Christine Vogeli, PhD

 Our study examines the impact of an emergency department (ED) patient navigation program for patients in a Medicaid accountable care organization across 3 hospitals in a large health system. Our program engages community health workers to (1) promote primary care engagement, (2) facilitate care coordination, and (3) identify and address patients’ health-related social needs.

 Our study was a retrospective analysis of health care utilization and costs in the 30 days following the index ED visit, comparing individuals receiving ED navigation and matched controls. The primary outcome of interest was all-cause return ED visits, and our secondary outcomes were hospital admissions and completed primary care appointments.

 Patients with ED visits who received navigation were matched to comparable patients with ED visits without an ED navigator interaction. Outcomes were analyzed using fixed effects logistic regression models adjusted for patient demographics, ED visit characteristics, and preceding utilization. Our primary outcome was odds of a return ED visit within 30 days, and our secondary outcomes were odds of a hospitalization within 30 days and odds of having primary care visit within 30 days.

 In our sample, there were 1117 ED visits by patients meeting our inclusion criteria with an ED navigator interaction, with 3351 matched controls. ED navigation was associated with 52% greater odds of a completed follow-up primary care appointment (odds ratio [OR], 1.52; 95% CI, 1.29-1.77). In patients with no ED visits in the preceding 6 months, ED navigation was associated with 32% decreased odds of repeat ED visits in the subsequent 30 days (OR, 0.68; 95% CI, 0.52-0.90). There was no statistically significant impact on return ED visits in those with higher baseline ED utilization.

 Our program demonstrates that high-intensity, short-term patient navigation in the ED can help reduce ED visits in those with low baseline ED utilization and facilitate stronger connections with primary care.

In this study, patient navigation in the emergency department (ED) was associated with 52% greater odds of a completed follow-up primary care appointment. In patients with no ED visits in the preceding 6 months, navigation was associated with 32% decreased odds of repeat ED visits in the subsequent 30 days.

Our program engages community health workers to promote primary care engagement, facilitate care coordination, and identify and address patients’ health-related social needs.

Our program demonstrates that high-intensity, short-term patient navigation can help reduce ED visits in those with low baseline ED utilization and facilitate stronger connections with primary care.

In the United States, approximately 33% of adults and 13% of children enrolled in Medicaid insurance programs report barriers to finding a doctor or delays in receiving care despite having a usual place of care.

 These barriers include lack of access to non–emergent care settings resulting from physical and economic barriers, as well as issues related to various social factors.

 Medicaid enrollees, in part because of these obstacles, have been shown to use the emergency department (ED) 6 to 7 times more often than privately insured patients.

Traditionally, EDs and systems of emergency care have not been designed to actively address barriers to care, the social conditions underlying many acute presentations, or patterns of frequent utilization for low-acuity conditions. To address these challenges, the ED Navigator Program was created in March 2018 by Mass General Brigham (MGB; formerly Partners Healthcare), a large health system based in Boston, Massachusetts, that provides care for 130,000 Medicaid members and more than 700,000 patients overall enrolled in its affiliated accountable care organizations (ACOs).

The ED Navigator Program was launched at 3 of the 8 general acute care hospitals’ EDs within our health system—Brigham and Women’s Hospital, Massachusetts General Hospital, and North Shore Medical Center—which were selected by population health management (PHM) leadership. Supported by the system’s Medicaid ACO and PHM program, each hospital hired and embedded 1 ED navigator, a layperson with experience in the health care or social services sector, into the hospital’s ED. The ED navigators serve as part of the ED care team and are tasked specifically with (1) promoting primary care engagement by scheduling a post–ED discharge appointment with a primary care physician (PCP) and addressing barriers to PCP access, (2) coordinating care for patients already engaged in a PHM program and referring patient candidates to appropriate PHM programs, and (3) identifying patients’ health-related social needs and facilitating connections to community-based resources. The ED navigators generally approach low- or moderate-acuity patients in person during the discharge planning phase of their ED visit. Their patient interactions are deliberately brief, given the nature of ED flow. Therefore, to be most effective during their encounter, the ED navigators typically review the patient’s record, prepare resource materials, and develop an engagement approach based on motivational interviewing and trauma-informed care techniques before approaching the patient in the ED. ED navigators use an internally developed social needs assessment instrument and refer patients to resources using shared resource guides, which are continually updated. Selected resources include housing, food, transportation, and employment support programs. Examples of specific referrals include connection to local housing advocates, food pantries, or public transportation assistance programs. ED navigators also use a translation phone line to ensure that they are able to address the needs of patients, regardless of language. Once a patient has been discharged, the ED navigator follows up with the patient telephonically within 72 hours or coordinates follow-up with a member of the patient’s longitudinal care team.

Several other health systems across the country have implemented different ED-based care coordination models, showing variable results with respect to health outcomes, cost, and utilization.

 However, community health worker interventions based in the ED have demonstrated some of the most promising results to date. Memorial Hermann Health System in Houston, Texas, in a quasi-experimental pre-post study with a comparison group, demonstrated that state-certified, bilingual community health workers working with uninsured and Medicaid patient populations reduced ED visits and generated cost savings ranging from $331 to $1369 per patient after 12 months.

 In a descriptive analysis, Boston Medical Center demonstrated the ability of health promotion advocates in the ED to increase referrals to social support resources.

 Likewise, Erlanger Health System in Chattanooga, Tennessee, performed a randomized controlled trial with patient navigators working with high ED utilizers and demonstrated a decrease in ED visits and costs.

 Given the prior experience of ED-based care coordination and community health worker interventions, we sought to examine the efficacy of our ED Navigator Program in facilitating linkages to primary care and reducing ED visits.

For this analysis, we used clinical and administrative data for patients enrolled in the MGB Medicaid ACO during the period of June 1, 2018 (when the ED Navigator Program was first fully implemented), to October 31, 2019.

Using electronic health records (EHRs), we identified treat-and-release ED visits to the 3 hospital EDs in which the ED Navigator Program was implemented. We excluded ED visits that were high acuity, defined as having an emergency severity index (ESI) score of 1 or 2, or having been triaged to the “acute” pod of the ED. At the hospital that does not use ESI, patients in the acute pod are determined to have a potential illness requiring immediate evaluation. ED visits occurring on Saturday and Sunday were also excluded from the analysis, as the ED Navigator Program is currently available only Monday through Friday. Overnight visits occurring during the week are often followed up by phone the following day, so these were included in the evaluation. We matched these data to Medicaid claims data and removed ED visits that we were unable to match, including ED visits occurring during date spans when patients were not aligned to the ACO or when Medicaid coverage lapsed, or ED visits for substance use services, which were not shared with the ACO. ED visits with multiple records or claims from the same day were collapsed into a single event. We further excluded ED visits for persons with fewer than 3 months of claims data available in the 6 months preceding the ED visit or fewer than 1 month of claims data in the month following the ED visit (

]). We then determined the number of ED visits, hospitalizations, and PCP visits in the preceding 6 months, also based on claims.

