Uniform Hospital Discharge Data Set Research Articles

The role of neighbourhood socioeconomic status in large for gestational age.

To determine whether neighbourhood socioeconomic status (SES) was associated with large for gestational age (LGA) while considering key sociodemographic and clinical confounding factors. All singleton infants whose parents were living in the city of Marseilles, France, between 2013 and 2016. Population-based study based on new-born hospital birth admission charts from the French National Uniform Hospital Discharge Data Set Database. LGA infants were compared to appropriate-for-gestational-age (AGA) infants. Multiple generalized logistic model analysis was used to examine factors associated with LGA. A total of 43,309 singleton infants were included, and 4,747 (11%) were born LGA. LGA infants were more likely to have metabolic and respiratory diseases and to be admitted to the neonatal intensive care unit. Multiparity, advanced maternal age, obesity and diabetes were associated with an increased risk of LGA. Lower neighbourhood SES was associated with LGA (aOR = 1.24, 95% CI: 1.14; 1.36; p<0.0001) independent of age, diabetes, obesity, maternal smoking and multiparity. The strength of this association increased with maternal age, reaching an aOR of 1.50 (95% CI: 1.26; 1.78; p<0.0001) for women > 35 years old. Neighbourhood SES could be considered an important factor for clinicians to better identify mothers at risk of having LGA births in addition to well-known risk factors such as maternal diabetes, obesity and age. The intensification of the association between SES and LGA with increasing maternal age suggests that neighbourhood disadvantage may act on LGA cumulatively over time.

140. Trends of Infective Endocarditis at a Northern New England Academic Medical Center, From 2011 to 2017: A Case for Improved Methods to Reliably Identify Associated Substance Use

BackgroundInfective endocarditis (IE) is a morbid and often lethal complication of injection drug use. There is an urgent need for accurate surveillance for IE related to substance use (SU) to support control strategies.MethodsWe conducted a retrospective comparative analysis of 3 datasets evaluating patients aged ≥16 years admitted to an academic medical center in New England with an ICD-9/10 discharge diagnosis of IE from April 2011 to December 2017. The 3 datasets included the hospital’s electronic medical record (EMR); the hospital’s Outpatient Parenteral Antibiotic Therapy (OPAT) program dataset; and the New Hampshire Uniform Hospital Discharge Data Set (UHDDS). We analyzed the number of admissions for IE per year, stratified by SU. We developed a SU composite measure by incorporating multiple sources of data from the EMR, and then verified accuracy of both the SU and IE diagnoses through manual chart review.ResultsThe EMR documented 472 hospital admissions for IE, representing 385 unique patients. The median age was 56 years and 59% were men. Admissions increased 67%, from 56 in 2012 to 84 in 2017. SU was coded as a discharge diagnosis in 27% of these admissions; however, based on our composite measure of SU, 45% IE admissions were possibly associated with SU. The proportion of IE patients who had evidence of SU increased from 20% in 2011 to 49% in 2017 (P = 0.002). Patients with SU compared with those without were younger (40.5 vs. 65.2 years, P < 0.001) and more likely to be on Medicaid (59% vs. 8%, P < 0.001). They had higher average charges ($146,633 vs. $107,223, P = 0.002) and lengths of stay (19.1 vs. 13.4 days, P < 0.001). The UHDDS and EMR datasets identified a similar numbers of patients with a diagnosis of IE; however, manual chart review revealed that IE was over-coded in ~one-fifth of admissions.ConclusionThe rate of IE in our hospital increased dramatically between 2011 and 2017, with a rising proportion associated with SU. Despite these trends, we found that discharge diagnosis coding alone substantially underestimated associated SU and overestimated IE disease burden. Our findings suggest public health administrative datasets, such as the UHDDS, can contribute to surveillance of IE disease burden with consideration of these important limitations, especially for assessing disease trends.Disclosures All authors: No reported disclosures.

