Coding of acute myocardial infarction. Clinical and policy implications

Administrative Data Use in National Registry Efforts: Blessing or Curse?

"Big data" refers to a growing field of large database research. Administrative data, a subset of big data, includes information from insurance claims, electronic medical records, and registries that can be useful for investigating novel research questions. While its use provides salient advantages, potential researchers relying on big data would benefit from knowing about how these databases are coded, common errors they may encounter, and how to best use large data to address various research questions. In the first section of this paper, Dr. Nicholas A. Bedard addresses the four major pitfalls to avoid with diagnosis and procedure codes in administrative data. In the next section, Dr. Jeffrey N. Katz et al. focus on the strengths and limitations of administrative data, suggesting methods to mitigate these limitations. Lastly, Dr. Elena Losina et al. review the uses and misuses of large databases for cost-effectiveness research, detailing methods for careful economic evaluations.

Sex Differences in 1-Year Rehospitalization for Heart Failure and Myocardial Infarction After Primary Percutaneous Coronary Intervention

It is unclear whether universal access to primary percutaneous coronary intervention (pPCI) may reduce sex differences in 1-year rehospitalization for heart failure (HF) and myocardial infarction (MI) after ST-elevation myocardial infarction (STEMI). We studied 7,597 consecutive STEMI patients (13.8% women, n = 1,045) who underwent pPCI from January 2007 to December 2013. Cox regression models adjusted for competing risk from death were used to assess sex differences in rehospitalization for HF and MI within 1 year from discharge. Compared with men, women were older (median age 67.6 vs 56.0 years, p < 0.001) with higher prevalence of co-morbidities and multivessel disease. Women had longer median door-to-balloon time (76 vs 66 minutes, p < 0.001) and were less likely to receive drug-eluting stents (19.5% vs 24.1%, p = 0.001). Of the medications prescribed at discharge, fewer women received aspirin (95.8% vs 97.6%, p = 0.002) and P2Y₁₂ antagonists (97.6% vs 98.5%, p = 0.039), but there were no significant sex differences in other discharge medications. After adjusting for differences in baseline characteristics and treatment, sex differences in risk of rehospitalization for HF attenuated (hazard ratio [HR] 1.05, 95% confidence interval [CI] 0.79 to 1.40), but persisted for MI (HR 1.68, 95% CI 1.22 to 2.33), with greater disparity in patients aged ≥60 years (HR 1.83, 95% CI 1.18 to 2.85) than those aged <60 years (HR 1.45, 95% CI 0.84 to 2.50). In conclusion, in a setting of universal access to pPCI, the adjusted risk of 1-year rehospitalization for HF was similar in both sexes, but women had significantly higher adjusted risk of 1-year rehospitalization for MI, especially older women.

Can hospital discharge data be used for monitoring indicators associated with postpartum hemorrhages? The HERA multicenter observational study

OBJECTIVES

The principal objective of this study was to assess the reliability of measuring the incidence of postpartum hemorrhages (PPH) from the national hospital discharge summary database (PMSI). The secondary objectives were to assess this reliability according to the maternity unit level and status and to assess the measurement of second-line procedures for PPH.

MATERIALS AND METHODS

This study compared PPH incidence rates from February through July 2011 in 131 maternity units, as measured in the PMSI and the prospective HERA study cohort, considered as the reference standard.

RESULTS

Compared with the cohort, PPH incidence was over-reported in the PMSI among vaginal deliveries (4.0% vs. 3.5; P<0.0001), but not cesareans (3.2 vs. 2.9%; P=0.1). For the second-line curative procedures, PMSI data underestimated the incidence of vessel embolization and transfusion (P<0.0001) among vaginal deliveries and of hypogastric ligation (P=0.002), other vessel ligation (P=0.005), and transfusion (P<0.0001) among cesareans.

CONCLUSION

Despite some coding inaccuracy in the PMSI, routinely collected data can provide acceptable estimates for maternity units and perinatal networks to use to improve quality of care through the monitoring of quality indicators. Improvements are nonetheless needed for international comparisons and other epidemiologic purposes.

Acute myocardial infarction hospital admissions and deaths in England: a national follow-back and follow-forward record-linkage study

BACKGROUND

Little information is available on how primary and comorbid acute myocardial infarction contribute to the mortality burden of acute myocardial infarction, the share of these deaths that occur during or after a hospital admission, and the reasons for hospital admission of those who died from acute myocardial infarction. Our aim was to fill in these gaps in the knowledge about deaths and hospital admissions due to acute myocardial infarction.

