What quantifies good primary care in the United States? A review of algorithms and metrics using real-world data

Researchers suggested that care coordination would improve when patients visit the same PCP for most of their primary care visits and when specialist referrals are determined by the likely responsible PCP [4, 10]. The issue of having multiple PCPs included increased medical services expenditures due to more office visits, prescriptions, and a more significant number of specialists seen for disease-specific populations [11]. Previous health services studies recommended that every patient see the same PCP [12]. However, in reality, patients may not have any responsible PCP, or they may have more than one. The claims-based “majority” and “plurality” algorithms (Table 1) were developed to identify the likely responsible PCP. The plurality provider is the provider who billed the highest number of Evaluation & Management (E&M) visits in one year, whereas the majority provider is the provider who billed for the plurality (greater than 50%) of all visits.

Several process measures have been documented in the literature, including medication intensification [13], lifestyle counseling documented in the notes of PCPs [14], the appropriate-care-model composite score for preventative care [15‐17], time from patient discharge to the first visit with PCP [18, 19], peer review and identification of negligent adverse events, and the duration of PCP consultation [20](1). Rather than the manual extraction, peer review, and identifications from the PCP notes, the process of these face-to-face encounters can be efficiently extracted from appropriate billing codes. For instance, medication intensification, defined as the initiation of a new medication or an increase in the dose of an existing medication [21], can be identified from the pharmacy data. Morrison et al. [13] used the EHR data from Brigham and Women’s and Massachusetts General Hospitals to identify the acute encounters for acute pain and infection of patients based on the availability of appropriate billing codes and ICD-9 diagnosis codes. The data linkage to patients’ pathology test data for elevated A1C, lipid profile, or blood pressure [13] provided the possibility for identifying the preventative care services the patient received. The use of billing codes [22] may accompany great uncertainty and does not necessarily reflect every single patient's clinical information when the billing codes are created in a way more relevant to providers.

There was an ongoing debate regarding the optimal duration of the consultation. Orton and colleagues demonstrated that longer consultations had significant beneficial effects on patients [23]; however, the Cochrane systematic review by Wilson and Childs showed no incremental benefits from having longer consultations [24]. The controversy led to the development of new measures. The long to short consultation ratio [25, 26] was a method to measure the quality of care for general practices. The PCPs with the fastest times (1^st quartile) were assigned as “faster” doctors; those in the last quartile with the longest times were described as “slower” doctors; and within each of the three doctor styles, the percentage distribution of consultation lengths was displayed and the ratio of long to short consultations calculated. The billing codes for claims data [27] also allowed for identifying the duration of the PCP consultations. Since 2003, the Centers for Medicare & Medicaid Services (CMS) collected claims data for Medicare and Medicaid patients across various categories and times, including Inpatient and Outpatient claims, Master Beneficiary Summary Files, and many other files. CMS continues to update a list of Current Procedural Terminology (CPT®) and Healthcare Common Procedure Coding System (HCPCS) codes, bill procedures, and services codes, which the American Medical Association primitively developed. In CPT codes, 99,211, 99,212, 99,213, 99,214, and 99,215 corresponded to 5 ~ , 10 ~ , 15 ~ , 25 ~ , and 40 ~ minute consultation, respectively [27]. Since COVID-19, telehealth has experienced a dramatic increase. Connection to PCP via telehealth was not novel but emerged during the COVID period as a promising model of healthcare. Some measures, for example, “no-shows” and “same-day” cancellations, “after-hours” care availability (i.e., care that is not Mon–Fri, 8 a.m. – 5 p.m.) from both EHR and claims data have evolved as quality-based metrics [28].

Distance to the medical offices and availability of primary care services in patients’ communities are known barriers to access to care, especially in rural areas [29]. The geographic access estimation, for example, number of PCPs/10,000 population and availability of PCPs within a distance of a 30-min drive time [30, 31], enables the assessment of the spatial distribution of primary care providers and helps inform the health planners of the possibility of optimizing the health resources in order to achieve efficient primary care [32, 33].

Another core principle of primary care is the continuity of care [34]. It illustrates an ongoing relationship between the PCP and the patient beyond a single episode of illness. A couple of studies have found a strong correlation [35‐38] of the continuity of PCP care with enhanced patient-physician relationships, which include trust-building, effective communications, a sense of responsibility over time, greater patient satisfaction and higher quality of care (e.g., better recognition of problems and diagnostic accuracy, effective management of patients with chronic conditions and maternity care outcomes), higher rates of compliance to medications, better performance of screening tests, timely receipt of preventive medicine services, less loss in follow-up visits, and a considerable reduction in hospitalizations, repeat hospitalizations, emergency department (ED) visits, and shorter length of hospital stay. There are multiple indices and algorithms to check the continuity of care from different dimensions (Table 1). From the review of existing evidence [39‐42], we have found convergences and divergences in a considerable spectrum of continuity of care measurements with different considerations of what is essential to measure. For instance, they can be described as measures of concentration (the proportion of consultations with one specific provider), dispersion (the number of different professionals consulted), distribution (the distribution of consultations between providers, giving higher scores to people who consult fewer providers); or sequence (whether each consultation was with the same provider as the previous consultation) [43]. Some measures are based on attributing scores to individuals (“individual measures”), while others attribute a score to each consultation (“visit measures”). Some approaches have focused on visit patterns only, whereas others have defined an individual provider as the “primary” provider for each patient. Evidently, no single measurement approach could fully capture the whole concept. Therefore, more emphasis should be given to developing and applying direct measures from the patient’s perspective, focusing on information sharing and care consistency between various organizations, and broadening the inclusion of continuity measurement in the multi-disciplinary team level [41].

