1 Introduction

The role of the private sector in developing better health systems and improving healthcare worldwide is increasingly recognized [2]. Galicia – a region in northwest Spain forming part of the Galicia-North Portugal cross-border Euroregion – is no exception, since 24 of the 36 hospitals that make up the Galician healthcare system (SERGAS) are private hospitals, accounting for 2,285 of a total of 9,809 beds and 73 of a total of 282 operating theatres [3, 4]. Moreover, 12 of the 187.6 specialist care doctors and 30 of the 340.5 nursing assistants per 100,000 inhabitants in Galicia are employed in private hospitals, while an annual average of 12.6% of diagnostic tests and 25.5% of surgical interventions are undertaken in private hospitals. Private hospitals thus play a significant role in SERGAS, despite the prevalence of a subsector with no profit motive and financed by agents who are not the main users of its services.

A crucial role in guaranteeing quality healthcare is played by advanced medical technologies, which are complex, costly, and usually limited to a few locations where they are managed by specialists [5]. Diagnostic technologies, in particular, are very important in guaranteeing healthcare quality, as they are essential to improving diagnostic precision and therapeutic decision-making [6]. Key diagnostic technologies are computed tomography (CT) and magnetic resonance imaging (MRI) scanners [7,8,9,10,11, 12]. Since their commercialization from the mid-1970s (CT) and the mid-1980s (MRI) [13], both have revolutionized medical diagnostics.Footnote 1 MRI scanners are used to evaluate blood vessels, breasts, and major organs, while CT scanners are used for clinical evaluation of the pelvis, thorax, abdomen, and colon, for the detection of tumours, pulmonary embolisms (CT angiography), abdominal aortic aneurysms, and spinal lesions, and for cardiology examinations, while CT enhancements (such as dose-modulated acquisition and iterative reconstruction algorithms) reduce X-ray dosage, improve hospital efficiency and clinical effectiveness, and reduce costs.

Key technology assessment agenciesFootnote 2 list CT and MRI as priority technologies, while a survey asking 387 internists to rank 30 medical innovations in terms of importance resulted in MRI and CT scanners being placed first and second, in that order [14]. Use of both technologies has increased rapidly – by an annual average of 5.6% and 6.8% for MRI and CT technologies, respectively, in the period 2014–2019 [15]. World Health Organization (WHO) objectives regarding systematic assessment of medical technologies refer to appropriate use, safety and acceptability, and practical and economic effectiveness [16], and highlight the usefulness of CT and MRI technologies for diagnosing severe injuries and substituting for more invasive and higher-risk approaches.

Nonetheless, in the interest of monitoring and controlling health spending [17, 18], identifying how and why these new technologies are adopted is timely and relevant. First, new healthcare needs require new diagnostic and therapeutic procedures that need to be analysed in terms of their use by hospitals. Second, new healthcare technologies are a key contributor to increased health spending in recent decades [19],indeed, several studies refer to a medical arms race, with advanced medical technologies deployed as a competitive strategy to attract patients and specialist physicians [20, 21]. Third, expected profits cannot be the only criterion that determines medical technology uptake by private hospitals [22].Footnote 3 Finally, a clear adoption pattern by private hospitals seems not to exist, as confirmed by annual variations in the number of CT and MRI scans (see Fig. 1 below).

Fig. 1
figure 1

Annual variation in the number of CT and MRI scanners in private hospitals (2006–2019)

Full size image

Our paper contributes to the literature on health technology evaluation with several findings. First, CT and MRI technology uptake by private hospitals as a substitute for more invasive and higher-risk medical interventions is mainly explained by hospital size, measured as the number of beds. Size can be understood as a proxy for hospital prestige and quality, since it is widely perceived that healthcare quality is better in larger hospitals, as confirmed by research into three overlapping areas that are near-synonymous with care in large institutions, namely specialization, centralization, and volume. Since each of these care areas produces significant profits, recognition and perceptions of prestige are on the side of larger hospitals with more beds [23,24,25]Footnote 4 Second, payment policies influence CT and MRI technology uptake by private hospitals, as already suggested by [26]. Because public sector Contracts influence the profitability of private hospitals, variable Contracts can be understood as a proxy for the profitability associated with new technologies.Footnote 5 Third, population density does not appear to significantly impact on CT and MRI technology uptake by private hospitals, suggesting that demand for CT and MRI scans is induced in private hospitals. Finally, the adoption rate of a new technology is slowed down when the innovation replaces older equipment in which professionals and institutions have invested substantial resources (time and money). Resistance to abandoning an obsolete or near-obsolete technology may be explained as a natural defence of past efforts to incorporate and master the technology, and often those who advocate innovation are also reluctant to abandon older technologies. This is especially relevant to MRI technology uptake, which seems to be affected by the fact that it is supposed to replace nuclear medicine technologies, e.g., gamma cameras and single-photon emission computerized tomography (SPECT) scanners.

