous results, which had described that, although women and older-aged persons do not express differences in there WTP for primary care services, they do present greater elasticity to price increases [10]. In relation to age, the absence of pharmaceutical co-payment to pensioners (generally older than 65) means that interpretation of the result of the age variable must be taken carefully. We found no relationship between WTP and the measurement of accessibility to the service. In other papers
271
higher WTP had been observed with greater accessibility [37], but in this case 60% of patients had an appointment on the same or following day, and 93.6% waited less than 30 min to be attended. In this framework, in which the clinical history of two thirds of the population census reveals reasons for visits in one year [2], and there is no significant delay in making an appointment, it may not be easy to evaluate the weight of accessibility in WTP. Some types of bias, such as strategic and hypothetical, are inherent to this methodology. Sometimes a service is overvalued before it actually has to be paid for. This can happen when the good is unknown (ex-ante studies) [31,38], or because there is the idea that the services will not really have to be paid for. It cannot be said that this has not occurred in this case, but it is worth noting that a different WTP has not been observed in subjects who do pay for other kinds of insurance. Another aspect that deserves consideration is the expression “zero-answers.” The question posed can show some rejection in a context of services without cost at time of use. During the field work it was pointed out that the possibility of the establishment of co-payment was not being researched, which produced a relatively low rate of “zeroanswers with respect to other studies that were looking for the declared value of health services [25,39,40]. The “zero-answers” characteristics may serve to draw a profile of the patient unwilling to buy the service in a hypothetical market, but this was not the objective of the paper. Assigning the value given by “similar subjects” to “zero-answers” does not substantially change the explicative capacity of the variables discussed above, confirming the solidity of the model. 5. Conclusion This paper reveals the attribution of economic value that users of primary care give to care received from the family physician, in a framework in which the monetary cost per visit is zero. It also shows the relation of this economic value with individual characteristics such as personal economic capacity, need and satisfaction with the care received. Knowing what determines the valuation users make of health services, which can guide health managers in care-quality incentivation policies, as well as the study of different dimensions of quality associated with a better perception by users and relating the characteristics of service providers and setting with user satisfaction, are lines of work that deserve to be explored. Conflict of interest None of the authors have any competing interests. Acknowledgements We are grateful to the patients, physicians and board of directors of the 2nd, 6th, 8th, 9th and 11th Primary Health Care Areas in Madrid (Spain), where the work was conducted. Funding: This work has been supported by the Fondo de Investigación Sanitaria, Grant Number 070514, Plan
272
J. Martín-Fernández et al. / Health Policy 94 (2010) 266–272
Nacional de Investigación Científica, Desarrollo e Innovación Tecnológica e Instituto de Salud Carlos III. References ˜ [1] Sánchez Martínez FI, Abellán Perpinán JM, Martínez Pérez JE. How should health and healthcare priorities be set and evaluated? Prioritization methods and regional disparities, 2008 SESPAS report. Gac Sanit 2008;(Suppl. 1):126–36. [2] Dirección General de Salud Pública y Alimentación. Informe 2007 del estado de salud de la población de la comunidad de Madrid. Available at: http://www.madrid.org/cs/Satellite?cid=1142404413760 &language=es&pagename=PortalSalud%2FPage%2FPTSA pintar ContenidoFinal&vest=1156329914003. [consulted on September 24, 2009]. [3] Fusté J, Séculi E, Brugulat P, Medina A, Juncà S. Public healthcare coverage versus double healthcare coverage: what are the differences? Gac Sanit 2005;19:15–21. [4] Olsen JA, Smith RD. Theory versus practice: a review of ‘willingnessto-pay’ in health and health care. Health Econ 2001;10:39–52. [5] Yasunaga H, Ide H, Imamura T, Ohe K. Willingness to pay for health care services in common cold, retinal detachment, and myocardiac infarction: an internet survey in Japan. BMC Health Serv Res 2006;6:12. [6] Busch S, Falba T, Duchovny N, Jofre-Bonet M, O’Malley S, Sindelar J. Value to smokers of improved cessation products: evidence from a willingness-to-pay survey. Nicotine Tob Res 2004;6:631–9. [7] Yasunaga H, Ide H, Imamura T, Ohe K. Analysis of factors affecting willingness to pay for cardiovascular disease-related medical services. Int Heart J 2006;47:273–86. [8] Phillips KA, Van Bebber S, Marshall D, Walsh J, Thabane L. A review of studies examining stated preferences for cancer screening. Prev Chronic Dis 2006;3:A75. [9] Yasunaga H. Willingness to pay for mass screening for prostate cancer: a contingent valuation survey. Int J Urol 2008;15:102–5. [10] Mataria A, Giacaman R, Khatib R, Moatti JP. Impoverishment and patients’ “willingness” and “ability” to pay for improving the quality of health care in Palestine: an assessment using the contingent valuation method. Health Policy 2006;75:312–28. [11] Borghi J, Jan S. Measuring the benefits of health promotion programmes: application of the contingent valuation method. Health Policy 2008;87:235–48. [12] O’Shea E, Gannon B, Kennelly B. Eliciting preferences for resource allocation in mental health care in Ireland. Health Policy 2008;88:359–70. [13] Pereira AG, Pearson SD. Patient attitudes toward continuity of care. Arch Intern Med 2003;163:909–12. [14] Kao AC, Krasny AJ. Preferences for patient cost sharing among medicare beneficiaries after HMO plan withdrawals. J Gen Intern Med 2002;17:446–50. ˜ de Salud. Dirección General de Atención al [15] Servicio Madrileno Paciente. Evaluación de la satisfacción de los usuarios de los Servicios de la Asistencia Sanitaria Pública de la Comunidad de Madrid. Atención Primaria. Available at: http://www.madrid.org/cs/Satellite?c= PTSA Multimedia FA&cid=1142537158446&pagename=PortalSalud% 2FPTSA Multimedia FA%2FPTSA documentoWebeditpro. [consulted on October 1, 2009]. [16] Borghi J. Aggregation rules for cost–benefit analysis: a health economics perspective. Health Econ 2008;17:863–75. [17] O’Brien B, Gafni A. When do the “dollars” make sense? Toward a conceptual framework for contingent valuation studies in health care. Med Decis Making 1996;16:288–99. [18] Bush K, Kivlahan DR, McDonell MB, Fihn SD, Bradley KA. The AUDIT alcohol consumption questions (AUDIT-C): an effective brief screening test for problem drinking. Ambulatory Care Quality Improvement Project (ACQUIP). Alcohol Use Disorders Identification Test. Arch Intern Med 1998;158:1789–95.