Continuity of care also has potential limitations. First, those measurements may not always have a straightforward meaning [43]; for example, the value between 0 (different doctors on every occasion) and 1 (all care from the same doctor) for the continuity of care (COC) index does not release a specific message by itself. Secondly, repeated contact with the same PCP may bring gain and loss. Having multiple PCPs [44] may allow confirmation of diagnoses and suggestion of additional directions for diagnoses after discussing with each other. Some British researchers shared their concerns that higher COC [45] might paradoxically impair the patients because familiarity may reduce the duration of each visit, prevent patients from discussing new problems, and lessen the PCP’s objectivity. It could frustrate PCPs [45] if their patients have insoluble issues, and the long-term continuity may lead to PCP burnout.

The temporal regularity of primary care visits has been a novel concept in recent years. This method of regularity score to assess the regularity of PCP was developed by Einarsdóttir et al. [46] as 1/[1 + Variance (\({\varnothing}i\))], where \({\varnothing}i\) was the time interval between the (i-l)th and ith PCP visits. It ranged from 0 to 1 (with 1 representing perfect regularity) and was divided into quartiles, each containing 25% of the study population. Any individual score did not represent a unique distribution of any set number of PCP visits; instead, an aggregate tendency toward even spread and lack of clustering. The two hypotheses [47] behind the regularity index was (1) that a more temporally regular pattern of visits reflected the higher quality of care at some sites, achieved through more effective efforts to manage patients proactively and conversely to prevent loss to follow-up, and (2) temporal regularity was capturing unmeasured patient-level variables that are associated with risk for poor outcomes, such as a patient’s propensity to attend scheduled appointments or to participate in other health-promoting activities. Researchers were generally positive about this new measurement because they have identified a high correlation between high temporal regularity of PCP visits and positive health outcomes in various study samples  [47‐51].

In this section, we present a case on operationalizing various metrics to guide the conceptualization and evaluation of emerging cancer survivorship care models involving PCPs [53, 54]. Foremost, the scope of the metrics can be compared against existing guidelines [55] or frameworks [56] to clearly define the role of PCPs. In the management of survivors of cancer, the relevant care domains include health surveillance, lifestyle modification, preventive care, management of physical, psychosocial issues, and chronic conditions, together with the use of survivorship care tools to ensure care continuity beyond the oncology to primary care setting, are highly compatible with the process and care continuity measures. It was postulated that a lack of regularity of interactions with the health system among cancer survivors compared to patients with other chronic conditions may have reduced opportunities for advocacy of self-management skills [57], a care gap that PCPs can address with existing skillsets [58]. Also, oncologists could conveniently apply algorithms to EHR data to identify the appropriate PCPs for survivors’ consideration, allaying reported concerns of disrupting existing doctor-patient relationships wherever possible [59, 60]. A clear delineation of the PCPs’ role could directly address existing divergent views reported in the literature [61] and demarcate roles and responsibilities [53] from tertiary providers. Furthermore, the clear links between these performance indicators and principles of primary care serve to reduce confidence-related hesitancy.

Besides helping to conceptualize PCPs’ roles, the library of measurements will allow a holistic and timely audit or evaluation of primary care involvement in new care models or programs using readily available EHR data in real time. By including metrics from different primary care aspects one can ensure comprehensive and robust outcomes to overcome the pitfalls of considering indicators in silo. Additionally, the granularity of the data from the metrics allows researchers to identify poor performing areas, facilitating the development of strategies to improve specific care aspects. Lastly, this case of cancer survivorship supported our recommendation on developing additional measures to assess information sharing and care consistency between various organizations. This endeavor will complement existing efforts by the American Society of Cancer Oncology to build oncology EHR for advancing and ensuring quality cancer care [62, 63].

We acknowledge that, in some cases, the available data may not be ideally suited for the objective measures described above. The implementation of these quality measures requires substantial investments in data collection, processing, and analysis. The digital transition of medical records to EHRs requires tremendous efforts, which may not be immediately achievable at the local, state, and national levels. Most delivery systems rely on technical support provided solely by a single EHR software vendor and only identify the care process for network PCPs, which results in a lack of interoperability across providers, networks, and vendors. Additionally, the adoption of “big data” in healthcare in the United States falls behind other industries under the regulatory environment of the Health Insurance Portability and Accountability Act, and the Health Information Technology for Economic and Clinical Health Act enacted to protect personal health information in the United States [64], compared to some other countries (e.g., the United Kingdom), where the detailed electronic primary care records, procedure registries, cause-specific hospitalization, mortality record, and census data are available at a national scale [65].

Moreover, there was a knowledge gap between the data-based algorithms versus the patient-reported quality of PCP care. Bentleret et al. [66] compared the claims data-based COC indices with self-reported National Health and Health Services Use questionnaires of 2,620 Medicare beneficiaries and found that most claims-based COC measures failed to coincide with patients’ subjective perceptions of continuous patient-provider relationships. Future studies may potentially gain insight into the performance of the EHR and claims-based metrics.

The COVID-19 pandemic significantly impacts patient care and the healthcare system, including primary care. The pandemic has dramatically changed how primary care is delivered. Providers defer elective and preventive visits, such as annual physical exams, and convert in-person visits to telemedicine/telehealth visits to prevent person-to-person transmission. Currently, CMS [67] allows more telehealth services for all clinicians by expanding the terms of CPT codes, including remote evaluations, virtual check-ins and e-visits, and remote patient monitoring, and temporarily waive Medicare and Medicaid’s requirements that physicians and non-physician practitioners be licensed in the state where they are providing services. All of these efforts could take effect but leave many uncertainties about the factual delivery of primary care. As more clinical data become collected and standardized across providers, we hope that integrating EHRs and administrative claims data will be a cost-effective way to assess the quality of primary care.