Introduction

The diffusion of a technology encompasses both the decision to adopt a technology (inter-firm diffusion) and the extent at which adopting firms use the capabilities of the technology (intra-firm diffusion) [1]. But studies on the diffusion of health information technology (health IT) among primary care providers (PCPs) typically focus on whether these providers adopt electronic health record (EHR) systems rather than on the use of these systems’ capabilities across providers [2, 3]. There exist approximately 250,000 physician practices in the United States [4]. They typically use a wide range of EHR systems, from basic to advanced EHR systems [3]. The basic EHR systems replace the traditional medical charts and collect demographic and clinical data about the patient, including medical history, medications taken, and laboratory results [2, 3]. The advanced EHR systems incorporate computerized provider order entry (CPOE) capabilities that allow providers to electronically enter medical orders for patient medications and diagnostic tests [2, 3]. They also allow providers to write clinical notes and provide treatment guidelines [2, 3]. Hence, the advanced EHR systems can both complement and substitute labor for medical practices, involving clinical decision making and complex communications with pharmacies, laboratories, and the patients [5]. This combination of technological and organizational capabilities is commonly referred to as “Meaningful Use” (MU) of EHRs [5].

In an effort to improve the quality and effectiveness of healthcare through the expanded use of EHRs in the United States, the Health Information Technology for Economic and Clinical Health (HITECH) Act, enacted as part of the 2009 American Recovery and Reinvestment Act (ARRA), was signed into law [6]. From 2011 to 2021, using funds available from the HITECH Act, the Centers for Medicare and Medicaid Services (CMS) funded the Medicaid EHR Incentive Program, later renamed the Medicaid Promoting Interoperability (MPI) program. In addition to adopting a certified EHR technology, this program required that PCPs use this technology in a “meaningful” way to ensure that health information is shared and exchanged to improve patient care [7]. At the end of 2010, CMS issued the criteria relating to how eligible PCPs can demonstrate “meaningful use” of EHR technology to qualify for incentive payments [8]. These criteria fall into four general categories: vocabulary, content exchange, privacy and security, and transport. More specifically, standards were set forth with regards to how PCPs should (a) record clinical summaries and prescriptions; (b) describe clinical procedures, laboratory tests, and medications; and (b) secure the transmission of information using the internet [8]. These standards included fifteen “core set” objectives that all PCPs must achieve and 10 “menu set” objectives, of which PCPs must select 5 [8, 9]. Overall, the goals were to improve safety, quality, and efficiency of healthcare and reduce health disparities, engage patients and families in their care, improve population and public health, protect privacy and security of health information, and reduce healthcare costs [8, 9]. The MPI program was voluntary for state Medicaid agencies; but all the 50 states and the District of Columbia participated. The federal government covered about 90 percent of the cost [8].

Eligible providers included physicians, nurse practitioners, certified nurse-midwives, dentists, and physician assistants. To qualify they must meet a minimum of 30% Medicaid encounter volume threshold over a 90-day period, but pediatricians could qualify with a 20% Medicaid encounter volume. Additional requirements included being in active status, having no outstanding state or federal sanctions, and not being in a hospital setting. Providers must first register at the CMS—Registration and Attestations System (R&A). Once registered, they can apply online using the Medical Assistance Provider Incentive Repository (MAPIR) system. Each provider then received an incentive payment of $21,250 for committing to "adopt, implement or upgrade" (AIU) a certified EHR system. In practice, “adoption” consists of acquiring or purchasing a certified EHR system; “implementation” consists of installing or starting to use a certified EHR system; and “upgrading” consists of improving an existing EHR to a certified EHR system. Enrollment through AIU phases ended in 2016.

Providers received a second incentive payment of $8500 for attesting to MU; but pediatricians that enrolled with a 20–29% Medicaid encounter volume received $4250 [8, 9]. Payments for attesting to MU for the first time after AIU enrollment ended in 2018. Pre-payment controls were built into the MAPIR system to detect inaccuracies in eligibility, reporting, and payments. Provider audits that targeted suspicious and anomalous data and random audits were performed to ensure that incentive payments were made to providers that met all program requirements. Providers were not required to participate in subsequent or consecutive years. It is worth noting that eligible providers could also enroll into the MPI program simply by attesting to MU of EHRs in a 90-day reporting period, but these providers were excluded from this study. Further details on the program can be found elsewhere [8, 9].

