Validating the ACG Case-mix System in a Spanish population setting: a cross-sectional study

Sicras-Mainar et al

a) Antoni Sicras-Mainar

b) Ruth Navarro-Artieda

c) Ignasi Ruano-Ruano

d) Josep Serrat-Tarrés

a) Josep Codes-Marco

a) Planning Management. Badalona Serveis Assistencials, SA. Badalona (Barcelona). Spain.

b) Medical Documentation Service. Hospital Germans Trías i Pujol. Badalona (Barcelona). Spain.

c) Planning Management. Barcelona’s Sanitary Region. CatSalut. Barcelona. Spain.

d) Public Health Management. Badalona Serveis Assistencials, SA. Badalona (Barcelona). Spain.


Purpose. Our objective was to validate the Johns Hopkins ACG case-mix system used in primary care and specialized care centres attending a defined population in Spain.

Methods. Population based cross-sectional study carried out applying the ACG case-mix system to the clinical records of the attended population in five primary care teams and two hospitals during the period of one year. Main measurements were: age, sex, clinical service (Family Medicine or Paediatrics), dependent variables (visits, episodes, primary care costs and total costs), and morbidity. A cost modelling for each patient was obtained by adding the fixed costs to the corresponding non-fixed costs. The determination coefficient (R2) was used to measure the explained variability by dividing intra-groups variance/total variance (ANOVA).

Results. 81,873 patients were included with an average number of 4.8±3.5 episodes and 8.0±8.1 visits/patient/year. The explained variance (R2) of ACG classification was 73.1% (75.5% log transformation) for episodes, 43.2% (54.0% log transformation) for visits, 19.6% (54.8% log transformation) for primary care costs, and 22.7% (48.3% log transformation) for total costs, p=0,000.

Conclusion. ACG System classified a defined population on the basis of morbidity and individual resources consumption. Moreover, ACG system has been useful to assess the clinical (co-morbidity) and economical information of each centre.

Key words: Adjusted clinical groups. Resource use. Management. Information system.


Patient classification systems have not been widely used in ambulatory care, even in the US where they were created1. Systems who classify patients according to resource consumption could be useful to compare the variability between individuals from the same population group. In this sense, ACG system classifies individuals in an unique group corresponding to a certain need for health care resources, according to their morbidity during a concrete time period. This system was developed by Starfield and Weiner2 with the aim of measuring population’s disease burden on the basis of co-morbidity levels, measured as a sum of diagnoses. It could be potentially used as a tool for risk adjustment in capitation budgets or for efficiency assessment in resource consumption3-7.

The ability of these tools to deal with clinical management needs remains uncertain in Spain. Some evidence about their theoretical properties is actually available, but information related with their practical applications is scarce in the literature8-10. In addition, no integration of clinical information and resource consumption coming from a unique geographical area has been developed. The aim of our study was to validate the Johns Hopkins ACG case-mix system used in primary care and specialized care centres attending a defined population in Spain.


Population based cross-sectional study carried out applying the ACG case-mix system to the clinical records of the attended population in five primary care teams (PCT) and two hospitals during the period of one year (2005). This geographical area has an assigned population of 107,720 inhabitants, 15.1% of them equal or greater than 65 years, in an urban setting, with a medium-low socioeconomic level and predominantly employed in industrial devices. The five PCT belonging to Badalona Serveis Assistencials (BSA), a health organization with similar characteristics to those operating in Catalonia. All patients attended during the year 2005, and previously assigned to the PCT, were included in the study. Patients changing their assignment to a PCT out of the study area during the year 2005 were excluded.

The collected variables were: age, sex, clinical service Family Medicine (equal or greater than 15 years) or Paediatrics (from 0 to 14 years), dependent variables (visits, episodes, primary care costs and total costs), and morbidity. Visits were defined as every contact of patients with the PCT due to a health problem carried out at the health centre or at patient’s home. Episodes were considered as equivalent to diagnosis and quantified by means of the International Classification of Primary Care (ICPC-2)11. A mapping conversion from ICPC-2 to International Classification of Diseases-9-Clinical Modification (ICD-9-CM) was made by a working group composed by an expert codifier, two clinicians and two technical advisors. ACG system requirement needs include age, sex and diagnosis codified according to ICD-9-CM. Its working algorithms have been described elsewhere2. At the end of the process, each patient is classified in a unique group of resource consumption.

A partial cost modelling was defined on the basis of the organization characteristics and the available information. Individual costs for each attended patient along the time period was used as a unit of analysis for the final calculation. Costs’ modelling for each patient was obtained by adding the fixed costs (structural and functional costs) to the corresponding non-fixed costs (resources utilization). The main expenditures on fixed costs were: salaries, taxes, social imbursements, fungible materials, technical devices, cleaning, maintenance, supplies and structural costs. These expenditures were imputed according to the total number of visits for each patient, as a proxy of general services utilization. Non-fixed costs were calculated on the basis of diagnostic tests (laboratory, radiology, etc.), therapeutic needs (drugs prescription) and referrals generated for each patient in the primary care. Specialized care costs were quantified by adding the expenditures from ambulatory visits, hospital discharges, emergency attendance and drugs prescription. These items were quantified on the basis of information from the analytical accountability system of the organization or according to the official fees established by the Catalonia Health Service12.