From these data, we identified all ED visits in which ED navigation occurred, limiting the data to the first ED visit with navigation for patients who had multiple encounters with the ED navigator. We developed a matched comparison population using the ED visits from patients who never used the ED navigator and selected 3 matched comparison patients for every intervention patient matched on the number of ED visits (0, 1, 2-3, or ≥ 4), inpatient stays (0, 1, or ≥ 2), and PCP visits (0, 1, or ≥ 2) in the 6 months prior to the ED visit. Only 1 episode per patient was selected. We obtained patient demographics from our system’s EHR data and Medicaid claims data.

For the evaluation, we conducted a retrospective matched analysis of health care utilization in the 30 days following the index ED visit, comparing persons who had navigation with those who did not. Our primary outcome of interest was any return ED visit (both treat-and-release ED visits and those resulting in a hospital admission) during the 30 days following the index ED visit. Our secondary outcomes of interest were any hospital admission or completed primary care appointment in the 30 days following the ED visit. We initially intended to explore missed appointments as a secondary outcome, but rates of missed appointments in both groups were markedly low, so we were unable to pursue this analysis. Thirty days was selected as the period of interest because it was felt to represent a period of time long enough for us to be able to observe an impact from the program but short enough that any observed effect could be reasonably attributable to this brief, limited intervention. This time period was also consistent with existing literature regarding ED return visits.

These binary outcomes were analyzed using fixed effects logistic regression models adjusted for patient age, sex, race, ethnicity, primary language, employment, behavioral health history, ED visit characteristics, hospital, and baseline utilization matching criteria in our logistic regression analysis of return ED visits and follow-up PCP visits within 30 days (

). In our logistic regression analysis of hospital admissions within 30 days, we adjusted only for age, sex, hospital, baseline utilization, ED visit characteristics, overnight ED visit, acuity, and history of mood disorder, which each showed independent correlation with the dependent variable so as to avoid overfitting the model, given the small number of 30-day admissions in the sample. ED visit characteristics include acuity level (defined by ESI or triage area), overnight visit (defined as arrival and discharge occurring between 10 pm and 7 am), ambulance arrival, and arrival as outside transfer. We used predetermined definitions based on diagnoses codes in claims data to incorporate indicators for patients’ socioeconomic status or related risk factors. Pediatric status was adjusted for using an age cutoff of 20 years, which is the age used at the included hospitals to triage patients to the separate pediatric section of the ED. Included covariates were selected a priori by the study team in consultation with ED providers, ED navigators, and program leaders to account for factors that may be associated with selection by the ED navigators or those that may be independently associated with our primary and secondary outcomes.

We used fixed effects for each of the 3 included hospitals to adjust for index ED visit hospital. A subgroup analysis was performed, stratifying patients on baseline levels of ED utilization to explore the potentially variable effect of the program on each of these populations. All included variables and subgroup analyses were defined by the study group a priori.

Our analysis was conducted as part of a routine program evaluation using a data repository approved by the MGB Institutional Review Board for retrospective program evaluation.

There were a total of 22,557 treat-and-release ED visits by MGB Medicaid ACO patients during the study period to the hospital EDs included in the ED Navigator Program. Of these, 12,113 visits met our criteria for inclusion and were able to be matched to claims data: 1315 with an associated ED navigator encounter and 10,798 potential control visits.

After matching, our final sample included 1117 intervention patients and 3351 comparison patients. ED visits with an ED navigator encounter were more likely to involve patients who were older, female, Black, married, and employed, and who also had documented anxiety and mood disorders (

). Intervention and comparison patients had similar claims availability in the prior 6 months (5.8 months among intervention patients and 5.8 months among comparison patients).

Overall, we were unable to detect a statistically significant difference in the odds of returning to the ED within 30 days for patients who received ED navigation compared with those who did not (odds ratio [OR], 0.88; 95% CI, 0.71-1.08) (

). However, among individuals with no ED visits in the preceding 6 months, the ED Navigator Program was associated with lower odds of subsequent 30-day ED presentation, with an OR of 0.68 (95% CI, 0.52-0.90) (

). However, the results were not statistically significant for individuals with 1, 2 to 3, or more than 3 visits in the preceding 6 months (adjusted ORs, 0.99 [95% CI, 0.64-1.52]; 1.15 [95% CI, 0.72-1.82]; and 1.21 [95% CI, 0.58-2.55], respectively).

With regard to PCP follow-up, after adjusting for included variables, patients who received ED navigation were significantly more likely to have a PCP visit within 30 days (OR, 1.52; 95% CI, 1.29-1.77). No significant difference was observed with regard to 30-day hospital admission (OR, 0.77; 95% CI, 0.47-1.27).

In our analysis, we found that ED navigator encounters were significantly associated with a reduction in return ED visits for patients with no prior visits in the preceding 6 months, as well as increased likelihood of completing a follow-up primary care appointment in the 30 days after the index ED visit for all patients. However, we were unable to identify a statistically significant difference in subsequent ED visits and hospital admission in the ED navigator group overall.

This study demonstrated that ED navigator encounters were significantly associated with increased rates of follow-up with primary care among a Medicaid ACO patient population. Specifically, with regard to primary care, patients with an ED navigator encounter had a statistically significant 52% increase in the odds of having a follow-up primary care appointment in the 30 days following their index ED visit. Notably, primary care practices in MGB already routinely engage in attempting to schedule follow-up primary care visits for patients seen in the ED. Therefore, the increased success of ED navigators in doing so suggests added incremental value in “capturing” patients at the point of care to reengage in long-term management.

With regard to acute care utilization, we found that the effect of the ED Navigator Program on reducing ED utilization was most pronounced among ED-naïve patients (ie, those with no prior ED visits in the preceding 6 months). In contrast, there was no significant effect on utilization on more frequent ED utilizers. This could be because of several factors. First, ED-naïve visitors may be less aware of the services provided in an ED vs urgent care or primary care and, once educated about other options by ED navigators, opt to use other forms of care for access. Secondly, those with higher baseline utilization are more likely individuals with ingrained patterns of ED utilization and care-seeking behaviors, as well as chronic conditions prompting recurrent presentation. These include both chronic health and mental health conditions, low health literacy, and a multitude of social factors, such as those related to housing or transportation.

 A brief 1-time encounter would be less likely to change well-established patterns of behavior or to definitively address the underlying social or medical problems precipitating frequent ED utilization.

Based on the results of this analysis, the MGB ED Navigator Program will continue to target and support patients with lower baseline levels of ED utilization. Additional research is needed to characterize patient and program factors that may affect the overall impact of the program. To better support patients with higher ED utilization patterns, we are collecting internal data to segment this population into distinct risk groups and to correspondingly create targeted referral pathways to longitudinal population health or care management programs. In addition, the magnitude of the difference in admission rates post intervention is notable, although no statistically significant difference was observed, which could potentially be due to small numbers of admissions in both groups. Additional research could aim to explore this potential relationship further with larger sample sizes.