[19] Mingote J, Moreno-Jiménez B, Rodríguez-Carvajal R, Gálvez M, Ruiz-López P. Psychometric validation of the Spanish version of the Patient–Doctor Relationship Questionnaire (PDRQ). Actas Esp Psiquiatr 2009;37:94–100. [20] Domingo Salvany A, Marcos Alonso J. Proposal of an indicator of “social class” based on the occupation. Gac Sanit 1989;3:320–6. [21] Drumond MF, O’Brien B, Stoddart GL, Torrance GW. Methods for the economic evaluation of health care programmes. 2nd ed. New York: Oxford University Press; 1997. p. 226. [22] Green WH, Hensher DA. A latent class model for discrete choice analysis: contrasts with mixed logit. Transport Res Part B: Method 2003;37(8):681–98. [23] Decreto 164/2008, de 17 de julio, por el que se establecen las tarifas de los servicios sanitarios prestados en los centros dependientes del Servicio Gallego de Salud y en las fundaciones públicas sanitarias. Diario Oficial de Galicia, no. 150, 5 de agosto de 2008, p. 15060-74. [24] ORDEN 629/2009, de 31 de agosto, de la Consejería de Sanidad, por la que se fijan los precios públicos por la prestación de los servicios y actividades de naturaleza sanitaria de la red de centros de la Comunidad de Madrid. B.O.C.M no. 215, de 15 de septiembre de 2009, p. 15–35. [25] Costa-Font J, Rovira-Forns J. Who is willing to pay for long-term care insurance in Catalonia? Health Policy 2008;86:72–84. [26] Lostao L, Regidor E, Geyer S, Aïach P. Patient cost sharing and social inequalities in access to health care in three western European countries. Soc Sci Med 2007;65:367–76. [27] Rückert IM, Böcken J, Mielck A. Are German patients burdened by the practice charge for physician visits (‘Praxisgebuehr’)? A crosssectional analysis of socio-economic and health related factors. BMC Health Serv Res 2008;12(8):232. [28] Smith RD. Construction of the contingent valuation market in health care: a critical assessment. Health Econ 2003;12:609–28. [29] Smith RD. Contingent valuation in health care: does it matter how the ‘good’ is described? Health Econ 2008;17:607–17. [30] Thomas R, Donaldson C, Torgerson D. Who answers ‘willingness to pay’ questions? J Health Serv Res Policy 2000;5:7–11. [31] Klose T. Contingent valuation methods in health care. Health Policy 1999;47:97–123. [32] Ryan M, Watson V. Comparing welfare estimates from payment card contingent valuation and discrete choice experiments. Health Econ 2009;18:389–401. [33] Heinzen RR, Bridges JF. Comparison of four contingent valuation methods to estimate the economic value of a pneumococcal vaccine in Bangladesh. Int J Technol Assess Health Care 2008;24:481–7. [34] Arrow K, Solow R, Portney PR, Leamer EE, Radner R, Schuman H. Report of the NOAA panel of contingent valuation. Federal Register 1993;58:4601–14. [35] Domínguez-Berjón F, Borrell C, Rodríguez-Sanz M, Pastor V. The usefulness of area-based socioeconomic measures to monitor social inequalities in health in Southern Europe. Eur J Public Health 2006;16:54–61. [36] Sáez M. Factors conditioning primary care services utilization. Empirical evidence and methodological inconsistencies. Gac Sanit 2003;17:412–9. [37] Cheraghi-Sohi S, Hole AR, Mead N, McDonald R, Whalley D, Bower P, et al. What patients want from primary care consultations: a discrete choice experiment to identify patients’ priorities. Ann Fam Med 2008;6:107–15. [38] Bergmo TS, Wangberg SC. Patients’ willingness to pay for electronic communication with their general practitioner. Eur J Health Econ 2007;8:105–10. [39] Bärnighausen T, Liu Y, Zhang X, Sauerborn R. Willingness to pay for social health insurance among informal sector workers in Wuhan, China: a contingent valuation study. BMC Health Services Res 2007;7:114. [40] Yeung R, Abdullah AS, McGhee SM, Hedley AJ. Willingness to pay for preventive travel health measures among Hong Kong Chinese residents. J Travel Med 2005;12:66–71.