The present study used data from the Florida—Medicaid Promoting Interoperability (FL-MPI) program to examine the impact of county-level factors on MU attestation rates among PCPs after receiving the first-year incentive payments. Its contribution to the literature is significant. Overall, the broad question posed was: which environmental factors influence the widespread use of advanced EHR functions by PCPs? The corresponding policy implications are whether tools can be developed to overcome existing county-level disparities. Nearly 9 out of 10 physician practices in the U.S. have adopted a basic EHR system [10]. But the rate of MU attestations has stalled among Medicaid providers, providing evidence of an emerging advanced use “digital divide”, contrary to the conclusion reached for hospitals [10–12]. Research indicates that the use of advanced EHR functions holds promise to address socioeconomic disparities for underserved populations [13, 14]. As health disparities are often geographically linked, it is important to obtain contextual insights about MU attestation rates among Medicaid providers.

Materials and methods

Theoretical framework

The theoretical framework for this study is the resource dependence theory (RDT), a well-developed organization theory in business management [15–17]. It offers relevant concepts to understand the diffusion of EHRs by identifying the importance of environmental factors [18–20]. RDT dictates that providers will do what they can to secure resources from their environments through a variety of actions or exchanges to reach their goals of attesting to MU of EHRs [17].

The three core constructs of RDT are environmental munificence, dynamism, and complexity [17]. Environmental munificence is concerned with the availability and accessibility of environmental resources [21, 22]. In this instance, it is tied to the availability of financial resources in the providers’ markets as they seek financial resources in the forms of direct payments from the patients and reimbursements from public and private insurers [18]. For example, providers in wealthier environments that cater to a more affluent and educated customer base might find it more beneficial to use the advanced functions of their EHR systems (patient portal; e-prescribe, etc.) to appeal to potential clients, typically more selective in their healthcare services [23]. The second core construct, environmental dynamism, reflects the rate of change in the environments, which increases the providers’ level of uncertainty [17, 22, 24]. Providers in uncertain markets must adapt quickly to survive as the need to strategize to secure resources from these markets becomes more pressuring [22, 24]. The last core construct, environmental complexity, reflects the level of competition in the markets [22, 25, 26]. In the face of competition, providers must compete for the scarce resources as well as secure their share of patients [18]. For example, they might find it beneficial to adopt a health technology in areas of high competition as they need to appeal to the patient population [18, 23].

Fig 1 depicts the theoretical model of this study. It is adapted from Kissam et al. [27] and Menachemi et al. [19] and incorporates the variables that characterize the core constructs of RDT.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Conceptual model: Resource Dependence Theory (RDT).

https://doi.org/10.1371/journal.pone.0295435.g001

Although studies that employ RDT in examining healthcare providers are increasingly common, there exists little consensus on how to operationalize the market environment in the context of healthcare. With the exception of education, all the variables used for the RDT constructs in Fig 1 above were selected using a systematic review conducted by Yeager and colleagues who reviewed twenty studies that relied on aspects of RDT to examine the healthcare market [28]. The authors summarized how the environment was measured in those studies and provided a list of variables that operationalized the RDT constructs [28]. Educational attainment, typically used in economic studies as a proxy for human capital and skills available in a market [1, 3], has also been used to operationalize environmental munificence in healthcare [27]. The characteristics of the providers are not themselves part of RDT although they might enable or hinder the providers’ ability to attest to MU [18, 29, 30]. They are thus considered as controls in this study.

Analytical approach

We conducted a 2011–2018 retrospective cohort study of 8415 providers that participated in the FL-MPI program. We narrowed our analysis to the counties that had a minimum of 10 providers to obtain enough grids for the county-level analysis [41, 42]. This reduced the number of counties from 67 to 53 counties. Because the dataset included only the dates the providers received payments for having attested to MU, it was not possible to determine the exact years of MU attestations. Hence, the unit of analysis was provider-years. Our sample included 2622, 2035, 1565, 777, 718, 698 providers that enrolled into the program, from 2011 to 2016, respectively. All the county ordinal variables were averaged over the study period. All data accessed complied with data protection regulation.