Epidemiology of Trauma in France: Mortality and Risk Factors Based on a National Administrative Health Database

Background: In France, the management of trauma exhibits several peculiarities, including a developed pre-hospital medicalisation and a large number of facilities. However, there is neither a national trauma system nor register. The billing system reflects the entire nation's hospital activity. The study aims at describing, using this database, the epidemiology of French trauma. Methods: A retrospective population-based cohort study has been conducted on trauma patients in France using the French National Uniform Hospital Discharge Data Set Database (PMSI-MCO) for 2016. Patients were selected using the Trauma Audit and Research Network (TARN) criteria, inspired by the UK trauma system. Sociodemographic and clinical information and hospital characteristics were collected. The main outcomes were 24-hour and 30-day mortality. Findings: Among 1 144 596 patients hospitalised in French hospitals for trauma in 2016, 144 058 patients were included in the study based on the TARN criteria. The mean age of the patients was 64 years. Females (50·8%) were over-represented among patients older than 75 years. The 24-hour and 30-day mortality was 1·4% and 5·9%, respectively. The severity of trauma differed according to the type of hospitals; university hospitals admitted severe trauma cases. In multivariate analysis, age, gender, injury body regions, co-morbidities, injury severity, transfusion, need for surgery, and ICU admission were independent factors of risk for both early and delayed mortality. Age and injury severity were the stronger predictors for mortality. Area-level deprivation was an adverse factor for 30-day mortality. Interpretation: The national burden of trauma care was assessed with administrative data in a country without a trauma system. The mortality associated with trauma in France was around 6%. Risk factors include age and severity, but other variables like the socioeconomic status influences the outcomes. Funding: None. Declaration of Interest: The authors have no conflicts of interest to disclose in relation with this manuscript.

Hospital Discharge Data for Guillain-Barré Syndrome and Influenza A (H1N1) Vaccine Adverse Events

To the Editor: As part of the public health response to the current pandemic (H1N1) 2009, surveillance for adverse events following vaccination for influenza A (H1N1) is a high priority (1). Surveillance for Guillain-Barre syndrome (GBS) has been of particular interest, because the syndrome was associated with the 1976–1977 swine influenza vaccine (1,2). To study this association, reliable ascertainment of recent incident cases of GBS is necessary. GBS is an acute, immune-mediated paralytic disorder of the peripheral nervous system (3–5) with an estimated annual incidence of 0.8–1.9/100,000 (6). Most cases are associated with an antecedent infection (6). Several surveillance systems are in place to monitor rates of postvaccination GBS (1–3), most of which include a component of electronic administrative record review for case detection. Analysis of computerized medical databases is a well-established method of monitoring for vaccine adverse events (7). Although the validity of such data varies, depending on the diagnosis and region, few studies have evaluated the use of hospital discharge data for GBS specifically (8,9). We reviewed the Tennessee Department of Health Uniform Hospital Discharge Dataset for all hospital discharge diagnoses in 4 major metropolitan regions of Tennessee in 2002–2003 with codes from the International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM), that might indicate acute GBS. Records with ICD-9-CM code 357.0 (acute infective polyneuritis) or other combinations suggestive of GBS within the top 10 diagnoses were requested. These data were compared with information on cases identified by directly requesting lists of patients with discharge diagnoses of GBS from hospital medical record departments. Charts of all reported cases were validated by chart review. Patients were classified as having acute GBS if they met Brighton Criteria Levels 1, 2, or 3 (10). A total of 344 records of possible cases of acute GBS were identified. Of these cases, 215 (63%) were identified through the state hospital discharge database, 315 (92%) were reported directly by hospitals, and 186 (54%) were identified by both systems. Among all suspected cases identified, only 103 (30%) met criteria for acute GBS (annual rate 2.1/100,000 population), 14 (4%) were out-of-state residents, 114 (33%) were nonacute cases that occurred before the study period and patients were readmitted for other reasons, 90 (26%) had no documentation of GBS in the medical record, 17 (5%) were duplicate reports, and 6 (2%) had insufficient information for further investigation. The predictive-value positive of a GBS diagnostic code from the statewide hospital discharge database representing acute GBS was only 30%. Of the 103 confirmed cases, 26 (25%) would have been missed if only the state hospital discharge database was used to identify potential cases. Of 103 cases, all were identified with ICD-9-CM diagnosis code 357.0; in 91 (88%) cases, this was the primary diagnosis. Other combinations of codes did not identify additional cases. Of cases of acute GBS identified, 32 (30%) met only clinical criteria (Brighton Level 1), 40 (39%) had either laboratory or electrophysiologic evidence (Brighton Level 2), and 32 (31%) had both (Brighton Level 3). Because the 2 surveillance systems we compared both relied on medical record discharge diagnoses, they were not independent, and we could not perform a capture/recapture analysis. Because GBS is a diagnosis for which the great majority of patients are hospitalized, and our overall incidence rate is within the range identified in other studies, it is likely that the combination of these methods is reasonably sensitive. The administrative hospital discharge database could not be relied on to confirm that all coded GBS cases were acute. Even if the 114 nonacute cases could easily have been identified and excluded from the initial list of 344 records, only 103 (45%) of the remaining 230 reports were identified as confirmed acute cases. Although the use of large hospital discharge databases may be useful as an adjunct for identification of GBS cases as part of public health surveillance, they lack sufficient sensitivity or specificity to be relied upon exclusively. The poor specificity of the system is particularly problematic for public health surveillance. A large investment of time and resources was necessary to perform manual chart reviews to confirm possible cases, two-thirds of which were ultimately found not to be cases at all. Statewide administrative hospital discharge diagnosis databases should not be solely relied on for GBS surveillance. Additional methods of reliable and efficient ascertainment and verification of cases are crucial to ensure valid data. Obtaining reliable methods is particularly important for urgent situations such as current surveillance for adverse events after pandemic (H1N1) 2009 virus vaccination, in which the detection of problems will have immediate public health effects.

