Healthcare 3 (2015) 89–96
Contents lists available at ScienceDirect
Healthcare journal homepage: www.elsevier.com/locate/hjdsi
Complexity of ambulatory care across disciplines David Katerndahl n, Robert Wood, Carlos Roberto Jaén Family & Community Medicine, University of Texas Health Science Center San Antonio, 7703 Floyd Curl Drive, San Antonio, TX 78229-3900, USA
art ic l e i nf o
a b s t r a c t
Article history: Received 11 August 2014 Received in revised form 15 January 2015 Accepted 3 February 2015 Available online 27 February 2015
Background: Complexity of care has implications for quality of care, health costs, medical errors, and patient and physician satisfaction. The objective was to compare complexity of ambulatory care across 14 medical specialties. Methods: This secondary analysis uses the 2010 National Ambulatory Medical Care Survey, which used a multistage probability design of primary sampling units throughout U.S. ambulatory practices across 14 specialties. Sampling weights enable results from 29,179 ambulatory visits to represent 878,653,561 visits. Data included symptoms, diagnoses, diagnostic procedures, and treatments provided. Measures of input, output and total encounter complexity and hourly complexity densities were computed. Results: Internal Medicine leads in total input and total encounter complexity with Family Medicine second in total encounter complexity. When duration-of-visit is considered, Family Medicine is the most complex discipline while Internal Medicine is the second most complex. Pediatrics lacks the complexity of Family Medicine and General Internal Medicine, and OB/GYN bears little similarity to Family Medicine or General Internal Medicine. Conclusions: Family Medicine and Internal Medicine encounters are the most complex overall, especially when duration-of-visit is considered. Implications: Revaluing payments based on complexity could bring better balance to cognitive and procedural services, and better meet the needs of people receiving insurance under the ACA. & 2015 Elsevier Inc. All rights reserved.
Keywords: Ambulatory care Complexity of care Interspecialty comparison System theory
1. Introduction Complexity of care is increasing. From 1997 to 2005, the mean number of clinical items addressed during an office visit increased from 5.4 to 7.1.1 From 1995 to 2005, the proportion of patients on medications increased from 21% to 44% while the proportion taking at least three drugs also increased.2 From 2002 to 2010, medications prescribed to adults rose 22%.3 From 1997 to 2005, office time spent per clinical item addressed declined.1 With health care reform, further change is anticipated. With the influx of newly-insured individuals, patient diversity and psychiatric comorbidity should increase. Increasing demands for documentation and electronic health records (EHR) use will not increase face-to-face durations-of-visit, but increase the time required to care for each patient. With increases in demand for care, physician shortages may become exaggerated, potentially shifting patients among existing specialties or decreasing durations-of-visit. Finally, primary care disciplines may uniquely experience increasing numbers of chronic medical problems,
n
Corresponding author. Tel.: þ 1 210 358 3200; fax: þ 1 210 223 6940. E-mail address: [email protected] (D. Katerndahl).
http://dx.doi.org/10.1016/j.hjdsi.2015.02.002 2213-0764/& 2015 Elsevier Inc. All rights reserved.
complexity of medication regimens, numbers of guidelineindicated services, demand for preventive services, and pressures for accountability and performance. Increased use of urgent care centers may leave primary care disciplines with even greater proportions of chronic care, multimorbidity visits. Such increasing complexity could affect the quality of health care. When medical care becomes more complex, practice guidelines are less effective4 and the rate of errors occurring rises.5 The risk of errors increases with seeing multiple patients with multiple conditions, use of multiple medications, and implementation of complex procedures.6 Complexity of care may partially explain why only 55% of adult patients receive recommended care.7 Poor quality of care was especially noted for time-intensive activities, such as history-taking, counseling, and patient education7 as well as screening and preventive medicine.8 In addition, not only is complex care at risk for inefficiencies, but to cope with perceived complexity, physicians may increase testing or lower the threshold at which they refer patients to specialists, adding to the cost of care. Finally, complexity of care could impact patient and physician perceptions of time adequacy and satisfaction. This could explain a decline in perceived autonomy and career satisfaction,9 with perceived autonomy poorer among primary care physicians.10 Yet, we know little about the differences in complexity of
90
D. Katerndahl et al. / Healthcare 3 (2015) 89–96
ambulatory care across disciplines; such differences have implications for the cost of care and its payment. Although previous work found that the complexity of ambulatory care in 2000 was the highest for Family Medicine compared with that of Cardiology and Psychiatry,11 no recent analysis has been conduct or involved more than three medical specialties. The purpose of this study was to compare the complexity of patient care across 14 specialties using data from the 2010 National Ambulatory Medical Care Survey (NAMCS).
2. Materials and methods Using a method designed to estimate complexity of ambulatory care derived from national databases,12 relative complexity can be estimated (computational details are provided as supplementary material). This study analyzed data from the most recentlyavailable NAMCS database (2010) for 14 medical specialties (Family Medicine, General Internal Medicine, Pediatrics, OBGYN, Cardiology, Dermatology, Neurology, Oncology, General Surgery, Orthopedics, Urology, Ophthalmology, Otorhinolaryngology (ENT), and Psychiatry). NAMCS used a multistage probability design of primary sampling units (PSUs) throughout the U.S., practices within PSUs, and patient visits within practices, designed specifically to be representative of all ambulatory visits in the United States. Trained physicians and office staff completed encounter data on patient visits selected, including patients’ symptoms, physicians’ diagnoses, diagnostic procedures, and treatments provided.
2.1. Computation of complexity of each input/output (see Fig 1) If we define “complexity” of a system as the amount of information needed to describe it or its behavior,5 then complexity (like error) is associated with volume, diversity, variability, and time limitations.6 It includes “cognitive complexity” (focused on the content of information flowing) and “relational complexity” (focused on the interactions by which information flows among agents).13 Cognitive complexity is measured in counts, while relational complexity is measured in variability. Any measure of complexity must reflect the range of problems and inputs seen by the physician.14 However, because a specialty is not defined by a single encounter, the measure of complexity needs to include inter-encounter variation as well. Whereas the complexity of an encounter includes the number of events occurring and the amount of information transferred, the complexity of a specialty must include the diversity (range of characteristics seen) and variability of events across encounters. Thus, “complexity of care” should reflect the complexity of the typical encounter and the complexity across encounters.15 Approaches such as case-mix measures, risk adjustment approaches, patient severity (or risk) measures, patient complexity estimates, and clinical problems per hour fail to measure complexity of care because they either equate severity with complexity, represent single patient measures, fail to capture all relevant dimensions, and/or assume linearity.16 Patient encounters and their components can be described by the quantity of information and services exchanged between patient and physician, by the visit-to-visit variability in quantity of these exchanges, and by their overall diversity.12 “Quantification” of visits included the per-patient number of reasons for visit, diagnoses, body systems examined and tests ordered, medications prescribed, other therapies and procedures ordered. Whether patients were new to the practice was also recorded. “Variability” was measured across visits within discipline by computing the coefficient of variation (COV)—a unit-free measure—from the
mean and standard deviation from quantities measured above. In addition, the COV of the age of patients seen was also computed. “Diversity” was defined as the proportion of categories needed to describe 95% of the visits for each specialty. The 95% proportion was chosen to minimize the impact of a rare or miscoded input/ output. Finally, patient demographic diversity was assessed as the proportion of categories within a gender X race X ethnicity matrix used by NAMCS needed to describe 95% of patients seen. The NAMCS databases provide a patient weight that allowed the 2010 sample of 31,229 visits to represent 1,008,802,005 visits that year in the United States.17 This patient visit weight was applied to the dataset so that estimates of complexity parameters produced by re-sampling techniques would better conform to national patterns of patient encounters. Visit input depended upon the reasons for visit, diagnoses, examination/testing, and patient characteristics. Visit output depended upon medications (including vaccines) and other therapies (including procedures, education, physical therapy, nutrition, etc.) prescribed, and visit disposition. In addition to the mean quantification for each variable, differences in discipline-specific duration-of-visit were used to determine an hourly complexity rate for each discipline. The complexity of each input/output was defined as the mean input/output quantity per clinical encounter weighted by its inter-encounter diversity and variability. To standardize the weightings and limit the impact of low diversity or variability on complexity, the weightings used were the Z-transformations of the diversity proportion and the COV, ranging between 0.5 and 1.0. 2.2. Computation of total complexity (see Fig 1) Total input and output complexities were calculated by summing their component complexities. However, calculation of total complexity was not merely the sum of input and output complexities. A fundamental principle of complex systems is that a logarithmic relationship exists between input and output, such that, as the information in the input increases linearly, the complexity of the system increases exponentially. To calculate total encounter complexity, total output complexity was multiplied by “2” raised to the power of the input complexity. Thus, total system complexity depends more heavily upon the complexity of the input.5 2.3. Complexity density (see Fig 1) The estimate of complexity of ambulatory care presented is a measure of the complexity of the typical clinical encounter. However, coping with complexity is time-dependent.13 An hourly complexity density estimate was derived by dividing the total complexity estimates by the duration-of-visit and then multiplying by 60. 2.4. Analysis The 95% confidence intervals were derived from bootstrap resampling procedures based on 500 samples, enabling comparisons across disciplines. Friedman’s test was used with multiple comparison posthoc testing to seek significant differences of complexity estimates among disciplines. Cluster analysis of component measures was used to group disciplines. First, to identify the number of clusters, the K-means method was used with the visit components to examine 2-, 3-, and 4-cluster models. Based upon analysis-of-variance (ANOVA) results, the 3-cluster solution produced the most consistent clear-cut differences. Then, to assign disciplines to clusters, Z-scores were computed for each component measure. The squared Euclidean distance approach was used because it is sensitive to differences in the magnitude of
D. Katerndahl et al. / Healthcare 3 (2015) 89–96
INPUTS*
OUTPUTS**
QUANTITY Mean number per visit of each input across all visits to that specialty
QUANTITY Mean number per visit of each output across all visits to that specialty
VARIABILITY (COV) SD for each input across all visits to that specialty divided by its mean quantity and then Z-transformed
VARIABILITY (COV) SD for each output across all visits to that specialty divided by its mean quantity and then Z-transformed
DIVERSITY Proportion of input used by all specialties needed to account for 95% of that specialty’s visits and then Z-transformed
DIVERSITY Proportion of output used by all specialties needed to account for 95% of that specialty’s visits and then Z-transformed
INPUT COMPLEXITIES Multiply each input mean quantity by its Z-transformed COV and diversity
91
OUTPUT COMPLEXITIES Multiply each output mean quantity by its Z-transformed COV and diversity
Patient Complexity is sum of Ztransformations of the proportion of new patients seen, COV of patient ages and gender/race/ethnicity diversity of patients TOTAL INPUT COMPLEXITY Sum of all component input complexities
TOTAL OUTPUT COMPLEXITY Sum of all component input complexities
TOTAL ENCOUNTER COMPLEXITY Product of Total Output Complexity and “2” raised to the power of the Total Input Complexity
Mean visit duration
INPUT, OUTPUT & TOTAL COMPLEIXTY DENSITIES Complexity densities computed by dividing Total Input, Total Output and Total Encounter Complexities by the mean visit duration and multiplying by 60 * “Input” includes reasons for visit, diagnoses, examination/testing, and patient characteristics ** “Output” includes medications and non-medication therapies prescribed, and visit disposition “SD” = Standard Deviation “COV” = Coefficient of Variation Fig. 1. Flow diagram of computations used.
