Accuracy of ICD-9-CM codes in hospital morbidity data, Victoria: implications for public health research

Identification of Patients With Metastatic Prostate Cancer With Natural Language Processing and Machine Learning

PURPOSE

Understanding treatment patterns and effectiveness for patients with metastatic prostate cancer (mPCa) is dependent on accurate assessment of metastatic status. The objective was to develop a natural language processing (NLP) model for identifying patients with mPCa and evaluate the model's performance against chart-reviewed data and an International Classification of Diseases (ICD) 9/10 code-based method.

METHODS

In total, 139,057 radiology reports on 6,211 unique patients from the Department of Veterans Affairs were used. The gold standard was metastases by detailed chart review of radiology reports. NLP performance was assessed by sensitivity, specificity, positive predictive value, negative predictive value, and date of metastases detection. Receiver operating characteristic curves was used to assess model performance.

RESULTS

When compared with chart review, the NLP model had high sensitivity and specificity (85% and 96%, respectively). The NLP model was able to predict patient-level metastasis status with a sensitivity of 91% and specificity of 81%, whereas sensitivity and specificity using ICD9/10 billing codes were 73% and 86%, respectively. For the NLP model, date of metastases detection was exactly concordant and within < 1 week in 55% and 58% of patients, compared with 8% and 17%, respectively, using the ICD9/10 billing codes method. The area under the curve for the NLP model was 0.911. A limitation is the NLP model was developed on the basis of a subset of patients with mPCa and may not be generalizable to all patients with mPCa.

CONCLUSION

This population-level NLP model for identifying patients with mPCa was more accurate than using ICD9/10 billing codes when compared with chart-reviewed data. Upon further validation, this model may allow for efficient population-level identification of patients with mPCa.

Savings of loss-of-life expectancy and lifetime medical costs from prevention of spinal cord injuries: analysis of nationwide data followed for 17 years

BACKGROUND

The objectives of this research were to determine the savings of loss-of-life expectancy (LE) and lifetime medical costs (LMC) from prevention of spinal cord injuries (SCI) in Taiwan.

METHODS

From the claims database of Taiwan National Health Insurance, we identified 6164 adult patients with newly diagnosed SCI with permanent functional disability from 2000 to 2015 and followed them until the end of 2016. We estimated survival function through the Kaplan-Meier method and extrapolated it to lifetime.

RESULTS

For the SCI cohort, the LE and loss-of-LE were 17.6 and 13.3 years, respectively, while those for SCI with coding of external causes (E-code) were 18.1 and 13.0 years, respectively. For the SCI cohort with E-code, the loss-of-LE of motor vehicle (MV)-related SCI was significantly higher than that of fall-related SCI. In young and middle-aged patients with SCI with E-code, the loss-of-LE of MV-related paraplegia was significantly higher than that of MV-related quadriplegia and fall-related SCI. With a 3% discount rate, the LMC for patients with SCI after diagnosis were US$82 772, while those for patients with SCI with E-code were US$81 473. The LMC and the cost per year for those living with quadriplegia were significantly higher than those for paraplegia in all age groups, possibly related to the higher frequencies of stroke, chronic lung disease and dementia.

CONCLUSIONS

We conclude that quadriplegia has a higher impact on medical costs than paraplegia, and MV-related SCI has a higher impact on loss-of-LE than fall-related SCI. We recommend comprehensive SCI prevention be established, including infrastructures of construction and transportation.

Towards a national perioperative clinical quality registry: The diagnostic accuracy of administrative data in identifying major postoperative complications

Accurately measuring the incidence of major postoperative complications is essential for funding and reimbursement of healthcare providers, for internal and external benchmarking of hospital performance and for valid and reliable public reporting of outcomes. Actual or surrogate outcomes data are typically obtained by one of three methods: clinical quality registries, clinical audit, or administrative data. In 2017 a perioperative registry was developed at the Alfred Hospital and mapped to administrative and clinical data. This study investigated the statistical agreement between administrative data (International Statistical Classification of Diseases and Related Health Problems (10th edition) Australian Modification codes) and clinical audit by anaesthetists in identifying major postoperative complications. The study population included 482 high-risk surgical patients referred to the Alfred Hospital anaesthesia postoperative service over two years. Clinical audit was conducted to determine the presence of major complications and these data were compared to administrative data. The main outcome was statistical agreement between the two methods, as defined by Cohen’s kappa statistic. Substantial agreement was observed for five major complications, moderate agreement for three, fair agreement for six and poor agreement for two. Sensitivity and positive predictive value ranged from 0 to 100%. Specificity was above 90% for all complications. There was important variation in inter-rater agreement. For four of the five complications with substantial agreement between administrative data and clinical audit, sensitivity was only moderate (61.5%–75%). Using International Statistical Classification of Diseases and Related Health Problems (10th edition) Australian Modification codes to identify postoperative complications at our hospital has high specificity but is likely to underestimate the incidence compared to clinical audit. Further, retrospective clinical audit itself is not a highly reliable method of identifying complications. We believe a perioperative clinical quality registry is necessary to validly and reliably measure major postoperative complications in Australia for benchmarking of hospital performance and before public reporting of outcomes should be considered.

Text preprocessing for improving hypoglycemia detection from clinical notes - A case study of patients with diabetes

BACKGROUND AND OBJECTIVE

Hypoglycemia is a common safety event when attempting to optimize glycemic control in diabetes (DM). While electronic medical records provide a natural ground for detecting and analyzing hypoglycemia, ICD codes used in the databases may be invalid, insensitive or non-specific in detecting new hypoglycemic events. We developed text preprocessing methods to improve automatic detection of hypoglycemia from analysis of clinical encounter text notes.

METHODS

We set out to improve hypoglycemia detection from clinical notes by introducing three preprocessing methods: stop word filtering, medication signaling, and ICD narrative enrichment. To test the proposed methods, we selected clinical notes from VA Maryland Healthcare System, based on various combinations of three criteria that are suggestive of hypoglycemia, including ICD-9 code of diabetes and hypoglycemia, laboratory glucose values < 70 md/dL, and text reference to a proximate hypoglycemia event. In addition, we constructed one dataset of 395 clinical notes from year 2009 and another of 460 notes from year 2014 to test the generality of the proposed methods. For each of the datasets, two physician judges manually reviewed individual clinical notes to determine whether hypoglycemia was present or absent. A third physician judge served as a final adjudicator for disagreements.

RESULTS

Each of the proposed preprocessing methods contributed to the performance of hypoglycemia detection by significantly increasing the F1 score in the range of 5.3∼7.4% on one dataset (p < .01). Among the methods, stop word filtering contributed most to the performance improvement (7.4%). Combining all the preprocessing methods led to greater performance gain (p < .001) compared with using each method individually. Similar patterns were observed for the other dataset with the F1 score being increased in the range of 7.7%∼9.4% by individual methods (p < .001). Nevertheless, combining the three methods did not yield additional performance gain.

CONCLUSION

The proposed text preprocessing methods improved the performance of hypoglycemia detection from clinical text notes. Stop word filtering achieved the most performance improvement. ICD narrative enrichment boosted the recall of detection. Combining the three preprocessing methods led to additional performance gains.

Utilization and Validity of the Dental Diagnostic System over Time in Academic and Private Practice

INTRODUCTION

To fill the void created by insufficient dental terminologies, a multi-institutional workgroup was formed among members of the Consortium for Oral Health Research and Informatics to develop the Dental Diagnostic System (DDS) in 2009. The adoption of dental diagnosis terminologies by providers must be accompanied by rigorous usability and validity assessments to ensure their effectiveness in practice.

OBJECTIVES

The primary objective of this study was to describe the utilization and correct use of the DDS over a 4-y period.

METHODS

Electronic health record data were amassed from 2013 to 2016 where diagnostic terms and Current Dental Terminology procedure code pairs were adjudicated by calibrated dentists. With the resultant data, we report on the 4-y utilization and validity of the DDS at 5 dental institutions. Utilization refers to the proportion of instances that diagnoses are documented in a structured format, and validity is defined as the frequency of valid pairs divided by the number of all treatment codes entered.

