Comparison of two prospective rate-setting models: the DRG and PIR models.

The article compares two statistical prospective hospital reimbursement models: the diagnosis-related group (DRG) model and the prospective individualized reimbursement (PIR) model. Both models are applied to the same variables from the same data set, a random sample of 10,000 hospital discharges in Maryland in 1983. For comparative purposes, the two statistical models are allowed to differ only in their treatment of the predictive variable, "patient age." The criteria of comparison and results (DRG and PIR, respectively) are: number of patient groups required (469 and 337); accuracy of prediction of length of stay (38 percent and 45 percent of the total variation is explained by the models); correction for sampling bias (0 and 2.4 percent additional explained variation); and accuracy of prediction of total charges ($526 and $262 average error per patient).     

Gender as a Moderator Between Having an Anxiety Disorder Diagnosis and Coronary Artery Bypass Grafting Surgery (CABG) Outcomes in Rural Patients

Purpose: This paper examines gender as a moderating variable between having an anxiety disorder diagnosis and coronary artery bypass grafting surgery (CABG) outcomes in rural patients. Methods: Using the 2008 Nationwide Inpatient Sample (NIS) database, 17,885 discharge records of patients who underwent a primary CABG surgery were identified. Independent variables included age, gender, race, median household income based on patient's ZIP code, primary expected payer, the Deyo, Cherkin, and Ciol Comorbidity Index, and an anxiety comorbidity diagnosis. Outcome variables included in‐hospital length of stay and patient disposition (routine and nonroutine discharge). A 2 × 2 analysis of variance and logistic regression analyses were used to assess the interaction between gender and an anxiety disorder diagnosis on in‐hospital length of stay and patient disposition. Findings: Twenty‐seven percent of rural patients undergoing a CABG operation had a comorbid anxiety diagnosis. Rural patients who had nonroutine discharge were more likely to have comorbid anxiety diagnosis compared to rural patients who had a routine discharge. There was a significant interaction effect between having an anxiety diagnosis and gender on length of hospital stay but not for patient disposition. Conclusions: Three findings were noteworthy. First, anxiety disorder is prevalent in rural patients who are undergoing a CABG operation. Second, anxiety was a significant independent predictor of both length of hospital stay and nonroutine discharge for patients receiving CABG surgery. Last, having an anxiety disorder diagnosis increased hospital stay for both males and females; however, females seemed to be impacted more than males.     

DRGs in Europe: a cross country analysis for cholecystectomy

Cholecystectomy is the surgical removal of the gallbladder. It is the most common method for treating symptomatic gallstones. Despite the existence of well-established treatment guidelines, the rate of cholecystectomy varies widely across Europe. We analyse patients in 10 countries that had undergone surgery for the treatment of symptomatic gallstones. We test the performance of three models in explaining variation in the (log of) cost of the inpatient stay (seven countries) or length of stay (three countries). The first model includes only the diagnosis-related group (DRG) variables to which cholecystectomy patients were coded (M(D)), the second uses a core set of patient characteristics and episode-specific explanatory variables (M(P)), and finally, the third model combines both sets of variables (M(F)). Countries vary both in the number of DRGs used to classify cholecystectomy patients (range: 2-8), and in the percentage of patients covered by a single DRG (range: 50%-92%). The ability of combining both DRGs and patient level variables to explain cost variation among patients ranges from 58% in Spain to over 81% in Finland. The comparison of models' performance suggests that incorporating relevant patient characteristics may significantly improve DRG systems.     

How well do diagnosis-related group systems group breast cancer surgery patients? Evidence from 10 European countries

We analysed patient-level data (n = 72,235) from 563 hospitals in 10 European countries to assess the ability of national diagnosis-related group (DRG) systems to account for patient-level variation in cost or lengths of stay of breast cancer surgery patients against a standard set of patient characteristics, treatment and quality variables. We find that European DRG systems use very different types of classification variables and numbers of DRGs (range: 3-7) to classify these patients. In 6 of 10 countries, the set of patient characteristics, treatment and quality variables, which we were able to define across countries, perform better than the set of national DRGs in accounting for patient-level variation in resource consumption. Moreover, there appear to be factors that are consistently significant determinants of cost/length of stay of breast cancer surgery cases but are not, or at least not fully, considered in European DRG systems. Our results therefore raise concerns as to whether all systems rely on the most appropriate classification variables. In several countries, policymakers should reevaluate the appropriateness of their DRG algorithm for breast cancer surgery and of specific DRG weights.     