variables. Ward’s method was used to assign each discipline to a cluster because it yields small, distinct clusters. Complexities were compared across clusters using ANOVA with REGWF posthoc testing, with p r0.05 deemed significant.
3. Results Excluding the “other medical specialties” (n¼ 2050), the 2010 NAMCS database included 29,179 visits. Based upon the weightings, the overall results represents 878,653,561 visits. Bootstrap
standard errors showed that complexity measures were highly stable for all disciplines. Looking at the quantities of care (Table 1), Family Medicine, Internal Medicine, Cardiology and Neurology report the most reasons-for-visit and diagnoses while General Surgery, Urology and OBGYN report the least. In addition, ENT, General Surgery and Neurology see the most new patients. In terms of outputs, Cardiology and Oncology prescribe the most medications while Pediatrics and Psychiatry use the most non-medication therapies. Variability in numbers of reasons-for-visit and diagnoses was consistent across disciplines (see Table 2). High variabilities in
92
D. Katerndahl et al. / Healthcare 3 (2015) 89–96
Table 1 Quantification of ambulatory care provided across disciplines. Medical specialty (n¼ number of visits)
Input
Output
Mean reasons per visit
Mean diagnoses per visit
Mean exams/tests Proportion of new patients per visit among all visits
Mean of medications per visit
Mean of other therapiesa per visit
1.90
1.33
0.11
3.06
1.09
Primary care General/family medicine 1.64 (n ¼6237) General internal 1.61 medicine (n¼ 2217) Pediatrics (n¼ 3501) 1.47
2.10
1.76
0.10
3.45
0.80
1.50
0.82
0.05
1.56
1.61
OBGYN (n¼2461)
1.37
1.43
2.28
0.09
1.34
1.01
Medical specialties Cardiology (n¼ 1793) Dermatology (n ¼1510) Neurology (n¼1556) Oncology (n¼ 1478)
1.51 1.37 1.61 1.50
2.25 1.72 1.84 1.81
1.31 1.00 0.58 2.01
0.13 0.22 0.27 0.07
4.94 1.78 2.92 3.56
1.08 1.23 0.98 0.71
1.26
1.46
0.95
0.27
1.96
1.08
1.42
1.55
0.64
0.23
1.24
1.30
Surgical specialties General surgery (n ¼1161) Orthopedic surgery (n ¼1460) Urology (n¼1614) Ophthalmology (n ¼1368) ENT (n¼ 1471)
1.36 1.43
1.83 1.91
1.40 0.36
0.17 0.15
2.64 2.13
0.87 0.96
1.47
1.67
0.35
0.31
1.49
0.98
Psychiatry (n¼ 1352)
1.41
1.60
0.19
0.05
2.19
1.56
Means are weighted means per visit for each input and output for that specialty. “n” reports the number of visits per specialty upon which the estimates are based. Data based on 2010 NAMCS Survey a
Other therapies include procedures, patient education, nutritional counseling, physical therapy and other interventions.
Table 2 Variability of ambulatory care provided across medical specialties. Medical Specialty
Input
Output
COV for reasons per visit
COV for diagnoses per visit
COV for exams/tests per visit
COV of age of patients among all visits
COV for medications COV for other per visit therapiesaper visit
0.47
0.46
1.46
0.50
0.84
1.10
Primary care General/family medicine General internal medicine Pediatrics
0.49
0.40
1.30
0.31
0.78
1.28
0.47
0.49
1.59
1.22
0.97
0.98
OBGYN
0.46
0.50
0.97
0.39
1.19
1.06
Medical specialties Cardiology Dermatology Neurology Oncology
0.50 0.45 0.48 0.52
0.38 0.47 0.46 0.48
1.32 0.67 1.93 0.86
0.22 0.41 0.37 0.24
0.56 1.22 0.90 0.83
1.20 0.94 1.04 1.24
0.45 0.46
0.49 0.48
1.41 1.27
0.32 0.38
1.28 1.54
0.93 0.83
0.45 0.46 0.47
0.46 0.44 0.46
0.84 2.16 2.87
0.28 0.37 0.58
1.09 1.22 1.35
0.96 1.07 0.95
0.48
0.47
3.58
0.46
0.75
0.81
Surgical specialties General Surgery Orthopedic surgery Urology Ophthalmology ENT Psychiatry
COV represents weighted coefficient of variation for each input and output per specialty across all specialty-specific visits, calculated as the standard deviation of the input/ output across all visits divided by the mean (from Table 1) for that input/output. Larger COVs represent very high variability, typically due to significant number of visits with zero for that input/output (e.g., no tests ordered). Note that the COVs for “Reasons” and “Diagnoses” are similar across disciplines, suggesting that specialties were not differentiated from each other bases upon the variability of these inputs. Data based on 2010 NAMCS Survey a
Other therapies include procedures, patient education, nutritional counseling, physical therapy and other interventions.
D. Katerndahl et al. / Healthcare 3 (2015) 89–96
93
Table 3 Diversity of ambulatory care provided across disciplines. Medical specialty
Input
Output
Proportion of reason for visit categories needed to account for 95% of visits
Proportion of diagnostic categories needed to account for 95% of visits
Proportion of exams/tests categories needed to account for 95% of visits
Proportion of the 8 possible gender by race/ethnicity categories needed to account for 95% of visits
Proportion of medication categories needed to account for 95% of visits
Proportion of other therapya categories needed to account for 95% of visits
Proportion of 5 possible visit dispositionsbneeded to account for 95% of visits
0.44
0.31
0.68
0.88
0.26
0.66
0.60
Primary care General/family medicine General internal medicine Pediatrics
0.38
0.27
0.64
0.75
0.18
0.61
0.60
0.28
0.21
0.57
0.75
0.10
0.53
0.60
OBGYN
0.18
0.13
0.68
0.75
0.13
0.57
0.60
Medical specialties Cardiology 0.14 Dermatology 0.12 Neurology 0.16 Oncology 0.25
0.10 0.07 0.15 0.20
0.68 0.36 0.79 0.64
0.88 0.75 0.75 0.75
0.13 0.16 0.15 0.17
0.68 0.95 0.70 0.56
0.60 0.40 0.60 0.60
0.19
0.72
0.63
0.12
0.78
0.60
Surgical specialties General 0.24 Surgery Orthopedic 0.11 Surgery Urology 0.12 Ophthalmology 0.06 ENT 0.17
0.10
0.61
0.88
0.09
0.52
0.60
0.08 0.05 0.12
0.65 0.75 0.89
0.63 0.75 0.75
0.14 0.13 0.12
0.57 0.63 0.46
0.40 0.60 0.60
Psychiatry
0.03
0.79
0.75
0.05
0.48
0.60
0.08
Weighted proportion of possible categories needed to include 95% of visits. “Diversity” is the fraction of the total number of diagnoses used by all specialties needed by that specialty to describe 95% of their visits. The 8 possible racial/ethnic categories include: male/female Non-Hispanic White/Non-Hispanic Black/Hispanic/Non-Hispanic Other. Data based on 2010 NAMCS Survey. a b
Other therapies include procedures, patient education, nutritional counseling, physical therapy and other interventions. Disposition 5 categories included: return as need, scheduled return visit, referred to another physician, send to emergency department, hospitalize.