Finally Present on Admission but Needs Attention

Momentum has finally gathered for a long-overdue addition to coding of hospital discharge diagnoses using the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM): indicating whether or not discharge diagnoses were when patients were admitted to hospitals. The present on (POA) indicator could greatly increase the utility of discharge diagnoses for various administrative, quality assurance, and research purposes. For many years, the primary rationale focused on using POA indicators for quality screening and monitoring, under the presumption that conditions that newly arise during hospital stays might possibly be more likely to represent complications and potential quality problems (yes, the wordy equivocation is intentional). POA indicators could particularly inform risk adjustment of hospital outcomes like mortality rates, specifically eliminating diagnoses representing hospitalacquired complications from risk-adjustment algorithms. In June 1992, the National Committee on Vital and Health Statistics (NCVHS) proposed multiple changes to the Uniform Hospital Discharge Data Set (UHDDS; the data required for all acute hospital discharges reimbursed by Medicare and Medicaid since 1974). Among several proposals for new UHDDS data elements, NCVHS suggested adding an alpha POA qualifier to each discharge diagnosis. NCVHS based this proposal on experiences at Mayo Clinic and in New York state, where POA indicators added modest additional to data collection.1 A dozen years later, NCVHS again recommended adding a discharge diagnosis modifier/flag for POA, this time specifically proposing modifications to UB-04 (the uniform bill submitted by hospitals) and the ANSI X12N Implementation Guide.2 The federal government never adopted NCVHS's proposals. Recent Medicare policy mandates have made adding POA indicators suddenly urgent. The Deficit Reduction Act of 2005 requires Medicare to stop paying hospitals for the costs of treating certain infections that occur in-hospital starting in October 2008. POA indicators offer the only feasible option to support this new mandate.3 Standard hospital bills already contain ICD-9-CM discharge diagnosis codes required to assign diagnosisrelated groups (DRGs), which determine Medicare hospital reimbursement levels. Effective March 1, 2007, UB-04 adds fields for POA indicators for discharge diagnoses, using standard coding guidelines.4 By definition, principal diagnoses are generally on admission inasmuch as they caused the hospitalization, but ICD-9-CM coding conventions require that a few conditions not POA be recorded as principal (eg, V codes for chemotherapy). Therefore, UB-04 has POA slots for both principal and secondary diagnoses. (The POA concept does not apply to certain ICD-9-CM codes, such as V codes for personal history of malignant neoplasm; coding guidelines exempt such codes from requiring POA indicators.3) These UB-04 changes apply to the paper form of the bill; modifications for electronic claims submission, used by the majority of providers, are still required. In this issue of Medical Care, Zhan et al5 demonstrate how POA indicators might alter Medicare payments for hospitalizations. Using hospital claims from California and New York, which have had POA indicators for over a decade, they assigned DRGs using all diagnoses and then again using only POA diagnoses. Relatively few cases (1 .4%) shifted to lower cost DRGs using only POA diagnoses for DRG assignment, and projected nationwide cost savings were

