Administrative data. Baby or bathwater?

Human Albumin Use in Adults in U.S. Academic Medical Centers

OBJECTIVE

To determine rates and predictors of albumin administration, and estimated costs in hospitalized adults in the United States.

DESIGN

Cohort study of adult patients from the University HealthSystem Consortium database from 2009 to 2013.

SETTING

One hundred twenty academic medical centers and 299 affiliated hospitals.

PATIENTS

A total of 12,366,264 hospitalization records.

INTERVENTIONS

Analysis of rates and predictors of albumin administration, and estimated costs.

MEASUREMENTS AND MAIN RESULTS

Overall the proportion of admissions during which albumin was administered increased from 6.2% in 2009 to 7.5% in 2013; absolute difference 1.3% (95% CI, 1.30-1.40%; p < 0.0001). The increase was greater in surgical patients from 11.7% in 2009 to 15.1% in 2013; absolute difference 3.4% (95% CI, 3.26-3.46%; p < 0.0001). Albumin use varied geographically being lowest with no increase in hospitals in the North Eastern United States (4.9% in 2009 and 5.3% in 2013) and was more common in bigger (> 750 beds; 5.2% in 2009 and 7.3% in 2013) compared to smaller hospitals (< 250 beds; 4.4% in 2009 to 6.2% in 2013). Factors independently associated with albumin use were appropriate indication for albumin use (odds ratio, 65.220; 95% CI, 62.459-68.103); surgical admission (odds ratio, 7.942; 95% CI, 7.889-7.995); and high severity of illness (odds ratio, 8.933; 95% CI, 8.825-9.042). Total estimated albumin cost significantly increased from $325 million in 2009 to $468 million in 2013; (absolute increase of $233 million), p value less than 0.0001.

CONCLUSIONS

The proportion of hospitalized adults in the United States receiving albumin has increased, with marked, and currently unexplained, geographic variability and variability by hospital size.

Developing Comprehensive Statewide Disability Information Systems

Enhancing information capacity that can be used to improve services for people with disabilities remains an ongoing public health challenge. Research that can lead to improved state data systems for this population is still in its infancy. In this exploratory study, we have identified common characteristics needed to manage program services within one state's provider network and proposed a conceptual framework for developing a statewide system. We identified three broad information areas: assessment, administration, and assurance. Linking data elements to a program's information deficiencies allows lead state agencies responsible for public health functions for persons with disabilities to facilitate administrative coordination between programs having mutual information and management needs. By implementing strategies that share information without jeopardizing program autonomy, states can improve services for people with disabilities. Incorporating this framework extends the boundaries of a comprehensive disability information system beyond survey monitoring.

Higher occurrence of nausea and vomiting after total hip arthroplasty using general versus spinal anesthesia: an observational study

BACKGROUND

Under the assumption that postoperative nausea and vomiting (PONV) may occur after total hip arthroplasty (THA) regardless of the anesthetic technique used, it is not clear whether general (GA) or spinal (SA) anesthesia has higher causal effect on this occurrence. Conflicting results have been reported.

METHODS

In this observational study, we selected all elective THA interventions performed in adults between 1999 and 2008 in a Swiss orthopedic clinic under general or spinal anesthesia. To assess the effect of anesthesia type on the occurrence of PONV, we used the propensity score and matching methods, which allowed us to emulate the design and results of an RCT.

RESULTS

Among 3922 procedures, 1984 (51 %) patients underwent GA, of which 4.1 % experienced PONV, and 1938 underwent SA, of which 3.5 % experienced PONV. We found that the average treatment effect on the treated, i.e. the effect of anesthesia type for a sample of individuals that actually received spinal anesthesia compared to individuals who received GA, was ATET = 2.00 % [95 % CI, 0.78-3.19 %], which translated into an OR = 1.97 [95 % CI 1.35; 2.87].

CONCLUSION

This suggests that the type of anesthesia is not neutral regarding PONV, general anesthesia being more strongly associated with PONV than spinal anesthesia in orthopedic surgery.

Pediatric patients with asthma: a high-risk population for subsequent hospitalization

OBJECTIVES

Asthma is one of the most common chronic conditions among children and is one of the leading causes for pediatric hospitalizations. More evidence is needed to clarify the risks of repeat hospitalization and the underlying factors contributing to adverse health outcomes among pediatric patients hospitalized with asthma. The purpose of this study was to examine the risk of subsequent hospitalizations among pediatric patients hospitalized with asthma compared to a reference cohort of children hospitalized for all other diagnoses.