Historically, care coordination programs have focused on patients with high levels of utilization, but this approach has been recently called into question.

 The results of the ED navigator intervention add to the existing literature by reflecting the potential value in targeting patients with low health care utilization but “rising risk” as well. In addition to redirecting care to more cost-effective environments in the short term, these interventions may also preempt eventual frequent utilization behavior and promote health in the long term by facilitating connections to primary care and to resources related to social determinants of health. They also illustrate that a brief 15-minute intervention can help change a patient’s overall care trajectory—at least in the short term—in contrast to more costly and time-intensive long-term programs.

The limitations of this study include generalizability, as this program and analysis are limited to a single health system and Medicaid ACO patients (not inclusive of patients who are dual eligible). The success of this program relies on being able to refer patients to existing programs and resources, which may not be available at more resource-constrained ACOs or institutions located in states with less of a robust social safety net than Massachusetts. A limitation of the program itself is that it is currently only available on weekdays. This programmatic limitation may affect the study because it is possible that the characteristics of patients visiting the ED on weekends may differ from those of patients visiting the ED during weekdays. Similarly, patients with low-acuity diagnoses seen overnight, who receive an intervention via phone, may have a different experience with the program than those who receive the face-to-face intervention.

With regard to methodology limitations, this study is a retrospective cohort analysis and not a randomized controlled trial. Therefore, it is possible that other factors may have confounded the results of our analysis. However, we accounted for this by matching patients based on baseline utilization and adjusting for potential confounders, including patient demographics, ED visit characteristics, health status, and socioeconomic status and related risk factors. Our adjustment for socioeconomic status was in part accomplished through the use of diagnosis codes in claims data. However, it is important to note that these socioeconomic diagnosis codes are infrequently used, and therefore, there is potential for misclassification.

Additionally, we were unable to match all visits recorded in our EHR with their associated claims. This is potentially because of patients churning in and out of Medicaid or our ACO (whose visits may have occurred in a brief period of lost eligibility); the exclusion of substance abuse visit claims from the data shared with our ACO; and, to a lesser extent, discrepancies in the ED dates of service as recoded in Epic and claims data (eg, as an ED encounter may span multiple days, it is possible that the date of ED arrival would not match a specific claim’s date of service). Regardless of these reasons, there would be no reason to believe that there would be any relationship between unmatched ED visits and our outcomes, unaccounted for by other included variables, or other systematic patterns of their distribution that may have biased our results, with the possible exception of substance use disorders (claims information regarding substance use disorder is not provided by the state to ACOs for privacy reasons). Patients in both the case and control groups had to have claims matched to ED records. Therefore, we do not believe this would have likely significantly biased our results. Although we did adjust our analysis for chart-documented history of substance use disorder and alcohol use disorder, because of this exception, these results may not necessarily be generalized to ED visits for these conditions. Similarly, given potential causes of inability to match records, it is possible that our findings cannot be generalized to patients who may rapidly churn in and out of an ACO.

Our analysis suggests that care coordination through the use of community health workers embedded as navigators in the ED with high-intensity, short-term interactions can help reduce ED visits among individuals with low levels of baseline ED utilization and facilitate stronger connections with primary care in a Medicaid ACO population. Our model highlights that short-term care coordination programs are successful, particularly for patients with lower levels of baseline ED utilization, and can ultimately promote primary care engagement, as well as assistance with health-related social needs. We believe that these results can provide important insights to health systems as they consider cost-effective program options for care management to strengthen primary care engagement and address the social determinants of health.

Salina Bakshi, MD, MPH, and Lucas C. Carlson, MD, MPH, contributed equally to this work and are listed as co–first authors.

The authors would like to acknowledge the contribution of Maryann Vienneau, Erin Maher, Alex Sheff, John Orav, Christin Price, Trancy Escobar, Katherine McLaughlin, Charline Gay, Elizabeth Fonseca, Kristen Risley, Deidra Smith-Horton, Laurie Isidro, Victoria Lo, Lindsay Jubelt, and Gregg Meyer to the development, implementation, and evaluation of this program. They would also like to recognize the ED navigators themselves for their dedication to their clients and regular efforts to support them across the continuum of care.

 Population Health Management, Mass General Brigham (SB, LCC, JG, PW, KH, CV, AOF), Boston, MA; Department of Medicine (SB, PW) and Department of Emergency Medicine (BJY), Massachusetts General Hospital, Harvard Medical School, Boston, MA; Department of Emergency Medicine (LCC) and Department of Medicine (AOF), Brigham and Women’s Hospital, Harvard Medical School, Boston, MA; The Mongan Institute, Massachusetts General Hospital (CV), Boston, MA.

 Funding for Medicaid accountable care organizations in part provided by state of Massachusetts.

 The authors report no relationship or financial interest with any entity that would pose a conflict of interest with the subject matter of this article.

 Concept and design (SB, LCC, KH, BJY, CV, AOF); acquisition of data (JG, KH, CV, AOF); analysis and interpretation of data (LCC, JG, PW, BJY, CV, AOF); drafting of the manuscript (SB, LCC, JG, PW, CV, AOF); critical revision of the manuscript for important intellectual content (SB, LCC, JG, PW, BJY, CV, AOF); statistical analysis (LCC, JG, CV); provision of patients or study materials (BJY, AOF); administrative, technical, or logistic support (SB, LCC, PW, KH); and supervision (SB, AOF).

 Priscilla Wang, MD, Massachusetts General Hospital, 15 Parkman St, Wang Ambulatory Care Center 634, Boston, MA 02114. Email: pgwang@partners.org.

1. Revisiting emergency department use in Medicaid. Medicaid and CHIP Payment and Access Commission. July 2014. Accessed June 12, 2020. https://www.macpac.gov/publication/mac-facts-revisiting-emergency-
department-use-in-medicaid/

2. Uscher-Pines L, Pines J, Kellermann A, Gillen E, Mehrotra A. Emergency department visits for nonurgent conditions: systematic literature review. 

3. Colligan EM, Pines JM, Colantuoni E, Howell B, Wolff JL. Risk factors for persistent frequent emergency department use in Medicare beneficiaries. 

. 2016;67(6):721-729. doi:10.1016/j.annemergmed.2016.01.033

4. Capp R, Misky GJ, Lindrooth RC, et al. Coordination program reduced acute care use and increased primary care visits among frequent emergency care users. 

. 2017;36(10):1705-1711. doi:10.1377/hlthaff.2017.0612

5. Katz EB, Carrier ER, Umscheid CA, Pines JM. Comparative effectiveness of care coordination interventions in the emergency department: a systematic review. 

