Effect of Physician Delegation to Other Healthcare Providers on the Quality of Care for Geriatric Conditions

The U.S. healthcare system is in a period of unprecedented change, driven by a rapidly aging population living with a greater burden of chronic conditions and broad consensus that the United States must optimize the value of care—that is “outcomes relative to costs.”1 The quality of care for outpatient management of common health problems in this population has consistently been shown to be inadequate.2-5 This gap between recommended and actual care provides an opportunity to improve the value of health care for older adults.6

Efforts to improve quality of care begin by defining quality. The Assessing Care of Vulnerable Elders (ACOVE) project developed quality indicators (QIs) to evaluate the processes of care provided to older Americans.7 The ACOVE QIs were derived by combining expert opinion with a systematic literature review for 22 common conditions, including dementia and memory loss, urinary incontinence (UI), and falls and dysmobility.2, 8 These QIs were updated and expanded to 26 conditions in 2007.9 In conjunction with the development of these QIs, the ACOVE investigators also developed a model of outpatient clinical practice change (the ACOVE 2 Model) to improve outpatient care processes though a structured intervention involving case finding, delegated clinical data collection, structured visit notes, physician and patient education, and linkage to community resources. Implementation of this model repeatedly demonstrated better quality of care for geriatric conditions in primary care.4, 5, 10-12

An important component of the ACOVE 2 intervention was the emphasis on team care, with office staff performing clinical work customized to their respective licenses and scopes of care, for tasks as varied as history taking, standing orders, problem-focused counseling, and referral to community-based organizations. In this manuscript, “delegation” is used as shorthand to describe which practitioners provide specific care processes. In the ACOVE studies, clinical sites determined how care processes would be implemented based on their own staffing, experience, needs, and decision-making processes. Thus, although the investigators were able to describe the extent to which these sites delegated care to different practitioners, they did not recommend how this should be accomplished. Instead, the investigators relied on the solutions and work flows that the sites developed themselves.

Prior work from two ACOVE studies demonstrated that integration of the ACOVE 2 intervention, as provided by advanced practice nurses, resulted in better care processes for dementia, depression, falls, heart failure, falls, UI, end-of-life care, and other conditions.4, 10 These findings suggest that a team-based approach, and one that encourages all care providers to work to the fullest extent of their licenses, may result in improved measures of quality. However, the independent relationship between delegation and quality of care has not been determined. The current analysis used pooled data from eight ACOVE studies to examine the effect of delegation on quality of care for three geriatric conditions: dementia, UI, and falls.

Methods

The ACOVE studies focused on whether people receive recommended processes of care for up to 26 target conditions. Care processes were categorically defined as preventive, diagnostic, treatment, or follow-up and were provided largely in primary care offices. (See prior ACOVE studies for a comprehensive list of conditions and QIs.)9 Eligibility criteria for QIs were based on individual study enrollment and QI criteria (detailed in Appendix S1 and methods of prior studies). For each ACOVE study, specially trained nurses reviewed all relevant medical records using reliable abstraction instruments to determine whether recommended care processes were received. If a provider offered a care process, the care process was scored as having been provided (and therefore the QI “passed”)—even if the individuals refused the care process.8

For this study, a subset of three geriatric conditions that were common in previous studies were evaluated: falls and fear of falling (12 QIs), cognitive impairment and dementia (19 QIs), and UI (16 QIs). To allow use of data collected over many years, ACOVE 1, ACOVE 2, and ACOVE 3 QIs were reconciled to align changes reflecting updates in best practices (see Appendix S1 for ACOVE 3 QIs). The institutional review boards at RAND, the University of California at Los Angeles, and when appropriate, the Veterans Affairs Healthcare System and local sites approved the ACOVE studies. The studies from which the data were obtained for this analysis are as follows (Table 1).

The ACOVE 1 study was an observational cohort study from a random sample of community-dwelling individuals age 65 and older enrolled in two managed care organizations. Individuals and QIs were identified using retrospective chart review during the study period.2

The ACOVE 2 study explored the effects of a multicomponent intervention on the quality of care for dementia, falls, and UI in two large, multisite practices using a quasi-experimental design.13 ACOVE 2 enrolled individuals aged 75 and older who screened positive for one of the target conditions.

The ACOVE Prime study used a quasi-experimental design to investigate the effects of a multicomponent intervention (modified version of the ACOVE 2 intervention model) for the care of falls and UI at five small and medium-sized community-based group practices nationwide.11

The ACOVE Alzheimer's Disease study used a pre- and postdesign to evaluate an intervention to improve dementia care using the ACOVE 2 model at two community-based practices. The intervention partnered with local Alzheimer's Association chapters to provide community-based support.5

The ACOVE Evercare study investigated the effects on quality of care of nurse care management in addition to standard care for a multimorbid population of older adults enrolled in a Medicare Advantage Special Needs Plan.14

The ACOVE UCLA Nurse Practitioner (NP) study used a quasi-experimental design to investigate the effects of NP co-management of older adults for depression, dementia, falls, heart failure, and UI in an academic geriatrics practice.4

The ACOVE UniHealth study used a case-study design to investigate the effects of NP co-management of dementia, depression, falls, and UI in two community-based practices.10

The Senior Health project was an observational evaluation of a population-based sample of Medicare beneficiaries in fee-for-service Medicare and a Medicare Advantage plan (unpublished data).

