Max Rady College of Medicine

Concept: Adjusted Clinical Groups® (ACG®) - Overview

 Printer friendly

Concept Description

Last Updated: 2019-01-17

Introduction to Adjusted Clinical Groups (ACG)

    The Johns Hopkins Adjusted Clinical Group® (ACG®) system is a population/patient case-mix adjustment system developed by researchers at The Johns Hopkins University School of Hygiene and Public Health in Baltimore, Maryland, U.S.A. The ACG system measures health status by grouping diagnoses into clinically cogent groups. The goal of the ACG system is to assign each individual a single, mutually exclusive ACG value, which is a relative measure of the individual's expected or actual consumption of health services. While our assessments suggest this measure has promise as a tool for comparing case-mix across populations, further analyses were done to determine how sensitive the ACG value is to other measures of population morbidity such as premature mortality or physician contact rate. We found that, in Manitoba, ACGs are closely related to premature mortality. (Reid et al, 2002).

    Some important introductory notes include:

    • The ACG system quantifies morbidity by grouping individuals based on their age and gender and all medical diagnoses that have been recorded over a defined period of time, typically one year.
    • International Classification of Disease, ICD-9 & ICD-10 diagnosis codes for similar conditions are clustered based on expected consumption of health care resources and short-term clinical outcomes.
    • ICD-9 and ICD-10 codes typically come from administrative data sets (eg. medical services/physician claims and hospital discharge abstract records).
    • The ACG system was designed to use 4-digit ICD-9 / ICD-10 diagnoses codes. However, work at MCHP has revealed that the ACG system performs well with less specific 3-digit ICD-9 codes. NOTE: Over time, the majority of ICD-9-CM codes recorded in the Medical Services /Physician Claims data are only 3-digits. In December 2018, a 5-digit diagnosis code variable was added to the Medical Services data. If recorded, the first 3-digits of both the 3-digit and 5-digit diagnosis code variables will match exactly. The 5-digit diagnosis code variable applies to records from 2015/16 forward.
    • Recent versions of the ACG system work with ICD-10 codes (ICD-10-CA truncated to the 4th digit).
    • A comparison of the ACG system in Manitoba and B.C. provided some indication that using 3-digit ICD-9 codes may lead to less resource intensive ACG scores than if the corresponding 4-digit ICD-9 codes are available (Reid R. et al., Performance of the ACG case mix system in two Canadian provinces. , 2001).

    The underlying assumption of the ACG system is that a measure of morbidity can help explain an individual's predicted consumption of medical care resources. For more information, please read The Johns Hopkins ACG System Web Site - Describing Morbidity Burden available on the Johns Hopkins web site.

ACG Values: Some Considerations

    ACG values represent the resource burden or general illness of an individual. There are several considerations when working with ACGs:

    • ACG values of 5100, 5110, 5200 represent non-users; they have no diagnosis in the time period.
    • Costs (if you are doing that) are out of skew with ACG values for new newborns, pregnancy, delivery and non-users.
    • When calculating ACG (or associated variables) all individuals must have at least 6 months of coverage and no more than one year. This generally means that you need to have access to the research registry and identify an index date of interest. An exception to this rule at MCHP has sometimes been made for individuals that died.
    • When using the medical services/physician claims data you must exclude diagnosis codes starting with A, B, or C. These are specific to chiropractic services and are not 'real' ICD-9-CM codes.
    • When using hospital discharge abstracts data, any 'E codes' should also be excluded since they are not used in the ACG grouping.
    • The ACG case mix system is based on ICD-9 - not ICD-9-CM codes. The differences are not great and work done by Reid R.J. et al., Do Some Physician Groups See Sicker Patients Than Others? Implications for Primary Care Policy in Manitoba., (2001) has shown consistent results with Manitoba data and other sources.
    • With the addition of variables on the delivery status of pregnant women and infant birth weight, there are now 92 ACG categories and sub-categories. Thus, ACGs are the mutually exclusive terminal groups of the ACG system and represent combinations of Aggregated Diagnosis Groups™ (ADGs®), age and gender categories.

Pregnancy and Childbirth

    ACG values beginning with 17 represent pregnancy and childbirth/delivery. This situation deserves special consideration.