Our findings can help decision-makers better plan investments in new medical technologies and to forestall possible premature diffusion without adequate knowledge of true efficacy, given that evaluation of costly technologies is crucial to monitoring and controlling health spending [17, 18].

Thus, starting from how private hospitals decide whether or not to acquire advanced medical technologies (probably based on expected profits, enhanced reputation, and/or competitive advantage), we explore other factors influencing the adoption decision. The uptake of advanced medical technology can potentially transform a fundamental healthcare resource into a symbol of power and prestige, or even an end in itself [27, 28]. Another issue is the technological imperative itself [29], which may stimulate technological innovation beyond genuine healthcare needs of diagnosing a disease and providing a cure [30, 31].

The remainder of the article is laid out as follows. Section 2 describes our data and analysis methods, Sect. 3 describes our results, Sect. 4 discusses the results and their implications, and Sect. 5 concludes.

2 Data and methods

Our research scope was Galician private hospitals equipped with CT and MRI technologies and the study period was 2006–2019. Data on CT and MRI scanners were provided by the Spanish Ministry of Health, while data on hospital inputs were obtained from specialist healthcare and in-patient healthcare establishment statistics as reported by SERGAS and hospitals and published by the Spanish Ministry of Health [32]. Measured as annual hospital inputs were number of beds, number of CT and MRI scanners, and number of other diagnostic and therapeutic medical technologies (gamma cameras, lithotripsy units, linear accelerator units, SPECT scanners, and haemodynamic facilities, henceforth referred to as other high-tech equipment). Sociodemographic data was obtained from the Galician Institute of Statistics (IGE)Footnote 6 for each hospital’s catchment area, as follows: number of inhabitants, population density (inhabitants per km2), and age as a factor explaining healthcare demand (population aged > 64 years).

Regarding adoption, associations with the different variables were analysed by means of binary logistic regression for CT and MRI, as follows:

$$\begin {aligned}CT\;Adoption= & \,{\beta }_{0}+{\beta }_{1}Beds+{\beta }_{2}Contract+{\beta }_{3}Inhabitants\\&+{\beta }_{4}Populationabove64years+{\beta }_{5}Populationdensity\\&+{\beta }_{6}Otherhightech+\varepsilon\end {aligned}$$
(1)
$$\begin {aligned}MRI\;Adoption= & \, {\beta }_{0}+{\beta }_{1}Beds+{\beta }_{2}Contract+{\beta }_{3}Inhabitants\\&+{\beta }_{4}Populationabove64years+{\beta }_{5}Populationdensity\\&+{\beta }_{6}Otherhightech+{\beta }_{7}CT+\varepsilon\end {aligned}$$
(2)

In both equations, the explanation variables refer to the number of hospital beds, the existence or otherwise of a Contract for CT or MRI services, the number of inhabitants, population density, the number of inhabitants aged above 64 years, and the existence or otherwise of other high-tech equipment, respectively. Note that CT technology predates MRI technology and that our sample has no case of adoption of MRI before CT technology. Thus, in Eq. (1), we do not include the prior existence of MRI technology as a possible explanatory variable for CT acquisition, and in Eq. (2), CT reflects whether or not a hospital has CT technology; this is because CT technology adoption/non-adoption is an additional explanatory variable for MRI uptake, given that CT technology predates and therefore may influence adoption of MRI technology [33]. Note that the Contract and other high-tech equipment variables in both equations are categorical variables.

The sample data used to perform the regression analyses were panel data [1] and statistical significance was established at p \(\le\) 0.05. Goodness-of-fit of the logistic regression models was evaluated using Pearson’s chi-square statistic, which compares observed and expected frequencies in a binomial scenario. Also calculated were odds ratio (OR) values. Given that the OR is around 1 for continuous independent variables, OR > 1 indicated that, as the continuous variable increases, a technology is more likely to be adopted, while the reverse is true for OR < 1, and OR = 1 indicates no change.

In order to overcome the risk of overfitting due to the small sample obtained (although it represents 100% of the universe of private hospitals in the region under study), the logistic regression was performed with a bootstrapping procedure. Bootstrapping is an internal validation method and is ideal for small samples [34]. The aim of bootstrapping is to replicate this procedure by sampling within the study population, with replacement to create more learning subgroups [35]. In each bootstrap sample, data are analysed as in the original study sample, repeating each step of model development, including predictor selection strategies. The bootstrap sample repetition process has been performed 1000 times.

3 Descriptive data and results

The population of the health areas in which the private hospitals are located has an average of 229,835 inhabitants (range 5,817–413,996), and the population is relatively aged, with inhabitants aged above 64 years numbering 35,546 on average (range 1,288–67,957). The population is also quite dispersed, with an average density of 412 inhabitants per km2 (range 57–1,005).