RESULTS

Nearly 10 million procedures ( n = 9,946,975) were documented at the 5 participating institutions between 2013 and 2016. There was a 1.5-fold increase in the number of unique diagnoses documented during the 4-y period. The utilization and validity proportions of the DDS had statistically significant increases from 2013 to 2016 ( P < 0.0001). Academic dental sites were more likely to document diagnoses associated with orthodontic and restorative procedures, while the private dental site was equally likely to document diagnoses associated with all procedures. Overall, the private dental site had significantly higher utilization and validity proportions than the academic dental sites.

CONCLUSION

The results demonstrate an improvement in utilization and validity of the DDS terminology over time. These findings also yield insight into the factors that influence the usability, adoption, and validity of dental terminologies, raising the need for more focused training of dental students.

KNOWLEDGE TRANSFER STATEMENT

Ensuring that providers use standardized methods for documentation of diagnoses represents a challenge within dentistry. The results of this study can be used by clinicians when evaluating the utility of diagnostic terminologies embedded within the electronic health record.

Investigation of Leukaemia and Lymphoma AR-DRGs at a Sydney Teaching Hospital

Using non-blinded methodology, this study checked the coding of acute leukaemia, non-acute leukaemia and lymphoma episodes assigned to the AR-DRGs R60 A, B, C and R61 A, B during the fiscal year 2000–2001 at a Sydney teaching hospital. The purpose was to investigate whether the assignment of fewer episodes of these diseases to the highest complexity AR-DRGs during that year compared to 1999–2000 was due to miscoding, or due to a true decrease in episodes. A check of all 242 episodes revealed a degree of miscoding (mainly under-coding) of complications and comorbidities that had caused a 15% DRG error rate; nevertheless, there was a true decrease in the highest complexity episodes. The error in DRG assignment may have caused some financial disadvantage to the hospital.

ICD-10 Mortality Coding and the NEIS: A Comparative Study

The collection and utilisation of mortality data are often hindered by limited access to contextual details of the circumstances surrounding fatal incidents. The National Coroners Information System (NCIS) can provide researchers with access to such information. The NCIS search capabilities have been enhanced by the inclusion of data supplied by the Australian Bureau of Statistics (ABS), specifically the ICD-10 Cause of Death codeset. A comparative study was conducted to identify consistencies and differences between ABS ICD-10 codes and those that could be generated by utilising the full NCIS record. Discrepancies between the two sets of codes were detected in over 50% of cases, which highlighted the importance of access to complete and timely documentation in the assignment of accurate and detailed cause of death codes.

The accuracy of International Classification of Diseases coding for dental problems not associated with trauma in a hospital emergency department

OBJECTIVES

Emergency department (ED) visits for nontraumatic dental conditions (NTDCs) may be a sign of unmet need for dental care. The objective of this study was to determine the accuracy of the International Classification of Diseases codes (ICD-10-CA) for ED visits for NTDC.

METHODS

ED visits in 2008-2099 at one hospital in Toronto were identified if the discharge diagnosis in the administrative database system was an ICD-10-CA code for a NTDC (K00-K14). A random sample of 100 visits was selected, and the medical records for these visits were reviewed by a dentist. The description of the clinical signs and symptoms were evaluated, and a diagnosis was assigned. This diagnosis was compared with the diagnosis assigned by the physician and the code assigned to the visit.

RESULTS

The 100 ED visits reviewed were associated with 16 different ICD-10-CA codes for NTDC. Only 2 percent of these visits were clearly caused by trauma. The code K0887 (toothache) was the most frequent diagnostic code (31 percent). We found 43.3 percent disagreement on the discharge diagnosis reported by the physician, and 58.0 percent disagreement on the code in the administrative database assigned by the abstractor, compared with what it was suggested by the dentist reviewing the chart.

CONCLUSION

There are substantial discrepancies between the ICD-10-CA diagnosis assigned in administrative databases and the diagnosis assigned by a dentist reviewing the chart retrospectively. However, ICD-10-CA codes can be used to accurately identify ED visits for NTDC.

Compliance with Surgical Care Improvement Project for Body Temperature Management (SCIP Inf-10) Is Associated with Improved Clinical Outcomes

Background:

In an effort to measure and improve the quality of perioperative care, the Surgical Care Improvement Project (SCIP) was introduced in 2003. The SCIP guidelines are evidence-based process measures designed to reduce preventable morbidity, but it remains to be determined whether SCIP-measure compliance is associated with improved outcomes.

Methods:

The authors retrospectively analyzed the electronic medical record data from 45,304 inpatients at a single institution to assess whether compliance with SCIP Inf-10 (body temperature management) was associated with a reduced incidence of morbidity and mortality. The primary outcomes were hospital-acquired infection and ischemic cardiovascular events. Secondary outcomes were mortality and hospital length of stay.

Results:

Body temperature on admission to the postoperative care unit was higher in the SCIP-compliant group (36.6° ± 0.5°C; n = 44,064) compared with the SCIP-noncompliant group (35.5° ± 0.5°C; n = 1,240) (P &lt; 0.0001). SCIP compliance was associated with improved outcomes in both nonadjusted and risk-adjusted analyses. SCIP compliance was associated with a reduced incidence of hospital-acquired infection (3,312 [7.5%] vs.160 [12.9%] events; risk-adjusted odds ratio [OR], 0.68; 95% CI, 0.54 to 0.85), ischemic cardiovascular events (602 [1.4%] vs. 38 [3.1%] events; risk-adjusted OR, 0.60; 95% CI, 0.41 to 0.92), and mortality (617 [1.4%] vs. 60 [4.8%] events; risk-adjusted OR, 0.41; 95% CI, 0.29 to 0.58). Median (interquartile range) hospital length of stay was also decreased: 4 (2 to 8) versus 5 (2 to 14) days; P &lt; 0.0001.

Conclusion:

Compliance with SCIP Inf-10 body temperature management guidelines during surgery is associated with improved clinical outcomes and can be used as a quality measure.

Validation of Diagnostic Groups Based on Health Care Utilization Data Should Adjust for Sampling Strategy

Supplemental Digital Content is available in the text.

Objective:

Valid measurement of outcomes such as disease prevalence using health care utilization data is fundamental to the implementation of a “learning health system.” Definitions of such outcomes can be complex, based on multiple diagnostic codes. The literature on validating such data demonstrates a lack of awareness of the need for a stratified sampling design and corresponding statistical methods. We propose a method for validating the measurement of diagnostic groups that have: (1) different prevalences of diagnostic codes within the group; and (2) low prevalence.

Methods:

We describe an estimation method whereby: (1) low-prevalence diagnostic codes are oversampled, and the positive predictive value (PPV) of the diagnostic group is estimated as a weighted average of the PPV of each diagnostic code; and (2) claims that fall within a low-prevalence diagnostic group are oversampled relative to claims that are not, and bias-adjusted estimators of sensitivity and specificity are generated.

Application:

We illustrate our proposed method using an example from population health surveillance in which diagnostic groups are applied to physician claims to identify cases of acute respiratory illness.

Conclusions:

Failure to account for the prevalence of each diagnostic code within a diagnostic group leads to the underestimation of the PPV, because low-prevalence diagnostic codes are more likely to be false positives. Failure to adjust for oversampling of claims that fall within the low-prevalence diagnostic group relative to those that do not leads to the overestimation of sensitivity and underestimation of specificity.