We calculated and presented descriptive statistics to characterize our sample. Next, we used one-way analysis of variance (ANOVA) to test for differences in MU attestation rates across each county variable. We then used pooled logistic regressions with generalized estimated equations (GEE) and the binomial family and logit link functions to examine the associations between MU attestations and the county-level variables, controlling for AIU phases and practices’ characteristics [43–45]. County and year dummies were also included. Using the delta methods, we estimated clustered standard errors to adjust for the non-independence of observations within counties [46–48]. In view of the large number of explanatory variables, we calculated the vector of variance inflator factors (VIF) to check for possible multicollinearity. We found that the highest VIF was below 5 and the mean VIF was 2.21, indicating that correlations between the covariates were relatively acceptable and that multi-collinearity was not a serious problem [48]. Odds ratios and corresponding 95% confidence intervals were derived using Stata 16 [49].

Results

Table 1 displays the descriptive statistics for the characteristics of the sample. The results indicate that, among the 8415 provider-years, physicians represented the largest group (52.1%), followed by nurse practitioners (23.6%), pediatricians (13.9%), dentists (8.1%), certified nurse-midwives (1.9%) and physician assistants (0.4%). Group practices were more common (63.4%) than solo practices (36.6%). For the AIU phases, we found that 36.6% of the providers in the sample adopted a certified EHR system; 28.5% implemented a certified EHR system; and 34.9% upgraded to a certified EHR system.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Descriptive statistics for the sample (N = 8415 provider-years, 2011–2018).

https://doi.org/10.1371/journal.pone.0295435.t001

Table 2 presents MU attestation rates for the overall program and by enrollment year. On the average, 41.9% of the sample achieved MU after receiving the first-year incentive. The result indicates notable decrease in MU attestation rates over time. More specifically, MU attestation rates decreased from 56.7% among the providers that enrolled in 2011 to 19.2% among those enrolled in 2016.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Meaningful use (MU) attestations rates (N = 8415 provider-years).

https://doi.org/10.1371/journal.pone.0295435.t002

For practical and computational purposes, we categorized counties as poor counties (poverty rates ≥20%) and non-poor counties (poverty rates < 20%) [50]. The other county rates were converted into three dichotomous variables corresponding to terciles (e.g., Q₁- tercile group that includes counties with the lowest rates was referred to as “low group”; Q₃-tercile group with the highest rates was the “high group”; and Q₂-tercile group was the “medium group”). The results for the bivariate analysis are presented in Table 3. They were all significant at p < .01, indicating that the MU attestation rates between the groups were not all equal. Namely, rural counties, non-poor counties, counties with higher high-school graduation rates, counties with lower HMO penetration rates, counties with larger changes in population, and counties with a lower number of PCPs per capita had higher MU attestation rates.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Unadjusted MU attestation rates of EHRs by county-level characteristics.

https://doi.org/10.1371/journal.pone.0295435.t003

Table 4 presents the multivariate results for the relationships between county variables and MU attestations. They indicate that providers located in rural counties were 80 percent more likely to attest to MU than providers in urban counties: OR = 1.80; 95% CI [1.30 2.50]. Providers located in poor counties were 32 percent less likely to attest to MU compared to providers in non-poor counties: OR = 0.68; 95% CI [0.55 0.84]. Providers located in counties with high HMO penetration rates were 34 percent less likely to attest to MU than providers in counties with low HMO penetration rates: OR = 0.66, 95% CI [0.57 0.77]. We also found that providers located in counties with moderate changes in their populations over the study period were 27% less likely to attest to MU compared to counties with lower changes in their populations: OR = 0.73, 95% CI [0.63 0.86]. Finally, the results indicate that providers located in counties with a medium number of PCPs per capita were 45 percent more likely to attest to MU compared to providers located in counties with a lower number of PCPs per capita: OR = 1.46, 95% CI [1.24 1.74].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Multivariate relationships between county-level factors and meaning use (MU) attestations of EHRs^a.

https://doi.org/10.1371/journal.pone.0295435.t004

Most of the control variables (not shown on Table 4) were significant in the multivariate logistic regressions. More specifically, compared to physicians, pediatricians were more likely to attest to MU while dentists and nurse practitioners were less likely to attest to MU. Group practices were more likely to attest to MU than solo practices. Finally, we found that providers that implemented or upgraded EHR systems were more likely to attest to MU than those that adopted EHR systems.

To check the robustness of the results, we use the approach used by Monestime et al. [30]. We divided the participants into early participants and late participants, which were subsets of providers who enrolled in 2011–2012 and in 2013–2016, respectively. The results of the logistic regressions remained robust in all the models.