    • A delivery is an event which consumes resources, and yet does not indicate an increased burden of disease. The ACG grouper was primarily designed to measure resource use, and it is only due to the fact that higher resource use generally occurs among sicker people that we are able to use the grouper the way we do.
    • A simple delivery, however, puts an otherwise healthy women into a category of resource use similar to someone with an actual disease.
    • Prior to a study using the ACG system, a decision should be made on whether the claims for pregnancy and childbirth should be included in the ACG input files. From past experience, the difference in resource use for these women is significant. In some MCHP studies these values have been dealt with separately, particularly, 1700 and 1710.

Resource Utilization Bands (RUB)

    To simplify things, the ACG System software will automatically assign a six-level (Low to High) simplified morbidity category termed Resource Utilization Bands, or RUB. The six RUBs are formed by combining the ACG mutually exclusive cells that measure overall morbidity burden. The six levels are listed below:
    RUB Category:
    0 - Non-users
    1 - Healthy Users
    2 - Low Morbidity
    3 - Moderate
    4 - High
    5 - Very High

Expanded Diagnosis Clusters ™ (EDCs™)

    The ACG system also includes Expanded Diagnosis Clusters (EDCs or Dino-Clusters) that can be used to categorize cases with similar diseases or conditions. The EDC methodology assigns ICD codes found in medical services/physician claims or hospital discharge abstract data to one of 264 EDCs, which are further organized into 27 categories called Major Expanded Diagnosis Clusters (MEDCs). As broad groupings of diagnosis codes, EDCs help to remove differences in coding behavior between practitioners. However, a significant number of diagnosis codes are not defined when creating EDCs. At MCHP, we have found that individuals who are not assigned an EDC ranges from 8% to 20% depending on who has done the work and the data source. Given that previous versions of the ACG System capture approximately 9,400 ICD codes, this is not unexpected. NOTE: there are around 14,000 ICD-9 codes.

Aggregated Diagnosis Groups (ADG)

    With the assistance of expert clinicians, the ACG system categorizes ICD-9 / ICD-10 diagnosis codes into one of 32 groups (see Note 1 below), called Aggregated Diagnosis Groups™ (ADGs®). The ACG method groups every medical diagnosis code, (with the exceptions noted above) assigned to a patient, and is based on the following five clinical and expected utilization criteria:

    1. duration of the condition (acute, recurrent, or chronic);
    2. severity of the condition (e.g., minor and stable versus major and unstable);
    3. diagnostic certainty (symptoms focusing on diagnostic evaluation versus documented disease focusing on treatment services);
    4. etiology of the condition (infectious, injury, or other); and
    5. specialty care involvement (medical, surgical, obstetric, haematology, etc.).

    If, over a defined interval (usually one year) an individual has at least one of the diagnoses in an Aggregated Diagnosis Group, they are assigned that ADG. A patient can be assigned as few as none and as many as 32 ADGs. The system further classifies ADGs as "major" or "minor".

    The ACG system then clusters the ADGs into 12 "Collapsed ADGs" (CADGs) and then combines CADGs into common patterns called "Major Ambulatory Categories" (MACs). MACs are then further partitioned and combined with relevant age and gender categories to form ACGs.


    1. ADGs are numbered 1 to 34, but two of these ADGs are labeled "No Longer in Use" (#15 and #19), hence there are currently 32 ADGs used in MCHP research. Two additional ADGs (#31 - Prevention/Administrative and #34 - Dental) are routinely excluded because they are not considered comorbidities. For a list of ADG numbers, labels and example ICD-9-CM codes please see the technical documentation example available in the Links section below (internal access only).
    2. Up until 2001 and the release of version 5, the ACG system was referred to as Ambulatory Care Groups (ACGs) and ADGs were referred to as Ambulatory Diagnostic Groups (ADGs). Ambulatory Diagnostic Groups were based on eight criteria. For more information on Ambulatory Diagnostic Groups, please see the glossary entry Ambulatory Diagnostic Groups (ADGs) .