METHODS

We used individually linked national hospital admission and mortality data for England from 2006 to 2010 to identify all primary and comorbid diagnoses of acute myocardial infarction during hospital stay and their associated fatality rates (during or within 28 days of being in hospital). Data were obtained from the UK Small Area Health Statistics Unit and supplied by the Health and Social Care Information Centre (now NHS Digital) and the Office of National Statistics. We calculated event rates (reported as per 100 000 population for relevant age and sex groups) and case-fatality rate for primary acute myocardial infarction diagnosed during the first physician encounter or during subsequent encounters, and acute myocardial infarction diagnosed only as a comorbidity. We also calculated what proportion of deaths from acute myocardial infarction occurred in people who had been in hospital on or within the 28 days preceding death, and whether acute myocardial infarction was one of the recorded diagnoses in such admissions.

FINDINGS

Acute myocardial infarction was diagnosed in the first physician encounter in 307 496 (69%) of 446 744 admissions with a diagnosis of acute myocardial infarction, in the second or later physician encounter in 52 374 (12%) admissions, and recorded only as a comorbidity in 86 874 (19%) admissions. Patients with comorbid diagnoses of acute myocardial infarction had two to three times the case-fatality rate of patients in whom acute myocardial infarction was a primary diagnosis. 135 950 deaths were recorded as being caused by acute myocardial infarction as the underlying cause of death, of which 66 490 (49%) occurred in patients who were in hospital on the day of death or in the 28 days preceding death. AMI was the primary diagnosis in 32 695 (49%) of these 66 490 patients (27 678 [42%] diagnosed in the first physician encounter and 5017 [8%] in a second or subsequent encounter), was a comorbid diagnosis in 12 118 (18%), and was not mentioned at all in the remaining 21 677 (33%). The most common causes of admission in people who did not have an acute myocardial infarction diagnosis but went on to die of acute myocardial infarction as the underlying cause of death were other circulatory conditions (7566 [35%] of 21 677 deaths), symptomatic diagnoses including non-specific chest pain, dyspnoea and syncope (1368 [6%] deaths), and respiratory disorders (2662 [12%] deaths), mainly pneumonia and chronic obstructive airways disease.

INTERPRETATION

As many acute myocardial infarction deaths occurring within 28 days of being in hospital follow a non-acute myocardial infarction admission as follow an acute myocardial infarction admission. These people are often diagnosed with other circulatory disorders or symptoms of circulatory disturbance. Further investigation is needed to establish whether there are symptoms and information that can be used to predict the risk of a fatal acute myocardial infarction in such patients, which can contribute to reducing the mortality burden of acute myocardial infarction.

FUNDING

Wellcome Trust, Medical Research Council, Public Health England, National Institute for Health Research.

Can Hospital Discharge Diagnosis be used for Surveillance of Bacteremia? A Data Quality Study of a Danish Hospital Discharge Registry

Objective:

To assess the data quality of septicemia and sepsis registration in a hospital discharge registry in the County of Northern Jutland, Denmark.

Design:

Comparison of data from the discharge registry of an 880-bed, public, urban hospital in the County of Northern Jutland with data from a computerized bac-teremia database at the regional department of clinical microbiology.

Setting:

Urban hospital with approximately 45,000 admissions per year.

Patients:

The study included 406 episodes of bac-teremia in the bacteremia database and 83 discharges with the diagnosis of septicemia registered in the hospital discharge registry between January 1, 1994, and December 31, 1994.

Interventions:

None.

Results:

Eighteen episodes were registered in both the hospital discharge registry and the bacteremia database. Using the bacteremia database as reference standard, the sensitivity for the diagnosis of septicemia in the hospital discharge registry was 4.4% (18/406; 95% confidence intervals [CI95, 2.4%-6.4%]). By review of hospital records, we estimated the positive predictive value of septicemia registration in the hospital discharge registry as 21.7% (18/83; CI95, 12.8%-30.5%). No blood culture had been obtained in 44.4% (36/81; CI95, 33.6%-55.3%) of the cases with a discharge diagnosis of septicemia. In 33.3% (27/81; CI95, 23.1%-43.6%), the discharge diagnosis of septicemia was given, although blood cultures were negative.