Contents lists available at ScienceDirect
Health Policy journal homepage: www.elsevier.com/locate/healthpol
Perception of the economic value of primary care services: A willingness to pay study Jesús Martín-Fernández a,∗ , Tomás Gómez-Gascón b , Juan Oliva-Moreno c , María Isabel del Cura-González d , Julia Domínguez-Bidagor e , Milagros Beamud-Lagos f , Teresa Sanz-Cuesta g a
San Martín de Valdeiglesias Health Center, 1st La Bola St., 28680 San Martin de Valdeiglesias, Madrid Health Service, Madrid, Spain Puerta Bonita II Health Center, Madrid Health Service. Faculty of Medicine, Complutense University of Madrid, Spain c Economic Analysis Department, Castilla la Mancha University, CIBERESP-CIBER in Epidemiology and Public Health, Spain d Research Unit, 9th Primary Care Area, Madrid Health Service. Rey Juan Carlos I University, Madrid, Madrid, Spain e Health Education, Quality Unit, 11th Primary Care Area, Madrid Health Service, Madrid, Spain f Research Unit, 11th Primary Care Area, Madrid Health Service. University Nursing School, Complutense University of Madrid, Madrid, Spain g Preventive Medicine and Public Health, Research Unit, 9th Primary Care Area, Madrid Health Service, Spain b
a r t i c l e
i n f o
Keywords: Primary health care Economics Patient satisfaction Health services research Health services needs and demand
a b s t r a c t Objective: Identify the economic value the user attributes to the visit to the family physician, in a setting of a National Health System, by the Willingness to Pay (WTP) expressed. Methods: Economic evaluation study, by the contingent valuation method. Questions were asked about WTP using a payment card format. Interviews were conducted with 451 subjects, in areas with different socioeconomic characteristics. An ordered probit was used to evaluate model’s validity. Results: Median WTP expressed was D 18 (interquartile range D 8–28), not including “zeroanswers” of thirty-four subjects (7.5%). This value represents 2% of average adjusted family incomes. Patients with higher incomes or with chronic illnesses presented a probability of 5–14 percentage points of expressing a high WTP. For every point of increase of patient satisfaction, the probability of presenting a WTP in the lowest range decreases 7.0 percentage points. Subjects with a low education level and those older than 65 expressed a lower WTP. Accessibility, risk perception, nationality and having private insurance were not related to the WTP expressed. Conclusions: Users of primary care have a clear perception of the economic value of care received from the family physician, even in a framework of providing services financed by taxes and without cost at the moment of use. This value increases in subjects with higher incomes, with greater need for care, or more satisfied. © 2009 Elsevier Ireland Ltd. All rights reserved.
1. Introduction The permanent debate on the proper allocation of economic resources to provide effective health care policies is also present in Spain [1]. The Spanish National Health Sys-
∗ Corresponding author. Tel.: +34 918610273; fax: +34 918612320. E-mail address: [email protected] (J. Martín-Fernández). 0168-8510/$ – see front matter © 2009 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.healthpol.2009.11.001
tem is financed fundamentally by taxes, and health care services have no monetary cost at all. Only prescribed pharmaceuticals are paid for, to a maximum of 40% of their total value, except pensioners and some persons with chronic illnesses who receive them free. In this context a high consumption of resources is observed, also in primary care (PC). In 2005 there were thirteen million PC visits in the Community of Madrid (CM) corresponding to 3,757,000 different people, almost 63% of the population census [2].
J. Martín-Fernández et al. / Health Policy 94 (2010) 266–272
Different uses of this public health system, and especially of PC, can be related to several factors in addition to health needs [3]. Therefore other elements, in addition to the need stated must be considered in health planning. Among these variables we found the user’s perception of value of the service to be fundamental. In the absence of a market, whereby society’s valuation of a health care service program could be expressed, the problem is: how do we measure perception of value in monetary terms? A possible applicable measure is the willingness to pay (WTP) for these services, calculated by the Contingent Valuation (CV) Method. Willingness to pay is a ‘theoretically correct’ approach, to the service valuation because of its foundation in welfare economics. In addition, WTP imposes no restrictions to which attributes of a program people are allowed to value, and allows comparability across other public services [4]. This methodology has been employed in different situations in health care that include the evaluation of the approach to various pathologies [5], the valuation of health technologies and preventive or curative treatments [4,6]. CV has also been used to study the “need” for certain services in prevalent illnesses [7], or preferences for preventive programs [8,9]. In the primary care setting, CV has been used to estimate the willingness to defray the costs of improvements in health care in systems in development [10], to evaluate health promotion programs [11], mental health care [12], the acceptability of payment for the continuation of care by the family physician [13], or the extension of certain kinds of health coverage [14]. Primary care services provided by the Public Health System in the CM are satisfactorily perceived [15], but the valuation that people make of them in scales comparable with other services considered necessary in a “welfare” society are unknown. Establishing the economic value attributed to this service could be valuable in the design of future health care policies [16], as well as to comparatively estimate the utility attributed to the service. The purpose of this paper is to study the economic value attributed by the user of the public health care system in the CM (Spain) to the service received at a visit to the family physician, by the WTP expressed. 2. Material and methods A cross-sectional study was designed that used the Contingent Valuation Method, in a framework of compensatory variation of the classification by O’Brien and Gafni [17]. Interviews were conducted with 451 subjects, over 35 years, who had just visited a family physician. These were randomly selected from the lists of visits at two rural and four urban centers of areas with different socioeconomic characteristics in the Community of Madrid (Spain). 2.1. Scenario and payment format Field work was conducted by a single trained interviewer, between December 2007 and March 2008. The interviewer explained that the study was an attempt to attribute value to a service whose price is not established and it was not at any time pursuing the estab-
267
lishment of direct payments for it. The following question was then asked: “Imagine for a moment that there is no public health care system. Suppose you have a health problem similar to what brought you here today to the visit and you have to be attended by the physician who received you today, but you must pay for this care directly. How much money would you be willing to pay for this visit?” The response was provided by a “double payment card” method. The first payment card only contained three values: less than D 20; between D 20 and D 40; more than D 40. These cutoff points were chosen following the valuation of the distribution of responses obtained in a pilot with twenty patients. The second card was fitted to the WTP expressed and contained the values: 0, 1–5, 6–10, 11–15, 16–20, 21–25, 26–30, 31–40, 41–50, 51–60, 61–70, 71–80, 81–90 and more than D 90. 2.2. Characteristics of the subjects and the service evaluated The subject’s characteristics collected included age, sex and place of birth, state of health, accessibility to the service, other types of insurance, risk perception, relation with the physician, socioeconomic situation and family income. The evaluation of health need was made noting the existence of chronic pathologies, hospital admissions in the last year, and number of visits to the family physician in the last year. The visual–analogical scale of EuroQol-5D was used to evaluate the subjective perception of the state of health. Accessibility to the service was studied by the time necessary to obtain an appointment and by the waiting time from time of appointment to the visit. To evaluate risk perception questions were asked about the existence of the following risky behavior: smoking habit, driving vehicles without obeying safety measures, risky sexual behavior, non-observation of job-safety measures and excessive alcohol consumption (AUDIT-C questionnaire) [18]. The relation with the physician was evaluated with the PDRQ questionnaire, adapted to Spanish [19]. To evaluate the socioeconomic situation questions were asked about the highest level of education completed (illiterate, no education, primary, secondary, superior), by job (class I for the most skilled job to V for the least skilled) [20], and by current employment situation. Adjusted family income was calculated adding up all the income of the family unit and weighting by family size N (average income = family income/N0.4 ). The exponent 0.4 reflects that the family economy is an economy of scale, in which each new member needs an increasingly smaller marginal contribution of resources to maintain the buying power of the family group. 2.3. Analysis The WTP declared was transformed into a continuous variable, applying to each interval its “middle value”, for the descriptive study of the distribution of the variable.