Mental and Major Physical ADGs

    Refinements to the ACG system resulted in the development of the concept of "Major ADGs". MCHP research has used the following ADGs to define Mental ADGs and Major Physical ADGs ( Martens et al., 2008, 2010 ):

      Mental ADGs

      • ADG 23 - Psychosocial: Time Limited, Minor
      • ADG 24 - Psychosocial: Recurrent or Persistent, Stable
      • ADG 25 - Psychosocial: Recurrent or Persistent, Unstable

      Major Physical ADGs

      • ADG 3 - Time Limited: Major
      • ADG 4 - Time Limited: Major - Primary Infection
      • ADG 9 - Likely to Recur: Progressive
      • ADG 11 - Chronic Medical: Unstable
      • ADG 16 - Chronic Specialty: Unstable - Orthopedic
      • ADG 22 - Injuries / Adverse Effects: Major
      • ADG 32 - Malignancy

ADGs in Pediatric Age Group

    The software classifies some ADGs as major and some as minor for each age group. In Currie et al., (2010), investigators identified that the ADGs excluded several diagnoses that are highly prevalent among children, and which are thought to have important effects. While the Johns Hopkins definition includes acute, unstable mental health conditions such as psychosis it excludes "stable" mental conditions such as Attention-Deficit Hyper Activity Disorder (ADHD) and Conduct disorders, two of the most common mental health conditions among children. It also excludes asthma and major injuries. Despite this, the researchers assert that the codes capture the majority of the chronic or acute major illnesses that occur during childhood.

    In accordance with the Johns Hopkins definition, Currie et al. (2010) used ADG codes 3, 9, 11, 12, 13, 18, 25, and 32 to determine the major illnesses for children ages 0-17. For children 18 and over, the ADG system defines major conditions as codes 3, 4, 9, 11, 16, 22, 25, and 32, but in this study the 0-17 definition was used for all participants in order to maintain consistency.

    Table A2 in Currie et al. (2010) lists the top 10 ICD-9 codes for children with Major Conditions, by age group.

    Congenital / Perinatal Conditions & ADG Codes
    Currie et al. (2010) used ICD-9 codes 740 - 779 to define congenital and perinatal problems. This definition counted only congenital and perinatal problems that are deemed major by the ADG system. Table A4 in Currie et al. (2010) lists the top 10 ICD-9 codes for congenital anomalies at each age.

    For a list of ICD-9- and ICD-10 codes used to identify congenital / perinatal conditions defined as specific major pediatric ADGs in more recent MCHP research, please see the MCHP Documentation - Congenital / Perinatal Conditions that are Major Pediatric ADGs .

Validation of ACGs

The ACG System Compared to Other Diagnoses Grouping Systems

    As opposed to other diagnosis grouping systems, the ACG system does not rely on identifying the principal or most responsible diagnosis.

    • It identifies common combinations of morbidities (related and unrelated) that build upon each other both additively and multiplicatively to determine an individual's need for health services. Thus, ACGs categorize people while most other grouping systems categorize events or episodes.
    • ACGs are not recommended as a tool for conducting individual patient analyses, for comparing individuals to a norm or for evaluating the uses of resources by an individual. ACGs are for the comparison of groups of members.
    • The ACG system provides, through its grouping strategies, two methods to quantify the burden of morbidity of individuals, for example, within a geographically defined area:

      1. use the ADG and demographic categories to create a morbidity "profile" for each individual. (e.g. an individual with 10 or more ADGs will have a larger burden than an individual with only one or two ADGs).
      2. use ACGs which provide a simplified method of categorizing persons into single morbidity categories. Either of these sets of variables can then be added to multivariate models to control for case-mix.

      In general, population-oriented analyses will have more flexibility and be more comprehensive if both users and non-users are included. The ACG system has been specifically designed to handle non-users. It encourages the use of two input files:

      1. a population file; and
      2. a medical claims/service utilization file. Persons in the population file but not in the claims/service utilization file will be assigned to a non-user ACG category.

      Note: Clinical Classifications Software (CCS) is another tool for clustering patient diagnoses and procedures into a manageable number of clinically meaningful categories. The CCS was developed for the Healthcare Cost and Utilization Project (HCUP), sponsored by the Agency for Healthcare Research and Quality (AHRQ). In the future, MCHP will attempt evaluate the utility of this grouper for research at the Centre. For now, more information on CCS is available on the HCUP web site: Healthcare Cost and Utilization Project (HCUP)- Clinical Classifications Software (CCS) for ICD-9-CM Fact Sheet

Use of the ACG System at MCHP

    MCHP has been using the ACG software for many years. Currently, we are running version 10 of the ACG system.

    Previous documentation for running different version of the ACG system is also available. See the SAS code and formats section below for information that was previously available in this concept for running version 6 and version 8 of the ACG system, as well as creating input files for running version 10 of the ACG system. (internal access only).

    The Links section below also provides access to current and historical information / MCHP documentation on ACGs, including:

    • an internal directory structure containing documentation on different versions of the ACG system ( (internal access only); and
    • an ACG Questions and Answers document, providing additional details for using the ACG system ( (internal access only).