Impact of Documentation Errors on Accuracy of Cause of Death Coding in an Educational Hospital in Southern Iran

Accurate cause of death coding leads to organised and usable death information but there are some factors that influence documentation on death certificates and therefore affect the coding. We reviewed the role of documentation errors on the accuracy of death coding at Shahid Mohammadi Hospital (SMH), Bandar Abbas, Iran. We studied the death certificates of all deceased patients in SMH from October 2010 to March 2011. Researchers determined and coded the underlying cause of death on the death certificates according to the guidelines issued by the World Health Organization in Volume 2 of the International Statistical Classification of Diseases and Health Related Problems-10th revision (ICD-10). Necessary ICD coding rules (such as the General Principle, Rules 1–3, the modification rules and other instructions about death coding) were applied to select the underlying cause of death on each certificate. Demographic details and documentation errors were then extracted. Data were analysed with descriptive statistics and chi square tests. The accuracy rate of causes of death coding was 51.7%, demonstrating a statistically significant relationship (p=.001) with major errors but not such a relationship with minor errors. Factors that result in poor quality of Cause of Death coding in SMH are lack of coder training, documentation errors and the undesirable structure of death certificates.

Maternal and hospital characteristics of non-medically indicated deliveries prior to 39 weeks

Non-medically indicated (NMI) deliveries prior to 39 weeks increase the risk of neonatal mortality, excess morbidity, and health care costs. The study's purpose was to identify maternal and hospital characteristics associated with NMI deliveries prior to 39 weeks. The study included 207,775 births to women without a previous cesarean and 38,316 births to women with a previous cesarean, using data from Florida's 2006-2007 linked birth certificate and inpatient record file. Adjusted risk ratios (ARR) and 95 % confidence intervals (CI) for characteristics were calculated using generalized estimating equation for multinomial logistic regression. Among women without a previous cesarean, NMI deliveries occurred in 18,368 births (8.8 %). Non-medically indicated inductions were more likely in women who were non-Hispanic white (ARR: 1.41, 95 % CI 1.31-1.52), privately-insured (ARR: 1.42, 95 % CI 1.26-1.59), and delivered in hospitals with <500 births per year. Non-medically indicated primary cesareans were more likely in women who were older than 35 years (ARR: 2.96, 95 % CI 2.51-3.50), non-Hispanic white (ARR: 1.44, 95 % CI 1.30-1.59), and privately-insured (ARR: 1.43, 95 % CI 1.17-1.73). Non-medically indicated primary cesareans were also more likely to occur in hospitals with <30 % nurse-midwife births, <500 births per year, and in large metro areas. Among women with previous cesarean, NMI repeat cesareans occurred in 16,746 births (43.7 %). Only weak risk factors were identified for NMI repeat cesareans. The risk factors identified varied by NMI outcome. This information can be used to inform educational campaigns and identify hospitals that may benefit from quality improvement efforts.

Profiling risk factors for chronic uveitis in juvenile idiopathic arthritis: a new model for EHR-based research

BACKGROUND

Juvenile idiopathic arthritis is the most common rheumatic disease in children. Chronic uveitis is a common and serious comorbid condition of juvenile idiopathic arthritis, with insidious presentation and potential to cause blindness. Knowledge of clinical associations will improve risk stratification. Based on clinical observation, we hypothesized that allergic conditions are associated with chronic uveitis in juvenile idiopathic arthritis patients.

METHODS

This study is a retrospective cohort study using Stanford's clinical data warehouse containing data from Lucile Packard Children's Hospital from 2000-2011 to analyze patient characteristics associated with chronic uveitis in a large juvenile idiopathic arthritis cohort. Clinical notes in patients under 16 years of age were processed via a validated text analytics pipeline. Bivariate-associated variables were used in a multivariate logistic regression adjusted for age, gender, and race. Previously reported associations were evaluated to validate our methods. The main outcome measure was presence of terms indicating allergy or allergy medications use overrepresented in juvenile idiopathic arthritis patients with chronic uveitis. Residual text features were then used in unsupervised hierarchical clustering to compare clinical text similarity between patients with and without uveitis.

RESULTS

Previously reported associations with uveitis in juvenile idiopathic arthritis patients (earlier age at arthritis diagnosis, oligoarticular-onset disease, antinuclear antibody status, history of psoriasis) were reproduced in our study. Use of allergy medications and terms describing allergic conditions were independently associated with chronic uveitis. The association with allergy drugs when adjusted for known associations remained significant (OR 2.54, 95% CI 1.22-5.4).

CONCLUSIONS

This study shows the potential of using a validated text analytics pipeline on clinical data warehouses to examine practice-based evidence for evaluating hypotheses formed during patient care. Our study reproduces four known associations with uveitis development in juvenile idiopathic arthritis patients, and reports a new association between allergic conditions and chronic uveitis in juvenile idiopathic arthritis patients.

Small area analysis of ambulatory care sensitive conditions in Victoria, Australia

Ambulatory care sensitive conditions (ACSCs) are used as a measure of access to primary health care. The purpose of this study was to identify factors associated with variation in ACSC admissions at a small area level in Victoria, Australia. The study was ecologic, using Victorian Primary Care Partnerships (PCPs) as the unit of analysis. Data sources were the Victorian Admitted Episodes Dataset, census data from the Australian Bureau of Statistics, and the Victorian Population Health Survey. Age- and sex-adjusted total ACSC admission rates were calculated, and weighted least squares multiple linear regression was used to examine the associations of total ACSC admission rates by various predictor variables. Key variables were categorized into 1 of 4 framework components for analyzing access and use of health care services: predisposing, enabling, need, or structural. Enabling characteristics explained 61.70% of the variation in ACSC admission rates across PCPs. Socioeconomic characteristics (income, education) and percentage with poor self-rated health were important factors in explaining variations in ACSC admissions at a small area-level [R(2)=0.77]. Community-level variables differentially affect access to primary health care, with significant variation by socioeconomic status. This analytical approach will assist researchers to identify community-level predicators of access across populations at locations, including factors that may be affected by policy change.

Patient characteristics associated with hospitalisations for ambulatory care sensitive conditions in Victoria, Australia

Background

Ambulatory Care Sensitive Conditions (ACSCs) are those for which hospitalisation is thought to be avoidable with the application of preventive care and early disease management, usually delivered in a primary care setting. ACSCs are used extensively as indicators of accessibility and effectiveness of primary health care. We examined the association between patient characteristics and hospitalisation for ACSCs in the adult and paediatric population in Victoria, Australia, 2003/04.

Methods

Hospital admissions data were merged with two area-level socioeconomic indexes: Index of Socio-Economic Disadvantage (IRSED) and Accessibility/Remoteness Index of Australia (ARIA). Univariate and multiple logistic regressions were performed for both adult (age 18+ years) and paediatric (age <18 years) groups, reporting odds ratios (OR) and 95% confidence intervals (CI) for a number of predictors of ACSCs admissions compared to non-ACSCs admissions.

Results

Predictors were much more strongly associated with ACSCs admissions compared to non-ACSCs admissions in the adult group than for the paediatric group with the exception of rurality. Significant adjusted ORs in the adult group were 1.06, 1.15, 1.13, 1.06 and 1.11 for sex, rurality, age, IRSED and ARIA variables, and 1.34, 1.04 and 1.09 in the paediatric group for rurality, IRSED and ARIA, respectively.

Conclusions

Disadvantaged paediatric and adult population experience more need of hospital care for ACSCs. Access barriers to primary care are plausible causes for the observed disparities. Understanding the characteristics of individuals experiencing access barriers to primary care will be useful for developing targeted interventions meeting the unique ambulatory needs of the population.

Using administrative health data to describe colorectal and lung cancer care in New South Wales, Australia: a validation study

Background

Monitoring treatment patterns is crucial to improving cancer patient care. Our aim was to determine the accuracy of linked routinely collected administrative health data for monitoring colorectal and lung cancer care in New South Wales (NSW), Australia.

Methods

Colorectal and lung cancer cases diagnosed in NSW between 2000 and 2002 were identified from the NSW Central Cancer Registry (CCR) and linked to their hospital discharge records in the NSW Admitted Patient Data Collection (APDC). These records were then linked to data from two relevant population-based patterns of care surveys. The main outcome measures were the sensitivity and specificity of data from the CCR and APDC for disease staging, investigative procedures, curative surgery, chemotherapy, radiotherapy, and selected comorbidities.