In standard clinical practice, physicians complete most ACOVE QIs, but many of these care processes, such as history taking and orthostatic vital signs, could be delegated to less highly trained providers. For the intervention groups in each study, each practice site determined whether and to whom a care process would be delegated. Using this site-determined approach, tasks were classified as being completed by physicians; NPs, physician assistants (PAs), or registered nurses (RNs); or medical assistants (MAs) or licensed vocational nurses (LVNs). Based on this classification, each QI was classified according to a site-specific intention to delegate based on the workflow in that clinic. (The concept of intention to delegate can be thought as analogous to intention to treat, but representing the intended care provider who should perform a task.) Intention to delegate is a practice workflow decision about how all individuals who meet certain clinical criteria should undergo a specific care process.

A simple ordinal scoring system was then created for this intention to delegate (delegation score). Scores were ordinal values between 1 and 3, with 1 representing tasks intended to be completed by physicians (no delegation); 2 for tasks intended to be performed by NPs, PAs, or RNs; and 3 if tasks were intended to be performed by LVNs or MAs. In short, the higher the delegation score, the less trained the provider was to whom the task was delegated. For example, a site where MAs asked patients about the basic history for a recent fall (ACOVE 3 QI: Fall 2) would receive a delegation score of 3; a clinic that relied on physicians to ask about the circumstances of a recent fall would receive a delegation score of 1. The level of delegation that a clinic or site used was assigned for each ACOVE QI, because delegation level at each clinic could be different for each QI (or differ within a clinic or site if the study had intervention and control arms). All patients eligible for a particular QI at a particular site were then assigned the QI-specific intended delegation score for that site. QIs for which researchers were unable to determine a delegation score were assigned a default value of 1 because this represented the usual practice of physicians being responsible for performing all elements of care.

There was concern that the ease of delegating the QI might be a confounder, related to whether the task was delegated and to the pass rate. By design, some QIs are easier to delegate than others. For example, reflexively ordering a urinalysis based on complaints of UI (ACOVE 3 QI: UI 6) is easy to delegate, but review of psychotropic medications or gait examination after a fall at home (ACOVE 3 QI: Falls 2 and 5) may require a higher level of training and expertise. To control for this possible confound, an additional variable was created for each QI to assess the ease of delegation. This covariate, QI maximum delegation, is a three-category variable, with a score of 3 indicating that at least one site in all the pooled ACOVE studies had a QI delegation score of 3 for that QI, a score of 2 indicating that at least one site had a QI delegation score of 2, and a score of 1 indicating that no site delegated this QI to nonphysician staff. This variable represents the maximum real-world delegation the QI had in any ACOVE study. Maximum delegation can be used to compare the choices of a practice (e.g., intention to delegate) with those of other practices or a theoretical “most efficient” practice to determine the effects of practice choices on quality.

Results

The eight studies conducted between 1998 and 2010 that are the basis of this pooled analysis included three observational cohorts, four intervention-only practices, and eight clinic sites with usual care and ACOVE 2 intervention practices. Thus, overall across the eight studies, there were 11 usual care practices and 12 intervention practices (Table 2). No site was involved in more than one study.

The participants in the ACOVE studies were all older Americans (aged ≥65) but were culturally and financially mixed. All were insured through Medicare (fee-for-service or a Medicare Advantage plan). Approximately two-thirds were female, and participants had a mean age of approximately 80. For studies that categorized ethnicity or race, 51% to 95% of patients were non-Hispanic white (Table 1). No practice cared for a veteran or indigent population. One practice was based at an academic health center. Two practices maintained relationships with academic institutions, with one loosely affiliated with a family medicine residency program and the other serving as a site for internal medicine residency and geriatric fellowship trainees. The practitioners at most sites were general internists or family medicine physicians, with a few having pursued additional geriatric fellowship or postresidency training. Five studies involved sites with NPs and/or physician assistants.

Across all sites, 4,776 patients were eligible for at least one QI, with 2,911 patients in the pooled intervention group and 1,865 in the pooled control group. Patients at intervention sites were eligible for 8,180 QIs and patients at control sites were eligible for 8,024 QIs. Of the 47 QIs studied, 37 (79%) involved tasks that at least one site delegated to a non-physician provider (level 2 or 3 delegation); this yielded 3,143 tasks that were delegated to an NP, PA, or RN, and 881 tasks there were delegated to a MA or LVN. There was high correlation between QI delegation and the ACOVE intervention (correlation coefficient (r) = 0.62, P < .001) and moderate correlation between QI delegation score and QI maximum delegation (r = 0.26, P < .001).