. 2012;60(1):12-23.e1. doi:10.1016/j.annemergmed.2012.02.025

6. Enard KR, Ganelin DM. Reducing preventable emergency department utilization and costs by using community health workers as patient navigators. 

7. Bernstein J, Dorfman D, Lunstead J, et al. Reaching adolescents for prevention: the role of pediatric emergency department health promotion advocates. 

. 2017;33(4):223-229. doi:10.1097/PEC.0000000000000662

8. Seaberg D, Elseroad S, Dumas M, et al. Patient navigation for patients frequently visiting the emergency department: a randomized, controlled trial. 

. 2017;24(11):1327-1333. doi:10.1111/acem.13280

9. Rising KL, LaNoue MD, Gerolamo AM, Doty AMB, Gentsch AT, Powell RE. Patient uncertainty as a predictor of 30-day return emergency department visits: an observational study. 

. 2019;26(5):501-509. doi:10.1111/acem.13621

10. Desai S, Gruber PF, Eiting E, et al. The effect of utilization review on emergency department operations. 

. 2017;70(5):623-631.e1. doi:10.1016/j.annemergmed.2017.03.043

11. Yeatts KB, Lippmann SJ, Waller AE, et al. Population-based burden of COPD-related visits in the ED: return ED visits, hospital admissions, and comorbidity risks. 

. 2013;144(3):784-793. doi:10.1378/chest.12-1899

12. Davis CI, Montgomery AE, Dichter ME, Taylor LD, Blosnich JR. Social determinants and emergency department utilization: findings from the Veterans Health Administration. 

. 2020;38(9):1904-1909. doi:10.1016/j.ajem.2020.05.078

13. Kirkpatrick S, Agana DFG, Lynch K, Carek PJ. Emergency department high utilizers among family medicine patients. 

. 2019;32(2):264-268. doi:10.3122/jabfm.2019.02.180184

14. Griffey RT, Kennedy SK, D’Agostino McGowan L, Goodman M, Kaphingst KA. Is low health literacy associated with increased emergency department utilization and recidivism? 

. 2014;21(10):1109-1115. doi:10.1111/acem.12476

15. Finkelstein A, Zhou A, Taubman S, Doyle J. Healthcare hotspotting – a randomized, controlled trial. 

. 2020;382(2):152-162. doi:10.1056/NEJMsa1906848

Articles

A navigation program demonstrated decreased odds of repeat emergency department (ED) visits in patients with low baseline ED utilization and increased odds of follow-up primary care appointments.

Improving Care Coordination and Reducing ED Utilization Through Patient Navigation

 Appropriate selection of patients is key to the success of care management programs (CMPs). Hybrid patient selection approaches, in which large data assets are culled to develop a list of patients for more targeted clinical review, are increasingly common. We sought to describe the patient and practice characteristics associated with high-risk patient identification and selection for a CMP during clinical review, and to explore variation across primary care practices. 

Standardized estimates of Medicare beneficiaries identified as high risk for poor outcomes and high medical expense, and appropriate for a CMP within a large Pioneer Accountable Care Organization, were developed using mixed effects logistic models. Study subjects were 2685 Medicare beneficiaries aged over 18 (includes individuals eligible for Medicare due to a disability) aligned to 35 primary care practices in 2013.

 Independent predictors of patient identification as high risk include older age; higher risk score; recent increases in medical conditions; higher numbers of medical hospitalizations, skilled nursing facility days, and primary care physician visits; and shorter relationships with the primary care physician. Older age, and lower income, but no prior hospice use were independently associated with patient selection for a CMP among the subset of patients identified as being high risk. Adjusted predicted percents of high-risk patients varied significantly across practices overall and for 5 of the 6 patient characteristics that were independently associated with identification as high risk. 

Inconsistency in high-risk patient identification and selection for a CMP may reflect differences in practice resources, but also highlight the need for continual training and feedback in order to protect against unintentional biases.

	Clinician review within hybrid case selection approaches includes decisions about which patients are high risk and, among high-risk patients, identification of the subset of patients most appropriate for a care management program (CMP). 

		Primary care physicians were more likely to recommend high-risk patients who are poorer, older, and not using hospice services for CMP. 

		Significant variations in adjusted proportions of patients selected for care management across practices may reflect differences in practice resources, but may also introduce bias into the process.

		Continual training and feedback to practices on their review decisions may help identify missed patient opportunities.

The observation that health expenditures are concentrated within a relatively small subset of the population is well known by now, with 5% of the population accounting for 50% of medical expense in any given year. One-third of the individuals in this group—frequently those with poor health status and multiple chronic conditions&mdash;remain high-cost the following year,

 making them appropriate for primary care—based care management approaches

 However, although effective care management targeted at the right individuals has been shown to moderate costs and utilization,

 inconsistent results have highlighted the need to appropriately select patients.

There are a host of available case-finding tools and algorithms that can be used to identify individuals appropriate for care management; nevertheless, there is also increasing understanding that these tools have limitations.

 Many case-finding tools are proprietary “black boxes” that create a single score to rank patients according to risk.

 Additionally, these tools can vary considerably in terms of their input criteria, and there is no standardization of definitions across tools

—most importantly, the input criteria cannot be modified to meet specific program characteristics and operational realities.

Data timing and completeness are a challenge for all case-finding algorithms, especially those based on paid claims. Delays in data capture can result in the identification of patients who are no longer enrolled within the health plan or practice, as well as identification of patients who no longer require intervention.

 Incomplete data, such as removal of substance use, HIV/AIDS, and in some states, mental health claims,

 paint incomplete pictures of patient needs. In addition, demographic and psychosocial characteristics are often incomplete or missing from readily available administrative data (eg, poverty, education, living situation). The absence of patient psychosocial data is particularly challenging, as these data could help to identify cohorts of patients who would be most likely helped by care management programs (CMPs)

; for example, addressing cost-related medication underuse among low-income patients may improve diabetes control and reduce inappropriate emergency department utilization.

 Although many healthcare systems do collect psychosocial data on patients, system alignment issues often preclude use of more granular and descriptive data for case finding.

Because of these limitations, case finding based on administrative data is typically not the only method used.

 Patient health-risk assessments and clinician referrals are techniques frequently used to identify patients for case management and can provide more detailed and actionable data, but they require time and resources to implement.

 Often, multiple case-finding methodologies, including predicted risk scores, utilization triggers/markers, medication adherence and gaps in care, and clinician referrals, are used in concert.

Case-finding approaches that rely on multiple methodologies are typically termed “hybrid approaches” and can vary widely, but generally refer to the use of multiple methodologies—often in sequence&mdash;to refine case finding. Within this evaluation, we studied a sequential approach to case finding that uses an algorithm to comb through administrative claims and identify a preliminary set of potentially high-risk patients. This preliminary set of patients is then reviewed by physicians and nurse care managers who make the final decision of whether to refer a patient for care management. This sequential hybrid approach is thought to be reliable

Clinician referral, within the boundaries of a pre-selected list, focuses the referral process and reduces the burden on clinicians, while at the same time ensuring primary care practice and physician (PCP) involvement. This involvement may increase buy-in of the program, which is especially important for programs embedded within physician practices. Clinician input can also help to identify patients with more modifiable health risks, including those with sub-optimal health literacy and health insight, social and home environment challenges, and/or inadequate coping skills and financial resources.