Patient classification and hospital costs of care for acute myocardial infarction in nine European countries

This study contributes to the literature on the performance of diagnosis-related groups (DRGs) for acute myocardial infarction (AMI) patients by evaluating in nine countries the factors--in addition to DRGs--that affect costs or length of stay and comparing the variation that can be explained with or without DRGs. We evaluate whether the existing DRGs for AMI patients would benefit from additional patient-related and treatment-related factors that are found in administrative data across countries. In most countries, the set of patient and quality variables performed better than the DRG variables. Our results suggest that DRG systems in all countries could be improved by including additional explanatory factors or by refining the existing DRGs. Our results suggest that for AMI and possibly for other related episodes, a refinement of DRGs to include information on patient severity, procedures and levels of complications could improve the ability of DRGs to explain resource use. It seems possible to improve DRG-like hospital payment systems through the inclusion of episode-specific variables.     

Do diagnosis-related groups appropriately explain variations in costs and length of stay of hip replacement? A comparative assessment of DRG systems across 10 European countries

This paper assesses the variations in costs and length of stay for hip replacement cases in Austria, England, Estonia, Finland, France, Germany, Ireland, Poland, Spain and Sweden and examines the ability of national diagnosis-related group (DRG) systems to explain the variation in resource use against a set of patient characteristic and treatment specific variables. In total, 195,810 cases clustered in 712 hospitals were analyzed using OLS fixed effects models for cost data (n=125,698) and negative binominal models for length-of-stay data (n=70,112). The number of DRGs differs widely across the 10 European countries (range: 2-14). Underlying this wide range is a different use of classification variables, especially secondary diagnoses and treatment options are considered to a different extent. In six countries, a standard set of patient characteristics and treatment variables explain the variation in costs or length of stay better than the DRG variables. This raises questions about the adequacy of the countries' DRG system or the lack of specific criteria, which could be used as classification variables.     

Predicting length of stay for Medicare patients at a teaching hospital.

This research examines how the patients' characteristics and clinical indicators affect length of stay for the top five Diagnosis-Related-Groups (DRGs) for Medicare patients at a teaching hospital in the United States. The top DRGs were selected on the basis of volume per year. Teaching hospitals in the United States devote a significant amount of their resources to research and teaching, while providing treatment for patients. The ability to predict length of stay can substantially improve a teaching hospital's capacity utilization, while ensuring that resources are available to meet the health care needs of the Medicare population. Multiple regression models are developed to predict the length of stay using the patients' characteristics and clinical indicators as independent variables. The results indicate that approximately 60 percent (R(2)) of the variance in the length of stay is explained by the patients' characteristics and clinical indicators for these DRGs. The Mortality and Severity indices are found to be the strongest predictors for length of stay in all DRGs. Other patients' characteristics and clinical indicators such as age, gender, race/ethnicity, marital status, admission type and admission source are also significant predictors for some DRGs. In addition, most of these variables affect the length of stay in the same manner as shown in previous studies, even though the previous studies do not have the DRG specificity of this study.     

Factors affecting hospital stay in psychiatric patients: the role of active comorbidity

ABSTRACT: BACKGROUND: Research on length of stay (LOS) of psychiatric inpatients is an under-investigated issue. In this naturalistic study we investigated factors which affect LOS in two groups of patients focusing particularly on the impact on LOS of medical comorbidity severe enough to require referral. METHODS: Active medical comorbidity was quantified using referral as the criterion. The study sample comprised of with 200 inpatients with the diagnosis of schizophrenia and 228 inpatients suffering from bipolar disorder (type I or II). Jonckheere and Mann-Whitney tests were used to estimate the influence of referrals on LOS, and regression analyses isolated variables associated with LOS separately for each group. RESULTS: Half of the patients needed one or more referrals for a non-psychiatric problem. The most common medical condition of patients with bipolar disorder was arterial hypertension. Inpatients with schizophrenia suffered mostly from an endocrine/metabolic disease - 12% of referrals were for Hashimoto's thyroiditis. A positive linear trend was found between LOS and number of referrals in both groups, the effect was greater for schizophrenia patients. The effect of referrals on LOS was verified by regression in both groups. Overall, referred patients showed greater improvement in GAF compared to controls. CONCLUSIONS: To our knowledge this is the first study to investigate physical comorbidity in psychiatric inpatients using the criteria of referral to medical subspecialties. Comorbidity severe enough to warrant referral is a significant determinant of hospital stay. This insight may prove useful in health care planning. The results show lack of effective community care in the case of schizophrenia and negative symptoms may be the cause of this. Our findings call for more attention to be paid to the general medical needs of inpatients with severe mental health and concurrent severe medical comorbidity.     