Conclusions:

The hospital discharge registry revealed numerous misclassifications, and the system was found not suited for surveillance of, or research in, bacteremia at present

Association of the 2011 ACGME resident duty hour reforms with mortality and readmissions among hospitalized Medicare patients

IMPORTANCE

Patient outcomes associated with the 2011 Accreditation Council for Graduate Medical Education (ACGME) duty hour reforms have not been evaluated at a national level.

OBJECTIVE

To evaluate the association of the 2011 ACGME duty hour reforms with mortality and readmissions.

DESIGN, SETTING, AND PARTICIPANTS

Observational study of Medicare patient admissions (6,384,273 admissions from 2,790,356 patients) to short-term, acute care, nonfederal hospitals (n = 3104) with principal medical diagnoses of acute myocardial infarction, stroke, gastrointestinal bleeding, or congestive heart failure or a Diagnosis Related Group classification of general, orthopedic, or vascular surgery. Of the hospitals, 96 (3.1%) were very major teaching, 138 (4.4%) major teaching, 442 (14.2%) minor teaching, 443 (14.3%) very minor teaching, and 1985 (64.0%) nonteaching.

EXPOSURE

Resident-to-bed ratio as a continuous measure of hospital teaching intensity.

MAIN OUTCOMES AND MEASURES

Change in 30-day all-location mortality and 30-day all-cause readmission, comparing patients in more intensive relative to less intensive teaching hospitals before (July 1, 2009-June 30, 2011) and after (July 1, 2011-June 30, 2012) duty hour reforms, adjusting for patient comorbidities, time trends, and hospital site.

RESULTS

In the 2 years before duty hour reforms, there were 4,325,854 admissions with 288,422 deaths and 602,380 readmissions. In the first year after the reforms, accounting for teaching hospital intensity, there were 2,058,419 admissions with 133,547 deaths and 272,938 readmissions. There were no significant postreform differences in mortality accounting for teaching hospital intensity for combined medical conditions (odds ratio [OR], 1.00; 95% CI, 0.96-1.03), combined surgical categories (OR, 0.99; 95% CI, 0.94-1.04), or any of the individual medical conditions or surgical categories. There were no significant postreform differences in readmissions for combined medical conditions (OR, 1.00; 95% CI, 0.97-1.02) or combined surgical categories (OR, 1.00; 95% CI, 0.98-1.03). For the medical condition of stroke, there were higher odds of readmissions in the postreform period (OR, 1.06; 95% CI, 1.001-1.13). However, this finding was not supported by sensitivity analyses and there were no significant postreform differences for readmissions for any other individual medical condition or surgical category.

CONCLUSIONS AND RELEVANCE

Among Medicare beneficiaries, there were no significant differences in the change in 30-day mortality rates or 30-day all-cause readmission rates for those hospitalized in more intensive relative to less intensive teaching hospitals in the year after implementation of the 2011 ACGME duty hour reforms compared with those hospitalized in the 2 years before implementation.

Lessons learned from the design and implementation of myocardial infarction adjudication tailored for HIV clinical cohorts

We developed, implemented, and evaluated a myocardial infarction (MI) adjudication protocol for cohort research of human immunodeficiency virus. Potential events were identified through the centralized Centers for AIDS Research Network of Integrated Clinical Systems data repository using MI diagnoses and/or cardiac enzyme laboratory results (1995-2012). Sites assembled de-identified packets, including physician notes and results from electrocardiograms, procedures, and laboratory tests. Information pertaining to the specific antiretroviral medications used was redacted for blinded review. Two experts reviewed each packet, and a third review was conducted if discrepancies occurred. Reviewers categorized probable/definite MIs as primary or secondary and identified secondary causes of MIs. The positive predictive value and sensitivity for each identification/ascertainment method were calculated. Of the 1,119 potential events that were adjudicated, 294 (26%) were definite/probable MIs. Almost as many secondary (48%) as primary (52%) MIs occurred, often as the result of sepsis or cocaine use. Of the patients with adjudicated definite/probable MIs, 78% had elevated troponin concentrations (positive predictive value = 57%, 95% confidence interval: 52, 62); however, only 44% had clinical diagnoses of MI (positive predictive value = 45%, 95% confidence interval: 39, 51). We found that central adjudication is crucial and that clinical diagnoses alone are insufficient for ascertainment of MI. Over half of the events ultimately determined to be MIs were not identified by clinical diagnoses. Adjudication protocols used in traditional cardiovascular disease cohorts facilitate cross-cohort comparisons but do not address issues such as identifying secondary MIs that may be common in persons with human immunodeficiency virus.