METHODS

The Washington State (WA) Comprehensive Hospital Abstract Reporting System (CHARS) was used to obtain data for the study. Data describing 81,946 hospitalized pediatric patients admitted from 2004 to 2008 were available. The risk of subsequent hospitalization among children admitted for asthma as compared to a reference cohort was examined.

RESULTS

The asthma cohort had a 33% (HR = 1.33 [99% confidence interval (CI) 1.21-1.46]; p < .001) increased risk of subsequent hospitalization from 2004 to 2008. Children in the asthma cohort under the age of 13 years demonstrated a significant increased risk of subsequent hospitalization as compared to the age-matched reference cohort of children without asthma. Those in the asthma cohort who were 3-5 years old demonstrated the highest risk (50%) of subsequent hospitalization (HR = 1.50 [99% CI 1.23-1.83]; p < .001).

CONCLUSIONS

Study results can be utilized in the development of appropriate interventions aimed at preventing and reducing hospital admissions, improving patient care, decreasing overall costs, and lessening complications among pediatric patients with asthma.

Evaluating comorbidities in total hip and knee arthroplasty: available instruments

Each year millions of patients are treated for joint pain with total joint arthroplasty, and the numbers are expected to rise. Comorbid disease is known to influence the outcome of total joint arthroplasty, and its documentation is therefore of utmost importance in clinical evaluation of the individual patient as well as in research. In this paper, we examine the various methods for obtaining and assessing comorbidity information for patients undergoing joint replacement. Multiple instruments are reliable and validated for this purpose, such as the Charlson Index, Index of Coexistent Disease, and the Functional Comorbidity Index. In orthopedic studies, the Charnley classification and the American Society of Anesthesiologists physical function score (ASA) are widely used. We recommend that a well-documented comorbidity index that incorporates some aspect of mental health is used along with other appropriate instruments to objectively assess the preoperative status of the patient.

The use of databases and registries to enhance colonoscopy quality

Administrative databases, registries, and clinical databases are designed for different purposes and therefore have different advantages and disadvantages in providing data for enhancing quality. Administrative databases provide the advantages of size, availability, and generalizability, but are subject to constraints inherent in the coding systems used and from data collection methods optimized for billing. Registries are designed for research and quality reporting but require significant investment from participants for secondary data collection and quality control. Electronic health records contain all of the data needed for quality research and measurement, but that data is too often locked in narrative text and unavailable for analysis. National mandates for electronic health record implementation and functionality will likely change this landscape in the near future.

The association of Medicare drug coverage with use of evidence-based medications in the Veterans Health Administration

BACKGROUND

Veterans with Medicare managed-care plans have access to pharmacy benefits outside the Veterans Health Administration (VA), but how this coverage affects use of medications for specific disease conditions within the VA is unclear.

OBJECTIVE

To examine patterns of pharmacotherapy among patients with diabetes mellitus, ischemic heart disease, and chronic heart failure enrolled in fee-for-service (FFS) or managed-care (HMO) plans and to test whether pharmacy benefit coverage within Medicare is associated with the receipt of evidence-based medications in the VA.

METHODS

A retrospective analysis of veterans dually enrolled in the VA and Medicare healthcare systems was conducted. We used VA and Medicare administrative data from 2002 in multivariable logistic regression analysis to determine the unique association of enrollment in Medicare FFS or managedcare plans on the use of medications, after adjusting for sociodemographic, geographic, and patient clinical factors.

RESULTS

A total of 369,697 enrollees met inclusion criteria for diabetes, ischemic heart disease, or chronic heart failure. Among patients with diabetes, adjusted odds ratios (ORs) of receiving angiotensin-converting enzyme (ACE) inhibitors and oral hypoglycemics in the FFS group were, respectively, 0.86 and 0.80 (p < 0.001). Among patients with ischemic heart disease, FFS patients were generally less likely to receive beta-blockers, antianginals, and statins. Among patients with chronic heart failure, adjusted ORs of receiving ACE inhibitors, angiotensin-receptor blockers, and statins in the FFS group were, respectively, 0.90, 0.78, and 0.79 (all p < 0.05). There were few systematic differences within HMO coverage levels.

CONCLUSIONS

FFS-enrolled veterans were generally less likely to be receiving condition-related medications from the VA, compared with HMO-enrolled veterans with lower levels of prescription drug coverage. Pharmacy prescription coverage within Medicare affects the use of evidence-based medications for specific disease conditions in the VA.