Outcomes Research: A Paradigm Shift for Nursing Research

Florence Nightingale was the first outcomes researcher in nursing. She used data to link mortality rates with patients’ diagnoses and treatments, and even developed the first uniform hospital discharge data set. Quite an amazing leader! An outcome can be described as an end result, or positive or negative change in a patient’s health status as a result of an intervention or treatment. So, the question is, why measure outcomes? Importantly enough, this topic has evolved with mixed ideas and discussions regarding what and where this would lead nursing and health care in the future.

Trends in surgery for gastroesophageal reflux disease: The effect of laparoscopic surgery on utilization

Background. In addition to substituting for open surgery, minimally invasive surgery may lower thresholds for intervention and thus increase overall utilization rates. The degree to which laparoscopy may have lowered the threshold for elective anti-reflux surgery is unknown. Methods. Using the Uniform Hospital Discharge Dataset and ICD-9 procedure and diagnosis codes, we identified all laparoscopic and open anti-reflux procedures performed on adults in Massachusetts, New Hampshire, and Vermont for each year from 1993 to 1998. We then examined secular trends and regional variation in the use of laparoscopic and open anti-reflux surgery. Results. The population-based rate of anti-reflux surgery more than doubled between 1993 (4.8 per 100,000) and 1998 (11.7 per 100,000). Laparoscopic anti-reflux procedures increased more than 6-fold between 1993 and 1998, from 1.2 to 8.9 procedures per 100,000 adults, with accompanying declines in overall length of stay and mortality. However, the number of open anti-reflux procedures decreased only modestly (22%) over this time period. In the year hospitals performed their first laparoscopic anti-reflux operation, procedure rates nearly tripled, on average, and then increased slowly in subsequent years. In 1997 and 1998, rates of anti-reflux surgery varied nearly 5-fold across hospital referral regions, ranging from 5.4 to 24.5 per 100,000. Conclusions. With the growth of minimally invasive surgery, rates of anti-reflux surgery have increased substantially, with wide regional variation in intervention rates. Further research is needed to determine the appropriate threshold for surgical treatment in patients with gastroesophageal reflux. (Surgery 2003;133:147-53.)

¿Existe relación entre el reingreso hospitalario y la calidad asistencial?

Caracterizar aquellos reingresos relacionados con problemas de calidad de la atención hospitalaria. Estudio descriptivo de las variables de los episodios de hospitalización (CMBD) de un hospital comarcal. Reingreso se define como episodio de hospitalización de un paciente que se produce en un período determinado tras uno previo que se denomina ingreso índice . El análisis se realiza según la relación clínica (coincidencia con la CIE-9-MC y GRD) y el tiempo que media entre los ingresos. El 20,9% de los reingresos están relacionados con problemas de calidad y la mitad se producen en el primer mes después del alta del ingreso índice; existen diferencias significativas con el grupo de los que reingresan por otros motivos: son 7,8 años más jóvenes y presentan menor número de diagnósticos y complicaciones. La duración de la estancia constituye un factor de riesgo. El número de pacientes que reingresa por problemas relacionados con la calidad es importante y constituye el primer grupo si se suman los de causa obstétrica por este mismo motivo. La mayoría de los diagnósticos se encuadran en el capítulo de complicaciones propias de ciertos procedimientos de la CIE-9-MC. El análisis de estos reingresos debe servir para diseñar estrategias de mejora de la calidad. To identify readmissions related to problems of quality of hospital care. Descriptive study of hospitalization episodes (uniform hospital discharge data set) in a district hospital. Readmission was defined as an episode of hospitalization of a patient occurring within a certain time period after an initial admission (index admission). The clinical relationship (agreement between International Classification of Diseases-ninth version-Clinical Modification [ICD-9-CM] and diagnosis-related group) and the time interval between hospitalization episodes were analyzed. A total of 20.9% of readmissions were found to be related to quality of care and approximately 50% were produced within 1 month of discharge after the index admission. Significant differences were found with patients readmitted for other reasons: these patients were 7.8 years younger and the number of diagnoses and complications was lower. Length of hospital stay was a risk factor. Problems of quality of care lead to a considerable number of readmissions and are the main cause of obstetric readmissions; most diagnoses are included under the heading of “iatrogenic” in the ICD-9-CM. Analysis of these readmissions may be useful to design strategies for improving quality of care.