The sample included the 24 private hospitals providing healthcare services to SERGAS, accounting, in 2019, for 17 CT and 12 MRI scanners, representing 30% of all CT and 46.3% of all MRI units in SERGAS. Among the private hospitals, 49% have CT technology, 32% have MRI technology, and 31% have at least one other high-tech equipment item. The private hospitals account for 2,040 beds in total, for an average of 85 beds per hospital (maximum 656 beds in a single hospital).

Figure 1, which depicts the evolution in the number of CT and MRI units installed in private hospitals during the study period, suggests that uptake does not seem to follow any particular pattern.

Tables 1 and 2 summarize results for CT and MRI technology uptake by private hospitals according to the logistic models proposed in Eqs. (1) and (2), respectively. Cox and Snell (C&S) R2 is a generalized coefficient of determination that estimates how much variance in the dependent variable is explained by the independent variables. The Hosmer and Lemeshow (H&L) test is a data goodness-of-fit test of logistic regression model, with p > 0.05 confirming the null hypothesis that the proposed model fits the observed data.

Table 1 Technology adoption by private hospitals: CT adoption as the dependent variable
Full size table
Table 2 Technology adoption by private hospitals: MRI adoption as the dependent variable
Full size table

Regarding CT adoption (Table 1), C&S R2 indicates that the independent variables explain 41.4% of the variance and the H&L test result of p > 0.05 confirms that the proposed model fits the observed data. The overall model classification rate for CT adoption is 80.97%. From the OR values, it can be inferred that the variable that most influences CT adoption by private hospitals, although by a small margin, is hospital size, proxied by the number of beds. Contrariwise, and rather surprisingly, none of the sociodemographic variables other than age above 64 years were significant in influencing CT technology uptake. The availability of other high-tech equipment and the existence of a Contract both reflect an inverse probability of acquiring a CT scanner, estimated as 2.66 and 8 times less likely, respectively.

Regarding MRI adoption (Table 2), Cox and Snell R2 indicates that the independent variables explain 44.2% of the variance and the Hosmer and Lemeshow test result of p > 0.05 confirms that the model fits the data well. The overall model classification rate for MRI adoption is 84.73%. The OR values indicate that, as happens with CT technology adoption, hospital size is the variable that most influences MRI adoption. All sociodemographic variables, with the exception of population density, influence the decision to adopt MRI technology – positively in the case of the number of inhabitants, and negatively in the case of the population aged > 64 years. The availability of high-tech equipment is not statistically significant, which would suggest that MRI is a complementary rather than a substitute technology. Availability of CT technology reflects an inverse probability of MRI scanner acquisition (estimated as 14.5 times less likely), suggesting that CT and MRI are substitute rather than complementary technologies. Finally, Contract funding of MRI scans reflects an inverse probability of acquiring an MRI scanner (estimated as 15.62 times less likely).

4 Discussion

Our results suggest that size, i.e., the number of beds, is the most influential variable in CT and MRI technology uptake by private hospitals. The expected profits of these hospitals is therefore not the only criterion determining their choice of medical technologies [22].

This finding is in accordance with previous literature on medical technology uptake. Folland et al. [36] reported greater technology uptake by larger hospitals, while Russell [37] found that larger hospitals were more likely to adopt new technologies, with both bed numbers and specialists and family physicians per bed positively influencing uptake. Likewise, Abedini et al. [38] concluded that the main factors influencing CT and MRI technology uptake were the number of beds and hospital location, Jae-Seok Hong [39] showed that the number of CT units reflected bed numbers and especially location in larger cities, and Hall [40] found that greater use of diagnostic technologies was associated with greater population density, more doctors per capita, a higher ratio of specialists to family physicians, and a higher percentage of doctors involved in teaching. Similar conclusions have also been reached by [41] regarding the influence of hospital size on CT and MRI technology uptake, by [42] on CT and MRI technology uptake in German hospitals, and by Zhu et al. [43] on medical technology uptake in US private hospitals. The fact that size is the most influential factor in technology uptake also corroborates the result obtained by [44] in a multiple regression study.

Matsumoto et al. [45] reported that geographical distribution of CT and MRI technology depends on spatial competition derived from market forces, arguing that scanners are first installed in the largest cities offering the highest expectations in terms of use, income, and profits. However, as more hospitals acquire the technology, competition for patients in smaller cities also grows, leading to uptake in increasingly less populated areas. From another perspective, Hofmann and Gransjøen [46] have studied geographic variations in outpatient CT and MRI studies and conclude that they affect quality of care and equity in access to technology; especially for the private sector.