Results from unadjusted analysis of QI pass probability and delegation demonstrated a strong association between QI pass rate and delegation. The QI pass probability was 0.36 for physician-performed tasks (delegation level 1), 0.55 for NP- or PA-delegated tasks (delegation level 2), and 0.61 for LVN- or MA-delegated tasks (delegation level 3). Relative to physician-performed tasks (delegation level 1), the probability ratio was 1.53 (95% confidence interval (CI) = 1.31–1.80, P < .001) for delegation level 2 and 1.69 (95% CI = 1.39–2.05, P < .001) for delegation level 3.

In the multivariable analyses (Table 2), the adjusted pass probability ratio versus no delegation was 1.37 (95% CI = 0.99–1.89, P = .05) for delegation level 2 and 1.06 (95% CI = 0.88–1.27, P = .55) for delegation level 3 . Of the covariates, the ACOVE intervention study groups were associated with a higher passing probability (pass probability ratio 1.73, P < .001) compared with the ACOVE usual care study groups. Passing probability varied according to QI condition, with fall QIs having a significantly higher passing probability than UI (reference group), although there was no statistically significant difference in the passing probability between dementia and UI QIs. Passing probability also varied according to QI domain, with diagnosis QIs having a significantly lower passing probability and screening-prevention QIs having a significantly higher passing probability than treatment QIs (reference group).

In sensitivity analysis, which included QI maximum delegation as an additional covariate, QI delegation score was no longer associated with QI pass probability, but maximum delegation score was significantly associated; the pass probability ratio for QI maximum delegation of 2 was 1.95 (95% CI = 1.55–2.44, P < .001). Results for the other covariates were similar to the base-case analysis (Table 2). Intervention sites had significantly higher QI passing probability than the ACOVE usual care sites (95% CI = 1.39–1.86, P < .001). Fall QIs had higher passing probability than UI QIs (95% CI = 1.11–1.4, P < .01), and screening and prevention QIs had higher passing probability than treatment QIs (95% CI = 1.26–1.97, P < .001).

Discussion

This pooled analysis of eight studies using ACOVE QIs found that delegation improved the quality of care provided for three common geriatric conditions. Moreover, the processes of care that were improved have been shown to correlate with better outcomes.15, 16 In addition, a secondary analysis examining the maximum delegatability (the lowest level of training to which any ACOVE study site delegated the QI) suggests that more delegation is possible and might result in higher quality scores.

These findings build on an emerging literature about the potential benefits of delegated, team-based outpatient care. For example, a recent metaanalysis for common medical conditions demonstrated that, for the management of hypertension, dyslipidemia, and diabetes mellitus, nurse-managed protocols result in small but significant improvements in secondary outcomes including blood pressure control, cholesterol levels, and glycosylated hemoglobin levels.17 Similarly, individual studies and metaanalyses suggest that NPs and physicians generally produce similar health outcomes for some common conditions,18-20 and co-management of depression has resulted in better mental health outcomes, care use, and cost.21-23

Demonstrations of patient-centered medical homes, another model for team-based care, have also shown modest improvements in care processes for preventive services in general medical population, although the effect for older adults with multimorbidity has been inconsistent.15, 24-26

These findings must be considered in light of the study's limitations. First, the analyses included observational and quasi-experimental studies. Second, there was high correlation between QI delegation and the ACOVE interventions, making the ascertainment of an independent effect of delegation more difficult. Moreover, delegation was only one element of a multicomponent intervention, and the intervention itself had a large independent, statistically significant effect on the probability of passing QIs. Third, sites were classified according to intention to delegate and not actual delegation at the level of individual patient. It is likely that some individual exceptions may have occurred in actual practice that differed from the intended delegation model. This may have resulted in some misclassification of actual delegation, which it is likely would attenuate the effect size in the models. Finally, the number of QIs that were delegated to LVNs or MAs was small, limiting the power to detect differences between levels 2 and 3 of delegation. Finally, this study examined only three geriatric conditions. The effect of delegation on the quality of care for other conditions remains to be determined.

Interventions in the process of care, including delegation, represent unique opportunities to enhance the quality of care within existing healthcare infrastructures. The ACOVE 2 intervention encourages all team members to work at the highest level commensurate with their training and that services (tasks) be delegated to the lowest level of professional training.27 This model attempts to leave persons with greater training or responsibility free to perform tasks or solve problems for which they are uniquely qualified.

Conclusion

The delegation of specific processes for the management of UI, dementia, and falls to lower-level providers is associated with higher quality care for these conditions. These findings suggest that efficiency and quality can be improved through team care with greater delegation of care processes. Additional research should aim to identify which tasks can be delegated for other conditions and determine the best ways to implement delegation in practices of various sizes.