 For example, primary care physicians are more likely to describe patients with poorly controlled diabetes, inadequate insurance, or with mental health or substance use disorders, as complex.

However, clinician referral may also introduce unanticipated variations in patient selection.

 Often, there are no specific criteria that physicians are asked to consider when identifying appropriate patients, leaving the door open to unintentional biases.

 Appropriate clinician referral into CMPs requires that clinicians have a clear understanding of the characteristics that place patients at higher risk for poor health outcomes, as well as the benefits the CMP could provide.

In this study, we sought to understand the unique patient characteristics associated with both patient identification as high risk for poor outcomes and subsequent clinician referral to care management. We also sought to describe variations in patient identification and selection for care management across PCPs overall, and for important patient characteristics. We examined the selection decisions during the second year of the CMP to avoid biases associated with the ramp-up of the program.

The Partners Healthcare Pioneer Accountable Care Organization (ACO)—1 of 4 Pioneer ACOs operating in eastern Massachusetts&mdash;is among the largest nationwide, and the only Pioneer ACO that exclusively aligns patients to primary care physicians. One of the signature initiatives within the Partners Pioneer ACO is the CMP. The CMP is a longitudinal program that relies on nurse care managers working in conjunction with primary care physicians to improve care quality and control healthcare utilization for referred patients. After a successful trial within the Medicare High Cost Beneficiary Demonstration Program,

 the CMP was implemented within all Partners Healthcare PCPs under the Medicare Pioneer ACO contract, starting in January 2012. Partners Healthcare uses a hybrid approach for patient selection, combining a claims-based algorithm with clinician review to identify high-risk patients who could benefit from the CMP.

The population for this study included traditional Medicare beneficiaries within the Partners Healthcare Pioneer ACO who were identified using a claims-based algorithm as potentially high risk for poor outcomes in the second year of the program (2013). By the second year, all practices had experience performing the clinical review and had 1 or more ACO patients who were actively engaged in CMP. In total, 7651 (12%) of 63,629 Medicare patients were identified by the Partners algorithm. Among individuals with a predictive risk score for future total medical expense via Optum ImpactPro software version 6.0 (Optum, Inc, Eden Prairie, Minnesota) that exceeds a minimum threshold, the Partners algorithm uses combinations of chronic conditions and patterns of healthcare utilization to identify individuals potentially at high risk and appropriate for subsequent clinical review. The Partners algorithm was developed with extensive input from an advisory committee of operational and clinical experts within the organization. During the clinical review process, PCPs receive detailed information on chronic conditions, utilization patterns, and predictive scores that triggered identification by the algorithm. PCPs had considerable latitude when selecting patients for care management, but were provided additional guidance through reference guides and regional medical directors who received orientation and training on the list review process.

In order to study clinician review decisions during the second year, this analysis excluded 2605 patients who had been identified and reviewed in the prior year, as well as 1267 patients who had moved, died before the review process, were not community-dwelling (eg, residing in long-term care), switched to a Medicare Advantage plan, or did not have a relationship with a Partners primary care clinician. Lists of the remaining 3779 patients (5.9% of 63,629 total Medicare ACO patients) were distributed to practice management organizations and, eventually, to PCPs and clinicians for final identification of individuals appropriate for the primary care—based CMP. We then further excluded 788 patients aligned to 44 PCPs that did not make differential review decisions (eg, identifying none or all patients as high risk for poor outcomes or requiring care management), as well as 306 patients who were not classified by the practice prior to the release of the 2014 list. Thus, the final study population consisted of 2685 patients who received primary care within 35 practices.

The dependent variables are the decisions made by primary care clinicians and their practices regarding whether patients were identified as high risk for poor outcomes and whether patients could benefit from CMP. 

 and include patient demographic characteristics, area-level poverty (American Community Survey 2012 5-Year Estimates

), prospective risk score and changes in the number of conditions from calendar year 2012 to the year directly preceding the first list review (May 1, 2012-April 30, 2013) (CMS-Hierarchical Condition Categories Risk Adjustment Model

), healthcare utilization (ED visits, medical hospitalizations, skilled nursing and hospice service use) in the 12 months prior, and characteristics of the patient’s relationship with their PCP. Practice characteristics include practice size, level of affiliation with the healthcare organization, whether the practice is a community healthcare center, years of experience with care management, CMP capacity, and level of practice readiness for patient-centered medical home transformation. Models were also adjusted for primary care physician characteristics, including primary care physician age, full-/part-time status, and duration of employment within the healthcare organization.

Descriptive analyses of the patient- and practice-level characteristics associated with patient identification as high risk for poor outcomes and selection for care management were explored using bivariate analyses. Standardized estimates of patient identification as high risk and selection for CMP were developed using logistic regression models with robust standard errors clustering at the practice. For regression models, we used the subset of explanatory variables that were independently predictive of either patient identification as high risk for poor outcomes or selection for care management at 

 <.20, and used a similar approach to develop standardized estimates at the practice level.

We explored the extent to which practices consistently identified patients with specific characteristics as high risk for poor outcomes. Specifically, we examined 6 patient characteristics that were independently associated with being considered high risk for poor outcomes and that were present in at least 10% of patients in our sample; similar models were used for each, as described above, to develop standardized practice-level estimates of identification as high risk among patients with and without the characteristic. Full interaction models did not enhance predictive ability and decreased the number of practices with stable estimates; therefore, we only included interactions between PCP and the characteristic of interest in our final models.

Global tests were used to test for differences in overall practice marginal estimates and interaction marginal estimates. Chi-square statistics were used to describe the relationship between overall practice marginal estimates and practice estimates for individuals with specific characteristics.

The study was approved by the Partners Institutional Review Board.

Among Medicare ACO patients identified by the claims-based algorithm and who also passed an administrative review (roughly 5.9% of total Medicare ACO patients), 86.3% were thought to be high risk for poor outcomes by their primary care providers; 77.2% of these patients were referred to a nurse-led CMP within the PCP.

A number of patient characteristics were independently associated with patient identification as high risk during the clinician review. Specifically, independent predictors of high risk include advanced age, higher prospective risk score, a recent increase in diagnosed conditions, and higher numbers of medical hospitalizations and days in skilled nursing facility (SNF) care in the prior year. Patients identified as high risk also were more likely to have more visits with their PCP in the prior 12 months, but shorter relationships with their PCP than patients not identified as high risk. The patient characteristics with the strongest independent association with being identified as high risk include age 85 years or older (93.1%), 2 or more hospitalizations in the prior year (91.5%), and 30 or more SNF days during the prior year (92.7%) (Table 1).