other descriptors were observed in those disciplines whose visits often do not include testing or therapies. High diversity for most inputs and medications was seen in Family Medicine and Internal Medicine while high diversity in testing was observed in ENT, Neurology and Psychiatry (see Table 3); high diversity in diagnoses for Family Medicine and Internal Medicine was found previously.18 However, low diversity in most inputs and outputs was reported in Psychiatry. Combining quantities, variability and diversities, complexity of encounter was computed (see Table 4). Primary care disciplines led in input (reasons-for-visit, diagnoses and patient characteristics) and output (other treatments) complexity components with General Internal Medicine leading in total input and total encounter complexity but Family Medicine a close second in total encounter complexity. When duration-of-visit is considered, Family Medicine becomes the most complex discipline with the highest total complexity density while General Internal Medicine is the second most complex. Using the rankings of input, output and total complexities and their densities, a significant difference across disciplines was found (Friedman’s ¼49.5, p o 0.001). Using multiple comparison procedure, Family Medicine was significantly (p r 0.05) more complex than Orthopedics, ENT and Psychiatry while General Internal Medicine and Cardiology were more complex than ENT and Psychiatry. Using ANOVA to determine intercluster differences, Table 5 shows that Cluster 1 consisted of Family Medicine, General Internal Medicine, Cardiology and Oncology while Cluster 2 included General Surgery, OB/GYN, Dermatology and Urology. Cluster 1 was distinguished by its high complexity of reasons-for-visit, diagnoses and medications with
resultant high input, output and total encounter complexity. Cardiology and Oncology bear numerous similarities with high medication quantities and complexity, low reason-for-visit and age variability, and low patient characteristic complexity. Cardiology is similar to Family Medicine in demographic diversity and output complexity, and is similar to Internal Medicine in quantity of diagnoses; Oncology is similar to Internal Medicine in quantity, variability and complexity of other therapies. Cluster 3 was unique in its minimal testing and total input complexities. Finally, Cluster 2 was distinguished by its moderate total input complexities. Although no cluster was unique in its input or output complexity density, Cluster 1 had the highest total complexity density followed by Cluster 2.
4. Discussion The high complexity of care for Family Medicine and General Internal Medicine is consistent with their role in managing multiple chronic problems, working with limited evidence, balancing multiple guidelines, registries and interdependent disorders, and polypharmacy. They coordinate care with multiple physicians, increasing the risk of error. When complex care is packed into the brief visits characteristic of primary care, the workload intensifies and face-to-face time with patients is limited.19,20 4.1. Implications The Medicare Physician Fee Schedule (PFS) is based on physician work, practice expense, and malpractice insurance. The payment of
94
D. Katerndahl et al. / Healthcare 3 (2015) 89–96
Table 4 Complexity of ambulatory care provided across medical specialties (weighted mean). Medical specialty Inputa
Outputa
Reasons Diagnoses Exam/ tests Primary care General/family 0.75 medicine General internal 0.71 medicine Pediatrics 0.61 OBGYN
0.53
Medical specialties Cardiology Dermatology Neurology Oncology Surgical specialties General surgery Orthopedic surgery Urology Ophthalmology ENT Psychiatry
Patients Total Medications Other therapy
Total Density encounterb Tota Disposition Total Visit durationa inputb
Tota outputal
Total complexityc
0.80
0.93
1.99
4.46
1.47
0.70
0.73
2.90
63.90
18.97
14.11
9.17
202.10
0.84
1.17
1.91
4.63
1.54
0.52
0.73
2.79
69.24
21.50
12.92
7.79
193.22
0.60
0.55
2.13
3.89
0.70
0.94
0.73
2.37
35.17
18.26
12.79
7.79
115.61
0.55
1.43
1.91
4.42
0.65
0.62
0.73
2.01
42.87
18.03
14.69
6.68
142.64
0.58 0.50 0.62 0.63
0.79 0.62 0.70 0.72
0.89 0.48 0.44 1.20
1.87 2.01 2.01 1.88
4.13 3.61 3.77 4.42
1.95 0.89 1.33 1.61
0.72 0.84 0.63 0.45
0.73 0.66 0.73 0.73
3.39 2.39 2.69 2.79
59.36 29.29 36.63 59.97
20.84 16.14 27.15 25.73
11.88 13.44 8.33 10.31
9.77 8.88 5.94 6.51
170.88 108.89 80.94 139.83
0.51 0.52
0.57 0.57
0.67 0.42
2.00 1.98
3.75 3.50
0.96 0.62
0.70 0.72
0.73 0.73
2.39 2.08
32.13 23.48
18.82 17.11
11.94 12.26
7.63 7.30
102.40 82.36
0.50 0.51 0.57
0.66 0.66 0.62
0.83 0.27 0.29
1.85 1.97 2.08
3.84 3.42 3.55
1.27 1.04 0.74
0.52 0.60 0.55
0.66 0.73 0.73
2.44 2.37 2.03
34.99 25.31 23.79
18.03 16.94 18.43
12.79 12.10 11.57
8.12 8.41 6.60
116.46 89.66 77.48
0.51
0.56
0.15
1.94
3.17
0.88
0.85
0.73
2.45
22.00
32.04
5.93
4.59
41.20
Data represent the specialty-specific calculated complexity (mean quantity [from Table 1] multiplied by Z-transformed COV [from Table 2] multiplied by Z-transformed diversity [from Table 3]) for each input/output and totals. “Patients” complexity is the sum of the Z-transformed new patient proportion [from Table 1], age COV [from Table 2] and demographic diversity [from Table 3]. “Total Input Complexity” is the sum of the input complexities across each specialty and “Total Output Complexity” is the sum of the output complexities across each specialty. “Total Encounter Complexity” is the product of the Total Output Complexity and “2” raised to the power of the Total Input Complexity. “Visit Duration” is the weighted mean duration-of-visit for each specialty. “Density” represents the total input, output and encounter complexity measures divided by the mean duration-of-visit and then multiplied by 60 to derive the hourly complexity density. a b c
Standard errors: Allo 0.01. Standard errors: Allo 0.05. Standard errors: Allo 0.15.
Table 5 Differences among component-based clusters. Measures
Complexity Input Reasons Diagnoses Exam/tests Patients Total Output Medications Other Therapy Disposition Total Total encounter Density Visit duration Input complexity Output complexity Total complexity
Cluster 1 Family medicine General internal medicine Cardiology Oncology
Cluster 2 General surgery OBGYN Dermatology Urology
Cluster 3 PediatricsENT Orthopedics Psychiatry Neurology Ophthalmology ENT
F(p) [cluster differences]
0.67 0.79 1.05 1.91 4.41
0.51 0.60 0.85 1.94 3.91
0.56 0.62 0.35 2.02 3.55
9.43(0.004) 16.83(0.000) 10.40(0.003) 3.14(0.083) 11.66(0.002)
1 42,3 1 42,3 1,2 43
1.64 0.60 0.73 2.97 63.12
0.94 0.67 0.70 2.31 34.82
0.89 0.72 0.73 2.33 27.73
12.54(0.001) 0.82(0.469) 3.93(0.052) 9.85(0.004) 46.59(0.000)
1 42,3
1 42,3 1 42,3
21.76 12.31 8.31 176.5
17.76 13.22 7.83 117.6
21.66 10.50 6.77 81.2
1.06(0.380) 2.14(0.164) 1.86(0.201) 19.61(0.000)
1 4243
1 4243
Data represent the cluster-specific mean complexities and visit-durations. “F(p)” represents the F-statistics and their p-values for each Oneway ANOVA comparing clusters. Significant (p r 0.05) intercluster differences as determined by REGWF posthoc testing are then described.