Opioid therapy for nonspecific low back pain and the outcome of chronic work loss

Outcomes of opioid therapy for noncancer pain remain to be more fully explored. Loss of work is among these outcomes. Especially when work loss becomes "chronic" (persists >or=90 days), it has profound psycho-social repercussions that compound suffering of those already in pain. Furthermore, costs escalate as work loss persists. We thus explored associations between opioid therapy for back pain and chronic work loss. Data consisted of workers compensation claims for nonspecific low back pain. We used multivariate analyses to control for diverse covariates. Workers with no opioid prescriptions constituted the reference group. Findings included the following: compared with the (no opioid) reference group, odds of chronic work loss were six times greater for claimants with schedule II ("strong") opioids; compared with the reference group, odds of chronic work loss were 11-14 times greater for claimants with opioid prescriptions of any type during a period of >or=90 days; and three years after injury, costs of claimants with schedule II opioids averaged $19,453 higher than costs of claimants in the reference group. Our analysis was not designed to ascertain antecedent causes, or why chronic work loss occurred in the first place. Rather, we focused on an ensuing consequence of opioid therapy, i.e., the outcome of chronic work loss, which occurred far removed in time (>or=90 days) after the worker's recorded date of back injury. The strong associations observed suggest that for most workers opioid therapy did not arrest the cycle of work loss and pain.

Demographic, clinical and occupational characteristics associated with early onset of delivery: findings from the Duke Health and Safety Surveillance System, 2001-2004

BACKGROUND

This cross-sectional study explores associations between preterm delivery and demographic, clinical and occupational characteristics of women employed within a university and health system.

METHODS

A comprehensive surveillance system linking individual-level data from Human Resources, medical insurance claims and a job-exposure matrix was used to identify women with a single live birth between 2001 and 2004 and describe maternal characteristics during pregnancy.

RESULTS

Preterm delivery occurred in 7.1% (n = 74) of the 1,040 women, a lower preterm delivery prevalence than observed in the general U.S. population. Nearly all (>99.5%) women utilized prenatal care services. Prevalence of preterm delivery was highest for inpatient nurses, nurses' aides and office staff. In multivariate analyses, preterm delivery was positively associated with several clinical conditions: placenta previa, diabetes and cardiovascular disorder/disease.

CONCLUSIONS

We observed associations between preterm delivery and several previously indicated clinical conditions. Further study of the effect of job characteristics on preterm delivery is warranted.

Functional outcome at 2.5, 5, 9, and 24 months after injury in the Netherlands

BACKGROUND

The collection of empirical data on the frequency, severity, and duration of functioning is a prerequisite to identify patient groups with long term or permanent disability.

METHODS

We fielded postal questionnaires in a stratified sample of 8,564 injury patients aged 15 years and older, who had visited an emergency department in the Netherlands. Measurements were at 2.5, 5, 9, and 24 months after the injury and included a generic health status classification (EQ-5D), socio-demographic, and medical information. We analyzed determinants of long-term functional outcome by multivariate regression analysis.

RESULTS

Five months after the injury health status of nonhospitalized injury patients was comparable to the general population's health (EQ-5D summary measure 0.87). Health status of patients admitted for 3 days or less improved until 9 months (0.82). For those admitted more than 3 days health status improved until 24 months (0.48 toward 0.67), but remained below population norms. Hospitalization, age and sex (females), type of injury (spinal cord injury, hip fracture, and lower extremity injury), and comorbidity were significant predictors of poor functioning in the long term.

CONCLUSIONS

Recovery patterns vary widely between nonhospitalized, shortly, and long hospitalized injury patients. Nonhospitalized injury patients recover within 5 months from an injury whereas a considerable group of hospitalized injury patients suffer from persistent health problems. Our study indicates the importance of health monitoring with an adapted longitudinal design for injury patients. The time intervals used should match the various stages of the recovery process, which depends on the severity of the injury studied.

GI Epidemiology: databases for epidemiological studies

BACKGROUND

Large databases are being increasingly used for examining the epidemiology and outcomes of digestive and liver disorders. The complexity and rigor of the methods used to conduct these studies are often underestimated.

AIMS

For the most commonly used databases, we provide a brief description of the contents, highlight strengths and weakness, and provide links for more detailed information. We also present a systematic approach to utilizing large databases for addressing research questions, highlighting commonly encountered study design issues, as well as strategies for resolving these issues.