Running the ACG System, Version 10 / Version 11 at MCHP

    The ACG System, both versions 10 and 11, operates in a Windows based environment and can be run interactively or in batch mode. Consider the following points:

    A. INPUT Files

    1. Population file - The population used in the ACG group should contain all of the individuals of interest that have at least 6 months of coverage. Typically, MCHP has used at least a time frame of 9 to 12 months. The population file should contain PHIN, AGE, SEX and other relevant demographic data. Additional information that the system may use includes low birth weight, pregnancy, and delivery status.
    2. Diagnoses file - The Diagnoses file should contain records with PHIN and a list of relevant diagnoses codes. ICD-9 and/or ICD-10 diagnoses codes (to the 3rd or 4th digit) are extracted for each individual from the Hospital Discharge Abstracts file. For Medical Services/Physician Claims data, limit the data to records with a prefix='7' and for efficiency, any non-ICD-9 codes should be excluded (e.g.: diag ^in: (' ','A','B','C')). Individuals must have at least 6 months of coverage and no more than one year. If an individual dies with less than 6 months then they are still retained and diagnosis codes are used. The database can have multiple records for each PHIN and a variable number of diagnoses codes.

    B. Create Diagnoses, Population and Drug Files Using the SAS Macro: %create_acg_diag()
    Use the MCHP SAS macro %create_acg_diag() to create the diagnosis, population, and drug raw text files, given the formatted files from MCHP, that are suitable for use with the Johns Hopkins ACG software. This includes diagnoses codes from the Hospital Discharge Abstracts data (both HAUM & DAD type files) and the Medical Services data. This program does some consistent processing around formatting the output files, formatting diagnoses codes, identifying deliveries, and low birth weight. The default output location is to the current SAS work directory.

    An example of the %create_acg_diag() macro code is available from the SAS code and formats section below (internal access only). More detailed documentation is available at the top of the macro program code. A current version of the macro code can be found in the MCHP system macro library.

    C. Running the ACG System Using the SAS Macro: %call_acg()

    Use the MCHP SAS macro %call_acg() to call the ACG software and run the ACG Grouper from within SAS using line commands. If the %create_acg_diag() macro was used to create the raw input files then a &workpath variable is available to identify the location of the files. This program generates various output files based on the ACG options.

    An example of the %call_acg() macro code is available from the SAS code and formats section below (internal access only). More detailed documentation is available at the top of the macro program code. A current version of the macro code can be found in the MCHP system macro library.

    NOTE: As of August, 2015, the old server has ACG software version 10 currently running, and this macro uses the v10 location. There is an option in the macro to identify the new v11 ACG software on the new server. The default location will be changed when most of the analysts are moved over to the new server.

    D. Output

    The warning file needs to be reviewed once the grouper is run. If the software does not generate the expected output, try running the software interactively as the warning and error messages generated provide additional information.
    For more information on using version 10 of the ACG System, please read the Data Analyst Meeting Summary Notes for November 27, 2013 (available internally only).

SAS Sample Code

    For examples of "old" SAS code used with different versions of the ACG Grouper, please see the SAS code and formats section below (internal access only).

Warning Codes from the ACG Grouper Software

    The ACG Grouper software will generate warning codes based on what it encounters. A list of the most common warning codes and descriptions are available (internal access only) in the Links section below.

Limitations and Cautions

  • Users should be aware of the limits on the license and use, as well as how to acknowledge the use of the ACG system properly. See The Johns Hopkins ACG® Case-Mix System acknowledgement statement under Acknowledgements in the More about the data repository section of our website.
  • MCHP does not use Drug Program Information Network (DPIN) data (pharmaceutical data) in the calculation of ACGs. However, for individuals, the number of ADGs is highly correlated with the number of different medications (ATC 4th level). (Kozyrskyj et al., (2005)).
  • When comparing costs/resource utilization, there are some issues with non-users, pregnancy/delivery, and newborns that should be considered.
  • If you discover issues with Expanded Diagnosis Clusters (EDCs), this may be caused by missing diagnoses.