As a previous step to the analysis, an exhaustive review of the obtained data was carried out to ensure the quality of clinical and administrative records. Costs variables, visits and episodes were log transformed. The percentage of explained variance was calculated by means of the determination coefficient (R2), obtained by dividing intra-groups variance/total variance (ANOVA). Variance homogeneity was assessed with the Cochrane test. Analyses were done using the statistical package SPSSW version 12.0.


Total number of assigned inhabitants was 107,720 at the end of the year 2005; 85.0% corresponding to Family Medicine and 15.0% to Paediatrics. Seventy-six per cent of total inhabitants were attended during the study period, amounting to 6.1visits/inhabitant/year for Family Medicine and 7.2 visits/inhabitant/year for Paediatrics.

General data and several indicators related with services utilization and costs are detailed in table 1. A total of 81,873 patients (76.0% of the assigned population) were included in the study, with an average number of 4.8±3.5 episodes/patient/year and 8.0±8.1 visits/patient/year. Mean age was higher for women, 42.6±23.0, than for men, 40.1±22.7, p=0.000. Proportion of men, average number of episodes and visits, and services utilization indicators are higher for the paediatric group.

The explained variance (R2) of ACG classification (n=106) was 73.1% (75.5% log transformation) for episodes, 43.2% (54.0% log transformation) for visits, 19.6% (54.8% log transformation) for primary care costs, and 22.7% (48.3% log transformation) for total costs, p=0,000 (table 2).


In our country, primary care is organized in a territorial assigned population model and the information system has achieved a high level of mechanization. Both features result in an ideal scenario to carry out geographical studies in a usual clinical practice setting. It’s noteworthy that our results must be interpreted with caution due to the lack of adequate standardization in the measure of variables across the different health centres. Despite of this question, to our knowledge, primary care centres have actually very similar clinical practice performance, protocol utilization and organization model and, therefore, the variability in health care results would be diminished13.

ACG grouper needs a limited number of variables (age, sex and diagnoses) to work. This simplicity has a good fit with some primary care characteristics such as the iterative assistance to the same patients along the time, the great volume of information daily managed, the limited time to attend each patient, and the multidisciplinary composition of PCT (physicians, nurses, and social workers). On the other hand, improvement chances appear in the specialized care related with the homogeneity of clinical, administrative and resource utilization records14-15.

Our percentage of explained variance is similar to the result of other published studies. Some authors2-10 have used different methods to calculate total costs, obtained by adding the costs coming from emergency visits, hospital discharges and specialists visits to the primary care costs. All of them agree, as our findings did, about one question: including in the analysis a variable related with morbidity leads to an increase in the proportion of explained variance; so, this is a practical conclusion that should be kept in mind for the quantification of health costs. For the whole dependent variables, the number of episodes achieved the higher proportion of explained variance, followed by the number of visits and, finally, by patient costs8-10,16-17. These results fully agree with the working sequence of ACG algorithm  and including the specialized care costs leads to an increase in the determination coefficient (R2) from 22.7% to 48.3% (log transformation). In our study, results are more consistent in Family Medicine than in Paediatrics, although there are no important differences comparing primary care adjusted costs with total adjusted costs (54.8% vs. 48.3%, respectively).

The most important limitations of the study are related with the quality and consistency of information systems, the precision in costs measurement, and the variability of diagnostic labelling by different physicians which could produce contaminant effects within ACG groups or lead to a low clinical sensitivity. ACG were designed to measure health care status and resource consumption in population groups, so they could be useful in community based future research applied to risk adjustment on a capitation budget model and to clinical management18-20.  We conclude that ACG are useful to classify individuals from a concrete population into groups on the basis of their morbidity and resource consumption (integrating primary care and specialized care data). Using ACG system jointly with a geographical information system provides an interesting scope of population co-morbidity. Moreover, ACG methodology has shown its usefulness to measure and integrate clinical and economical information from the different centres providing health care.


1. Elola J, Daponte A, Navarro V. Health indicators and the organisation of Health care Systems in Western Europe. Am J Public Health. 1995;85:1397-1401.

2. Starfield B, Weiner J, Mumford L, Steinwachs D. Ambulatory Care Groups: a categorization of diagnoses for research and management. Health Ser Res. 1991;26:53-74.