Discussion

While unexpedted, the findings that providers in rural counties had higher MU attestation rates than providers in urban counties were also found in more recent studies [51, 52]. A plausible explanation is the presence of regional extension centers (RECs) in the rural counties [51]. The REC program was established by the HITECH Act and administered through the office of National Coordinator (ONC) for Health Information Technology. The program included 62 grantee organizations throughout the U.S. that promoted EHR adoption and the optimal use of EHR technology through outreach and the provision of technical assistance [53]. Florida was among the 3 states with the highest number of RECs (4) granted by the ONC [53]. The REC program prioritized small primary care practices in rural areas [53, 54]. Previous research showed that these providers had lower rates of EHR adoption than larger organizations and that they generally lacked resources to adopt and maintain EHRs [55]. The REC program was found to be very successful in engaging providers in rural areas [6, 51, 56].

The findings that a county’s poverty rate is negatively associated with MU of EHRs are consistent with those of most previous studies [57, 58]. They suggests that PCPs in these counties might face challenges to EHR diffusion related to tight operating budgets and limited capacity for integrating EHR training into the workflow [57]. But it is to note that other researchers found mixed results [59, 60]. The results also indicate that HMO penetration rate was negatively associated MU attestations. While others found mixed results [61, 62], this finding is consistent with another study conducted among a large number of physicians in Florida [63]. It is also congruent with the medical technology literature indicating that higher managed care penetrations in geographically defined healthcare markets are associated with lower adoption of new technologies [61, 64]. We also found that moderate changes in a county’s population size was associated with lower MU attestation rates. This finding is consistent with other studies indicating that uncertainty delayed the decision to pursue expensive strategies [22, 65]. Finally, we found that a medium or moderate increase in the number of physicians per capita (10,000) increased the probability of attesting to MU. This is in line with other studies in both healthcare and other industries that suggest that in the face of competition, organizations can adopt strategic decision making, become more innovative, and create competitive advantage [63, 66–68]. Hence, in competitive markets, physicians might use their EHRs to brand their services as high quality and high technology to facilitate referrals [23].

Florida ranks among the states with the highest number of Medicaid recipients (over 6.7 million recipients as of 2023) [69]. It is the third-largest state by population size (22.2 million in 2022) [70]. Florida in 2022 was 77.05% White, 17.02% Black, and 5.93% other [70]. Nearly half (45%) of working Floridians are living in or near poverty [69]. As of 2022, 30 of the Florida’s 67 counties were qualified as rural counties with about 750,000 residents [69]. But more than 1 million persons live in the rural portions of Florida’s 37 urban counties [69]. Hence, Florida’s varied socioeconomic status provides an ideal setting for contextual insights about meaningful use of EHRs among Medicaid providers that serve traditionally underserved communities.

Meaningful use of EHRs requires EHR capabilities that include, among other things, patient portals, patient-monitoring system, and computerized provider order entry (CPOE) as discussed earlier. Leveraging advanced use, EHRs hold great promise to improve poor health outcomes and health equity [9, 10, 71]. The Medicaid Promoting Interoperability Program has greatly contributed to the uptake of technology in clinical care practices in the U.S., as nearly 9 out of 10 offices have adopted a basic EHR system [10]. However, the rate of MU of EHRs has stalled among Medicaid providers [10]. In Florida specifically the majority of Medicaid providers do not use the EHR system to their full capabilities [30]. Our study provides critical insights into the structural barriers (regional, local, economic, and resource-related issues) that hinder the optimal diffusion of health IT for low-resourced providers [71].

Our results might provide useful information to other initiatives to increase EHR diffusion in underserved communities. For example, the Health Resource and Services Administration is supporting several initiatives of EHR diffusion in community health centers and rural health clinics, which are important safety net providers of care for traditionally underserved populations [10]. Our study can provide insights to these initiatives and the like across otherwise similar areas. These findings can also provide insights into the larger field of health informatics regarding factors that can influence an unintended “digital divide” for providers and healthcare institutions in medically underserved communities and regions [13]. This study has limitations related to the datasets used. First, the data are from the state of Florida and, as a result, the generalizability to other states should be done with caution. Second, it was not possible to determine the exact year providers attested to MU. Hence, the sample dataset was constructed as a cross-section with multiple years of data rather than a longitudinal dataset. Nonetheless, the pooled logistic regressions with GEE have been shown to provide consistent estimates of effects similar to those obtained with the use of time-dependent Cox analyses [72, 73].

Acknowledgments

The authors thank all the attendants of the American Society of Health Economists conference who provided feedback to help improve this paper. We also thank Nathina St Felix for helping with the bibliography and references.