CONCLUSIONS

1. Research using large databases requires the same essential skills needed to conduct research studies using other data sources. These include a rigorous study design, expertise in analytic methods, and relevant research questions. 2. The completeness and accuracy of information contained in the database must be assessed. Methods for improving the quality and completeness of this information should be considered. 3. Despite similarities among large databases, gaining insight and experience into the structure and content of each database is essential. Key points *Large databases can be a powerful source of information to examine the clinical epidemiology and outcomes of digestive and liver disorders. * Research using large databases requires the same essential skills needed to conduct research studies using other data sources. These include a rigorous study design, expertise in analytic methods, and relevant research questions. * The completeness and accuracy of information contained in the database must be assessed. Methods for improving the quality and completeness of this information should be considered. * Despite similarities among large databases, gaining insight and experience into the structure and content of each database is essential. * Examples of commonly used large databases are presented with a synopsis of information contained in the database, as well as strengths and limitations of using the database for research.

Adapting the Rx-Risk-V for Mortality Prediction in Outpatient Populations

OBJECTIVES

We sought to operationalize, test, and validate an outpatient pharmacy-based case-mix adjuster.

METHODS

Outpatients from the Department of Veterans Affairs (VA) prescribed a nonsteroidal anti-inflammatory drug (NSAID) or cyclooxygenase-2 selective drug during 2002 were identified. We updated and extended the Rx-Risk-V by adding 26 additional disease categories and mapping them to VA drug-class codes; derived empirical weights for each from a logistic model of 1-year mortality; adjusted for age, race and sex; and scored the weights into 1 measure of comorbidity. We compared the weighted score to the Deyo diagnosis-based comorbidity index and validated it in a national cohort of 260,321 outpatients with chronic heart failure (CHF).

RESULTS

One-year mortality among the 724,270-outpatient NSAID cohort was 1.6% (n = 11,766). Using a baseline model of age, race, and gender (c-index = 0.716), we found that the Deyo measure improved the prediction of mortality (c-index = 0.765), and the pharmacy comorbidity score further improved the prediction (c-index = 0.782), an increase of 25.8%. Using both, we found further improvement (c-index = 0.792). Among the CHF cohort, 9.7% (n = 25,251) died within 1 year. Performance of the baseline model controlling for age, race, and gender (c index = 0.620) improved with addition of the pharmacy comorbidity score (c index = 0.689), compared with the addition of the Deyo measure (c index = 0.651), an increase of 55.1%. Together, they slightly improved prediction in CHF patients (c index = 0.695).

CONCLUSIONS

The updated and extended Rx-Risk-V is useful for case-mix adjustment of mortality in an outpatient population.

Revision total knee arthroplasty complication rates in Northern Illinois

A paucity of population-based data exist which describe the rapid growth of revision total knee arthroplasties, changes in patient characteristics, or the association of hospital volume with complication rates. We analyzed whether inpatient complications for 2986 revision knee arthroplasties done on patients admitted to 63 hospitals in northern Illinois from 1993-1999 were correlated with volume of revision total knee arthroplasties. Coded complication rates for hospitals with less than seven, seven to 14, or greater than 14 annual procedures were compared using logistic regression to control for clinical and demographic characteristics of patients, hospital teaching status, and the proportion of the hospitals' patients discharged to rehabilitation facilities. Revision total knee arthroplasties increased 59%, and the overall complication rate declined from 9.3% during 1993-1996 to 7.3% during 1997-1999 (p = .04). When compared with the lowest volume hospitals, medium-volume hospitals had higher complication rates, whereas the highest volume hospitals were not significantly different. The absence of volume-outcome effects may be related to the relatively high volume of primary knee arthroplasties done at almost all area hospitals, surgeon group coverage across multiple hospitals, and the small annual number of revision total knee arthroplasties done during these years.

How safe is primary knee replacement surgery? Perioperative complication rates in Northern Illinois, 1993-1999

OBJECTIVE

To describe inpatient complications for primary total knee replacement (TKR) in a period of rapidly growing orthopedic surgery capacity, declining length of stay, and more frequent discharge to rehabilitation facilities.

METHODS

Complication incidence according to published coding algorithms was estimated for 35,531 primary TKR admissions of northern Illinois residents to 65 Illinois hospitals. Complication odds were estimated as a function of patients' clinical and sociodemographic status, hospital volume, residency training, TKR length of stay, International Classification of Diseases, Ninth Revision (ICD-9) coding intensity, and discharges to skilled nursing or rehabilitation facilities.