services is based on an expanded coding list known as the Healthcare Procedures Coding System (HCPCS). The work relative value unit (RVU) assigned to ambulatory visits is identical across
specialties with the assumption that the work is equivalent for each specialty. This assumption is faulty; General Internal Medicine, Family Medicine, Cardiology and Oncology have significantly
D. Katerndahl et al. / Healthcare 3 (2015) 89–96
higher complexity of care and therefore workload than other specialties. Adult primary care EM services are generally undervalued. The range of EM services is not fully captured by CPT codes and procedures are overvalued. There is no measure of complexity of care, and decision-making about EM payment is delegated to a select group of specialty societies with little primary care involvement. Because Family Medicine and General Internal Medicine cluster with Cardiology and Oncology, one might expect that all four disciplines would be paid similarly; however, these specialties are paid differently with the majority of payment in the latter two specialties coming from procedures (Cardiology) and chemotherapy (Oncology). Revaluing RVUs would create incentives to better balance their services between cognitive and procedural work. Increasing documentation demands, guideline requirements and EHR use coupled with shortages of primary care physicians have significant implications. The Affordable Care Act (ACA) should not only intensify these trends,21 but result in more low-income patients seeking primary care, increasing patient diversity and demand for preventive and chronic care, and a shift of focus from inpatient to outpatient settings. These newly-insured patients are anticipated to be more complex than usual, magnifying current physician shortages and decreasing durations-of-visit. Quality of care may decline due to physicians’ inability to meet increased complexity demands of comprehensive and guideline-based care. Such consequences require adjustments in primary care delivery that can only be accomplished with an expanded primary care workforce and transformation of the primary care offices teams into models such as the patient-centered medical home attending to the work both within the consultation and around the consultation.22 These changes require adequate payment reform. Health costs may increase for primary care as complexity and density increase. Time-sensitive activities and preventive services may further decrease as primary care physicians attempt to compensate for complexity burden by reducing such activities. But total cost of care may still be reduced if the use of primary care within the system is optimized.23 Finally, physicians in high complexity disciplines (e.g. Family Medicine and General Internal Medicine) may experience declining career satisfaction and recruitment of new physicians, resulting from growing perceptions of time inadequacy and physician shortages.9 Increased financial input into primary care can translate into savings in other settings.24 In addition, improved primary care payment support could enable new models of care, improving both practice function and physician satisfaction.22 4.2. Issues and limitations This methodology evaluates large numbers of visits and is not suitable to measure complexity of individual encounters. Reliance upon the NAMCS database may minimize possible input complexity because only three complaints and three diagnoses can be included. Prior studies in Family Medicine settings suggest that visits often address more than three problems.25 In addition, physicians may overestimate the duration of visit, leading to underestimates of complexity density. Content of self-report cards used in NAMCS may not accurately reflect the visit, distorting complexity measures. The 1985–1991 NAMCS surveys could not differentiate between primary and specialty care disciplines based upon ten primary care criteria.26 However, good agreement exists between content recorded and physician rankings,27 and studies of physician-rated work found consistent, reproducible estimates.28 The method used here also does not differentiate among diagnoses in terms of impact on complexity (URI weights the same as schizophrenia); however, the effect of particular diagnoses may depend upon the training and experience of the physician treating it. Though the minimum number of visits per discipline was 1161,
95
variability measures could be affected by use of varying numbers of visits per discipline. Definitions of “meaningful” levels of complexity have not been determined. Finally, this study did not assess care delivered outside of the office visit or complexity embedded in clinical problems due to social and contextual dependence on family and relational issues. In addition, methodological decisions potentially impacted results. First, to test whether including post-operative visits with other visits (as done in this study) distorted relative complexity, we excluded post-operative visits from complexity calculations. Only four disciplines changed their total encounter complexities by more than 3% and total encounter complexity did not differ significantly with or without post-operative visits (paired t ¼0.011, p¼ 0.991). Overall rankings of specialties with and without postoperative care changed little for either total encounter complexity (rs ¼ 0.995, p ¼0.000) and total complexity density (rs ¼0.996, p¼ 0.000). Second, Psychiatry’s reliance on a different diagnostic classification system (DSM-V) than other disciplines may falsely simplify its observed complexity.13 Psychosocial factors were not included in this measure, supporting a falsely-low estimate for Psychiatry. However, Psychiatry was low in reasons-for-visit and testing complexity, suggesting that diagnosis complexity is not the whole story. Finally, use of COV as the statistic of variability resulted in the paradox that Pediatrics (with its disciplinedefined limits on age) had the most variability in patient age. Using the entire sample’s mean age instead would result in a 20% reduction in total encounter complexity for Pediatrics alone, but such reduction minimizes complexity caused by differences in developmental stage as health complexity changes most rapidly during these years. The high rates of errors29 and adverse drug events30 suggest that Pediatrics is more than minimally complex.
5. Conclusions Family Medicine and General Internal Medicine encounters are most complex. Psychiatry has the lowest input complexities with long visits further decreasing complexity densities. While Pediatrics has some diversity similarities to other primary care disciplines, its lack of chronic care visits results in complexities similar to Psychiatry and surgical specialties. Finally, OB/GYN has the lowest output complexity and clusters with General Surgery and Urology, baring few similarities to primary care disciplines. These findings have important implications for discrepancies in payment. Revaluing payments based on complexity could bring better balance to cognitive and procedural services, and better meet needs of people receiving insurance under the ACA.
Conflict of interest disclosure statement This statement accompanies the article “Complexity of ambulatory care across disciplines,” authored (co-authored) by “David Katerndahl” (“Robert Wood and Carlos Roberto Jaén”) and submitted to Healthcare as an original article. Below all authors have disclosed relevant commercial associations that might pose a conflict of interest: Consultant arrangements: None Stock/other equity ownership: None Patent licensing arrangements: None Grants/research support: None Employment: None Speakers' bureau: None Expert witness: None Other: None
96
D. Katerndahl et al. / Healthcare 3 (2015) 89–96
Acknowledgments We wish to acknowledge that this study was funded through a contract from the American Academy of Family Physicians. We wish to thank Drs. Robert Phillips and Sandra Burge for their critique of the manuscript. Appendix A. Supporting information Supplementary data associated with this article can be found in the online version at http://dx.doi.org/10.1016/j.hjdsi.2015.02.002. References 1. Abbo ED, Zhang Q, Zelder M, Huang ES. Increasing number of clinical items addressed during the time of adult primary care visits. J Gen Intern Med. 2008;23:2058–2065. 2. Carter LM, McHenry DS, Godlington FL, Meechan JG. Prescribed medication taken by patients attending general dental practice: changes over 20 years. Br Dent J 2007;2003:E8, 10.1038/bdj.2007.629. Accessed 07.07.14. 3. Chai G, Goverale L, McMahon AW, Trinidad JP, Staffa J, Murphy D. Trends of outpatient prescription drug utilization in US children. Pediatrics. 2012;130: 23–31. 4. Petros P. Non-linearity in clinical practice. J Eval Clin Pract. 2003;9(2):171–178. 5. Bar-Yam Y. Dynamics of Complex Systems. Reading, MA: Perseus Books; 1997. 6. Croskerry P, Shapiro M, Campbell S, et al. Profiles in patient safety. Acad Emerg Med. 2004;11:289–299. 7. McGlynn EA, Asch SM, Adams J, et al. Quality of health care delivered to adults in the United States. N Engl J Med. 2003;348:2635–2645. 8. Asch SM, Kerr EA, Keesey J, et al. Who is at greatest risk for receiving poor quality of health care? N Engl J Med. 2006;354:1147–1156. 9. Katerndahl D, Parchman M, Wood R. Perceived complexity of care, perceived autonomy, and career satisfaction among primary care physicians. J Am Board Fam Med. 2009;22:24–33. 10. Burdi MD, Baker LC. Physicians’ perceptions of autonomy and satisfaction in California. Health Affairs. 1999;18:134–145. 11. Katerndahl DA, Wood R, Jaen CR. Family medicine outpatient encounters are more complex than those of cardiology and psychiatry. J Am Board Fam Med. 2011;24:6–15.
12. Katerndahl DA, Wood R, Jaen CR. A method for estimating relative complexity of ambulatory care. Ann Fam Med. 2010;8:341–347. 13. Boisot M, Child J. Organizations as adaptive systems in complex environments. Organ Sci. 1999;10:237–252. 14. Rosen AK, Reid R, Broemeling AM, Rakovski CC. Applying a risk-adjustment framework to primary care. Ann Fam Med. 2003;1:44–51. 15. Reza FM. Introduction to Information Theory. New York: Dover Publications, Inc; 1994. 16. Green LA. Implications of measuring complexity (ed). Ann Fam Med. 2010;8:291–292. 17. National Center for Health Statistics. National Ambulatory Medical Care Survey. Hyattsville, MD: U.S. Department of Health and Human Services; 2010. 18. Edwards ST, Mafi JN, Landon BE. Trends and quality of care in outpatient visits to generalist and specialist physicians delivering primary care in the United States, 1997–2010. J Gen Intern Med. 2014;29:947–955. 19. Horner RD, Szaflarski JP, Ying J, et al. Physician work intensity among medical specialties. Med Care. 2011;49:1007–1011. 20. Jacobson Jr CJ, Bolon S, Elder N, et al. Temporal and subjective work demands in office-based patient care. Med Care. 2011;49:52–58. 21. Hoff T. Embracing a diversified future for US primary care. Am J Manag Care. 2013;19(1):e9–e13. 22. Sinsky CA, Willard-Grace R, Schutzbank AM, Sinsky TA, Margolis D, Bodenheimer T. In search of joy in practice: a report of 23 high-functioning primary care practices. Ann Fam Med. 2013;11:272–278. 23. Starfield Barbara. Is primary care essential?. Lancet. 1994;344.8930:1129–1133. 24. Neuhausen K, Grumbach K, Bazemore A, Phillips RL. Integrating community health centers into organized delivery systems can improve access to subspecialty care. Health Affairs. 2012;31:1708–1716. 25. Goodson JD. Unintended consequences of resource-based relative value scale reimbursement (ed). JAMA. 2007;298:2308–2310. 26. Franks P, Clancy CM, Nutting PA. Defining primary care: empirical analysis of the National Ambulatory Medical Care Survey. Med Care. 2014;35:655–668. 27. Rethans JJ, Martin E, Metsemakers J. To what extent do clinical notes by general practitioners reflect actual medical performance. Br J Gen Pract. 1994;44: 153–156. 28. Hsiao WC, Braun P, Yntema D, Becker ER. Estimating physicans’ work for a resource-based relative-value scale. N Engl J Med. 1988;319:835–841. 29. Neuspiel DR, Stubbs EH. Patient safety in ambulatory care. Ped Clin N Am. 2012;59:1341–1354. 30. Tache SV, Sonnichsen A, Ashcroft DM. Prevalence of adverse drug events in ambulatory care. Ann Pharmacother. 2011;45:977–989.