268
J. Martín-Fernández et al. / Health Policy 94 (2010) 266–272
“Zero-answers” were analyzed separately. We believe that a zero valuation of the service does not correspond to the real perception of the person interviewed. In fact, there are other options approaching zero that enable expressing more realistically a low-value perception. We therefore assume that these responses represented a rejection to the question posed (“zero protests”), or the inability to pay, and were therefore treated separately. The theoretical validity of the WTP was studied analyzing its relation to need, satisfaction, socioeconomic position and average income based on the conceptual framework of the “welfare theory” [17,21]. We therefore expected a higher WTP in people with more health needs, with greater economic resources or better social situation (level of superior studies, higher skilled work) and in those more satisfied with the service received. Given that the dependent variable is not observable a model was constructed in which the latent variable y* is not observed (it is the “real valuation” of the individual), and depends upon a linear combination of explicative variables. An ordered probit was utilized, a model that enables grasping the ordered nature and intensity of the dependent variable [22]. Thus, the higher the value of our latent variable, the higher the probability will be that the individual will state a higher category in the valuation scale. The dependent variable acquires a value of 1–4 depending on the intensity of the order. Thus y = 1 corresponds to the lowest valuation, WTP less than D 10, y = 2 represents a mid-low WTP, between D 11 and 20, y = 3 a mid-high WTP, between D 21 and 40, and y = 4 the highest WTP, more than de D 40. We can express our model as follows: y∗ = ˇ0 + ˇ1 x1 + · · · + ˇn xn + ε where x1 , xn are the explicative variables, ˇ0 is a constant, ˇ1 . . .ˇn are the coefficients that adjust the model and ε is a coefficient that represents a random error with normal distribution. The explicative variables were chosen depending on the requirements of the model, eliminating those that presented colinearity (they could be partially explained by the linear combination of other selected variables). The model was initially constructed without accounting for “zero-answers.” Subsequently, there is a sensitivity analysis with all the subjects, substituting the “zeros” by their “estimated WTP.” These values were predicted based on the responses observed in individuals who revealed their WTP. Thus, the “zero values” were substituted for each individual by the average values of the WTP observed in the similar individuals whose WTP > 0. To reduce possible biases, a matching was established between each individual with WTP = 0 and more similar individuals with WTP > 0. Age, gender, level of studies, and level of income as control variables were taken into consideration. 3. Results An invitation to participate was extended to 487 subjects; thirty-six (7.4%) did not accept, twenty-seven for not having time to be interviewed, two for lack of inter-
est in the subject, two who gave no reason and in five cases for other reasons. Of those who did not accept participating ten were men and twenty-six women, with an average age of 51.1 years (57.3 years for the group included p = 0.013) and no differences were found with respect to the rest of the group as far as area of residence, number of chronic illnesses or frequency of utilization of the service. Table 1 shows the characteristics of the 451 subjects that finally consented to participate in the study. The mean value of the WTP was D 19.6 (CI 95%, D 18.0–21.2), for those subjects who expressed values greater than zero. Median distribution was D 18 (interquartile range D 8–28). Looking at the average number of visits in a year, this value represents approximately 2% of average adjusted family incomes. Thirty-four subjects (7.5%) expressed a WTP equal to zero. The profile of these subjects is a person of middle age somewhat older than the rest of the group (63.4 vs 56.8 years; p = 0.011), with a better perception of his/her quality of life (EVA-EuroQol-5D 74.5 vs 63.6; p = 0.004), who makes the same use of the system (same number of admissions, visits, and waiting time), with higher aggregation of illiterate or uneducated subjects (44.1% vs 19.4%, p < 0.001) and with lower adjusted family income (D 1011.6 vs 1311.4, p = 0.080). Table 2 shows the results of the empirical model without “zero-answers.” The reference category is a male, between 35 and 49 years of age, Spanish nationality, without chronic illnesses, not hospitalized in the prior year, with less than ten annual visits, with a wait at the last visit of less than 15 min, without private health insurance, without recognized risky behavior, with secondary studies and a monthly adjusted family income level ≤D 600. Patient satisfaction with the relationship with the physician was introduced as a continuous variable. Table 3 shows the results incorporating the estimates effected on the WTP by “zero-answers.” These results are expressed in marginal terms, that is, in comparison with the category of reference. In the case of satisfaction, the result refers to a change of one unit on its measurement scale. Age is a variable associated with WTP. While there are no differences in intermediate ages, those older than 65 have a higher probability of 18.2 percentage points of indicating a low WTP. Women tend to show a lower WTP than men (p < 0.10). Having been diagnosed with chronic illnesses is associated with greater WTP. The probability of presenting a high WTP in these subjects is 5.0 percentage points greater, and 13.9 percentage points less of expressing a low WTP. For every unit of increase on the Likert scale in the PDRQ-9 questionnaire, which indicates higher satisfaction, the probability of presenting a WTP in the lowest range decreases 7.0 percentage points. Those persons who have not completed any level of education present a higher probability of expressing a lower WTP than those who have secondary studies (22.2 percentage points higher). There is no difference between superior and secondary studies. Family income level is also associated positively with a higher WTP. This effect is observed at the threshold of