Results

Data for 2917 colorectal and 1580 lung cancer cases were analysed. Unknown disease stage was more common for lung cancer in the administrative data (18%) than in the survey (2%). Colonoscopies were captured reasonably accurately in the administrative data compared with the surveys (82% and 79% respectively; 91% sensitivity, 53% specificity) but all other colorectal or lung cancer diagnostic procedures were under-enumerated. Ninety-one percent of colorectal cancer cases had potentially curative surgery recorded in the administrative data compared to 95% in the survey (96% sensitivity, 92% specificity), with similar accuracy for lung cancer (16% and 17%; 92% sensitivity, 99% specificity). Chemotherapy (~40% sensitivity) and radiotherapy (sensitivity≤30%) were vastly under-enumerated in the administrative data. The only comorbidity that was recorded reasonably accurately in the administrative data was diabetes.

Conclusions

Linked routinely collected administrative health data provided reasonably accurate information on potentially curative surgical treatment, colonoscopies and comorbidities such as diabetes. Other diagnostic procedures, comorbidities, chemotherapy and radiotherapy were not well enumerated in the administrative data. Other sources of data will be required to comprehensively monitor the primary management of cancer patients.

Patient, physician, encounter, and billing characteristics predict the accuracy of syndromic surveillance case definitions

BACKGROUND

Syndromic surveillance systems are plagued by high false-positive rates. In chronic disease monitoring, investigators have identified several factors that predict the accuracy of case definitions based on diagnoses in administrative data, and some have even incorporated these predictors into novel case detection methods, resulting in a significant improvement in case definition accuracy. Based on findings from these studies, we sought to identify physician, patient, encounter, and billing characteristics associated with the positive predictive value (PPV) of case definitions for 5 syndromes (fever, gastrointestinal, neurological, rash, and respiratory (including influenza-like illness)).

METHODS

The study sample comprised 4,330 syndrome-positive visits from the claims of 1,098 randomly-selected physicians working in Quebec, Canada in 2005-2007. For each visit, physician-facilitated chart review was used to assess whether the same syndrome was present in the medical chart (gold standard). We used multivariate logistic regression analyses to estimate the association between claim-chart agreement about the presence of a syndrome and physician, patient, encounter, and billing characteristics.

RESULTS

The likelihood of the medical chart agreeing with the physician claim about the presence of a syndrome was higher when the treating physician had billed many visits for the same syndrome recently (ORper 10 visit, 1.05; 95% CI, 1.01-1.08), had a lower workload (ORper 10 claims, 0.93; 95% CI, 0.90-0.97), and when the patient was younger (ORper 5 years of age, 0.96; 95% CI, 0.94-0.97), and less socially deprived (ORmost versus least deprived, 0.76; 95% CI, 0.60-0.95).

CONCLUSIONS

Many physician, patient, encounter, and billing characteristics associated with the PPV of surveillance case definition are accessible to public health, and could be used to reduce false-positive alerts by surveillance systems, either by focusing on the data most likely to be accurate, or by adjusting the observed data for known biases in diagnosis reporting and performing surveillance using the adjusted values.

Alcohol outlet density and harm: comparing the impacts on violence and chronic harms

INTRODUCTION AND AIMS

A number of studies have previously identified relationships between the density of alcohol outlets and rates of violence, with different types of outlets related to violence in different locations. The previous work in Australia has been limited to studies based on police data, which are subject to numerous biases. This study extends the previous work by utilising hospital admissions as a less biased outcome measure, incorporating a 14 year longitudinal design and by developing comparative models for violence and rates of alcohol use disorders.

DESIGN AND METHODS

The study examines trends in postcode-level hospital admission data for assault and for alcohol use disorders over a 14 year period (n = 186) and their relationship with the density of three kinds of alcohol outlets. Fixed-effects models are developed to control for the differences between postcodes and for the overall trends in outlet density and morbidity rates.

RESULTS

The results of this study suggest that the density of alcohol outlets where the main activity is alcohol consumption (i.e. pubs) is positively related to rates of assault-related hospital admissions, while the density of off-premise alcohol outlets is related to the rate of alcohol use disorders.

DISCUSSION AND CONCLUSIONS

These findings have significant implications for alcohol policies in Victoria, in particular pointing to the significant contribution of packaged alcohol outlets to both acute and chronic alcohol-related harm.

Systemic factors of errors in the case identification process of the national routine health information system: a case study of Modified Field Health Services Information System in the Philippines

Background

The quality of data in national health information systems has been questionable in most developing countries. However, the mechanisms of errors in the case identification process are not fully understood. This study aimed to investigate the mechanisms of errors in the case identification process in the existing routine health information system (RHIS) in the Philippines by measuring the risk of committing errors for health program indicators used in the Field Health Services Information System (FHSIS 1996), and characterizing those indicators accordingly.

Methods

A structured questionnaire on the definitions of 12 selected indicators in the FHSIS was administered to 132 health workers in 14 selected municipalities in the province of Palawan. A proportion of correct answers (difficulty index) and a disparity of two proportions of correct answers between higher and lower scored groups (discrimination index) were calculated, and the patterns of wrong answers for each of the 12 items were abstracted from 113 valid responses.

Results

None of 12 items reached a difficulty index of 1.00. The average difficulty index of 12 items was 0.266 and the discrimination index that showed a significant difference was 0.216 and above. Compared with these two cut-offs, six items showed non-discrimination against lower difficulty indices of 0.035 (4/113) to 0.195 (22/113), two items showed a positive discrimination against lower difficulty indices of 0.142 (16/113) and 0.248 (28/113), and four items showed a positive discrimination against higher difficulty indices of 0.469 (53/113) to 0.673 (76/113).

Conclusions

The results suggest three characteristics of definitions of indicators such as those that are (1) unsupported by the current conditions in the health system, i.e., (a) data are required from a facility that cannot directly generate the data and, (b) definitions of indicators are not consistent with its corresponding program; (2) incomplete or ambiguous, which allow several interpretations; and (3) complete yet easily misunderstood by health workers.

Taking systemic factors into account, the case identification step needs to be reviewed and designed to generate intended data in health information systems.

Health reform: is routinely collected electronic information fit for purpose?

OBJECTIVE

Little has been reported about the completeness and accuracy of data in existing Australian clinical information systems. We examined the accuracy of the diagnoses of some chronic diseases in an ED information system (EDIS), a module of the NSW Health electronic medical record (EMR), and the consistency of the reports generated by the EMR.

METHODS

A list of ED attendees and those admitted was generated from the EDIS, using specific (e.g. angina) and possible clinical terms (e.g. chest pain) for the selected chronic diseases. This EDIS list was validated with an audit of discharge summaries, and compared with a list generated, using similar specific and possible Systematized Nomenclature of Medicine-Clinical Terms (SNOMED-CT), from the underlying EMR database.

RESULTS

Of the 33,115 ED attendees, 2559 had diabetes mellitus (DM), cardiovascular disease or asthma/chronic obstructive pulmonary disease; of these 2559, 876 were admitted. Discharge summaries were missing for 12-15% of patients. Only three-quarters or fewer of the diagnoses were confirmed by the discharge summary audit, best for DM and worst for cardiovascular disease. Proportion of agreement between the lists generated from the EDIS and EMR was best for DM and worst for asthma/chronic obstructive pulmonary disease. Possible reasons for this discrepancy are technical, such as use of different extraction terms or system inconsistency; or clinical, such as data entry, decision-making, professional behaviour and organizational performance.

CONCLUSIONS

Variations in information quality and consistency of the EDIS/EMR raise concerns about the 'fitness for purpose' of the information for care and planning, information sharing, research and quality assurance.