The duration of obsolete technology was not found to affect CT and MRI technology uptake to a statistically significant degree. This finding contradicts [47], who concluded that the adoption rate of new technology is reduced if the innovation is a replacement for equipment with a relatively lengthy remaining useful life, and also contradicts [48], who suggested that there is a natural defence of past investments in time, effort, and money made by professionals and institutions in incorporating and mastering the corresponding technologies. Similar results are reported by [49] in a study on the relationship between adoption and abandonment of medical technologies.

Beyond financial incentives and hospital size, [50, 51] argue that acquiring an innovation enhances the prestige of a hospital and its physicians, and, as suggested by [52], can additionally be perceived as an ind icator of care quality. However, our study shows that the availability of other high-tech equipment does not significantly influence uptake of CT and MRI technologies. The explanation may be found in game theory applied to the hospital sector [53]. In competitive advantage terms, any hospital that innovates potentially obtains more referrals and more demand for its services and so forces other hospitals to adopt the new technology. While this situation removes the competitive advantage for both hospitals, the hospital that does not innovate will be left at a competitive disadvantage. Rei and Hirotaka [54] report a positive correlation between CT and MRI technology uptake and a competitive private hospital market. However, [42], in their study of private and public German hospitals, show that CT and MRI technology uptake decreases as nearby hospitals adopt the innovations, and [55] report similar results for cancer treatment technology uptake.

The high and positive coefficients that we found for hospital size, inhabitants, and population density in relation to CT and MRI technology uptake by private hospitals corroborate [56], who reported that increased diffusion of medical technologies was related to private but not public insurance, with insurance coverage understood as an indicator of medical technology diffusion in the private sector. Hillman and Schwartz [57], however, reported that a system that remunerates services on the basis of prospective funding, i.e., the system used for Contracts, implies greater investment recovery risk than a system based on retrospective funding, i.e., the system used for public hospitals. Indeed, according to [58], the main factors influencing the decision to adopt MRI technology are acquisition cost and the investment recovery rate. However, for [38], the insurance companies that ultimately fund CT and MRI scanners play a limited role in diffusion of these technologies. In this respect, authors such as [26] argue that a retrospective financing system, based on costs incurred, generates incentives for the adoption (and use) of new technologies by guaranteeing cost recovery.

Our study also points to the lack of influence of public subsidies in CT and MRI private sector technology uptake, corroborating the findings of both [42] regarding adoption of CT, MRI, and positron emission tomography (PET) in German private hospitals, and Sorenson et al. [59] on the role of hospital payments in new medical technology uptake in different countries. Hillman et al. [57] point out that the prospective funding system, as used by SERGAS for Contractcs, implies a greater risk for investment recovery than the retrospective funding system used for public hospitals. The fact that cost recovery is guaranteed generates incentives for the adoption and use of new technologies, and also explains the competitive relationship between public and private healthcare.

The non-significance of sociodemographic variables may support the hypothesis that medical technology demand is induced from the private sector, as suggested by Calcott [60]. Our finding that population density is significant, yet has no influence on the probability of a private hospital acquiring a CT or MRI scanner, corroborates [61], who suggested that the non-significance of sociodemographic variables in models of demand for health services may point to the existence of induced demand. Sandoval et al. [41] also pointed to demand induction for CT and MRI technologies, especially in small hospitals. However, our finding contradicts [62], who showed that areas of between 50,000 and 100,000 inhabitants are more likely to incur private healthcare expenditure than areas with smaller populations.

Finally, the positive value of the population density coefficient suggests that the likelihood of technology uptake by a hospital is greater, the more densely populated its catchment area. This finding corroborates [37], who observed that technology uptake is broader and faster in areas of strong population growth, reflecting market expansion, but contradicts [63], who found no significant relationship between the degree of market concentration and medical technology uptake.

5 Conclusions

We analysed the factors that influence CT and MRI technology uptake by private hospitals, using a logistic regression model to explore the impact of hospital size, various population characteristics, Contracts for the provision of services, and hospital access to other diagnostic and therapeutic technologies.

The variable that most influenced the decision by private hospitals to adopt medical technologies, whether CT or MRI, is hospital size, measured by the number of beds. Regarding CT technology uptake, demand is induced in private hospitals, other high-tech equipment (e.g., gamma camera, lithotripsy, linear accelerator, etc.) behaves as a substitute, and Contracts have a counterfactual effect. As for MRI technology uptake sociodemographic variables affect the adoption decision, positively in the case of the number of inhabitants and negatively in the case of the population aged > 64 years; the availability of a CT scanner reduces the probability of a private hospital adopting MRI technology, suggesting that MRI is a substitute rather than a complementary technology in private hospitals; Contracts have a negative influence on MRI technology uptake.

Our study is framed within the need to provide information on new medical technology trends to healthcare policy-makers so that they can address deficits in implantation and accessibility, and also evaluate costs, given the potential impact on health spending. Our research also contributes to behavioural studies of early adopters of innovations and of possiblly premature diffusion of technologies prior to full understanding of their efficacy.