3. Reid RJ, MacWilliam L, Verhulst L, Roos N, Atkinson M. Performance of the ACG case-mix system in two Canadian provinces. Med Care. 2001;39:86-99.4. Hughes JS, Averill RF, Eisenhandler J, Goldfield NI, Muldoon J, Neff JM, Gay JC. Clinical Risk Groups (CRGs): a classification system for risk adjusted capitation-based payment and health care management. Med Care. 2004;42:81-90.5. Meenan RT, Goodman MJ, Fishman PA, Hornbrook MC, O'Keeffe-Rosetti MC, Bachman DJ. Using risk-adjustment models to identify high-cost risks. Med Care. 2003;41:1301-12.6. Rosen AK, Loveland SA, Rakovski CC, Christiansen CL, Berlowitz DR. Do different case-mix measures affect assessments of provider efficiency?. Lessons from the Department of Veterans Affairs. J Ambul Care Manage. 2003;26:229-42.7. Adams EK, Bronstein JM, Raskind-Hood C. Adjusted clinical groups: predictive accuracy for Medicaid enrollees in three states. Health Care Financ Rev. 2002;24:43-61.

8. Juncosa S, Bolíbar B, Roset M, Tome R. Performance of an ambulatory case mix system in primary care in Spain: Ambulatory Care Groups (ACGs). European J Public Health. 1999;9:27-35.

9. Bolaños-Carmona V, Ocaña-Riola R, Prados-Torres A, Gutierrez-Cuadra P. Variations in health services utilization by primary care patients. Health Serv Manage Res. 2002;15:116-25.

10. Sicras-Mainar A, Serrat-Tarres J. Measurement of relative cost weights as an effect of the retrospective application of adjusted clinical groups in primary care. Gac Sanit. 2006;20:132-41.11. Lamberts H, Wood M. International Classification of Primary Care (ICPC-2). Classification of consultation reasons.Barcelona: Masson/SG; 1990.12. Garcia-Cardona F, Molins-Perez G, Farre-Pradell J, Martin-Sanchez A, Pane-Mena O, Gallego-Español R. Cost accountability in primary care: list of services. Aten Primaria. 1995;16:141-5.

13. Sackett D, Rosenberg W, Gray J, Haynes RB, Richardson WS. Evidence based medicine: what it is and what it isn't. BMJ 1996;312:71-72.

14. Orueta JF, Urraca J, Berraondo I, Darpon J, Aurrekoetxea JJ. Adjusted Clinical Groups (ACGs) explain the utilization of primary care in Spain based on information registered in the medical records: a cross-sectional study. Health Policy. 2006;76:38-48.

15. Pietz K, Ashton CM, McDonell M, Wray NP. Predicting healthcare costs in a population of veterans affairs beneficiaries using diagnosis-based risk adjustment and self-reported health status. Med Care. 2004;42:1027-35.16. Engstrom SG, Carlsson L, Ostgren CJ, Nilsson GH, Borgquist LA. The importance of comorbidity in analysing patient costs in Swedish primary care. BMC Public Health. 2006;6:36.17. Carlsson L, Strender LE, Fridh G, Nilsson G. Types of morbidity and categories of patients in a Swedish county. Applying the Johns Hopkins Adjusted Clinical Groups System to encounter data in primary health care. Scand J Prim Health Care. 2004;22:174-9. 18. Fishman PA, Goodman MJ, Hornbrook MC, Meenan RT, Bachman DJ, O'Keeffe Rosetti MC. Risk adjustment using automated ambulatory pharmacy data: the RxRisk model. Med Care. 2003;41:84-99.19. Wahls TL, Barnett MJ, Rosenthal GE. Predicting resource utilization in a veteran’s health administration primary care population: comparison of methods based on diagnoses and medications. Med Care. 2004;42:123-8.

20. Wong ST, Kao C, Crouch JA, Korenbrot CC. Rural American Indian Medicaid health care services use and health care costs in California. Am J Public Health. 2006;96:363-70.

Table 1. General features of included patients and costs results

Variables distribution

Family Medicine




     Number of patients




     Number of episodes




     Number of visits




     Mean age (SD)




     Sex (% women)




     Average number of episodes by patient (SD)




     Average number of visits by patient (SD)




     Average number of episodes by visit (SD)





     Primary care costs (euros)




     Total costs (euros)




     Average Primary care costs by patient (SD)




     Average Total costs by patient (SD)




SD: Standard deviation

Table 2. Distribution explained variance results          

Variables distribution

Family Medicine



Explained variance (R2) of ACG classification without log transformation(1)

     Number of visits




     Number of episodes




     Number of ADG




     Referrals to specialists costs




     Drug prescription costs




     Primary care costs




     Total costs




Explained variance (R2) of ACG classification with log transformation(1)

     Ln number of visits




     Ln number of episodes




     Ln Primary care costs




     Ln Total costs




R2: determination coefficient (Fisher-Snedecor); ACG: Adjusted Clinical Groups (n=106); (1):statistical signification, p<0,001; ADG: Ambulatory Diagnostic Groups; Ln: log transformation


Copyright Priory Lodge Education Limited 2008

First Published February 2008

Home • Journals • Search • Rules for Authors • Submit a Paper • Sponsor us
Rules for Authors
Submit a Paper
Sponsor Us

Google Search

Advanced Search



Default text | Increase text size