Comparing the coding of complications in Queensland and Victorian admitted patient data

Objective. To examine differences between Queensland and Victorian coding of hospital-acquired conditions and suggest ways to improve the usefulness of these data in the monitoring of patient safety events. Design. Secondary analysis of admitted patient episode data collected in Queensland and Victoria. Methods. Comparison of depth of coding, and patterns in the coding of ten commonly coded complications of five elective procedures. Results. Comparison of the mean complication codes assigned per episode revealed Victoria assigns more valid codes than Queensland for all procedures, with the difference between the states being significantly different in all cases. The proportion of the codes flagged as complications was consistently lower for Queensland when comparing 10 common complications for each of the five selected elective procedures. The estimated complication rates for the five procedures showed Victoria to have an apparently higher complication rate than Queensland for 35 of the 50 complications examined. Conclusion. Our findings demonstrate that the coding of complications is more comprehensive in Victoria than in Queensland. It is known that inconsistencies exist between states in routine hospital data quality. Comparative use of patient safety indicators should be viewed with caution until standards are improved across Australia. More exploration of data quality issues is needed to identify areas for improvement. What is known about the topic? Routine data are low cost, accessible and timely but the quality is often questioned. This deters researchers and clinicians from using the data to monitor aspects of quality improvement. Previous studies have reported on the quality of diagnosis coding in Australia but not specifically on the quality of use of the condition-onset flag denoting hospital-acquired conditions. What does this paper add? Few studies have tested the consistency of the data between Australian states. No previous studies have evaluated the comprehensiveness of the coding of hospital-acquired conditions using routine data. This paper compares two states to highlight the differences in the coding of complications, with the aim of improving routine data to support patient safety. What are the implications for practitioners? The results imply more work needs to be done to improve the coding and flagging of complications so the data are valid and comprehensive. Further research should identify problem areas responsible for differences in the data so that training and audit strategies can be developed to improve the collection of this information. Practitioners may then be more confident in using routine coded inpatient data as part of the process of monitoring patient safety.

Correlates of undefined cause of injury coded mortality data in Australia

The objective of this research was to identify the level of detail regarding the external causes of death in Australia and ascertain problematic areas where data quality improvement efforts may be focused. The 2003 national mortality dataset of 12,591 deaths with an external cause of injury as the underlying cause of death (UCOD) or multiple cause of death (MCOD) based on ICD-10 code assignment from death certificate information was obtained. Logistic regression models were used to examine the precision of coded external cause of injury data. It was found that overall, accidents were the most poorly defined of all intent code blocks with over 30% of accidents being undefined, representing 2,314 deaths in 2003. More undefined codes were identified in MCOD data than for UCOD data. Deaths certified by doctors were more likely to use undefined codes than deaths certified by a coroner or government medical office. To improve the quality of external cause of injuries leading to or associated with death, certifiers need to be made aware of the importance of documenting all information pertaining to the cause of the injury and the intent behind the incident, either through education or more explicit instructions on the death certificate and accompanying instructional materials. It is important that researchers are aware of the validity of the data when they make interpretations as to the underlying causes of fatal injuries and causes of injury associated with deaths.

Development of a validation algorithm for 'present on admission' flagging

BACKGROUND

The use of routine hospital data for understanding patterns of adverse outcomes has been limited in the past by the fact that pre-existing and post-admission conditions have been indistinguishable. The use of a 'Present on Admission' (or POA) indicator to distinguish pre-existing or co-morbid conditions from those arising during the episode of care has been advocated in the US for many years as a tool to support quality assurance activities and improve the accuracy of risk adjustment methodologies. The USA, Australia and Canada now all assign a flag to indicate the timing of onset of diagnoses. For quality improvement purposes, it is the 'not-POA' diagnoses (that is, those acquired in hospital) that are of interest.

METHODS

Our objective was to develop an algorithm for assessing the validity of assignment of 'not-POA' flags. We undertook expert review of the International Classification of Diseases, 10th Revision, Australian Modification (ICD-10-AM) to identify conditions that could not be plausibly hospital-acquired. The resulting computer algorithm was tested against all diagnoses flagged as complications in the Victorian (Australia) Admitted Episodes Dataset, 2005/06. Measures reported include rates of appropriate assignment of the new Australian 'Condition Onset' flag by ICD chapter, and patterns of invalid flagging.

RESULTS

Of 18,418 diagnosis codes reviewed, 93.4% (n = 17,195) reflected agreement on status for flagging by at least 2 of 3 reviewers (including 64.4% unanimous agreement; Fleiss' Kappa: 0.61). In tests of the new algorithm, 96.14% of all hospital-acquired diagnosis codes flagged were found to be valid in the Victorian records analysed. A lower proportion of individual codes was judged to be acceptably flagged (76.2%), but this reflected a high proportion of codes used <5 times in the data set (789/1035 invalid codes).

CONCLUSION

An indicator variable about the timing of occurrence of diagnoses can greatly expand the use of routinely coded data for hospital quality improvement programmes. The data-cleaning instrument developed and tested here can help guide coding practice in those health systems considering this change in hospital coding. The algorithm embodies principles for development of coding standards and coder education that would result in improved data validity for routine use of non-POA information.

Incidence of traumatic long-bone fractures requiring in-hospital management: a prospective age- and gender-specific analysis of 4890 fractures

BACKGROUND

Musculoskeletal trauma represents a considerable global health burden; however, reliable population-based incidence data are lacking. Thus, we prospectively investigated the age- and sex-specific incidence patterns of long-bone fractures in a defined population.

METHODS

A 4-year prospective study of all long-bone fractures in a defined Norwegian population was carried out. The demographic data, as well as data on fracture type and location and mode of treatment were collected using recognised classification (e.g., AO/OTA - Arbeitsgemeinschaft für Osteosynthesefragen/Orthopaedic Trauma Association; Gustilo-Anderson (GA) for open fractures). Age- and sex-adjusted incidences were calculated using population statistics.

RESULTS

During the study period, 4890 long-bone fractures were recorded. The overall incidence per 100,000 per year was 406 with a 95% confidence interval (95%CI) of 395-417. The age-adjusted incidence for those <16 years (339; 95%CI: 318-360) was lower than that for those >or=16 years (427; 95%CI: 414-440). The overall male incidence (337; 95%CI: 322-355) was lower than the female (476; 95%CI: 459-493), but the male:female ratio was 2:1 among those <50 years, and 1:3 in those >or=50 years. The upper limb fractures had an overall incidence of 159 (95%CI: 152-166), whereas the lower limb fracture incidence was 247 (95%CI: 238-256). Open fractures occurred in 3%, with an incidence of 13 (95%CI: 11-15). Paediatric fractures were more likely to be treated conservatively with only 8% requiring internal fixation, compared to 56% internal fixation in those >or=16 years of age. An increase in the use of angular stable plates occurred during the study period.

CONCLUSION

This prospectively collected study of long-bone fractures in a defined population recognises age- and gender-specific fracture patterns. Boys predominate in the younger age group for which treatment is basically conservative. In the senior population, women and operative treatment predominate.

The Risk and Consequences of Clinical Miscoding Due to Inadequate Medical Documentation: A Case Study of the Impact on Health Services Funding

As coded clinical data are used in a variety of areas (e.g. health services funding, epidemiology, health sciences research), coding errors have the potential to produce far-reaching consequences. In this study the causes and consequences of miscoding were reviewed. In particular, the impact of miscoding due to inadequate medical documentation on hospital funding was examined. Appropriate reimbursement of hospital revenue in the casemix-based (output-based) funding system in the state of Victoria, Australia relies upon accurate, comprehensive, and timely clinical coding. In order to assess the reliability of these data in a Melbourne tertiary hospital, this study aimed to: (a) measure discrepancies in clinical code assignment; (b) identify resultant Diagnosis Related Group (DRG) changes; (c) identify revenue shifts associated with the DRG changes; (d) identify the underlying causes of coding error and DRG change; and (e) recommend strategies to address the aforementioned. An internal audit was conducted on 752 surgical inpatient discharges from the hospital within a six-month period. In a blind audit, each episode was re-coded. Comparisons were made between the original codes and the auditor-assigned codes, and coding errors were grouped and statistically analysed by categories. Changes in DRGs and weighted inlier-equivalent separations (WIES) were compared and analysed, and underlying factors were identified. Approximately 16% of the 752 cases audited reflected a DRG change, equating to a significant revenue increase of nearly AU$575,300. Fifty-six percent of DRG change cases were due to documentation issues. Incorrect selection or coding of the principal diagnosis accounted for a further 13% of the DRG changes, and missing additional diagnosis codes for 29%.The most significant of the factors underlying coding error and DRG change was poor quality of documentation. It was concluded that the auditing process plays a critical role in the identification of causes of coding inaccuracy and, thence, in the improvement of coding accuracy in routine disease and procedure classification and in securing proper financial reimbursement.