Contents lists available at ScienceDirect
Healthcare journal homepage: www.elsevier.com/locate/hjdsi
Complexity of ambulatory care across disciplines David Katerndahl n, Robert Wood, Carlos Roberto Jaén Family & Community Medicine, University of Texas Health Science Center San Antonio, 7703 Floyd Curl Drive, San Antonio, TX 78229-3900, USA
art ic l e i nf o
a b s t r a c t
Article history: Received 11 August 2014 Received in revised form 15 January 2015 Accepted 3 February 2015 Available online 27 February 2015
Background: Complexity of care has implications for quality of care, health costs, medical errors, and patient and physician satisfaction. The objective was to compare complexity of ambulatory care across 14 medical specialties. Methods: This secondary analysis uses the 2010 National Ambulatory Medical Care Survey, which used a multistage probability design of primary sampling units throughout U.S. ambulatory practices across 14 specialties. Sampling weights enable results from 29,179 ambulatory visits to represent 878,653,561 visits. Data included symptoms, diagnoses, diagnostic procedures, and treatments provided. Measures of input, output and total encounter complexity and hourly complexity densities were computed. Results: Internal Medicine leads in total input and total encounter complexity with Family Medicine second in total encounter complexity. When duration-of-visit is considered, Family Medicine is the most complex discipline while Internal Medicine is the second most complex. Pediatrics lacks the complexity of Family Medicine and General Internal Medicine, and OB/GYN bears little similarity to Family Medicine or General Internal Medicine. Conclusions: Family Medicine and Internal Medicine encounters are the most complex overall, especially when duration-of-visit is considered. Implications: Revaluing payments based on complexity could bring better balance to cognitive and procedural services, and better meet the needs of people receiving insurance under the ACA. & 2015 Elsevier Inc. All rights reserved.
Keywords: Ambulatory care Complexity of care Interspecialty comparison System theory
1. Introduction Complexity of care is increasing. From 1997 to 2005, the mean number of clinical items addressed during an office visit increased from 5.4 to 7.1.1 From 1995 to 2005, the proportion of patients on medications increased from 21% to 44% while the proportion taking at least three drugs also increased.2 From 2002 to 2010, medications prescribed to adults rose 22%.3 From 1997 to 2005, office time spent per clinical item addressed declined.1 With health care reform, further change is anticipated. With the influx of newly-insured individuals, patient diversity and psychiatric comorbidity should increase. Increasing demands for documentation and electronic health records (EHR) use will not increase face-to-face durations-of-visit, but increase the time required to care for each patient. With increases in demand for care, physician shortages may become exaggerated, potentially shifting patients among existing specialties or decreasing durations-of-visit. Finally, primary care disciplines may uniquely experience increasing numbers of chronic medical problems,
n
Corresponding author. Tel.: þ 1 210 358 3200; fax: þ 1 210 223 6940. E-mail address: [email protected] (D. Katerndahl).
http://dx.doi.org/10.1016/j.hjdsi.2015.02.002 2213-0764/& 2015 Elsevier Inc. All rights reserved.
complexity of medication regimens, numbers of guidelineindicated services, demand for preventive services, and pressures for accountability and performance. Increased use of urgent care centers may leave primary care disciplines with even greater proportions of chronic care, multimorbidity visits. Such increasing complexity could affect the quality of health care. When medical care becomes more complex, practice guidelines are less effective4 and the rate of errors occurring rises.5 The risk of errors increases with seeing multiple patients with multiple conditions, use of multiple medications, and implementation of complex procedures.6 Complexity of care may partially explain why only 55% of adult patients receive recommended care.7 Poor quality of care was especially noted for time-intensive activities, such as history-taking, counseling, and patient education7 as well as screening and preventive medicine.8 In addition, not only is complex care at risk for inefficiencies, but to cope with perceived complexity, physicians may increase testing or lower the threshold at which they refer patients to specialists, adding to the cost of care. Finally, complexity of care could impact patient and physician perceptions of time adequacy and satisfaction. This could explain a decline in perceived autonomy and career satisfaction,9 with perceived autonomy poorer among primary care physicians.10 Yet, we know little about the differences in complexity of
90
D. Katerndahl et al. / Healthcare 3 (2015) 89–96
ambulatory care across disciplines; such differences have implications for the cost of care and its payment. Although previous work found that the complexity of ambulatory care in 2000 was the highest for Family Medicine compared with that of Cardiology and Psychiatry,11 no recent analysis has been conduct or involved more than three medical specialties. The purpose of this study was to compare the complexity of patient care across 14 specialties using data from the 2010 National Ambulatory Medical Care Survey (NAMCS).
2. Materials and methods Using a method designed to estimate complexity of ambulatory care derived from national databases,12 relative complexity can be estimated (computational details are provided as supplementary material). This study analyzed data from the most recentlyavailable NAMCS database (2010) for 14 medical specialties (Family Medicine, General Internal Medicine, Pediatrics, OBGYN, Cardiology, Dermatology, Neurology, Oncology, General Surgery, Orthopedics, Urology, Ophthalmology, Otorhinolaryngology (ENT), and Psychiatry). NAMCS used a multistage probability design of primary sampling units (PSUs) throughout the U.S., practices within PSUs, and patient visits within practices, designed specifically to be representative of all ambulatory visits in the United States. Trained physicians and office staff completed encounter data on patient visits selected, including patients’ symptoms, physicians’ diagnoses, diagnostic procedures, and treatments provided.
2.1. Computation of complexity of each input/output (see Fig 1) If we define “complexity” of a system as the amount of information needed to describe it or its behavior,5 then complexity (like error) is associated with volume, diversity, variability, and time limitations.6 It includes “cognitive complexity” (focused on the content of information flowing) and “relational complexity” (focused on the interactions by which information flows among agents).13 Cognitive complexity is measured in counts, while relational complexity is measured in variability. Any measure of complexity must reflect the range of problems and inputs seen by the physician.14 However, because a specialty is not defined by a single encounter, the measure of complexity needs to include inter-encounter variation as well. Whereas the complexity of an encounter includes the number of events occurring and the amount of information transferred, the complexity of a specialty must include the diversity (range of characteristics seen) and variability of events across encounters. Thus, “complexity of care” should reflect the complexity of the typical encounter and the complexity across encounters.15 Approaches such as case-mix measures, risk adjustment approaches, patient severity (or risk) measures, patient complexity estimates, and clinical problems per hour fail to measure complexity of care because they either equate severity with complexity, represent single patient measures, fail to capture all relevant dimensions, and/or assume linearity.16 Patient encounters and their components can be described by the quantity of information and services exchanged between patient and physician, by the visit-to-visit variability in quantity of these exchanges, and by their overall diversity.12 “Quantification” of visits included the per-patient number of reasons for visit, diagnoses, body systems examined and tests ordered, medications prescribed, other therapies and procedures ordered. Whether patients were new to the practice was also recorded. “Variability” was measured across visits within discipline by computing the coefficient of variation (COV)—a unit-free measure—from the
mean and standard deviation from quantities measured above. In addition, the COV of the age of patients seen was also computed. “Diversity” was defined as the proportion of categories needed to describe 95% of the visits for each specialty. The 95% proportion was chosen to minimize the impact of a rare or miscoded input/ output. Finally, patient demographic diversity was assessed as the proportion of categories within a gender X race X ethnicity matrix used by NAMCS needed to describe 95% of patients seen. The NAMCS databases provide a patient weight that allowed the 2010 sample of 31,229 visits to represent 1,008,802,005 visits that year in the United States.17 This patient visit weight was applied to the dataset so that estimates of complexity parameters produced by re-sampling techniques would better conform to national patterns of patient encounters. Visit input depended upon the reasons for visit, diagnoses, examination/testing, and patient characteristics. Visit output depended upon medications (including vaccines) and other therapies (including procedures, education, physical therapy, nutrition, etc.) prescribed, and visit disposition. In addition to the mean quantification for each variable, differences in discipline-specific duration-of-visit were used to determine an hourly complexity rate for each discipline. The complexity of each input/output was defined as the mean input/output quantity per clinical encounter weighted by its inter-encounter diversity and variability. To standardize the weightings and limit the impact of low diversity or variability on complexity, the weightings used were the Z-transformations of the diversity proportion and the COV, ranging between 0.5 and 1.0. 2.2. Computation of total complexity (see Fig 1) Total input and output complexities were calculated by summing their component complexities. However, calculation of total complexity was not merely the sum of input and output complexities. A fundamental principle of complex systems is that a logarithmic relationship exists between input and output, such that, as the information in the input increases linearly, the complexity of the system increases exponentially. To calculate total encounter complexity, total output complexity was multiplied by “2” raised to the power of the input complexity. Thus, total system complexity depends more heavily upon the complexity of the input.5 2.3. Complexity density (see Fig 1) The estimate of complexity of ambulatory care presented is a measure of the complexity of the typical clinical encounter. However, coping with complexity is time-dependent.13 An hourly complexity density estimate was derived by dividing the total complexity estimates by the duration-of-visit and then multiplying by 60. 2.4. Analysis The 95% confidence intervals were derived from bootstrap resampling procedures based on 500 samples, enabling comparisons across disciplines. Friedman’s test was used with multiple comparison posthoc testing to seek significant differences of complexity estimates among disciplines. Cluster analysis of component measures was used to group disciplines. First, to identify the number of clusters, the K-means method was used with the visit components to examine 2-, 3-, and 4-cluster models. Based upon analysis-of-variance (ANOVA) results, the 3-cluster solution produced the most consistent clear-cut differences. Then, to assign disciplines to clusters, Z-scores were computed for each component measure. The squared Euclidean distance approach was used because it is sensitive to differences in the magnitude of
D. Katerndahl et al. / Healthcare 3 (2015) 89–96
INPUTS*
OUTPUTS**
QUANTITY Mean number per visit of each input across all visits to that specialty
QUANTITY Mean number per visit of each output across all visits to that specialty
VARIABILITY (COV) SD for each input across all visits to that specialty divided by its mean quantity and then Z-transformed
VARIABILITY (COV) SD for each output across all visits to that specialty divided by its mean quantity and then Z-transformed
DIVERSITY Proportion of input used by all specialties needed to account for 95% of that specialty’s visits and then Z-transformed
DIVERSITY Proportion of output used by all specialties needed to account for 95% of that specialty’s visits and then Z-transformed
INPUT COMPLEXITIES Multiply each input mean quantity by its Z-transformed COV and diversity
91
OUTPUT COMPLEXITIES Multiply each output mean quantity by its Z-transformed COV and diversity
Patient Complexity is sum of Ztransformations of the proportion of new patients seen, COV of patient ages and gender/race/ethnicity diversity of patients TOTAL INPUT COMPLEXITY Sum of all component input complexities
TOTAL OUTPUT COMPLEXITY Sum of all component input complexities
TOTAL ENCOUNTER COMPLEXITY Product of Total Output Complexity and “2” raised to the power of the Total Input Complexity
Mean visit duration
INPUT, OUTPUT & TOTAL COMPLEIXTY DENSITIES Complexity densities computed by dividing Total Input, Total Output and Total Encounter Complexities by the mean visit duration and multiplying by 60 * “Input” includes reasons for visit, diagnoses, examination/testing, and patient characteristics ** “Output” includes medications and non-medication therapies prescribed, and visit disposition “SD” = Standard Deviation “COV” = Coefficient of Variation Fig. 1. Flow diagram of computations used.