RESULTS

Primary TKR admissions increased 36% between 1993 and 1999, length of stay declined 43%, average ICD-9 code use increased 31%, and rehabilitation discharges increased 68%. Major complication rates declined 44% (12.4% to 6.9%; P < 0.0001) over this period, reflecting a 50% reduction in the adjusted odds of complication between 1993 and 1999. There was no association of procedure volume and outcome.

CONCLUSION

It is likely that the reduction in complications reflects true safety improvements as well as reduced length of stay.

Evaluating a Medicaid home and community-based physical disability waiver

State Medicaid waivers have fostered innovative health delivery systems for persons with special needs. Yet their overall cost effectiveness remains poorly understood. Changes were recently analyzed in Medicaid health services and costs for persons newly enrolled in the Kansas Physical Disability (PD) Waiver and found that ambulatory services increased, consistent with meeting enrollees' unmet medical needs and their access to enriched services. Home health, transportation, and personal care services also rose. Though not significant, hospital inpatient, outpatient, and long-term care services declined. This movement towards community-based service use in the short term reflects improved self-directed care and possible long-term cost savings.

[Obstetrical inpatient care and hospitalization risks in hospitals of Brazil]

OBJECTIVE

To analyze variations in early neonatal mortality, stillbirth rates, and a set of indicators collected from obstetric hospitals affiliated to the Brazilian National Unified Health System (SUS) for their monitoring through the Hospital Data System (SIH/SUS) and Live Births Data System (SINASC).

METHODS

One-hundred and thirty five hospitals in the state of Rio de Janeiro were assessed in 1997. Factor analysis was conducted using principal components. Score distribution for the first two components were established, which allowed to classify hospitals according to maternal risk profile and care outcomes.

RESULTS

Hospitals affiliated to SUS were responsible for 77.8% of all deliveries in the state of Rio de Janeiro and 23% of them performed fewer than 100 deliveries a year. Among hospitals of extreme high maternal risk and low performance, there were several units considered as referral centers for high-risk pregnancy. It was also observed that 5% of hospital units with low complexity infrastructures showed a profile of high maternal risk and questionable care outcomes.

CONCLUSIONS

The Hospital Information Data System affiliated to the National Unified Health System has proven to be an important information source for monitoring hospital stillbirth and early neonatal mortality rates as well as for planning surveillance actions for health services providing obstetric and/or neonatal care.

Effect of definition of mortality on hospital profiles

BACKGROUND

Hospitals are ranked based on risk-adjusted measures of postoperative mortality, but definitions differ about which deaths following surgery should be included.

OBJECTIVE

To determine whether varying the case definition of deaths following surgery that are included in coronary artery bypass surgery quality assessment affects the identification of outlier hospitals.

RESEARCH DESIGN

The study used a prospective cohort design.

SUBJECTS

A total of 15,288 patients undergoing coronary artery bypass surgery without other cardiac procedures from October 1993 to March 1996 at all (N = 43) Veterans Affairs hospitals that conduct cardiac surgery.

MEASURES

The first measure included any death occurring within 30 days after surgery, regardless of cause, in or out of the hospital (30-day mortality). The second measure included 30-day mortality plus any death occurring 30 days to 6 months after surgery that was judged to be a direct result of a perioperative complication of the surgery (all procedure-related mortality).

RESULTS

Hospital performance as assessed by the two different definitions of death varied substantially. The rankings of hospitals differed for 86% (37/43) of hospitals. Twenty-one percent (9/43) changed their quartile of rank, and five hospitals changed their outlier status. The correlation of observed-to-expected ratios was high (r = 0.96), but there was disagreement of outlier status (kappa = 0.71).

CONCLUSIONS

Judgments regarding the quality of a hospital's performance of coronary artery bypass surgery vary depending on the definition of postoperative mortality that is used. Further research is needed to assess what definition is most appropriate to identify quality of care problems.

Lack of integration, and seasonal variations in demand explained performance problems and waiting times for patients at emergency departments: a 3 years evaluation of the shift of responsibility between primary and secondary care by closure of two acute hospitals