Trends in percutaneous coronary interventions in new South Wales, Australia

This is the first detailed study on percutaneous coronary intervention (PCI) in New South Wales (NSW), Australia. Hospital data for PCIs carried out between 1 July 1990 and 30 June 2002 are analysed. The study explores trends in PCI rates by selected socio-demographic factors, the utilisation of angioplasties vis-a-vis stents, emergency admissions, and selected coexisting conditions which determine the disease status of PCI patients. Logistic regression models are used to study the medical conditions that require both PCI and coronary artery bypass graft (CABG). The PCI rate has grown rapidly at 12.1% per annum, with a particularly rapid increase for persons aged 75+. The rate of multiple stent utilisation increased at 4.6% per annum. Pacific-born and Middle-Eastern-born patients are more than twice as likely as the Australian-born to have diabetes. Factors affecting failure of PCI requiring CABG include perforation and multi-vessel disease. PCI services in public hospitals need to be increased to facilitate the availability of these procedures to all segments of the population, as do targeted community-level programmes to educate high-risk groups in the control of heart diseases.

The incidence and cost of cardiac surgery adverse events in Australian (Victorian) hospitals 2003-2004

The aim of this study was to estimate the incidence of adverse events in acute surgical admissions for cardiac disease in admitted episodes in the year 2003-2004 and to estimate the cost of these complications to the Victorian health system. Cardiac surgery adverse events are among the most frequent and significant contributors to the morbidity, mortality and cost associated with hospitalisation. Patient-level costing data set for major Victorian public hospitals in 2003-2004 was analysed for adverse events using C-prefixed markers, denoting complications that arose during the course of hospital treatment for cardiac surgery diagnosis related groups (DRGs). The cost of adverse events was estimated by linear regression modelling, adjusted for age and co-morbidity. A total of 16,766 multi-day cardiac disease cases were identified, of whom 6,181 (36.85%) had at least one adverse event. Patients with adverse events stayed approximately 7 days longer and had four times the case fatality rate than those without. After adjustment for age and co-morbidity, the presence of an adverse event adds AUS$5,751. The sum of the total cost of adverse events for each DRG was AUS$42.855 million, representing 21.6% of total expenditure on cardiac surgery and adding 27.5% in broad terms to the cardiac surgery budget.

Effect of an acute medical admission unit on key quality indicators assessed by funnel plots

OBJECTIVE

To determine the effect of the introduction of an acute medical admissions unit (AMAU) on key quality efficiency and outcome indicator comparisons between medical teams as assessed by funnel plots.

METHODS

A retrospective analysis was performed of data relating to emergency medical patients admitted to St James' Hospital, Dublin between 1 January 2002 and 31 December 2004, using data on discharges from hospital recorded in the hospital in-patient enquiry system. The base year was 2002 during which patients were admitted to a variety of wards under the care of a named consultant physician. In 2003, two centrally located wards were reconfigured to function as an AMAU, and all emergency patients were admitted directly to this unit. The quality indicators examined between teams were length of stay (LOS) <30 days, LOS >30 days, and readmission rates.

RESULTS

The impact of the AMAU reduced overall hospital LOS from 7 days in 2002 to 5 days in 2003/04 (p<0.0001). There was no change in readmission rates between teams over the 3 year period, with all teams displaying expected variability within control (95%) limits. Overall, the performance in LOS, both short term and long term, was significantly improved (p<0.0001), and was less varied between medical teams between 2002 and 2003/04.

CONCLUSIONS

Introduction of the AMAU improved performance among medical teams in LOS, both short term and long term, with no change in readmissions. Funnel plots are a powerful graphical technique for presenting quality performance indicator variation between teams over time.

Quality of diagnosis and procedure coding in ICD-10 administrative data

OBJECTIVES

The International Classification of Disease, 10th Revision (ICD-10) was introduced worldwide beginning in the late 1990s. Because there have been no published data on the quality of coding using ICD-10, the aim of our analysis is to assess the quality of ICD-10 coding in routinely collected hospital discharge data from Australia, which began using ICD-10 in 1998.

METHODS

Audit data from the years 1998-1999 (n = 7004) and 2000-2001 (n = 7631), excluding same-day chemotherapy and dialysis cases, were used in data analysis. Quality measures included prevalence comparisons, sensitivity, positive predictive value (PPV), and the kappa statistic.

RESULTS

Comparison of the audit sample to public hospital discharges showed little difference in age and gender, with audited cases more likely to be overnight stays. There was no difference in the median number of hospital assigned diagnosis and procedure codes per discharge. Agreement of the principal diagnosis code was 85% at the 3-digit level and 79% at the 4-digit level in 1998-1999; this rate had improved to 87% and 81% in 2000-2001. Principal procedure code agreement was 85% in 1998-1999 and 83% in 2000-2001 at the 5-digit level, and 81% and 80% at the 7-digit level, respectively. Specific major diagnoses, comorbid diagnoses, major procedures, and minor procedures showed good-to-excellent coding quality.

CONCLUSIONS

The transition to ICD-10 has occurred with no loss of data quality, with data showing a high level of reliability and adherence to coding standards. When consideration is given to the nature of the analysis, administrative data can provide highly reliable population-based estimates of hospitalization rates.

The epidemiology of hospitalised wrist fractures in older people, New South Wales, Australia

The epidemiology and trends in wrist fracture admissions to public and private acute hospitals in New South Wales (NSW), Australia, between July 1993 and June 2003 were examined using routinely collected hospital separations statistics. During the study period, the number of hospital separations for wrist fractures increased by 71% in men, an average yearly increase of 6.5%, and by 43% in women, an average yearly increase of 3.9%. A modest, but significant, increase in age-specific and age-standardised hospitalisation rates for wrist fractures was also observed. Whilst the majority of wrist fractures were due to falls, the proportion of falls-related wrist fractures decreased significantly over time. This decrease was more pronounced in males and was accompanied by a rise in the proportion of wrist fractures resulting from high energy mechanisms such as transport, violence and machinery-related incidents. The difference in hospitalised wrist fracture rates between men and women could not be explained solely on the basis of the role played by osteoporosis, indicating the need for more research to improve our understanding of the underlying factors of this type of fracture in older people.

Accuracy of injury coding under ICD-9 for New Zealand public hospital discharges

OBJECTIVE

To determine the level of accuracy in coding for injury principal diagnosis and the first external cause code for public hospital discharges in New Zealand and determine how these levels vary by hospital size.

METHOD

A simple random sample of 1800 discharges was selected from the period 1996-98 inclusive. Records were obtained from hospitals and an accredited coder coded the discharge independently of the codes already recorded in the national database.

RESULTS

Five percent of the principal diagnoses, 18% of the first four digits of the E-codes, and 8% of the location codes (5th digit of the E-code), were incorrect. There were no substantive differences in the level of incorrect coding between large and small hospitals.

CONCLUSIONS

Users of New Zealand public hospital discharge data can have a high degree of confidence in the injury diagnoses coded under ICD-9-CM-A. A similar degree of confidence is warranted for E-coding at the group level (for example, fall), but not, in general, at higher levels of specificity (for example, type of fall). For those countries continuing to use ICD-9 the study provides insight into potential problems of coding and thus guidance on where the focus of coder training should be placed. For those countries that have historical data coded according to ICD-9 it suggests that some specific injury and external cause incidence estimates may need to be treated with more caution.

The incidence and cost of adverse events in Victorian hospitals 2003–04

OBJECTIVES

To determine the incidence of adverse events in patients admitted in the year 2003-04 to selected Victorian hospitals; to identify the main hospital-acquired diagnoses; and to estimate the cost of these complications to the Victorian and Australian health system.