variables. Ward’s method was used to assign each discipline to a cluster because it yields small, distinct clusters. Complexities were compared across clusters using ANOVA with REGWF posthoc testing, with p r0.05 deemed significant.
3. Results Excluding the “other medical specialties” (n¼ 2050), the 2010 NAMCS database included 29,179 visits. Based upon the weightings, the overall results represents 878,653,561 visits. Bootstrap
standard errors showed that complexity measures were highly stable for all disciplines. Looking at the quantities of care (Table 1), Family Medicine, Internal Medicine, Cardiology and Neurology report the most reasons-for-visit and diagnoses while General Surgery, Urology and OBGYN report the least. In addition, ENT, General Surgery and Neurology see the most new patients. In terms of outputs, Cardiology and Oncology prescribe the most medications while Pediatrics and Psychiatry use the most non-medication therapies. Variability in numbers of reasons-for-visit and diagnoses was consistent across disciplines (see Table 2). High variabilities in
92
D. Katerndahl et al. / Healthcare 3 (2015) 89–96
Table 1 Quantification of ambulatory care provided across disciplines. Medical specialty (n¼ number of visits)
Input
Output
Mean reasons per visit
Mean diagnoses per visit
Mean exams/tests Proportion of new patients per visit among all visits
Mean of medications per visit
Mean of other therapiesa per visit
1.90
1.33
0.11
3.06
1.09
Primary care General/family medicine 1.64 (n ¼6237) General internal 1.61 medicine (n¼ 2217) Pediatrics (n¼ 3501) 1.47
2.10
1.76
0.10
3.45
0.80
1.50
0.82
0.05
1.56
1.61
OBGYN (n¼2461)
1.37
1.43
2.28
0.09
1.34
1.01
Medical specialties Cardiology (n¼ 1793) Dermatology (n ¼1510) Neurology (n¼1556) Oncology (n¼ 1478)
1.51 1.37 1.61 1.50
2.25 1.72 1.84 1.81
1.31 1.00 0.58 2.01
0.13 0.22 0.27 0.07
4.94 1.78 2.92 3.56
1.08 1.23 0.98 0.71
1.26
1.46
0.95
0.27
1.96
1.08
1.42
1.55
0.64
0.23
1.24
1.30
Surgical specialties General surgery (n ¼1161) Orthopedic surgery (n ¼1460) Urology (n¼1614) Ophthalmology (n ¼1368) ENT (n¼ 1471)
1.36 1.43
1.83 1.91
1.40 0.36
0.17 0.15
2.64 2.13
0.87 0.96
1.47
1.67
0.35
0.31
1.49
0.98
Psychiatry (n¼ 1352)
1.41
1.60
0.19
0.05
2.19
1.56
Means are weighted means per visit for each input and output for that specialty. “n” reports the number of visits per specialty upon which the estimates are based. Data based on 2010 NAMCS Survey a
Other therapies include procedures, patient education, nutritional counseling, physical therapy and other interventions.
Table 2 Variability of ambulatory care provided across medical specialties. Medical Specialty
Input
Output
COV for reasons per visit
COV for diagnoses per visit
COV for exams/tests per visit
COV of age of patients among all visits
COV for medications COV for other per visit therapiesaper visit
0.47
0.46
1.46
0.50
0.84
1.10
Primary care General/family medicine General internal medicine Pediatrics
0.49
0.40
1.30
0.31
0.78
1.28
0.47
0.49
1.59
1.22
0.97
0.98
OBGYN
0.46
0.50
0.97
0.39
1.19
1.06
Medical specialties Cardiology Dermatology Neurology Oncology
0.50 0.45 0.48 0.52
0.38 0.47 0.46 0.48
1.32 0.67 1.93 0.86
0.22 0.41 0.37 0.24
0.56 1.22 0.90 0.83
1.20 0.94 1.04 1.24
0.45 0.46
0.49 0.48
1.41 1.27
0.32 0.38
1.28 1.54
0.93 0.83
0.45 0.46 0.47
0.46 0.44 0.46
0.84 2.16 2.87
0.28 0.37 0.58
1.09 1.22 1.35
0.96 1.07 0.95
0.48
0.47
3.58
0.46
0.75
0.81
Surgical specialties General Surgery Orthopedic surgery Urology Ophthalmology ENT Psychiatry
COV represents weighted coefficient of variation for each input and output per specialty across all specialty-specific visits, calculated as the standard deviation of the input/ output across all visits divided by the mean (from Table 1) for that input/output. Larger COVs represent very high variability, typically due to significant number of visits with zero for that input/output (e.g., no tests ordered). Note that the COVs for “Reasons” and “Diagnoses” are similar across disciplines, suggesting that specialties were not differentiated from each other bases upon the variability of these inputs. Data based on 2010 NAMCS Survey a
Other therapies include procedures, patient education, nutritional counseling, physical therapy and other interventions.
D. Katerndahl et al. / Healthcare 3 (2015) 89–96
93
Table 3 Diversity of ambulatory care provided across disciplines. Medical specialty
Input
Output
Proportion of reason for visit categories needed to account for 95% of visits
Proportion of diagnostic categories needed to account for 95% of visits
Proportion of exams/tests categories needed to account for 95% of visits
Proportion of the 8 possible gender by race/ethnicity categories needed to account for 95% of visits
Proportion of medication categories needed to account for 95% of visits
Proportion of other therapya categories needed to account for 95% of visits
Proportion of 5 possible visit dispositionsbneeded to account for 95% of visits
0.44
0.31
0.68
0.88
0.26
0.66
0.60
Primary care General/family medicine General internal medicine Pediatrics
0.38
0.27
0.64
0.75
0.18
0.61
0.60
0.28
0.21
0.57
0.75
0.10
0.53
0.60
OBGYN
0.18
0.13
0.68
0.75
0.13
0.57
0.60
Medical specialties Cardiology 0.14 Dermatology 0.12 Neurology 0.16 Oncology 0.25
0.10 0.07 0.15 0.20
0.68 0.36 0.79 0.64
0.88 0.75 0.75 0.75
0.13 0.16 0.15 0.17
0.68 0.95 0.70 0.56
0.60 0.40 0.60 0.60
0.19
0.72
0.63
0.12
0.78
0.60
Surgical specialties General 0.24 Surgery Orthopedic 0.11 Surgery Urology 0.12 Ophthalmology 0.06 ENT 0.17
0.10
0.61
0.88
0.09
0.52
0.60
0.08 0.05 0.12
0.65 0.75 0.89
0.63 0.75 0.75
0.14 0.13 0.12
0.57 0.63 0.46
0.40 0.60 0.60
Psychiatry
0.03
0.79
0.75
0.05
0.48
0.60
0.08
Weighted proportion of possible categories needed to include 95% of visits. “Diversity” is the fraction of the total number of diagnoses used by all specialties needed by that specialty to describe 95% of their visits. The 8 possible racial/ethnic categories include: male/female Non-Hispanic White/Non-Hispanic Black/Hispanic/Non-Hispanic Other. Data based on 2010 NAMCS Survey. a b
Other therapies include procedures, patient education, nutritional counseling, physical therapy and other interventions. Disposition 5 categories included: return as need, scheduled return visit, referred to another physician, send to emergency department, hospitalize.