J. Martín-Fernández et al. / Health Policy 94 (2010) 266–272
269
Table 1 Characteristics of the subjects participating in the study. Mean (CI 95%) Age Sex (male/female) Nationality (Spanish/other) Other insurance (yes/no) Chronic illness (yes/no) VAS-EuroQol-5D Visits to the doctor per year Hospital admissions (yes/no)
Median (IQ range)
57.3 (56.0–58.7)
Percentages over total
57.0 (45.0–70.0) 36.6%/63.4% 89.6%/10.4% 23.5%/76.5% 71.2%/28.8%
64.4 (62.5–66.4) 15.3 (14.3–16.4)
50.0 (60.0–80.0) 13 (7–20) 20.0%/80.0%
Time in making appointment Same day One day Two days Three days More than three days
26.6% 33.0% 19.1% 5.1% 16.2%
Waiting time at visit Less than 15 min Between 16 and 30 min Between 31 and 60 min More than 1 h
70.5% 23.1% 5.5% 0.9%
Relation with family physician (1 worst, 5 best possible) Number of risky behaviors
4.4 (4.3–4.5) 0.5 (0.4–0.6)
4.8 (4.0–5.0) 0.0 (0.0–1.0)
Education level Illiterate No education Primary education Secondary education Superior education
0.9% 20.4% 37.9% 25.5% 15.3%
Social group Managers, directors Intermediate positions Skilled non-manual worker Skilled manual worker Partially skilled manual worker Unskilled manual worker
12.9% 16.0% 13.5% 40.8% 2.7% 14.3%
Adjusted family income (D )
1288.8 (1200.0–1377.6)
966.6 (682.1–1591.5)
CI 95%: Confidence interval 95%; IQ range: interquartíle range (percentile 25 to percentile 75); VAS-EuroQol-5D: visual analog scale of EuroQol-5D questionnaire. D 1000, where there is a higher inclination to refer to a high or mid-high WTP and a lower inclination to refer to a low WTP. As far as the model’s non-explicative variables, these were nationality, having private insurance or not, number of visits, waiting time before the visit and not communicating risky behavior. The incorporation of the abovementioned values for the “zero-answers” signifies passing from a sample of 417 to 451 individuals. The joint significance of the model improves, and the estimated coefficients remain relatively constant. Only gender and hospital admission lose significance, while the value of the estimated coefficients for the significant variables is very similar in both models.
4. Discussion Despite monetary cost being non-existent for the user of visits to the family physician in the Spanish National Health System, the latter has a clear perception of the economic value of the service received, which is expressed in a WTP of around D 18, approximately 2% of average adjusted family disposable income. The value of D 18 for each visit
represents approximately 60% of the value of the rates the public health system charges other insurers for repeated visits to a primary care physician, which range between D 30 and D 39 [23,24]. The difference does not appear too great taking into account that the second figure refers to the price and the first to the value attributed and considering the complete lack of experience of users with respect to the real cost of the service, given that co-payment for the care of the family physician does not exist in the Spanish Health System. The WTP declared in this paper is in the range of the WTP for dependency insurance in our country (about D 18/month of 1999, was an acceptable price for 40% of those who accepted paying for the concept) [25]; or the WTP for a prostate cancer screening test, about $15 of 2006 [9]; or the WTP for all the care required by a common illness like a cold, D 29 of 2005 [5]. It is important to note that this study is conducted in a context of high consumption of this health resource, as in 2005 almost 63% of the population census in the Community of Madrid utilized primary care services [2]. These circumstances signal that the value of the WTP offered is quite realistic, despite there being no direct payment and the existence of a situation of relative overuse,
270
J. Martín-Fernández et al. / Health Policy 94 (2010) 266–272
Table 2 Ordered probit model not including answers “zero”. Variables
Age 50–65 years Age > 65 years Female Foreigner Chronic illness Hospital admissions 10–19 visits/year >20 visits/year Waiting time 15–30 min Waiting time > 30 min Private insurance Satisfaction One risky behavior More risky behaviors No education Primary education Superior education AFI D 600–1000 AFI D 1000–1500 AFI D 1500–3000 AFI > D 3000 Probability (Y)
WTP > D 40
WTP: D 20–40
WTP: D 11–19
WTP: D 0–10
dy/dx
Std. err.
dy/dx
Std. err.
dy/dx
Std. err.
dy/dx
Std. err.
0.0016 −0.0639 −0.0358 −0.0300 0.0496 −0.0311 0.0092 0.0160 0.0235 0.0474 0.0255 0.0290 −0.0063 −0.0197 −0.0666 −0.0489 −0.0168 0.0440 0.1265 0.0686 0.1408
0.0223 0.0213 0.0200 0.0212 0.0178 0.0168 0.0198 0.0229 0.0220 0.0444 0.0231 0.0101 0.0181 0.0246 0.0191 0.0209 0.0224 0.0279 0.0482 0.0360 0.0759
0.0026 −0.1212 −0.0556 −0.0583 0.0933 −0.0583 0.0150 0.0254 0.0360 0.0619 0.0389 0.0483 −0.0108 −0.0364 −0.1443 −0.0859 −0.0302 0.0656 0.1292 0.0924 0.1226
0.0367 0.0428 0.0289 0.0478 0.0361 0.0348 0.0317 0.0349 0.0308 0.0453 0.0324 0.0168 0.0310 0.0501 0.0485 0.0378 0.0431 0.0370 0.0306 0.0389 0.0312
−0.0004 0.0033 0.0105 0.0016 −0.0040 0.0032 −0.0024 −0.0045 −0.0073 −0.0192 −0.0079 −0.0070 0.0014 0.0025 −0.0116 0.0080 0.0027 −0.0141 −0.0598 −0.0259 −0.0721
0.0055 0.0093 0.0078 0.0059 0.0069 0.0042 0.0055 0.0074 0.0084 0.0232 0.0089 0.0042 0.0038 0.0026 0.0175 0.0063 0.0027 0.0116 0.0285 0.0178 0.0452
−0.0038 0.1818 0.0809 0.0867 −0.1390 0.0862 −0.0218 −0.0369 −0.0522 −0.0901 −0.0565 −0.0704 0.0157 0.0536 0.2224 0.1268 0.0443 −0.0955 −0.1959 −0.1351 −0.1912
0.0535 0.0657 0.0416 0.0727 0.0544 0.0519 0.0460 0.0505 0.0445 0.0665 0.0467 0.0236 0.0453 0.0749 0.0788 0.0558 0.0638 0.0536 0.0484 0.0570 0.0593
0.0763
0.3080
0.3163
0.2994
Characteristics of the model: N 417, LR Chi2 84.11, p < 0.0001, pseudo-R2 0.0768. In bold p < 0.05. In italics p < 0.10; AFI: adjusted family income.