DESIGN

The patient-level costing dataset for major Victorian public hospitals, 1 July 2003-30 June 2004, was analysed for adverse events by identifying C-prefixed diagnosis codes denoting complications, preventable or otherwise, arising during the course of hospital treatment. The in-hospital cost of adverse events was estimated using linear regression modelling, adjusting for age and comorbidity.

MAIN OUTCOME MEASURES

Cost of each patient admission ("admitted episode"), length of stay and mortality.

RESULTS

During the designated timeframe, 979,834 admitted episodes were in the sample, of which 67,435 (6.88%) had at least one adverse event. Patients with adverse events stayed about 10 days longer and had over seven times the risk of in-hospital death than those without complications. After adjusting for age and comorbidity, the presence of an adverse event adds dollar 6826 to the cost of each admitted episode. The total cost of adverse events in this dataset in 2003-04 was dollar 460.311 million, representing 15.7% of the total expenditure on direct hospital costs, or an additional 18.6% of the total inpatient hospital budget.

CONCLUSION

Adverse events are associated with significant costs. Administrative datasets are a cost-effective source of information that can be used for a range of clinical governance activities to prevent adverse events.

Factors influencing the costs of emergency medical admissions to an Irish teaching hospital

This study examined whether there is a relationship between coded diseases at the time of hospital discharge and costs of hospital re-admission. We carried out a systematic review of data relating to emergency medical patients admitted to St. James' Hospital in Dublin between 1 January 2002 and 31 October 2004. Data on discharges from hospital were analyzed as recorded in the hospital in-patient enquiry (HIPE) system. Of 15,876 episodes recorded among 11,201 patients admitted the number of re-admissions numbered up to 43. Age, year of admission, and frequency of admission were factors associated with increased hospital costs. HIPE coding at first discharge predicted increased costs: codes related to heart failure, pneumonia, stroke, diabetes, malignancy, psychiatric, and anaemia-related codes. Clinical coding using the HIPE database thus strongly predicted hospital costs.

Ten-year absolute risk of osteoporotic fractures according to BMD T score at menopause: the Danish Osteoporosis Prevention Study

In the non-HRT arms of the DOPS study, 10-year fracture risk was higher at each level of T score than predicted by the Kanis algorithm. Under-reporting of fractures in registers and inclusion of HRT users are probable explanations for inappropriately low fracture risk estimates for younger women.

INTRODUCTION

International recommendations highlight the importance of absolute fracture risk in establishing intervention thresholds. The available estimates of long-term risk have been derived by combining relative risks from meta-analyses with U.S. normative BMD data and Swedish fracture incidence records. We validated the 2001 Kanis risk algorithm using incident fractures observed in untreated women in the first 10 years of the Danish Osteoporosis Prevention Study (DOPS). Comparisons were also made with the relative risks derived from a recent meta-analysis of 12 cohort studies.

MATERIALS AND METHODS

We analyzed DXA of the spine and hip from 872 women who were enrolled in the non-hormone replacement therapy (HRT) arms of the study and had not received HRT, bisphosphonates, or raloxifene. We collected verified reports of fractures at each visit. We focused on fractures of the hip, spine, shoulder, and forearm to provide risks comparable with the Kanis algorithm. Accordingly, asymptomatic radiographic vertebral fractures were not included.

RESULTS

Seventy-eight women (9%) sustained relevant fractures. The risk of fracture increased by 1.32 (95% CI, 1.02; 1.70) for each unit decrease in femoral neck T score and by 1.30 (95% CI, 1.06; 1.58) for each unit decrease in lumbar spine T score at baseline. Absolute fracture risk was higher than expected from the Kanis algorithm at all T score levels. The difference was greatest for participants in the higher range of T scores. At T = -1, the observed risk was 10.9% as opposed to an expected risk of 5.7%. Relative risk gradients were similar to those of the recent meta-analysis.

CONCLUSIONS

In healthy women, examined in the first year or two after menopause, 10-year fracture risk was higher at each level of BMD T score than expected from the model by Kanis et al. Inclusion of HRT users in the cohorts used may have led to higher BMD values and lower absolute fracture risk in the Kanis model. These longitudinal data can be used directly in estimating absolute fracture risk in untreated north European women from BMD at menopause.

Do consultants differ? Inferences drawn from hospital in-patient enquiry (HIPE) discharge coding at an Irish teaching hospital

OBJECTIVE

To find out if there was a difference between hospital consultants, all trained in acute general medicine, in length of stay (LOS), re-admission rates, resource utilisation, and diagnostic coding, among patients admitted as emergencies to St James' Hospital (SJH) Dublin.

METHODS

A retrospective analysis was performed of data on discharges from hospital, recorded in the hospital in-patient enquiry (HIPE) system, relating to 9204 episodes among 6968 emergency medical patients admitted to SJH between 1 January 2002 and 31 October 2003. For comparative analysis, four physician groups were defined consisting of gastroenterology (GI, n = 4), respiratory (n = 3), general internal medicine (GIM, n = 2), or specialty (n = 5).

RESULTS

GIM consultants had the shortest LOS (median 5 days); GIM and respiratory consultants were less likely to have long stay patients (> 30 days, p<0.0001). Patients re-admitted under the same consultant had a longer LOS than those re-admitted under a different consultant (p<0.0001). Endoscopy and GI radiology investigations were used most by GI consultants, computed tomography of the thorax by respiratory, ECHO by respiratory and specialty, and computed tomography of brain by GIM and specialty consultants. GI diagnostic codings were more frequent with GI consultants (p<0.0001), respiratory diagnoses and malignancy with respiratory (p<0.0001 for both), diabetes and hypertension with specialty (p = 0.0017), and heart failure more with GIM consultants (p = 0.001).

CONCLUSIONS

This study found that the HIPE database was very powerful in predicting differences between hospital consultants in LOS, re-admission rates, resource utilisation, and disease coding. It would be of interest to examine the extent to which protocols and guidelines could reduce such variations.

Epidemiology of Scalds in Vulnerable Groups in New South Wales, Australia, 1998/1999 to 2002/2003

In this study, the recently introduced International Classification of Disease, 10th revision, code for hot tap water scalds was used to examine the epidemiology of these cases and other scalds injuries in children younger than 5 years of age and adults aged 65 years and older. Although the trunk was the most common area in which scalds occurred, young children were more likely to sustain head and neck scalds (15%, 95% confidence interval 10.8-18.3) because of hot tap water than older people (2%, 95% confidence interval 0.2-4.4). Hospital separation rates for hot water scalds decreased significantly during the study period in both boys (chi(2) = 15.6, df = 1, P < .001) and girls (chi(2) = 5.6, df = 1, P < .001) who were younger than 5 years of age, which might be attributable to the introduction of new standards regulating the provision of hot tap water to various buildings. The severity of scalds cases did not seem to be correlated with the length of hospital stay, which remained unchanged in both age groups.

Accuracy of ICD-9-CM codes for identifying cardiovascular and stroke risk factors

OBJECTIVES

We sought to determine which ICD-9-CM codes in Medicare Part A data identify cardiovascular and stroke risk factors.

DESIGN AND PARTICIPANTS

This was a cross-sectional study comparing ICD-9-CM data to structured medical record review from 23,657 Medicare beneficiaries aged 20 to 105 years who had atrial fibrillation.

MEASUREMENTS

Quality improvement organizations used standardized abstraction instruments to determine the presence of 9 cardiovascular and stroke risk factors. Using the chart abstractions as the gold standard, we assessed the accuracy of ICD-9-CM codes to identify these risk factors.

MAIN RESULTS

ICD-9-CM codes for all risk factors had high specificity (>0.95) and low sensitivity (< or =0.76). The positive predictive values were greater than 0.95 for 5 common, chronic risk factors-coronary artery disease, stroke/transient ischemic attack, heart failure, diabetes, and hypertension. The sixth common risk factor, valvular heart disease, had a positive predictive value of 0.93. For all 6 common risk factors, negative predictive values ranged from 0.52 to 0.91. The rare risk factors-arterial peripheral embolus, intracranial hemorrhage, and deep venous thrombosis-had high negative predictive value (> or =0.98) but moderate positive predictive values (range, 0.54-0.77) in this population.