Can we improve trauma mortality in a state with a voluntary trauma system

Background: Arizona has no organized statewide trauma system. We looked at the1997 and the 1998 Uniform Hospital Discharge Data Set (UHDDS) for the State of Arizona, and examined the trauma mortality data at both trauma hospitals and nontrauma hospitals. Methods: All qualifying mortalities based on hospital data from 1997 through 1998 were reviewed for the State of Arizona. Trauma deaths from 32 nontrauma hospitals were examined and compared with that of 7 level 1 trauma centers within the state. Results: In this time period, there were 375 qualified mortalities from nontrauma centers and 761 qualified mortalities from level 1 trauma facilities. Only 29 (8%) of nontrauma hospital deaths were found to be due to motor vehicle accidents. Only 8 (4%) mortalities at nontrauma centers were due to firearms. Conclusion: The data suggest that patients are arriving at the appropriate facility for definitive care despite the absence of a formal statewide trauma system.

An overview of the injury severity score and the new injury severity score

Objective—The research was undertaken to describe the injury severity score (ISS) and the new injury severity score (NISS) and to illustrate their statistical properties. Design—Descriptive analysis and assessment of the...

Estimation of completeness of AIDS case reporting in Massachusetts.

One of the most important aspects of any surveillance system is degree of completeness. We conducted a multiple source capture-recapture study using the 1994 Massachusetts Uniform Hospital Discharge Data Set (UHDDS) and Medicaid claims data to evaluate the completeness of the state's AIDS registry. We used encrypted social security numbers as the primary link to ensure confidentiality. For cases that did not link in the first round owing to missing social security numbers, we linked using gender and date of birth. Staff reviewed unmatched records from the Uniform Hospital Discharge Data Set and Medicaid datasets to determine subjects' AIDS case status. Using the Uniform Hospital Discharge Data Set, the AIDS registry was 92.6% complete (95% confidence interval (CI) = 91.6-93.5). The Medicaid claims dataset suggested the AIDS registry to be 94.5% complete (95% confidence interval = 93.7-95.3). The completeness of reporting to the state AIDS registry continues to be high, but there are differences by gender and mode of transmission of the virus. The continued assessment of completeness of reporting is important to ensure quality of the surveillance database over time.

Massachusetts weapon-related injury surveillance system

Context: Surveillance data on nonfatal weapon-related injuries—particularly those treated only in the emergency department (ED)—have been largely unavailable. Objective: To develop a surveillance system for fatal and nonfatal gunshot wounds and sharp instrument assaults. Design: The Massachusetts Department of Public Health (DPH) developed an ED-based reporting system for weapon-related injuries. Inpatient discharges and mortality data were linked to ED data, and police data sources were evaluated. Setting: Statewide. Participants: All 84 acute care hospital emergency departments. Uniform Hospital Discharge Data Set (UHDDS), mortality, and police data were obtained from existing systems. Main Outcome Measures: System sensitivity, predictive value positive (PVP), representativeness, flexibility, usefulness, acceptability, and sustainability. Results: The ED reporting system captured 82% of firearm-related injuries and 74% of reportable sharp instrument wounds from 1994 to 1996. Case reporting by demographic characteristics of the victim was representative and largely complete. Reporting by characteristics of the incident (e.g., type of gun) was less reliable. Police data were used to augment the system. From 1994 to 1996, ED-treated gunshot wounds dropped 41%, from 662 in 1994 to 393 in 1996. Unintentional and self-inflicted gunshot wounds showed no declines. For every firearm-related homicide, 4.7 nonfatal firearm assaults were treated; 85% of self-inflicted gunshot wounds were fatal. Conclusions: The system has proven timely (1996 ED data were available for release in March 1997), flexible (the reporting form has been revised several times), useful (DPH responds to 150 weapon injury data requests annually), acceptable (reporting is voluntary and no hospital declined participation), and sustainable (state funding is currently supporting the ED reporting system).