A novel method for detecting inpatient pediatric asthma encounters using administrative data

Multiple methods for detecting asthma encounters are used today in public surveillance, quality reporting, and clinical research. Failure to detect asthma encounters can make it difficult to measure the scope and effectiveness of hospital or community-based interventions important in comparative effectiveness research and accountable care. Given the pairing of asthma with certain respiratory conditions, the objective of this study was to develop and test an asthma detection algorithm with specificity and sensitivity using 2 criteria: (1) principal discharge diagnosis and (2) asthma diagnosis code position. A medical record review was conducted (n=191) as the gold standard for identifying asthma encounters given objective criteria. The study team observed that for certain principal respiratory diagnoses (n=110), the observed odds ratio that encounters were for asthma when asthma was coded in the second or third code position was not significantly different than when asthma was coded as the principal diagnosis, 0.36 (P=0.42) and 0.18 (P=0.14), respectively. In contrast, the observed odds ratio was significantly different when asthma was coded in the fourth or fifth positions (P<.001). This difference remained after adjusting for covariates. Including encounters with asthma in 1 of the 3 first positions increased the detection sensitivity to 0.84 [95% confidence interval (CI): 0.76-0.92] while increasing the false positive rate to 0.19 [95% CI: 0.07-0.31]. Use of the proposed algorithm significantly improved the reporting accuracy [0.83 95%CI:0.76-0.90] over use of (1) the principal diagnosis alone [0.55 95% CI:0.46-0.64] or (2) all encounters with asthma 0.66 [95% CI:0.57-0.75]. Bed days resulting from asthma encounters increased 64% over use of the principal diagnosis alone. Given these findings, an algorithm using certain respiratory principal diagnoses and asthma diagnosis code position can reliably improve asthma encounter detection for population-based health impact measurement.

Teaching hospital five-year mortality trends in the wake of duty hour reforms

BACKGROUND

The Accreditation Council for Graduate Medical Education (ACGME) implemented duty hour regulations for residents in 2003 and again in 2011. While previous studies showed no systematic impacts in the first 2 years post-reform, the impact on mortality in subsequent years has not been examined.

OBJECTIVE

To determine whether duty hour regulations were associated with changes in mortality among Medicare patients in hospitals of different teaching intensity after the first 2 years post-reform.

DESIGN

Observational study using interrupted time series analysis with data from July 1, 2000 to June 30, 2008. Logistic regression was used to examine the change in mortality for patients in more versus less teaching-intensive hospitals before (2000-2003) and after (2003-2008) duty hour reform, adjusting for patient comorbidities, time trends, and hospital site.

PATIENTS

Medicare patients (n = 13,678,956) admitted to short-term acute care non-federal hospitals with principal diagnoses of acute myocardial infarction (AMI), gastrointestinal bleeding, or congestive heart failure (CHF); or a diagnosis-related group (DRG) classification of general, orthopedic, or vascular surgery.

MAIN MEASURE

All-location mortality within 30 days of hospital admission.

KEY RESULTS

In medical and surgical patients, there were no consistent changes in the odds of mortality at more vs. less teaching intensive hospitals in post-reform years 1-3. However, there were significant relative improvements in mortality for medical patients in the fourth and fifth years post-reform: Post4 (OR 0.88, 95 % CI [0.93-0.94]); Post5 (OR 0.87, [0.82-0.92]) and for surgical patients in the fifth year post-reform: Post5 (OR 0.91, [0.85-0.96]).

CONCLUSIONS

Duty hour reform was associated with no significant change in mortality in the early years after implementation, and with a trend toward improved mortality among medical patients in the fourth and fifth years. It is unclear whether improvements in outcomes long after implementation can be attributed to the reform, but concerns about worsening outcomes seem unfounded.

The difference between critical care initiation anion gap and prehospital admission anion gap is predictive of mortality in critical illness

OBJECTIVE

We hypothesized that the delta anion gap defined as difference between critical care initiation standard anion gap and prehospital admission standard anion gap is associated with all cause mortality in the critically ill.

DESIGN

Observational cohort study.

SETTING

Two hundred nine medical and surgical intensive care beds in two hospitals in Boston, MA.

PATIENTS

Eighteen thousand nine hundred eighty-five patients, age ≥18 yrs, who received critical care between 1997 and 2007.