Among high-risk patients, advanced age was also an independent predictor of patient referral to a CMP. However, when determining whether high-risk patients would benefit from care management, clinicians were also more likely to select individuals without prior hospice use and those living in lower-income areas. The adjusted proportions of high-risk patients 85 years or older selected for care management was 80.1%, and between 81.2% and 81.9% among patients residing in areas with more than 10% of individuals living in poverty (Table 1).

Lists of patients identified by the claims-based algorithm were reviewed within 35 practices where patients received their primary care. On average, practices reviewed 76 algorithm-identified Medicare ACO beneficiaries (range = 5-248) in 2013. Adjusted proportions of patients identified as high risk and referred to a CMP varied according to practice characteristics. Specifically, after adjusting for patient characteristics and other practice characteristics, practices with 6 or more physicians (88.8%), non—community health centers (87.3%), and practices without the capacity to provide care management to all patients (93.5%) had the largest adjusted proportions of algorithm-identified ACO patients identified as high risk during clinical review. Among high-risk patients, adjusted proportions selected for care management were highest within practices employed by the health delivery organization (80.8%) and non–community health center practices (78.4%) (

After adjusting for patient and primary care physician characteristics and controlling for the clustering of patients within PCPs, practices varied significantly in terms of both the adjusted proportion of patients identified as high risk (36% to 99%; 

 <.001) and the proportion of patients referred to the CMP (18% to 96% ; 

We further explored the extent to which patients with specific characteristics were consistently categorized as high risk across practices. For this exploratory analysis, we selected the 6 patient characteristics that were both independently associated with patient identification as high risk and were present in at least 10% of the patient population. Practices varied significantly in how they categorized patients for 5 of 6 patient characteristics studied (

Hybrid patient identification approaches, that use algorithms to identify a subset of patients who warrant more careful clinical review, hold promise as a feasible method for incorporating clinician judgment and knowledge of patient psychosocial needs into case findings. However, we know little about the clinical review process, including the criteria used by PCPs to identify patients as high risk, how practices identify which high-risk patients would be appropriate for care management, and how these decisions differ across PCPs. 

Consistent with findings from our interviews of primary care physicians and care managers,

 we found that primary care teams used additional information about their patients’ clinical profiles to decide whether a patient on an algorithm-generated list was high risk. Patients identified as high risk tended to be older, sicker, and with greater healthcare utilization, including primary care visits, than patients not identified as such. Patients with shorter relationships with their primary care physician were more likely to be identified as high risk, suggesting that PCPs might either have different thresholds for identifying patients who are new to their practice as high risk or may not have had the opportunity to establish necessary supports for their newer patients. Lack of a long-term patient—PCP relationship may compromise effective communication with patients and their caregivers, which is particularly important in older, complex patients.

As a second step, primary care teams were asked to identify which high-risk patients were appropriate for care management, and when selecting, patient clinical need was far less of a consideration. In addition to patient age and no prior hospice use, patient area—level poverty was independently associated with selection of high-risk individuals for CMP services. Thus, although predicted risk and prior utilization are important for identifying patients who are high risk, they are not the primary factors associated with the decision to refer patients to a CMP.

Our study extends prior work by clearly differentiating clinical assessments of whether patients are high risk from determinations of whether high-risk patients would be appropriate for CMP services. Additionally, our findings support other studies that have found that one of the benefits of including clinician input is the ability to consider nonclinical factors, such as modifiable health risks, patient health literacy, social and home environment challenges, social support, or inadequate coping skills and financial resources.

 Patients appropriate for care management often have issues with medical decision making and care coordination in addition to poorly controlled disease.

 Physicians may also preferentially select patients who they feel are more willing to participate in a CMP.

We found considerable variation in the overall adjusted proportions of patients identified as high risk across practices, as well as significant variations in practice-level identification rates for specific cohorts of patients, suggesting that practices may not use the same criteria to identify high-risk patients. Although practices were provided with central guidance on how to perform the clinical review, patient-by-patient decisions are left to the practices.  Some of this diversity in identification of high-risk patients may appropriately reflect unmeasured differences in practice resources and access to specialty care and services, such as behavioral health specialists and addiction counselors, for their patient population. Differences could also reflect variations in program implementation, such as nurse skill levels or comfort with different patient populations. Care management, by nature, is a diverse intervention for individuals with complex healthcare needs.

 The multifaceted nature of care management not only complicates the identification of patients, but may bias selection toward patients whose needs can be met by the skills of the care manager available to the practice. Training for care managers and feedback to practices regarding the benefits and limitations of CMP may help minimize unintentional disparities in patient selection.

Hybrid approaches allow for clinician input into case finding for care management, but training and monitoring is required to protect against unintentional biases.

Implementing a Hybrid Approach to Select Patients for Care Management: Variations Across Practices

Hybrid approaches to case finding show promise as a method to increase the success of care management programs (CMPs). A large healthcare system implemented a hybrid approach in which clinicians review algorithm-generated lists of potential high-risk patients within their practice and select the patients most appropriate for the CMP.  We sought to understand the criteria clinicians used when selecting patients.

 We conducted 20 semi-structured interviews with a convenience sample of primary care clinicians and their care managers from a diverse set of practices.

 Two independent reviewers thematically coded interview responses through an iterative process.

 In addition to clinical need (eg, disease severity or multiple comorbidities), interviewees considered a number of nonclinical patient characteristics that they felt placed patients at risk for suboptimal healthcare utilization and poor adherence. These include patients’ predisposing (eg, health literacy or navigation challenges, physical vulnerabilities, insight regarding their health) and enabling characteristics (eg, social and home environment, coping skills, financial resources). Interviewees also considered patients’ existing linkages with the health system and whether other clinicians were already meeting a patient’s care needs.

In selecting patients for a CMP, clinicians considered patient characteristics that are not easily captured in standard clinical and administrative data. A hybrid approach to case finding concentrates clinician review on patients who meet standard clinical and healthcare utilization criteria, and allows clinicians to incorporate knowledge of patients’ predisposing and enabling characteristics that are not readily maintained in clinical data.

	This study identified the factors that primary care clinicians and care managers considered as they reviewed algorithm-generated patient lists to identify a final set of patients appropriate for care management. 

		In addition to the complexity of patients’ medical conditions, clinicians considered patients’ health literacy, physical vulnerabilities, social and home environment, and coping skills when selecting patients for care management. 

		Clinicians also considered the extent to which patients’ care needs were met by strong existing linkages to care providers outside of the primary care practice.

Healthcare delivery systems are increasingly entering into risk contracts which require healthcare providers to manage the utilization of defined patient populations. Care management programs (CMPs), particularly those based in primary care practices, have become a common strategy to meet the cost containment and quality targets in these contracts. Due to the concentration of costs in persons with multiple chronic conditions, these persons are often selected for care management interventions that aim to enhance the coordination of health services and improve patients’ ability to navigate the health system.