In 1996, Stockholm County decided to reduce the costs of health care in order to release resources for upcoming medical needs. The method was both structural changes and transfer of low technology care from hospitals to other settings. The effects of interventions on service quality for patients and organisational performance of departments of internal medicine, orthopaedics and surgery were evaluated. Three cross-sectional studies were performed for comparison over time. Details on all individuals who visited A&E departments during 1 week in May 1997, May 1998 and May 1999 were recorded prospectively, and 16246 visits were registered. From 1995 to 1999 the total number of visits increased by 21% according to annual statistics. The utilisation of emergency care rose by 40/1000, and was not associated with the growth of population by 4.5%. Hospitals responded to peaks of demand by stringent prioritisation. Median waiting times were unchanged, but mean waiting times were prolonged over time, in particular for younger, not-referred patients. There was a direct correlation between waiting times and number of visitors. Total length of stay at A&E departments was related mainly to the waiting time for the first examination by a physician and cycle time for X-rays. Increased number of visits strained the capacity of hospitals and led to temporary loss of service quality for patients. The expected chain-reaction of integrated care did not take place, since providers outside hospitals were resistant to the shift of responsibility. Hospitals were utilised as primary care centres. The sub-optimal integration and fragmented care with an inappropriate balance between providers seems a universal problem.

Continuous training as a key to increase the accuracy of administrative data

The aim of this study was to evaluate the impact of a program of training, education and awareness on the accuracy of the data collected from hospital discharge abstracts. Four random samples of hospital discharge abstracts relating to four different periods were studied. The evaluation of the impact of systematic training and education activities was performed by checking the quality of abstracting information from the medical records. The analysis was carried out at the Istituto Dermopatico dell'Immacolata, a research hospital (335 beds) in Rome, Italy, which specializes in dermatology, plastic and vascular surgery. Error rates in discharge abstracts were subdivided into six categories: selection of the wrong principal diagnosis (type A); low specificity of the principal diagnosis (type B); incomplete reporting of secondary diagnoses (type C); selection of the wrong principal procedure (type D); low specificity of the principal procedure (type E); incomplete reporting of procedures (type F). A specific rate for errors modifying classification in diagnosis related groups (DRG) was then estimated and the effect of re-abstracting on the case-mix index evaluated. Error types A, B, C, E and F dropped from 8.5% to 2%, 15.8 to 4.9, 31.8 to 13.1,4.1 to 0.3 and 22 to 2.6%, respectively. Error type D was 0.7 both in the first (the baseline) and fourth periods of analysis. All differences in error types were statistically significant. In 1999 8.3% of cases were assigned to a different DRG after re-abstracting as compared with 24.3% in the third quarter of 1994, 23.8% in the first quarter of 1995 and 5.5% in September-October 1997. Continuous training and feedback of information to departments have shown to be successful in improving the quality of abstracting information at patient level from the medical record. These positive results were facilitated by the introduction of a prospective payment system to finance inpatient hospital activity. The effort to increase administrative data quality at hospital level facilitates the use of those data sets for internal quality management activities.

Strategies for improving comorbidity measures based on Medicare and Medicaid claims data

Claims-based measures of comorbid illness severity have generally relied on the diagnoses listed for a single hospitalization. Unfortunately, such diagnostic information is often limited because patients have not been hospitalized during periods of interest, because of incomplete coding of diagnoses on claims forms, or because listed diagnoses represent complications of the hospitalization rather than pre-existing comorbid conditions. To address these limitations, we developed and tested four comorbidity index scores for patients with breast cancer, each based on different sources of health services claims from Medicare and Medicaid: hospitalization for breast cancer surgery; outpatient care prior to the hospitalization; other inpatient care prior to the hospitalization; and all sources combined. Varying the number and type of sources of diagnostic information yielded only very small improvements in the prediction of mortality at 1 and 3 years. Surprisingly, even simpler measures of comorbidity (crude number of diagnoses) and of prior health care utilization (total days spent in the hospital) performed at least as well in predicting mortality as did the more complex index scores which assigned points and weights for specific conditions. The greatest improvement in explanatory power was observed when another source of clinical information (cancer stage derived from a population-based cancer registry) was used to supplement claims information. Expanding the source of claims diagnoses and focusing on time periods prior to an index hospitalization are insufficient for substantially improving the explanatory power of claims-based comorbidity indices. Other improvements suggested by our results should include: increasing the completeness and accuracy of claims diagnoses; supplementing diagnoses with health care utilization information in claims data; and supplementing claims data with other sources of clinical information.

From research to policy: what have we learned?

The Manitoba Centre for Health Policy and Evaluation has now had eight years of experience as an academic research unit interfacing with policymakers. Most of our research has focused on the determinants of health and on the delivery of health care from a population perspective. Each project that we have undertaken has made its own contribution and reinforced or built on the contribution of others. By communicating closely with policymakers at all levels, while maintaining an arm's-length relationship and the right of publication, MCHPE acts as a knowledgeable non-stakeholder with a commitment to inform the broader public.