other descriptors were observed in those disciplines whose visits often do not include testing or therapies. High diversity for most inputs and medications was seen in Family Medicine and Internal Medicine while high diversity in testing was observed in ENT, Neurology and Psychiatry (see Table 3); high diversity in diagnoses for Family Medicine and Internal Medicine was found previously.18 However, low diversity in most inputs and outputs was reported in Psychiatry. Combining quantities, variability and diversities, complexity of encounter was computed (see Table 4). Primary care disciplines led in input (reasons-for-visit, diagnoses and patient characteristics) and output (other treatments) complexity components with General Internal Medicine leading in total input and total encounter complexity but Family Medicine a close second in total encounter complexity. When duration-of-visit is considered, Family Medicine becomes the most complex discipline with the highest total complexity density while General Internal Medicine is the second most complex. Using the rankings of input, output and total complexities and their densities, a significant difference across disciplines was found (Friedman’s ¼49.5, p o 0.001). Using multiple comparison procedure, Family Medicine was significantly (p r 0.05) more complex than Orthopedics, ENT and Psychiatry while General Internal Medicine and Cardiology were more complex than ENT and Psychiatry. Using ANOVA to determine intercluster differences, Table 5 shows that Cluster 1 consisted of Family Medicine, General Internal Medicine, Cardiology and Oncology while Cluster 2 included General Surgery, OB/GYN, Dermatology and Urology. Cluster 1 was distinguished by its high complexity of reasons-for-visit, diagnoses and medications with
resultant high input, output and total encounter complexity. Cardiology and Oncology bear numerous similarities with high medication quantities and complexity, low reason-for-visit and age variability, and low patient characteristic complexity. Cardiology is similar to Family Medicine in demographic diversity and output complexity, and is similar to Internal Medicine in quantity of diagnoses; Oncology is similar to Internal Medicine in quantity, variability and complexity of other therapies. Cluster 3 was unique in its minimal testing and total input complexities. Finally, Cluster 2 was distinguished by its moderate total input complexities. Although no cluster was unique in its input or output complexity density, Cluster 1 had the highest total complexity density followed by Cluster 2.
4. Discussion The high complexity of care for Family Medicine and General Internal Medicine is consistent with their role in managing multiple chronic problems, working with limited evidence, balancing multiple guidelines, registries and interdependent disorders, and polypharmacy. They coordinate care with multiple physicians, increasing the risk of error. When complex care is packed into the brief visits characteristic of primary care, the workload intensifies and face-to-face time with patients is limited.19,20 4.1. Implications The Medicare Physician Fee Schedule (PFS) is based on physician work, practice expense, and malpractice insurance. The payment of
94
D. Katerndahl et al. / Healthcare 3 (2015) 89–96
Table 4 Complexity of ambulatory care provided across medical specialties (weighted mean). Medical specialty Inputa
Outputa
Reasons Diagnoses Exam/ tests Primary care General/family 0.75 medicine General internal 0.71 medicine Pediatrics 0.61 OBGYN
0.53
Medical specialties Cardiology Dermatology Neurology Oncology Surgical specialties General surgery Orthopedic surgery Urology Ophthalmology ENT Psychiatry
Patients Total Medications Other therapy
Total Density encounterb Tota Disposition Total Visit durationa inputb
Tota outputal
Total complexityc
0.80
0.93
1.99
4.46
1.47
0.70
0.73
2.90
63.90
18.97
14.11
9.17
202.10
0.84
1.17
1.91
4.63
1.54
0.52
0.73
2.79
69.24
21.50
12.92
7.79
193.22
0.60
0.55
2.13
3.89
0.70
0.94
0.73
2.37
35.17
18.26
12.79
7.79
115.61
0.55
1.43
1.91
4.42
0.65
0.62
0.73
2.01
42.87
18.03
14.69
6.68
142.64
0.58 0.50 0.62 0.63
0.79 0.62 0.70 0.72
0.89 0.48 0.44 1.20
1.87 2.01 2.01 1.88
4.13 3.61 3.77 4.42
1.95 0.89 1.33 1.61
0.72 0.84 0.63 0.45
0.73 0.66 0.73 0.73
3.39 2.39 2.69 2.79
59.36 29.29 36.63 59.97
20.84 16.14 27.15 25.73
11.88 13.44 8.33 10.31
9.77 8.88 5.94 6.51
170.88 108.89 80.94 139.83
0.51 0.52
0.57 0.57
0.67 0.42
2.00 1.98
3.75 3.50
0.96 0.62
0.70 0.72
0.73 0.73
2.39 2.08
32.13 23.48
18.82 17.11
11.94 12.26
7.63 7.30
102.40 82.36
0.50 0.51 0.57
0.66 0.66 0.62
0.83 0.27 0.29
1.85 1.97 2.08
3.84 3.42 3.55
1.27 1.04 0.74
0.52 0.60 0.55
0.66 0.73 0.73
2.44 2.37 2.03
34.99 25.31 23.79
18.03 16.94 18.43
12.79 12.10 11.57
8.12 8.41 6.60
116.46 89.66 77.48
0.51
0.56
0.15
1.94
3.17
0.88
0.85
0.73
2.45
22.00
32.04
5.93
4.59
41.20
Data represent the specialty-specific calculated complexity (mean quantity [from Table 1] multiplied by Z-transformed COV [from Table 2] multiplied by Z-transformed diversity [from Table 3]) for each input/output and totals. “Patients” complexity is the sum of the Z-transformed new patient proportion [from Table 1], age COV [from Table 2] and demographic diversity [from Table 3]. “Total Input Complexity” is the sum of the input complexities across each specialty and “Total Output Complexity” is the sum of the output complexities across each specialty. “Total Encounter Complexity” is the product of the Total Output Complexity and “2” raised to the power of the Total Input Complexity. “Visit Duration” is the weighted mean duration-of-visit for each specialty. “Density” represents the total input, output and encounter complexity measures divided by the mean duration-of-visit and then multiplied by 60 to derive the hourly complexity density. a b c
Standard errors: Allo 0.01. Standard errors: Allo 0.05. Standard errors: Allo 0.15.
Table 5 Differences among component-based clusters. Measures
Complexity Input Reasons Diagnoses Exam/tests Patients Total Output Medications Other Therapy Disposition Total Total encounter Density Visit duration Input complexity Output complexity Total complexity
Cluster 1 Family medicine General internal medicine Cardiology Oncology
Cluster 2 General surgery OBGYN Dermatology Urology
Cluster 3 PediatricsENT Orthopedics Psychiatry Neurology Ophthalmology ENT
F(p) [cluster differences]
0.67 0.79 1.05 1.91 4.41
0.51 0.60 0.85 1.94 3.91
0.56 0.62 0.35 2.02 3.55
9.43(0.004) 16.83(0.000) 10.40(0.003) 3.14(0.083) 11.66(0.002)
1 42,3 1 42,3 1,2 43
1.64 0.60 0.73 2.97 63.12
0.94 0.67 0.70 2.31 34.82
0.89 0.72 0.73 2.33 27.73
12.54(0.001) 0.82(0.469) 3.93(0.052) 9.85(0.004) 46.59(0.000)
1 42,3
1 42,3 1 42,3
21.76 12.31 8.31 176.5
17.76 13.22 7.83 117.6
21.66 10.50 6.77 81.2
1.06(0.380) 2.14(0.164) 1.86(0.201) 19.61(0.000)
1 4243
1 4243
Data represent the cluster-specific mean complexities and visit-durations. “F(p)” represents the F-statistics and their p-values for each Oneway ANOVA comparing clusters. Significant (p r 0.05) intercluster differences as determined by REGWF posthoc testing are then described.