Adapting a Statewide Patient Database for Comparative Analysis and Quality Improvement

In 1991 the Michigan Health and Hospital Association established the Michigan Patient Outcome Measures Program to support quality improvement activities among member hospitals. Data submitted by each hospital are based on the Uniform Hospital Discharge Data Set (UHDDS). Each participating hospital is provided with tabulations for each of the 18 indicators, along with comparative data for each of five hospital peer groups. FACILITATING DEVELOPMENT OF A CLINICAL PATHWAY: Review of data on stroke patients indicated that patients discharged to home represented the single largest group of patients and that their length of stay was lower than for other groups of stroke patients. Hospital A used a clinical pathway to achieve reduction in length of stay of almost a day for this group. CASE STUDY 2: IMPROVING OBSTETRICS AND GYNECOLOGIC SERVICES THROUGH EDUCATION: The Department of Obstetrics and Gynecological Surgery at Hospital B achieved improvement in birth trauma, hysterectomy, cesarean section (C-section) births, and vaginal births after C-section. Program participants are considering adopting a birth trauma indicator. CASE STUDY 3: ESTABLISHING A HOSPITALWIDE QI PROGRAM: The Newborn Team at Hospital C initiated several QI initiatives aimed at reducing the newborn mortality rate, including guidelines for earlier identification of high-risk obstetric cases and for earlier intervention by the perinatalogist. Data that are not risk or severity adjusted have value in assessing hospital procedures and systems and can be used to educate and effect change in practice patterns.

Improving the E coding of hospitalizations for injury: do hospital records contain adequate documentation?

Incomplete external cause of injury (E) coding limits the usefulness of hospital discharge data sets for injury surveillance and research. Hospital medical records were examined to determine whether they contained adequate cause of injury documentation to allow for more complete E coding of injury discharges. Medical records for a sample of discharges involving a principal diagnosis of injury from the Uniform Hospital Discharge Data Set for Rhode Island were selected. We assigned E codes to these discharges and compared our E codes with those of the discharge data set. Documentation of cause of injury in the medical records was sufficient to allow assignment of a specific E code to 70% of the injuries for which no E codes or vague E codes were submitted on the Uniform Hospital Discharge Data Set. It was estimated that specific cause of injury documentation is available in the medical records of 80% of all injury discharges in Rhode Island; for approximately 90%, an E code describing at least the broad cause of injury could be assigned. Rates of E coding can be substantially increased by making better use of existing documentation in medical records.

Improving the Accuracy of Death Certificates

To the Editor. —Because of death certificate inaccuracies and other reasons, including the trend toward electronic document filing, the National Center for Health Statistics has called meetings of a steering committee to reengineer the death registration process. 1,2 The suggestion was made that the Uniform Hospital Discharge Data Set (UHDDS) might contain information that could improve or augment the death certificate because the UHDDS is often available quickly after death, it contains a list of diagnoses used for coding a patient's conditions, and it is a telegraphic distillation of the medical record. On the surface the idea sounds good, but an important problem needs to be addressed. The UHDDS captures the admitting diagnoses, principal diagnoses, and other diagnoses in the medical record. However, it is possible that major diagnoses that would serve as the underlying cause of death (for the death certificate) might escape capture in the UHDDS. For example,

Methodological Issues in Using Hospital Discharge Data to Determine the Incidence of Hospitalized Injuries

Estimates of the incidence of hospitalized injuries based on hospital discharge data are inconsistent because of variations in 1) the definition of injury and 2) the criteria for excluding repeat admissions for the same injury event. Using 1983 data from the Uniform Hospital Discharge Data Set for the state of Rhode Island, the authors demonstrate the effects on injury incidence rates associated with the various definitions and exclusion criteria used in previous studies. The overall injury rate (11.9/1,000 population) was substantially reduced (rate difference, 1.7; 95% confidence interval 1.2-2.1) when adverse effects or complications of medical and surgical care ("medical injuries"), not usually defined as injuries by injury researchers, were excluded. Estimates of the incidence of "true" injury hospitalizations (excluding medical injuries, late effects, and complications of care) ranged from 9.9/1,000, when repeat admissions identified as transfers from another acute facility were excluded, to 7.2/1,000, when repeat admissions identified as elective admissions were excluded. Marked variability in incidence rate estimates by age, sex, nature of injury, and state of residence of victim was also noted. The addition to hospital discharge data sets of a separate variable to identify readmissions for the same problem would result in more valid incidence estimates for injury research and surveillance.