without being able to contribute to the debate on the need or not of establishing systems of co-payment to finance health systems [26,27]. A clear perception exists of the value of the service, without valuing the preferences with respect to the mode of financing (by taxes alone or accompanied by co-payments). These preferences, moreover, could not be deduced from exclusive view of users of the system, given that the value of use versus non-use would be over-represented [16].
A discussion of the potential problems of the method chosen is called for, which relate fundamentally to the scenario presented and the formulation of risk, payment vehicle chosen, and administration of the survey [28]. Given that the description of the scenario may have implications for the WTP expressed [29], we chose as realistic a description as possible, making reference to a good immediately obtained. The payment card format causes
Table 3 Ordered probit model assigning value to answers “zero”. Variables
WTP > 40D
WTP: 20–40D
WTP: 11–19D
WTP: 0–10D
dy/dx
Std. err.
dy/dx
Std. err.
dy/dx
Std. err.
dy/dx
Std. err.
Age 50–65 years Age > 65 years Female Foreigner Chronic illness Hospital admissions 10–19 visits/year >20 visits/year Waiting time 15–30 min Waiting time > 30 min Private insurance Satisfaction One risky behavior More risky behaviors No education Primary education Superior education AFI D 600–1000 AFI D 1000–1500 AFI D 1500–3000 AFI > D 3000
−0.0001 −0.0664 −0.0238 −0.0253 0.0456 −0.0254 0.0036 0.0117 0.0193 0.0362 0.0278 0.0254 −0.0002 −0.0105 −0.0664 −0.0388 −0.0180 0.0405 0.1121 0.0661 0.1373
0.0192 0.0194 0.0168 0.0189 0.0154 0.0150 0.0170 0.0200 0.0188 0.0381 0.0210 0.0091 0.0163 0.0231 0.0169 0.0184 0.0194 0.0245 0.0431 0.0334 0.0718
−0.0001 −0.1373 −0.0421 −0.0537 0.0948 −0.0519 0.0066 0.0209 0.0332 0.0551 0.0466 0.0470 −0.0003 −0.0206 −0.1572 −0.0755 −0.0362 0.0677 0.1329 0.0982 0.1339
0.0354 0.0402 0.0283 0.0457 0.0339 0.0335 0.0310 0.0345 0.0301 0.0477 0.0318 0.0287 0.0301 0.0476 0.0442 0.0369 0.0419 0.0366 0.0330 0.0393 0.0339
<0.0001 −0.0073 0.0041 −0.0024 0.0036 −0.0008 −0.0005 −0.0019 −0.0039 −0.0113 −0.0063 −0.0032 < 0.0001 0.0005 −0.0270 0.0014 <0.0001 −0.0091 −0.0481 −0.0208 −0.0682
0.0024 0.0118 0.0047 0.0080 0.0082 0.0048 0.0025 0.0043 0.0055 0.0174 0.0073 0.0232 0.0020 0.0018 0.0210 0.0054 0.0037 0.0088 0.0256 0.0160 0.0442
0.0002 0.2110 0.0617 0.0813 −0.1440 0.0781 −0.0097 −0.0307 −0.0486 −0.0800 −0.0681 −0.0693 0.0005 0.0307 0.2506 0.1130 0.0541 −0.0992 −0.1969 −0.1435 −0.2030
0.0522 0.0638 0.0411 0.0713 0.0525 0.0512 0.0456 0.0504 0.0437 0.0684 0.0458 0.0235 0.0444 0.0717 0.0755 0.0553 0.0637 0.0530 0.0492 0.0569 0.0593
Probability (Y)
0.0676
0.2868 2
0.3296 2
0.3160
Characteristics of the model: N 451, LR Chi 100.18, p < 0.0001, pseudo-R 0.0852. In bold p < 0.05. In italics p < 0.10; AFI: adjusted family income.