Predicting the revolving door phenomenon among patients with schizophrenic, affective disorders and non-organic psychoses

OBJECTIVE: The aim of the study was to identify the variables that predict the revolving door phenomenon in psychiatric hospital at the moment of a second admission. METHODS: The sample consisted of 3,093 patients who have been followed during 5 to 24 years after their first hospital admission due to schizophrenia, and affective or psychotic disorders. Those who had had four or more admissions during the study period were considered as revolving door patients. Logistic regression analyses were used to assess the impact of gender, age, marital status, urban conditions, diagnosis, mean period of stay on the first admission, interval between the first and second admissions on the patterns of hospitalization. RESULTS: The variables with the highest predictive power for readmission were the interval between first and second admissions, and the length of stay in the first admission. CONCLUSIONS: These data may help public health planners in providing optimal care to a small group of patients with more effective utilization of the available services.     

Predicting length of stay in an acute care hospital: the role of psychosocial problems.

In an acute care hospital, a major performance indicator is patient length of stay. This study, in a large university teaching acute care hospital in Canada, examined the effect of psychosocial problems on length of stay, controlling for patient demographics and medical condition. Average days stay for Diagnostic Related Groups (DRGs) was used as a proxy variable for severity of medical condition, and the Person-in-Environment (PIE) classification system was used to measure psychosocial problems. Data were collected on a sample of 160 patients; 78 in psychiatry and 82 in medical/surgical wards. In a regression analysis, the severity of the patient's psychosocial problem was a more significant predictor of length of stay than the DRG variable. The identification of psychosocial problems and their severity add an important and complementary dimension to research into the effectiveness of social workers in reducing length of stay. Workers found clients had significantly more problems related to their social role functioning than problems in the environment.     

Case-mix adjustment using objective measures of severity: the case for laboratory data.

OBJECTIVE. We evaluate the use of routinely gathered laboratory data to subclassify surgical and nonsurgical major diagnostic categories into groups homogeneous with respect to length of stay (LOS). DATA SOURCES AND STUDY SETTING. The source of data is the Combined Patient Experience database (COPE), created by merging data from computerized sources at the University of California San Francisco (UCSF) Medical Center and Stanford University Medical Center for a total sample size of 73,117 patient admissions. STUDY DESIGN. The study is cross-sectional and retrospective. All data were extracted from COPE consecutive admissions; the unit of analysis is an admission. The outcome variable LOS proxies hospital resource utilization for an inpatient stay. Nine (candidate) predictor variables were derived from seven lab tests (WBC, Na, K, C02, BUN, ALB, HCT) by recording the whole-stay minimum or maximum test result. DATA COLLECTION/EXTRACTION METHODS. Patient groups were formed by first assigning to major diagnostic categories (MDCs) all 73,117 admissions. Each MDC was then partitioned into medical and surgical subgroups (sub-MDCs). The 13 sub-MDCs selected for study define a study population of 32,599 patients that represents approximately 45 percent of inpatients. Within each of the 13 sub-MDCs, patients were randomly assigned to one of two data sets in a ratio of 2:1. The first set was used to create, the second to validate, three different LOS predictors. Predictive accuracies of individual DRG classes were compared with those of two alternative classification schemes, one formed by recursive partitioning (the sub-MDC) using only lab test results, the other by partitioning with both lab test results and individual DRGs. PRINCIPAL FINDINGS. For the eight largest sub-MDCs (81 percent of study population), individual DRGs explained 23 percent of the within sub-MDC variance in LOS, laboratory data classes explained 31 percent, and classes derived by considering individual DRGs and laboratory data explained 37 percent. (Each result is a weighted average R2. The average number of LOS classes into which the eight largest sub-MDCs were partitioned were 20, 10, and 10, respectively. Within six of the eight, partitioning on the basis of laboratory data alone explained more within sub-MDC variance than did partitioning into individual DRGs. CONCLUSIONS. Routine lab test data improve the accuracy of LOS prediction over that possible using DRG classes. We note that the improvements do not result from overfitting the data, since the numbers of LOS classes we use to predict LOS are considerably fewer than the numbers of individual DRGs.     