 Appropriate case finding, or the identification of patients most likely to benefit from care management, is key to maximizing the cost-effectiveness and sustainability of CMPs.

Many CMPs rely on quantitative approaches to high-risk patient identification, such as risk prediction models based on prior healthcare utilization and clinical conditions. These approaches typically rely on administrative data, such as billing claims, to predict future healthcare utilization, but are often unable to include more nuanced patient characteristics, including socioeconomic and psychosocial circumstances.

 Claims-based approaches often fall short of identifying the patients who are most likely to engage in and benefit from care management because they are unable to detect factors such as patient willingness to participate or barriers to participation.

 Moreover, use of data-driven algorithms often introduces a time lag between when patients need help and when they are identified for the program.

An alternative strategy is to rely on direct referrals from primary care physicians (PCPs) to identify appropriate patients. This strategy is based on the premise that PCPs know their patients best and can consider patient characteristics not routinely available in administrative data or medical records.

 However, PCPs vary widely in their referral patterns for such programs,

 and concerns have been raised about underselection of patients with poor existing linkages with the healthcare system.

 Finally, physicians with a poor understanding of care management goals or approaches may refer inappropriate patients to these programs.

Hybrid approaches that combine predictive modeling and clinician input have been suggested as a way to overcome the biases and limitations of the individual approaches.

 Prior research has explored the extent to which primary care clinician case finding agrees with predictive modeling case finding,

 but has not examined the added benefit of clinical review of a list of potentially high-risk patients generated from a predictive algorithm. We investigated the contribution of clinician input in the selection of high-risk patients within an active primary care-based CMP using semi-structured interviews of PCPs and their nurse care managers (CM). We were particularly interested in the criteria and patient factors clinicians used in practice when determining which potential high-risk patients within their panel should be targeted for care management. Specifically, we examined the extent to which clinicians considered: 1) patient characteristics that are and are not routinely available in clinical and billing data; and 2) existing patient linkages to other providers within the healthcare system when reviewing lists of potential high-risk patients, and identifying patients appropriate for practice-based care management. We hypothesized that clinician involvement enhanced the case selection process by allowing the inclusion of information about the patients not readily available in administrative data.

Description of Care Management Program and Patient Selection

The CMP is a primary care-based, nurse-led longitudinal care program designed to help high-risk patients (ie, chronically ill, medically complex patients) better manage their health and healthcare utilization. Approximately 400,000 Medicare and commercially insured patients 18 years or older were eligible for the CMP in 2013. CMP patient selection starts with the use of a billing claims-based predictive algorithm to identify an initial list of high-risk patients. As a first step, chronic conditions, utilization patterns and a predictive risk score for future total medical expense are generated using the prior year of paid claims using Optum ImpactPro version 6.0 software (Optum Inc, Eden Prairie, Minnesota). Combinations of specific chronic conditions and patterns of healthcare utilization, as well as overall risk score, are then used to identify the initial set of high-risk patients using an algorithm developed with input from an advisory committee of operational and clinical experts within the healthcare organization. Approximately 5% of patients are identified as potentially high risk based on the claims-based algorithm. These patients are then sorted into PCP-specific lists for clinical review. Lists provided to PCPs included the chronic conditions, utilization, and predictive risk score for each patient, as well as patient name and medical number so clinicians could easily recognize their patients. Each PCP, together with their CM, was responsible for reviewing between 1 and 39 potentially high-risk patients for 2 insurance-risk contracts in 2013, the second year of the program. Overall, clinicians selected approximately 61% of patients on the algorithm-generated list for the CMP. Clinicians were asked to classify patients on their list according to whether they were appropriate for the CMP, or they were not appropriate either because they were not considered high-risk or because the patient’s care needs were already being met. PCPs were given considerable latitude in how they classified and selected their patients for the CMP.

Using semi-structured interviews of PCP and CM dyads, we elicited and analyzed the patient characteristics that clinicians considered when selecting patients for CMP. This project was undertaken as a quality improvement initiative, and as such, was not formally supervised by Partners Healthcare’s Institutional Review Board per their policies.

We identified a convenience sample of 20 PCP and CM dyads from 18 ambulatory practices within an integrated healthcare system consisting of 270 primary care practices located in eastern Massachusetts. Physician organization leaders identified participants with the intention of maximizing variation in practice characteristics, such as the level of affiliation with the delivery system, location, and type (

). The CMP director then confirmed willingness to participate.

A physician researcher (CH) and project manager (VH) conducted all interviews, either by phone or in person, at the respective practices. Interviews lasted 45 to 60 minutes and were audio recorded with consent from all participants. All interviews were completed between May and September 2013. A research assistant (LM) transcribed the audiotapes verbatim.

Interviews followed a semi-structured, open-ended interview guide developed through an iterative process by the research team, and designed to obtain general programmatic quality improvement information, as well as information on the characteristics of patients selected for care management. Relevant interview questions included (see the 

] for the complete interview guide): 1) “How do you define a ‘high-risk’ patient and what criteria do you use to determine if a patient is ‘high-risk’?” 2) “What types of patients did you remove because they were ‘not high-risk’?” 3) “What types of patients did you remove because they have their care needs met elsewhere?”

The Andersen Model of Health Services Utilization

 provided a basic framework for organizing interview responses according to whether they reflected patient predisposition to use services (predisposing factors), characteristics that enable or impede healthcare use (enabling factors), and patient need for healthcare (need). We coded and analyzed interview responses using QSR International’s software, NVivo version 9, a qualitative data management and analysis software tool (QSR International Pty Ltd, Doncaster, Australia). In the first level of analysis, an interviewer and a research assistant (VH and LM) independently read and thematically coded all interview transcripts—a process that allows coders to organize and evaluate similar quotations by themes.

 Through an iterative process, content-related minor themes or factors were collapsed into major factors, then minor and major factors were discussed until a mutual agreement was reached. We met with clinician researchers and operational leaders to solicit feedback, present disconfirming evidence, and validate the coding scheme.

 After coding the data into this framework, we retained and summarized the major and minor predisposing, enabling, and need factors mentioned in at least 20% of interviews. Exploratory analyses examined variations in themes by practice affiliation and model.

Although limited to a single large integrated delivery system, our convenience sample of interviewed practices represents a broad range of practice affiliations, locations, and models (Table 1). The major and minor factors that PCPs and CMs considered in deciding which patients would be appropriate for care management within their primary care practice are summarized in 

 and exemplary quotations are provided in 

Most interviewees (85%) considered a patient’s ability to comprehend medical instructions and navigate the healthcare system when determining whether patients were appropriate for the CMP. Specifically, more than half (55%) of interviewed PCPs and CMs reported that they considered patient need for education about their diseases and appropriate management as they selected high-risk patients. Almost half of interviewees suggested that activated patients (ie, patients who are good “advocates for themselves,” “well-versed and knowledgeable about what they need,” and “can take care of themselves”) may not benefit from the additional resources provided by care management. On the other hand, interviewees felt that patients with linguistic or cultural barriers, as well as individuals with impaired executive function, would be more likely to benefit from the services rendered by CMs.