J. Martín-Fernández et al. / Health Policy 94 (2010) 266–272
the user to behave as if he or she would be in a setting in which the same product was being sold at different prices [30]. Its suitability, and the advantages and inconveniences compared with other formats and other methods of estimating the WTP have been widely discussed [28,31,32], but it is a commonly accepted tool. To minimize any possible bias that the range presented in the card might introduce [28,31,33], we asked a question in two phases, the first with a wide range and the second attempting to determine the WTP. As recommended in the principal guides, the survey was administered by a trained interviewer [34], outside the health care process, immediately after the visit, which may favor the response rate. To evaluate the validity of the response we then studied the explicative variables in the framework of an election process that situates the user as a rational consumer. WTP was expected to be higher in those subjects with higher economic capacity, with greater need and with a higher perception of quality of the care received. As expected, rise in family income is associated with a higher probability of increasing the WTP, especially above the threshold of D 1000, and a positive elasticity can be intuited with respect to income depending on demand for the service studied. Education level was more explicative than social group in the model chosen, although there is a high correlation between social class, level of education and economic capacity [35]. Need is expressed several ways. The existence of chronic illnesses, as in other studies [5,9,10], is related to higher WTP. However, no association with the number of visits to the physician in the last year is revealed. The relation between this variable and health need may not be immediate in a services-provider system in which the direct cost is zero. That the number of visits depends on other factors, besides objective need, in our health system, has already described [36]. The relation between WTP and any indicator of quality perceived in the service evaluated is crucial. The personal style of the physician and the capacity to involve the patient in decision-making have been found to be characteristics for which the patient expressed a higher WTP in primary care [37]. Here, these qualities in the perception of a better-quality relation with the family physician are synthesized. Age and gender should not have a predetermined influence in the theoretical framework. We observed that women and those older than 65 showed a lower WTP. Women are known to visit the family physician more frequently and the same occurs with the older years in life, independently of morbidity [36]. In both cases there is lower disposable income, which is congruent with previous results, which had described that, although women and older-aged persons do not express differences in there WTP for primary care services, they do present greater elasticity to price increases [10]. In relation to age, the absence of pharmaceutical co-payment to pensioners (generally older than 65) means that interpretation of the result of the age variable must be taken carefully. We found no relationship between WTP and the measurement of accessibility to the service. In other papers
271
higher WTP had been observed with greater accessibility [37], but in this case 60% of patients had an appointment on the same or following day, and 93.6% waited less than 30 min to be attended. In this framework, in which the clinical history of two thirds of the population census reveals reasons for visits in one year [2], and there is no significant delay in making an appointment, it may not be easy to evaluate the weight of accessibility in WTP. Some types of bias, such as strategic and hypothetical, are inherent to this methodology. Sometimes a service is overvalued before it actually has to be paid for. This can happen when the good is unknown (ex-ante studies) [31,38], or because there is the idea that the services will not really have to be paid for. It cannot be said that this has not occurred in this case, but it is worth noting that a different WTP has not been observed in subjects who do pay for other kinds of insurance. Another aspect that deserves consideration is the expression “zero-answers.” The question posed can show some rejection in a context of services without cost at time of use. During the field work it was pointed out that the possibility of the establishment of co-payment was not being researched, which produced a relatively low rate of “zeroanswers with respect to other studies that were looking for the declared value of health services [25,39,40]. The “zero-answers” characteristics may serve to draw a profile of the patient unwilling to buy the service in a hypothetical market, but this was not the objective of the paper. Assigning the value given by “similar subjects” to “zero-answers” does not substantially change the explicative capacity of the variables discussed above, confirming the solidity of the model. 5. Conclusion This paper reveals the attribution of economic value that users of primary care give to care received from the family physician, in a framework in which the monetary cost per visit is zero. It also shows the relation of this economic value with individual characteristics such as personal economic capacity, need and satisfaction with the care received. Knowing what determines the valuation users make of health services, which can guide health managers in care-quality incentivation policies, as well as the study of different dimensions of quality associated with a better perception by users and relating the characteristics of service providers and setting with user satisfaction, are lines of work that deserve to be explored. Conflict of interest None of the authors have any competing interests. Acknowledgements We are grateful to the patients, physicians and board of directors of the 2nd, 6th, 8th, 9th and 11th Primary Health Care Areas in Madrid (Spain), where the work was conducted. Funding: This work has been supported by the Fondo de Investigación Sanitaria, Grant Number 070514, Plan
272
J. Martín-Fernández et al. / Health Policy 94 (2010) 266–272
Nacional de Investigación Científica, Desarrollo e Innovación Tecnológica e Instituto de Salud Carlos III. References ˜ [1] Sánchez Martínez FI, Abellán Perpinán JM, Martínez Pérez JE. How should health and healthcare priorities be set and evaluated? Prioritization methods and regional disparities, 2008 SESPAS report. Gac Sanit 2008;(Suppl. 1):126–36. [2] Dirección General de Salud Pública y Alimentación. Informe 2007 del estado de salud de la población de la comunidad de Madrid. Available at: http://www.madrid.org/cs/Satellite?cid=1142404413760 &language=es&pagename=PortalSalud%2FPage%2FPTSA pintar ContenidoFinal&vest=1156329914003. [consulted on September 24, 2009]. [3] Fusté J, Séculi E, Brugulat P, Medina A, Juncà S. Public healthcare coverage versus double healthcare coverage: what are the differences? Gac Sanit 2005;19:15–21. [4] Olsen JA, Smith RD. Theory versus practice: a review of ‘willingnessto-pay’ in health and health care. Health Econ 2001;10:39–52. [5] Yasunaga H, Ide H, Imamura T, Ohe K. Willingness to pay for health care services in common cold, retinal detachment, and myocardiac infarction: an internet survey in Japan. BMC Health Serv Res 2006;6:12. [6] Busch S, Falba T, Duchovny N, Jofre-Bonet M, O’Malley S, Sindelar J. Value to smokers of improved cessation products: evidence from a willingness-to-pay survey. Nicotine Tob Res 2004;6:631–9. [7] Yasunaga H, Ide H, Imamura T, Ohe K. Analysis of factors affecting willingness to pay for cardiovascular disease-related medical services. Int Heart J 2006;47:273–86. [8] Phillips KA, Van Bebber S, Marshall D, Walsh J, Thabane L. A review of studies examining stated preferences for cancer screening. Prev Chronic Dis 2006;3:A75. [9] Yasunaga H. Willingness to pay for mass screening for prostate cancer: a contingent valuation survey. Int J Urol 2008;15:102–5. [10] Mataria A, Giacaman R, Khatib R, Moatti JP. Impoverishment and patients’ “willingness” and “ability” to pay for improving the quality of health care in Palestine: an assessment using the contingent valuation method. Health Policy 2006;75:312–28. [11] Borghi J, Jan S. Measuring the benefits of health promotion programmes: application of the contingent valuation method. Health Policy 2008;87:235–48. [12] O’Shea E, Gannon B, Kennelly B. Eliciting preferences for resource allocation in mental health care in Ireland. Health Policy 2008;88:359–70. [13] Pereira AG, Pearson SD. Patient attitudes toward continuity of care. Arch Intern Med 2003;163:909–12. [14] Kao AC, Krasny AJ. Preferences for patient cost sharing among medicare beneficiaries after HMO plan withdrawals. J Gen Intern Med 2002;17:446–50. ˜ de Salud. Dirección General de Atención al [15] Servicio Madrileno Paciente. Evaluación de la satisfacción de los usuarios de los Servicios de la Asistencia Sanitaria Pública de la Comunidad de Madrid. Atención Primaria. Available at: http://www.madrid.org/cs/Satellite?c= PTSA Multimedia FA&cid=1142537158446&pagename=PortalSalud% 2FPTSA Multimedia FA%2FPTSA documentoWebeditpro. [consulted on October 1, 2009]. [16] Borghi J. Aggregation rules for cost–benefit analysis: a health economics perspective. Health Econ 2008;17:863–75. [17] O’Brien B, Gafni A. When do the “dollars” make sense? Toward a conceptual framework for contingent valuation studies in health care. Med Decis Making 1996;16:288–99. [18] Bush K, Kivlahan DR, McDonell MB, Fihn SD, Bradley KA. The AUDIT alcohol consumption questions (AUDIT-C): an effective brief screening test for problem drinking. Ambulatory Care Quality Improvement Project (ACQUIP). Alcohol Use Disorders Identification Test. Arch Intern Med 1998;158:1789–95.