MEASUREMENTS

The exposure of interest was delta anion gap and categorized a priori as <0, 0-5, 5-10, and >10 mEq/L. Logistic regression examined death by days 30, 90, and 365 postcritical care initiation and in-hospital mortality. Adjusted odds ratios were estimated by multivariable logistic regression models. The discrimination of delta anion gap for 30-day mortality was evaluated using receiver operator characteristic curves performed for a subset of patients with all laboratory data required to analyze the data via physical chemical principles (n = 664).

INTERVENTIONS

None.

RESULTS

Delta anion gap was a particularly strong predictor of 30-day mortality with a significant risk gradient across delta anion gap quartiles following multivariable adjustment: delta anion gap <0 mEq/L odds ratio 0.75 (95% confidence interval 0.67-0.81; p < 0.0001); delta anion gap 5-10 mEq/L odds ratio 1.56 (95% confidence interval 1.35-1.81; p < 0.0001); delta anion gap >10 mEq/L odds ratio 2.18 (95% confidence interval 1.76-2.71; p < 0.0001); and all relative to patients with delta anion gap 0-5 mEq/L. Similar significant robust associations post multivariable adjustments are seen with death by days 90 and 365 as well as in-hospital mortality. Correcting for albumin or limiting the cohort to patients with standard anion gap at critical care initiation of 10-18 mEq/L did not materially change the delta anion gap-mortality association. Delta anion gap has similarly moderate discriminative ability for 30-day mortality in comparison to standard base excess and strong ion gap.

CONCLUSION

An increase in standard anion gap at critical care initiation relative to prehospital admission standard anion gap is a predictor of the risk of all cause patient mortality in the critically ill.

Case definitions for acute myocardial infarction in administrative databases and their impact on in-hospital mortality rates

OBJECTIVE

To identify validated ICD-9-CM/ICD-10 coded case definitions for acute myocardial infarction (AMI).

DATA SOURCES

Ovid Medline (1950-2010) was searched to identify studies that validated acute myocardial infarction (AMI) case definitions. Hospital discharge abstract data and chart data were linked to validate identified AMI definitions.

STUDY DESIGN

Systematic literature review, chart review, and administrative data analysis.

DATA COLLECTION/EXTRACTION METHODS

Data on sensitivity/specificity/positive and negative predictive values (PPV and NPV) were extracted from previous studies to identify validated case definitions for AMI. These case definitions were validated in administrative data through chart review and applied to hospital discharge data to assess in-hospital mortality.

PRINCIPAL FINDINGS

Of the eight ICD-9-CM definitions validated in the literature, use of ICD-9-CM code 410 to define AMI had the highest sensitivity (94 percent) and specificity (99 percent). In our data, ICD-9-CM/ICD-10 codes 410/I21-I22 in all available coding fields had high sensitivity (83.3 percent/82.8 percent) and PPV (82.8 percent/82.2 percent). The in-hospital mortality among AMI patients identified using this case definition was 7.6 percent in ICD-9-CM data and 6.6 percent in ICD-10 data.

CONCLUSIONS

We recommend that ICD-9-CM 410 or ICD-10 I21-I22 in the primary diagnosis coding field should be used to define AMI. The use of a consistent validated case definition would improve comparability across studies.

Implementation of ICD-10 in Canada: how has it impacted coded hospital discharge data?

Background

The purpose of this study was to assess whether or not the change in coding classification had an impact on diagnosis and comorbidity coding in hospital discharge data across Canadian provinces.

Methods

This study examined eight years (fiscal years 1998 to 2005) of hospital records from the Hospital Person-Oriented Information database (HPOI) derived from the Canadian national Discharge Abstract Database. The average number of coded diagnoses per hospital visit was examined from 1998 to 2005 for provinces that switched from International Classifications of Disease 9^th version (ICD-9-CM) to ICD-10-CA during this period. The average numbers of type 2 and 3 diagnoses were also described. The prevalence of the Charlson comorbidities and distribution of the Charlson score one year before and one year after ICD-10 implementation for each of the 9 provinces was examined. The prevalence of at least one of the seventeen Charlson comorbidities one year before and one year after ICD-10 implementation were described by hospital characteristics (teaching/non-teaching, urban/rural, volume of patients).