More than half (60%) of interviewees considered whether their patients had physical vulnerabilities, including a high risk of falling, frailty, or advanced age, when determining appropriateness for CMP. Clinicians reported that, in some cases, they would select patients with relatively stable functional status if they had physical vulnerabilities because they felt care management might help facilitate patient care in the event of decompensation.

Patient insight regarding his/her health. 

Just over half (55%) of interviewees considered the extent to which patients accurately perceived their health status and their need for assistance. This included elderly patients who were reluctant to lose their independence, younger patients in denial of their medical conditions, or patients who minimize their health needs. Interviewees felt these types of patients lacked an accurate perception of their limitations and needs, potentially leading them to refuse certain types of medical care or assistance. Interviewees stated that they would include these patients in the program, but that it might be challenging to engage and impact these types of patients.

 All interviews mentioned that they considered their patients’ social and home environment when assessing appropriateness for the CMP. Interviewees observed that relatives often serve as patients’ advocates and caretakers, helping patients navigate the health system, monitoring medication regimens, and noting changes in the patient’s health status. A large majority (90%) of interviews considered the absence of patients’ familial and social supports as reasons to include patients in the CMP. Instability in a patient’s home environment was mentioned as a consideration in a large majority (75%) of interviews, with issues ranging from addiction and verbal abuse to hoarding, as well as spousal illness, hospitalization, or recent death. Some (30%) interviewees mentioned the social isolation of living alone as an added risk factor for elderly and frail chronically ill patients. Finally, lack of access to transportation, home delivery of food, or other community resources were also considerations for CMP inclusion (30%).

Over half (65%) of interviewees noted that they would select patients with underlying mild psychological concerns, such as health anxiety, who often seek reassurance from medical professionals, whether through frequent phone calls or urgent care visits. Such patients may benefit from a CM, who could act as a point-person to help them manage anxiety related to their medical conditions without consuming limited office resources.

A minority (35%) of interviewees mentioned socioeconomic instability (ie, chronic unemployment or unstable housing), and its effect on a patient’s ability to manage his or her chronic illness(es). Medication adherence issues related to financial instability were also considered.

All interviewees considered their patients’ chronic conditions, including the severity or stage of the patient’s disease, and the patient’s complexity or combination of conditions. Interviewees considered whether complicated medication regimens placed their patients at higher risk for decline, complications, and hospitalization. Interviewees specifically highlighted patients with co-occurring medical and psychiatric disorders, such as depression, which they felt further complicated patient care and the patient’s ability to manage his or her medical condition(s).

In identifying patients appropriate for CMP, interviewees considered factors in isolation, but also described the interplay between patient factors, and how the combination of these factors either supported or impeded a patient’s ability to effectively manage his or her health.

Once clinicians determined whether patients were high-risk, interviewees then considered whether the CMP would be the right program for the patient. Interviewees considered whether the patient was already connected with other clinicians, and if the resources available through the CMP would address the patient’s needs. Specifically, they reported considering existing relationships with specialist physicians (eg, psychiatrists, cardiologists) or specialty clinics (eg, transplant, oncology), private CMs, or institutional care providers (eg, group homes, assisted living). Decisions regarding whether care needs were met relied on clinicians’ perceptions of whether there were additional resources or services the CMP could provide, such as transitions to end-of-life care. Moreover, interviewees’ understanding of the CMP also impacted their decisions about whether to include certain subgroups of patients. For instance, some interviewees excluded patients whose primary diagnosis was a psychiatric condition or substance abuse because they felt the CMP did not yet have the resources to meet these patients’ needs.

Our study evaluated the factors clinicians considered important in determining whether potentially high-risk patients within their practice would benefit from a practice-based CMP. Similar to others, we found that clinicians consider patients’ predisposing factors, including health literacy and patient self-efficacy, as well as enabling factors, such as home and family supports, when identifying patients appropriate for care management.

 We found that these factors remained important, even in our hybrid approach to case finding, in which clinician decision making is limited to a set of risk contract patients with prior high-cost care patterns and complex chronic illness. Additionally, clinicians in our investigation considered whether existing medical and community resources were already addressing patient needs and whether CMP would enhance or interfere with existing patient care networks.

Many of the factors clinicians considered in their decision making are not routinely captured in clinical or billing data, and are therefore unavailable for case finding.

 For example, patient home environment and social support, as well as patient health literacy, were factors considered by a large majority of clinicians when identifying patients appropriate for CMP. This information is not contained in billing records and is not routinely captured in electronic health records. 

Previous research found only a small subgroup of patients concordantly selected by PCPs and predictive modeling for care management.

 Compared with PCP-based case selection, a predictive modeling approach tends to identify patients who are at a higher risk of healthcare utilization and have a higher morbidity burden,

 but are less likely to be helped by care management, either because their utilization is not preventable or they are unwilling or unable to participate.

The hybrid approach in this study builds on these findings. Starting with a predictive model to identify a subset of patients that may be appropriate for CMP effectively limits the review burden on clinicians, and focuses the clinical review on the subset of patients who are enrolled in a risk contract and have chronic conditions and patterns of prior healthcare utilization that may be amenable to care management.

More systematic capture of clinical and demographic data may become standard in newer electronic medical records or health information exchanges,

 but many of the nonclinical, nuanced factors clinicians reported considering are unlikely to become standard elements of clinical documentation, either because of privacy concerns or because of their limited relevance to standard clinical care.

Incorporating clinical input in patient selection addresses some of the limitations associated with the use of claims or medical record-based algorithms for case finding, but might also introduce new biases. Physicians and practices may vary in their threshold for CMP referral or how they weigh specific criteria. For example, some practices may have existing programs or supports, such as special programs for patients with heart failure or diabetes, or close ties with cancer centers that might impact their patient selection. Variations in decision making may also be related to differences in knowledge or acceptance of team approaches to patient care. For instance, we found that some clinicians were unsure whether the CMP would have the necessary resources for patients with co-occurring mental health and substance use. More training and education as to the types of patients who would be a good fit for CMP may improve the patient selection process.

Clinicians consider a number of patient predisposing and enabling characteristics not typically available in clinical data systems when selecting high-risk patients for care management.

Christine Vogeli, PhD

Articles

Improving Care Coordination and Reducing ED Utilization Through Patient Navigation

Implementing a Hybrid Approach to Select Patients for Care Management: Variations Across Practices

Clinician Considerations When Selecting High-Risk Patients for Care Management

Assessing Outcomes in Child Psychiatry