CONCLUSIONS

Using ICD-9-CM codes alone, heart failure, coronary artery disease, diabetes, hypertension, and stroke can be ruled in but not necessarily ruled out. Where feasible, review of additional data (eg, physician notes or imaging studies) should be used to confirm the diagnosis of valvular disease, arterial peripheral embolus, intracranial hemorrhage, and deep venous thrombosis.

Childhood femur fractures, associated injuries, and sociodemographic risk factors: a population-based study

OBJECTIVE

The objectives of this study were to determine the incidence of femur fractures in Colorado children, to assess underlying causes, to determine the prevalence and predictors of associated injuries, and to identify potentially modifiable risk factors.

METHODS

The study population included all Colorado residents who were aged 0 to 17 years at the time of injury between January 1, 1998, and December 31, 2001. Cases of femur fracture were ascertained using the population-based Colorado Trauma Registry and International Classification of Diseases, Ninth Revision, Clinical Modification codes 820.0 to 821.39. Associated injuries with an Abbreviated Injury Scale of 2 or higher were classified into 5 categories. Poisson regression, small area analysis, and multivariate logistic regression were used to identify predictors of femur fractures and associated injuries, respectively.

RESULTS

During the study period, 1139 Colorado children (795 boys, 344 girls) sustained femur fractures, resulting in the incidence of 26.0 per 100000 person-years. Rates were higher in boys than in girls in all age groups (overall risk ratio: 2.19; 95% confidence interval: 1.92-2.47) but did not differ by race/ethnicity. Femur fractures that were caused by nonaccidental trauma showed more distal and combined shaft + distal pattern; their incidence did not differ by gender or race but was higher in census tracts with more single mothers and less crowded households. Associated injuries were present in 28.6% of the cases, more often in older children. Fatalities occurred only among children with associated injuries. Children who were involved in nonaccidental trauma, motor vehicle crashes, or auto-pedestrian accidents were 16 to 20 times more likely to have associated injuries than those with femur fractures as a result of a fall. In small-area analysis, the incidence of femur fractures in infants and toddlers was higher in census tracts characterized by higher proportion of Hispanics, single mothers, and more crowded households. Among children 4 to 12 years of age, the incidence was higher in census tracts with fewer single-family houses and more crowded households. Finally, the incidence of femur fractures among teenagers was higher in rural tracts and those with a higher proportion of Hispanics.

CONCLUSIONS

Femur fractures and associated injuries remain a major cause of morbidity in children. Predictors of femur fractures change with age; however, the risk is generally higher among children who live in the areas with lower socioeconomic indicators.

The use of injury surveillance databases to identify emerging injury hazards

Among the many valuable uses of injury surveillance is the potential to alert health authorities and societies in general to emerging injury trends, facilitating earlier development of prevention measures. Other than road safety, to date, few attempts to forecast injury data have been made, although forecasts have been made of other public health issues. This may in part be due to the complex pattern of variance displayed by injury data. The profile of many injury types displays seasonality and diurnal variance, as well as stochastic variance. The authors undertook development of a simple model to forecast injury into the near term. In recognition of the large numbers of possible predictions, the variable nature of injury profiles and the diversity of dependent variables, it became apparent that manual forecasting was impractical. Therefore, it was decided to evaluate a commercially available forecasting software package for prediction accuracy against actual data for a set of predictions. Injury data for a 4-year period (1996 to 1999) were extracted from the Victorian Emergency Minimum Dataset and were used to develop forecasts for the year 2000, for which data was also held. The forecasts for 2000 were compared to the actual data for 2000 by independent t-tests, and the standard errors of the predictions were modelled by stepwise hierarchical multiple regression using the independent variables of the standard deviation, seasonality, mean monthly frequency and slope of the base data (R = 0.93, R(2) = 0.86, F(3, 27) = 55.2, p < 0.0001). Significant contributions to the model included the SD (beta = 1.60, p < 0.001), mean monthly frequency (beta = -0.72, p < 0.002), and the seasonality of the data (beta = 0.16, p < 0.02). It was concluded that injury data could be reliably forecast and that commercial software was adequate for the task. Variance in the data was found to be the most important determinant of prediction accuracy. Importantly, automated forecasting may provide a vehicle for identifying emerging trends.

Underestimating Injury Mortality Using Statewide Databases

BACKGROUND

This study examines the potential for misclassifying injury-related deaths reported in Vital Statistics and assesses the rate of postdischarge death among injured patients released from hospital, emergency department (ED), and emergency medical services (EMS) care.

METHODS

Statewide death certificate, inpatient, ED, and EMS databases for 1996 through 1997 were probabilistically linked and information in each database compared.

RESULTS

One thousand two hundred ninety-four injured inpatients or ED patients were matched with a death certificate record that listed an injury (56.3%) or illness (43.7%) as the primary cause of death. Injured decedents with an illness-coded cause of death were older (p < 0.001), with causes of death indicative of chronic medical conditions. Few deaths occurred within 30 days of inpatient discharge (6%); however, 38% and 9% of deaths in ED and EMS databases occurred after discharge from health care, respectively. Many deaths among EMS and ED patients occur in subsequent phases of care.

CONCLUSION

Estimates of injury mortality substantially increase when using multiple independent databases.

Patient and disease profile of emergency medical readmissions to an Irish teaching hospital

OBJECTIVE

To determine whether there was a relationship between coded diseases at the time of hospital discharge, a pattern of ordering investigations, and hospital readmission in a major teaching hospital.

DESIGN

Systematic review of data relating to emergency medical patients admitted to St James' Hospital Dublin between 1 January and 31 December 2002.

DATA SOURCES AND METHODS

Data on discharges from hospital recorded in the Hospital In-Patient Enquiry (HIPE) system. The value of HIPE data in describing the relationship between the pattern of resource utilisation, diagnostic related groups, and hospital readmission has not previously been examined.

RESULTS

Of 5038 episodes recorded among 4050 patients admitted, the number of readmissions was up to 15. Age and male gender were factors associated with readmission, and readmitted patients remained in hospital for longer. No particular test request predicted readmission, but computed tomography of the brain was associated with a reduced readmission rate. Discharge diagnostic related group coding at first discharge predicted readmission-codes related to heart failure, respiratory system, alcohol, malignancy, and anaemia.

CONCLUSIONS

It was found that clinical coding using the HIPE database strongly predicted hospital readmission. It may be argued that early hospital readmission reflects unsatisfactory patient care, alternatively that many readmissions are not preventable, representing either new events in elderly patients with chronic illnesses and frequent co-morbidity or related to social factors. The utility of specific interventions, in patients at high risk for hospital readmission, could be explored.

International Classification of Diseases-9th revision coding for preeclampsia: how accurate is it?

OBJECTIVE

The purpose of this study was to evaluate the accuracy of the International Classification of Diseases-9th revision codes for preeclampsia and eclampsia.

STUDY DESIGN

The University of Illinois Medical Center at Chicago discharge database was used to identify 135 women from 1999 through 2001 whose disease was coded as having preeclampsia or eclampsia. With American College of Obstetrics and Gynecology criteria as the gold standard, the diagnosis that was determined through chart review was compared with the International Classification of Diseases-9th revision code that was present in the discharge database. Patients were classified as true cases if the International Classification of Diseases-9th revision code matched the American College of Obstetricians and Gynecologists diagnosis; the positive predictive value of the code was then calculated.

RESULTS

The overall positive predictive value for the complete sample was only 54%, but the positive predictive value for severe preeclampsia was 84.8%, which was high compared with mild preeclampsia (45.3%) and eclampsia (41.7%). Diagnostic (clinician) error was the most common reason for miscoding error.

CONCLUSION

The findings suggest that International Classification of Diseases-9th revision codes for preeclampsia/eclampsia vary greatly in their accuracy of diagnosis. Therefore, a review of medical records is required when data are being gathered on the incidence of preeclampsia and eclampsia.