services is based on an expanded coding list known as the Healthcare Procedures Coding System (HCPCS). The work relative value unit (RVU) assigned to ambulatory visits is identical across
specialties with the assumption that the work is equivalent for each specialty. This assumption is faulty; General Internal Medicine, Family Medicine, Cardiology and Oncology have significantly
D. Katerndahl et al. / Healthcare 3 (2015) 89–96
higher complexity of care and therefore workload than other specialties. Adult primary care EM services are generally undervalued. The range of EM services is not fully captured by CPT codes and procedures are overvalued. There is no measure of complexity of care, and decision-making about EM payment is delegated to a select group of specialty societies with little primary care involvement. Because Family Medicine and General Internal Medicine cluster with Cardiology and Oncology, one might expect that all four disciplines would be paid similarly; however, these specialties are paid differently with the majority of payment in the latter two specialties coming from procedures (Cardiology) and chemotherapy (Oncology). Revaluing RVUs would create incentives to better balance their services between cognitive and procedural work. Increasing documentation demands, guideline requirements and EHR use coupled with shortages of primary care physicians have significant implications. The Affordable Care Act (ACA) should not only intensify these trends,21 but result in more low-income patients seeking primary care, increasing patient diversity and demand for preventive and chronic care, and a shift of focus from inpatient to outpatient settings. These newly-insured patients are anticipated to be more complex than usual, magnifying current physician shortages and decreasing durations-of-visit. Quality of care may decline due to physicians’ inability to meet increased complexity demands of comprehensive and guideline-based care. Such consequences require adjustments in primary care delivery that can only be accomplished with an expanded primary care workforce and transformation of the primary care offices teams into models such as the patient-centered medical home attending to the work both within the consultation and around the consultation.22 These changes require adequate payment reform. Health costs may increase for primary care as complexity and density increase. Time-sensitive activities and preventive services may further decrease as primary care physicians attempt to compensate for complexity burden by reducing such activities. But total cost of care may still be reduced if the use of primary care within the system is optimized.23 Finally, physicians in high complexity disciplines (e.g. Family Medicine and General Internal Medicine) may experience declining career satisfaction and recruitment of new physicians, resulting from growing perceptions of time inadequacy and physician shortages.9 Increased financial input into primary care can translate into savings in other settings.24 In addition, improved primary care payment support could enable new models of care, improving both practice function and physician satisfaction.22 4.2. Issues and limitations This methodology evaluates large numbers of visits and is not suitable to measure complexity of individual encounters. Reliance upon the NAMCS database may minimize possible input complexity because only three complaints and three diagnoses can be included. Prior studies in Family Medicine settings suggest that visits often address more than three problems.25 In addition, physicians may overestimate the duration of visit, leading to underestimates of complexity density. Content of self-report cards used in NAMCS may not accurately reflect the visit, distorting complexity measures. The 1985–1991 NAMCS surveys could not differentiate between primary and specialty care disciplines based upon ten primary care criteria.26 However, good agreement exists between content recorded and physician rankings,27 and studies of physician-rated work found consistent, reproducible estimates.28 The method used here also does not differentiate among diagnoses in terms of impact on complexity (URI weights the same as schizophrenia); however, the effect of particular diagnoses may depend upon the training and experience of the physician treating it. Though the minimum number of visits per discipline was 1161,
95
variability measures could be affected by use of varying numbers of visits per discipline. Definitions of “meaningful” levels of complexity have not been determined. Finally, this study did not assess care delivered outside of the office visit or complexity embedded in clinical problems due to social and contextual dependence on family and relational issues. In addition, methodological decisions potentially impacted results. First, to test whether including post-operative visits with other visits (as done in this study) distorted relative complexity, we excluded post-operative visits from complexity calculations. Only four disciplines changed their total encounter complexities by more than 3% and total encounter complexity did not differ significantly with or without post-operative visits (paired t ¼0.011, p¼ 0.991). Overall rankings of specialties with and without postoperative care changed little for either total encounter complexity (rs ¼ 0.995, p ¼0.000) and total complexity density (rs ¼0.996, p¼ 0.000). Second, Psychiatry’s reliance on a different diagnostic classification system (DSM-V) than other disciplines may falsely simplify its observed complexity.13 Psychosocial factors were not included in this measure, supporting a falsely-low estimate for Psychiatry. However, Psychiatry was low in reasons-for-visit and testing complexity, suggesting that diagnosis complexity is not the whole story. Finally, use of COV as the statistic of variability resulted in the paradox that Pediatrics (with its disciplinedefined limits on age) had the most variability in patient age. Using the entire sample’s mean age instead would result in a 20% reduction in total encounter complexity for Pediatrics alone, but such reduction minimizes complexity caused by differences in developmental stage as health complexity changes most rapidly during these years. The high rates of errors29 and adverse drug events30 suggest that Pediatrics is more than minimally complex.
5. Conclusions Family Medicine and General Internal Medicine encounters are most complex. Psychiatry has the lowest input complexities with long visits further decreasing complexity densities. While Pediatrics has some diversity similarities to other primary care disciplines, its lack of chronic care visits results in complexities similar to Psychiatry and surgical specialties. Finally, OB/GYN has the lowest output complexity and clusters with General Surgery and Urology, baring few similarities to primary care disciplines. These findings have important implications for discrepancies in payment. Revaluing payments based on complexity could bring better balance to cognitive and procedural services, and better meet needs of people receiving insurance under the ACA.
Conflict of interest disclosure statement This statement accompanies the article “Complexity of ambulatory care across disciplines,” authored (co-authored) by “David Katerndahl” (“Robert Wood and Carlos Roberto Jaén”) and submitted to Healthcare as an original article. Below all authors have disclosed relevant commercial associations that might pose a conflict of interest: Consultant arrangements: None Stock/other equity ownership: None Patent licensing arrangements: None Grants/research support: None Employment: None Speakers' bureau: None Expert witness: None Other: None
96
D. Katerndahl et al. / Healthcare 3 (2015) 89–96
Acknowledgments We wish to acknowledge that this study was funded through a contract from the American Academy of Family Physicians. We wish to thank Drs. Robert Phillips and Sandra Burge for their critique of the manuscript. Appendix A. Supporting information Supplementary data associated with this article can be found in the online version at http://dx.doi.org/10.1016/j.hjdsi.2015.02.002. References 1. Abbo ED, Zhang Q, Zelder M, Huang ES. Increasing number of clinical items addressed during the time of adult primary care visits. J Gen Intern Med. 2008;23:2058–2065. 2. Carter LM, McHenry DS, Godlington FL, Meechan JG. Prescribed medication taken by patients attending general dental practice: changes over 20 years. Br Dent J 2007;2003:E8, 10.1038/bdj.2007.629. Accessed 07.07.14. 3. Chai G, Goverale L, McMahon AW, Trinidad JP, Staffa J, Murphy D. Trends of outpatient prescription drug utilization in US children. Pediatrics. 2012;130: 23–31. 4. Petros P. Non-linearity in clinical practice. J Eval Clin Pract. 2003;9(2):171–178. 5. Bar-Yam Y. Dynamics of Complex Systems. Reading, MA: Perseus Books; 1997. 6. Croskerry P, Shapiro M, Campbell S, et al. Profiles in patient safety. Acad Emerg Med. 2004;11:289–299. 7. McGlynn EA, Asch SM, Adams J, et al. Quality of health care delivered to adults in the United States. N Engl J Med. 2003;348:2635–2645. 8. Asch SM, Kerr EA, Keesey J, et al. Who is at greatest risk for receiving poor quality of health care? N Engl J Med. 2006;354:1147–1156. 9. Katerndahl D, Parchman M, Wood R. Perceived complexity of care, perceived autonomy, and career satisfaction among primary care physicians. J Am Board Fam Med. 2009;22:24–33. 10. Burdi MD, Baker LC. Physicians’ perceptions of autonomy and satisfaction in California. Health Affairs. 1999;18:134–145. 11. Katerndahl DA, Wood R, Jaen CR. Family medicine outpatient encounters are more complex than those of cardiology and psychiatry. J Am Board Fam Med. 2011;24:6–15.
12. Katerndahl DA, Wood R, Jaen CR. A method for estimating relative complexity of ambulatory care. Ann Fam Med. 2010;8:341–347. 13. Boisot M, Child J. Organizations as adaptive systems in complex environments. Organ Sci. 1999;10:237–252. 14. Rosen AK, Reid R, Broemeling AM, Rakovski CC. Applying a risk-adjustment framework to primary care. Ann Fam Med. 2003;1:44–51. 15. Reza FM. Introduction to Information Theory. New York: Dover Publications, Inc; 1994. 16. Green LA. Implications of measuring complexity (ed). Ann Fam Med. 2010;8:291–292. 17. National Center for Health Statistics. National Ambulatory Medical Care Survey. Hyattsville, MD: U.S. Department of Health and Human Services; 2010. 18. Edwards ST, Mafi JN, Landon BE. Trends and quality of care in outpatient visits to generalist and specialist physicians delivering primary care in the United States, 1997–2010. J Gen Intern Med. 2014;29:947–955. 19. Horner RD, Szaflarski JP, Ying J, et al. Physician work intensity among medical specialties. Med Care. 2011;49:1007–1011. 20. Jacobson Jr CJ, Bolon S, Elder N, et al. Temporal and subjective work demands in office-based patient care. Med Care. 2011;49:52–58. 21. Hoff T. Embracing a diversified future for US primary care. Am J Manag Care. 2013;19(1):e9–e13. 22. Sinsky CA, Willard-Grace R, Schutzbank AM, Sinsky TA, Margolis D, Bodenheimer T. In search of joy in practice: a report of 23 high-functioning primary care practices. Ann Fam Med. 2013;11:272–278. 23. Starfield Barbara. Is primary care essential?. Lancet. 1994;344.8930:1129–1133. 24. Neuhausen K, Grumbach K, Bazemore A, Phillips RL. Integrating community health centers into organized delivery systems can improve access to subspecialty care. Health Affairs. 2012;31:1708–1716. 25. Goodson JD. Unintended consequences of resource-based relative value scale reimbursement (ed). JAMA. 2007;298:2308–2310. 26. Franks P, Clancy CM, Nutting PA. Defining primary care: empirical analysis of the National Ambulatory Medical Care Survey. Med Care. 2014;35:655–668. 27. Rethans JJ, Martin E, Metsemakers J. To what extent do clinical notes by general practitioners reflect actual medical performance. Br J Gen Pract. 1994;44: 153–156. 28. Hsiao WC, Braun P, Yntema D, Becker ER. Estimating physicans’ work for a resource-based relative-value scale. N Engl J Med. 1988;319:835–841. 29. Neuspiel DR, Stubbs EH. Patient safety in ambulatory care. Ped Clin N Am. 2012;59:1341–1354. 30. Tache SV, Sonnichsen A, Ashcroft DM. Prevalence of adverse drug events in ambulatory care. Ann Pharmacother. 2011;45:977–989.