Related concepts 

Related terms 



  • Currie J, Stabile M, Manivong P, Roos LL. Child health and young adult outcomes. 2008;National Bureau of Economic Research Working Paper 14482. [Abstract] [Related Links] (View)
  • Currie J, Stabile M, Manivong P, Roos LL. Child health and young adult outcomes. Journal of Human Resources 2010;45(3):517-548. [Abstract] (View)
  • Kozyrskyj A, Lix L, Dahl M, Soodeen R. High-cost Users of Pharmaceuticals: Who are They? Winnipeg, MB: Manitoba Centre for Health Policy, 2005. [Report] [Summary] (View)
  • Martens P, Fransoo R, The Need to Know Team, Burland E, Prior H, Burchill C, Romphf L, Chateau D, Bailly A, Ouelette C. What Works? A First Look at Evaluating Manitoba's Regional Health Programs and Policies at the Population Level. Winnipeg, MB: Manitoba Centre for Health Policy, 2008. [Report] [Summary] [Additional Materials] (View)
  • Martens PJ, Bartlett J, Burland E, Prior H, Burchill C, Huq S, Romphf L, Sanguins J, Carter S, Bailly A. Profile of Metis Health Status and Healthcare Utilization in Manitoba: A Population-Based Study. Winnipeg, MB: Manitoba Centre for Health Policy, 2010. [Report] [Summary] [Updates and Errata] [Additional Materials] (View)
  • Powe NR, Weiner JP, Starfield B, Stuart M, Baker A, Steinwachs DM. Systemwide provider performance in a Medicaid program. Profiling the care of patients with chronic illnesses. Medical Care 1996;34(8):798-810. [Abstract] (View)
  • Reid R, MacWilliam L, Verhulst L, Roos NP, Atkinson M. Performance of the ACG case mix system in two Canadian provinces. Med Care 2001;39(1):86-99. [Abstract] (View)
  • Reid RJ, Roos NP, MacWilliam L, Frohlich N, Black C. Assessing population health care need using a claims-based ACG morbidity measure: a validation analysis in the Province of Manitoba. Health Serv Res 2002;37(5):1345-1364. [Abstract] (View)
  • Reid RJ, Bogdanovic B, Roos NP, Black C, MacWilliam L, Menec V. Do Some Physician Groups See Sicker Patients Than Others? Implications for Primary Care Policy in Manitoba. Winnipeg, MB: Manitoba Centre for Health Policy and Evaluation, 2001. [Report] [Summary] (View)
  • Reid RJ, MacWilliam L, Roos NP, Bogdanovic B, Black C. Measuring Morbidity in Populations: Performance of the Johns Hopkins Adjusted Clinical Group (ACG) Case-Mix Adjustment System in Manitoba. Winnipeg, MB: Manitoba Centre for Health Policy and Evaluation, 1999. [Report] [Summary] (View)
  • Starfield B, Weiner J, Mumford L, Steinwachs D. Ambulatory care groups: a categorization of diagnoses for research and management. Health Serv Res 1991;26(1):53-74. [Abstract] (View)
  • The Johns Hopkins University Bloomberg School of Public Health. The Johns Hopkins ACG Case-Mix System Documentation & Application Manual Version 5.0. In: Jonathan P. Weiner (ed). Baltimore, MD: Johns Hopkins University; 2001. 0-0.(View)
  • The Johns Hopkins University Bloomberg School of Public Health. The Johns Hopkins ACG Case-Mix System Version 6.0 Release Notes (Systems Documentation). In: Jonathan P. Weiner (ed). Baltimore, MD: Johns Hopkins University; 2003. 0-0.(View)
  • Weiner JP, Starfield BH, Steinwachs DM, Mumford LM. Development and application of a population-oriented measure of ambulatory care case-mix. Med Care 1991;29(5):452-472. [Abstract] (View)
  • Weiner JP, Starfield BH, Lieberman RN. Johns Hopkins ambulatory care groups (ACGs). A case-mix system for UR, QA and capitation adjustment. HMO Practice 1992;6(1):13-19. [Abstract] (View)
  • Weiner JP, Dobson A, Maxwell SL:Coleman, K, Starfield B, Anderson GF. Risk-adjusted medicare capitation rates using ambulatory and inpatient diagnoses. Health Care Financing Review 1996;17(3):77-99. [Abstract] (View)


  • Health Measures
  • Health Services
  • health status
  • morbidity

Contact us

Manitoba Centre for Health Policy
Community Health Sciences, Max Rady College of Medicine,
Rady Faculty of Health Sciences,
Room 408-727 McDermot Ave.
University of Manitoba
Winnipeg, MB R3E 3P5 Canada