Factors Associated with Extended Length of Stay for Patients Presenting to an Urban Psychiatric Emergency Service: A Case-Control Study

This study assessed factors associated with extended length of stay (ELOS) for patients presenting to a psychiatric emergency service (PES). Two hundred six subjects with a length of stay of 24 h or longer were compared with time-matched controls (patients that presented directly after the ELOS patient). Binary logistic regression was used to identify risk factors for ELOS. ELOS was associated with suicidal ideation, disposition to an inpatient unit, homicidal ideation, lack of insurance, homelessness, male gender, past history of psychiatric hospitalization, diagnosis of substance abuse, significant psychiatric co-morbidity (represented by three or more Axis I diagnoses), and diagnosis of a psychotic disorder. Lack of insurance, suicidal ideation, disposition to inpatient unit, and homicidal ideation all made nonredundant contributions to predicting stays of 24 h or longer.     

Patient classification systems and hospital costs of care for knee replacement in 10 European countries

Knee replacement is a common surgical procedure performed to relieve pain and disability from degenerative osteoarthritis. This study evaluates the ability of ten European diagnosis-related group (DRG) systems to explain variations in costs or in length of stay for knee replacements. We assessed three different models in predicting variation of cost and length of stay. The first model, M(D), included only DRG groups as explanatory variables; the second, M(P), used a set of patient-level variables; and the third, M(F), included all variables from both M(D) and M(P). The total number of DRGs used to group knee replacement is low, ranging from two to six. All DRG systems except one differentiate between primary knee replacement and revision surgery. Considerable differences exist in the rate of revision surgery. There is also high variation in mean cost (from € 3809 to € 8158) and in mean length of stay (LoS) (from 4.2 to 13.6 days). The explanatory power of DRGs varies from 21.5 to 72.5% with values of around 40% in most countries of the study. Findings suggest that DRG systems could be enhanced either by the inclusion of patient-level variables, by the use of measures of clinical outcome or by improving cost and administrative information.     

Predicting the cost of hospital stay for stroke patients: the use of diagnosis related groups

In order to provide tailor-made care, governments are considering the implementation of output-pricing based on hospital case-mix measures, such as diagnosis related groups (DRG). The question is whether the current DRG classification system can provide a satisfactory prediction of the variance of costs in stroke patients and if not, in what way other variables may enhance this prediction. In this study, data from 731 stroke patients hospitalized at University Hospital Maastricht during 1996-1998 are used in the cost analysis. The DRG classification for this group uses information--in addition to the DRG classification operation or no operation--on the patient's age combined with discharge status. The results of regression analysis show that using DRGs, the variance explained in the costs amounts to 34%. Adding other variables to the DRGs, the variance explained increases to about 61%. Additional factors highly correlating with inpatient costs are the level of functioning after stroke, comorbidity, complications, and 'days of stay for non-medical reasons'. Costs decreased for stroke patients discharged during the latter part of the years studied, and if stroke patients happened to die during their hospital stay. The results do suggest that future implementation of output-pricing based on the DRG case-mix measures is feasible for stroke patients only if it is enhanced with information on complications and the level of functioning.     

Casemix-based funding of Northern Territory public hospitals: adjusting for severity and socio-economic variations

The Northern Territory intends to make use of Australian National Diagnosis Related Groups (DRGs) and their cost relativities as the basis for the allocation of budgets among public hospitals. The study reported here attempted to assess the extent to which there are variations in severity of illness and socio-economic status which are not adequately explained by DRG alone and, if so, to develop a DRG payment adjustment index by use of routinely available data items. The investigation was undertaken by use of a database containing all discharges between July 1992 and June 1995. Hospital length of stay was used as a proxy for cost. Multivariate analysis was undertaken and it was found that several variables were associated with cost variations within DRGs. Stepwise multiple linear regression was used to develop a model in which 14 variables were able to explain 45% of the variations. Index values were subsequently computed from the regression model for each of eight categories of admitted patient episodes which are the intersections of three binary variables: Aborigine or non-Aborigine, rural or urban usual place of residence of the patient and hospital type (teaching or other). It is intended that these index values will be used to compute differential funding rates for each hospital in the Territory. © 1998 John Wiley & Sons, Ltd.     