Classifying Patients with Burns for Hospital Reimbursement: Diagnosis-Related Groups and Modifications for Severity

This study was designed to evaluate the relative severity and resource consumption of hospitalized patients with burns in a national cross section of hospitals, both with and without burn centers. We investigated to determine whether clinical variables or severity of illness measures not recorded in the Uniform Hospital Discharge Data Set are significant in explaining variation in length of stay, total cost, and mortality for patients with burns. The ability of the six burn diagnosis-related groups (DRGs) to explain variation in patients' length of stay was 20% and their ability to predict total costs was 24%. For the same patient population, the explanatory power of the DRGs improved to 54% for length of stay and 44% for costs when these variables were adjusted by the Severity of Illness Index. We also investigated whether hospitals with burn centers treated a more severely ill population of patients with burns than did hospitals without such centers. Significantly higher levels of severely ill patients with burns (p less than or equal to 0.0001) were found at burn center hospitals. Other patients or treatment variables, combined with a case-mix severity measure, were evaluated for their ability to further increase the explanatory power of DRGs. We also discuss here the use of the study results for reevaluating reimbursement policy.

Injury Surveillance Using Hospital Discharge Abstracts Coded by External Cause of Injury (E Code)

Current knowledge of the patterns of injury in the United States derive principally from mortality statistics that constitute less than 0.1% of all injuries reaching medical attention. There presently exists no national system for the surveillance of nonfatal injuries. To illustrate the usefulness and feasibility of conducting injury surveillance using E-coded hospital discharge data, we examined the surveillance data from the Massachusetts Statewide Childhood Injury Prevention Program. By using E-coded hospital discharge data, we increased the number of cases available for analysis by 40-fold over deaths, and we were able to describe the epidemiologic characteristics of the important causes of nonfatal childhood injuries. We therefore propose the development of a national injury surveillance system based on the Uniform Hospital Discharge Data Set coded by both the nature of the injury (N Code) and external cause (E Code).

Injury Surveillance Using Hospital Discharge Abstracts Coded by External Cause of Injury (E Code)

Current knowledge of the patterns of injury in the United States derive principally from mortality statistics that constitute less than 0.1% of all injuries reaching medical attention. There presently exists no national system for the surveillance of nonfatal injuries. To illustrate the usefulness and feasibility of conducting injury surveillance using E-coded hospital discharge data, we examined the surveillance data from the Massachusetts Statewide Childhood Injury Prevention Program. By using E-coded hospital discharge data, we increased the number of cases available for analysis by 40-fold over deaths, and we were able to describe the epidemiologic characteristics of the important causes of nonfatal childhood injuries. We therefore propose the development of a national injury surveillance system based on the Uniform Hospital Discharge Data Set coded by both the nature of the injury (N Code) and external cause (E Code).

Refining DRGs

Congress exempted certain types of specialty hospitals from the Medicare Prospective Payment System because it was recognized that Diagnosis-Related Groups (DRGs) may not properly define the case mixes in such institutions. This study is part of a larger investigation into case mix and payment in children's hospitals, one category of exempted institutions. A national sample of approximately 500,000 cases was developed with intentional oversampling of children's and university hospitals to allow detection of specialized types of cases. Five case-mix classification schemes--DRGs, Disease Staging, Patient Management Categories, Severity of Illness Score, and Pediatric Diagnosis System groups--were applied to the data set, and data items not included in the Uniform Hospital Discharge Data Set were collected. A set of Children's Diagnosis-Related Groups (CDRGs), based on modification of the current DRG system, resulted from the study. When CDRGs were applied to an independent sample of children's hospital data, length of stay variance was reduced by 47.6% compared with 32.3% for DRGs (length of stay outliers removed). These results suggest that incremental approaches to DRG refinement in other clinical areas where current definitions are inadequate may be better than rejecting the large amounts of statistical and clinical analyses existing in the DRG system. Similar methods can be used to correct problems brought on by changes in medical practice.