[19] Mingote J, Moreno-Jiménez B, Rodríguez-Carvajal R, Gálvez M, Ruiz-López P. Psychometric validation of the Spanish version of the Patient–Doctor Relationship Questionnaire (PDRQ). Actas Esp Psiquiatr 2009;37:94–100. [20] Domingo Salvany A, Marcos Alonso J. Proposal of an indicator of “social class” based on the occupation. Gac Sanit 1989;3:320–6. [21] Drumond MF, O’Brien B, Stoddart GL, Torrance GW. Methods for the economic evaluation of health care programmes. 2nd ed. New York: Oxford University Press; 1997. p. 226. [22] Green WH, Hensher DA. A latent class model for discrete choice analysis: contrasts with mixed logit. Transport Res Part B: Method 2003;37(8):681–98. [23] Decreto 164/2008, de 17 de julio, por el que se establecen las tarifas de los servicios sanitarios prestados en los centros dependientes del Servicio Gallego de Salud y en las fundaciones públicas sanitarias. Diario Oficial de Galicia, no. 150, 5 de agosto de 2008, p. 15060-74. [24] ORDEN 629/2009, de 31 de agosto, de la Consejería de Sanidad, por la que se fijan los precios públicos por la prestación de los servicios y actividades de naturaleza sanitaria de la red de centros de la Comunidad de Madrid. B.O.C.M no. 215, de 15 de septiembre de 2009, p. 15–35. [25] Costa-Font J, Rovira-Forns J. Who is willing to pay for long-term care insurance in Catalonia? Health Policy 2008;86:72–84. [26] Lostao L, Regidor E, Geyer S, Aïach P. Patient cost sharing and social inequalities in access to health care in three western European countries. Soc Sci Med 2007;65:367–76. [27] Rückert IM, Böcken J, Mielck A. Are German patients burdened by the practice charge for physician visits (‘Praxisgebuehr’)? A crosssectional analysis of socio-economic and health related factors. BMC Health Serv Res 2008;12(8):232. [28] Smith RD. Construction of the contingent valuation market in health care: a critical assessment. Health Econ 2003;12:609–28. [29] Smith RD. Contingent valuation in health care: does it matter how the ‘good’ is described? Health Econ 2008;17:607–17. [30] Thomas R, Donaldson C, Torgerson D. Who answers ‘willingness to pay’ questions? J Health Serv Res Policy 2000;5:7–11. [31] Klose T. Contingent valuation methods in health care. Health Policy 1999;47:97–123. [32] Ryan M, Watson V. Comparing welfare estimates from payment card contingent valuation and discrete choice experiments. Health Econ 2009;18:389–401. [33] Heinzen RR, Bridges JF. Comparison of four contingent valuation methods to estimate the economic value of a pneumococcal vaccine in Bangladesh. Int J Technol Assess Health Care 2008;24:481–7. [34] Arrow K, Solow R, Portney PR, Leamer EE, Radner R, Schuman H. Report of the NOAA panel of contingent valuation. Federal Register 1993;58:4601–14. [35] Domínguez-Berjón F, Borrell C, Rodríguez-Sanz M, Pastor V. The usefulness of area-based socioeconomic measures to monitor social inequalities in health in Southern Europe. Eur J Public Health 2006;16:54–61. [36] Sáez M. Factors conditioning primary care services utilization. Empirical evidence and methodological inconsistencies. Gac Sanit 2003;17:412–9. [37] Cheraghi-Sohi S, Hole AR, Mead N, McDonald R, Whalley D, Bower P, et al. What patients want from primary care consultations: a discrete choice experiment to identify patients’ priorities. Ann Fam Med 2008;6:107–15. [38] Bergmo TS, Wangberg SC. Patients’ willingness to pay for electronic communication with their general practitioner. Eur J Health Econ 2007;8:105–10. [39] Bärnighausen T, Liu Y, Zhang X, Sauerborn R. Willingness to pay for social health insurance among informal sector workers in Wuhan, China: a contingent valuation study. BMC Health Services Res 2007;7:114. [40] Yeung R, Abdullah AS, McGhee SM, Hedley AJ. Willingness to pay for preventive travel health measures among Hong Kong Chinese residents. J Travel Med 2005;12:66–71.