How well do diagnosis-related groups explain variations in costs or length of stay among patients and across hospitals? Methods for analysing routine patient data

We set out an analytical strategy to examine variations in resource use, whether cost or length of stay, of patients hospitalised with different conditions. The methods are designed to evaluate (i) how well diagnosis-related groups (DRGs) capture variation in resource use relative to other patient characteristics and (ii) what influence the hospital has on their resource use. In a first step, we examine the influence of variables that describe each individual patient, including the DRG to which the patients are assigned and a range of personal and treatment-related characteristics. In a second step, we explore the influence that hospitals have on the average cost or length of stay of their patients, purged of the influence of the variables accounted for in the first stage. We provide a rationale for the variables used in both stages of the analysis and detail how each is defined. The analytical strategy allows us (i) to identify those factors that explain variation in resource use across patients, (ii) to assess the explanatory power of DRGs relative to other patient and treatment characteristics and (iii) to assess relative hospital performance in managing resources and the characteristics of hospitals that explain this performance.     

Inpatient resource use: a comparison of family medicine and internal medicine physicians.

BACKGROUND. It is not known whether differences exist between the use of inpatient resources by family medicine and internal medicine physicians when patient demographic and complexity variables are statistically controlled. METHODS. The study population was all patients in 13 higher volume diagnosis-related groups (DRGs) discharged from the family medicine (n = 306) and internal medicine services (n = 2374) of the University of Cincinnati Hospital during 1985 and 1986. The dependent variables were length of stay and inpatient readmission within 2 weeks. Stratification by DRGs was used to control for the effects of age and case mix on these variables. RESULTS. With the exception of findings regarding one DRG, the results do not indicate that differences exist in average length of stay between patients of family medicine and internal medicine physicians after adjustment for other variables. Furthermore, almost all of the explained variance in length of stay was attributed to patient complexity and not to physician specialty or patient race or sex. For all discharges, the proportion of patients readmitted within 2 weeks was about 4% higher for the internal medicine service. However, multivariate analysis did not support the importance of physician specialty (family medicine or internal medicine) as a predictor of whether readmission occurred within 2 weeks. CONCLUSIONS. General indicators of resource use (such as length of stay or readmission occurrence) without adjustment for patient case mix, demographics, and complexity are inadequate for comparison of health care providers. Further research regarding interspecialty differences should use longitudinal data from large populations, which would permit more detailed examination of resource utilization.     

Psychiatric illness in patients presenting for obesity treatment

The prevalence of DSM-III-R Axis I psychiatric disorders was investigated in a sample of 54 obese patients at the time of presentation for weight reduction treatment. Patients were interviewed using the Structured Clinical Interview for DSM-III-R Axis I diagnosis (SCID-I) designed to probe for the major Axis I syndromes. High rates of affective disorders, particularly those with depressive symptomatology, were found. Twenty-six percent of patients were in the midst of a current Axis I affective disorder at the time of examination. Forty-eight percent had a history of affective disorder, and 57% had at least one lifetime Axis I diagnosis. Only one patient met criteria for an Axis I eating disorder. Patients with psychiatric diagnoses could not be discriminated from those without on the basis of Body Mass Index. However, the former had significantly more impairment as measured by the Beck Depression Inventory and the Global Assessment of Functioning. These findings may have implications for the assessment and treatment of obese patients in clinical and research settings.     

How well do diagnosis-related groups for appendectomy explain variations in resource use? An analysis of patient-level data from 10 European countries

Appendectomy is a common and relatively simple procedure to remove an inflamed appendix, but the rate of appendectomy varies widely across Europe. This paper investigates factors that explain differences in resource use for appendectomy. We analysed 106,929 appendectomy patients treated in 939 hospitals in 10 European countries. In stage 1, we tested the performance of three models in explaining variation in the (log of) cost of the inpatient stay (seven countries) or length of stay (three countries). The first model used only the diagnosis-related groups (DRGs) to which patients were coded, the second model used a core set of general patient-level and appendectomy-specific variables, and the third model combined both sets of variables. In stage two, we investigated hospital-level variation. In classifying appendectomy patients, most DRG systems take account of complex diagnoses and comorbidities but use different numbers of DRGs (range: 2 to 8). The capacity of DRGs and patient-level variables to explain patient-level cost variation ranges from 34% in Spain to over 60% in England and France. All DRG systems can make better use of administrative data such as the patient's age, diagnoses and procedures, and all countries have outlying hospitals that could improve their management of resources for appendectomy.