Results

Nine Canadian provinces switched from ICD-9-CM to ICD-I0-CA over a 6 year period starting in 2001. The average number of diagnoses coded per hospital visit for all code types over the study period was 2.58. After implementation of ICD-10-CA a decrease in the number of diagnoses coded was found in four provinces whereas the number of diagnoses coded in the other five provinces remained similar. The prevalence of at least one of the seventeen Charlson conditions remained relatively stable after ICD-10 was implemented, as did the distribution of the Charlson score. When stratified by hospital characteristics, the prevalence of at least one Charlson condition decreased after ICD-10-CA implementation, particularly for low volume hospitals.

Conclusion

In conclusion, implementation of ICD-10-CA in Canadian provinces did not substantially change coding practices, but there was some coding variation in the average number of diagnoses per hospital visit across provinces.

Low serum 25-hydroxyvitamin D at critical care initiation is associated with increased mortality

OBJECTIVE

We hypothesized that deficiency in 25-hydroxyvitamin D at critical care initiation would be associated with all-cause mortalities.

DESIGN

Two-center observational study.

SETTING

Two teaching hospitals in Boston, MA.

PATIENTS

The study included 1,325 patients, age ≥ 18 yrs, in whom 25-hydroxyvitamin D was measured 7 days before or after critical care initiation between 1998 and 2009.

MEASUREMENTS

25-hydroxyvitamin D was categorized as deficiency in 25-hydroxyvitamin D (≤ 15 ng/mL), insufficiency (16-29 ng/mL), and sufficiency (≥ 30 ng/mL). Logistic regression examined death by days 30, 90, and 365 postcritical care initiation and in-hospital mortality. Adjusted odds ratios were estimated by multivariable logistic regression models.

INTERVENTIONS

None.

KEY RESULTS

25-hydroxyvitamin D deficiency is predictive for short-term and long-term mortality. Thirty days following critical care initiation, patients with 25-hydroxyvitamin D deficiency have an odds ratio for mortality of 1.85 (95% confidence interval 1.15-2.98; p = .01) relative to patients with 25-hydroxyvitamin D sufficiency. 25-hydroxyvitamin D deficiency remains a significant predictor of mortality at 30 days following critical care initiation following multivariable adjustment for age, gender, race, Deyo-Charlson index, sepsis, season, and surgical vs. medical patient type (adjusted odds ratio 1.94; 95% confidence interval 1.18-3.20; p = .01). Results were similarly significant at 90 and 365 days following critical care initiation and for in-hospital mortality. The association between vitamin D and mortality was not modified by sepsis, race, or neighborhood poverty rate, a proxy for socioeconomic status.

CONCLUSION

Deficiency of 25-hydroxyvitamin D at the time of critical care initiation is a significant predictor of all-cause patient mortality in a critically ill patient population.

Validity of the diagnosis of acute myocardial infarction in korean national medical health insurance claims data: the korean heart study (1)

Background and Objectives

Medical insurance claims (MIC) data are one of the largest sources of outcome data in the form of International Classification of Diseases (ICD) codes. We evaluated the validity of the ICD codes from the Korean National MIC data with respect to the outcomes from acute myocardial infarction (AMI) in the Korean Heart Study.

Subjects and Methods

Baseline information was obtained from health examinations conducted from 1994 to 2001. Outcome information regarding the incidence of AMI came from hospital admission discharge records from 1994 to 2007. Structured questionnaires were sent to 98 hospitals. In total, 107 cases of AMI with ICD codes of I21- (93 men, 26-73 years of age) were included in the final analyses. ICD code accuracy and reliability (kappa) for AMI were calculated.

Results

A large number of AMI cases were from hospitals located in the Seoul area (75.9%). The accuracy of AMI was 71.4%, according to World Health Organization criteria (1997-2000, n=24, kappa=0.46) and 73.1% according to the European Society of Cardiology/American College of Cardiology (ESC/ACC) criteria (2001-2007, n=83, kappa=0.74). An age of 50 years or older was the only factor related to inaccuracy of codes for AMI (odds ratio, 4.6; 95% confidence interval, 1.2-17.7) in patients diagnosed since January 2001 using ESC/ACC criteria (n=83).

Conclusion

The accuracy for diagnosing AMI using the ICD-10 codes in Korean MIC data was >70%, and reliability was fair to good; however, more attention is required for recoding ICD codes in older patients.

Red cell distribution width and all-cause mortality in critically ill patients

OBJECTIVE

Red cell distribution width is a predictor of mortality in the general population. The prevalence of increased red cell distribution width and its significance in the intensive care unit are unknown. The objective of this study was to investigate the association between red cell distribution width at the initiation of critical care and all cause mortality.

DESIGN

Multicenter observational study.

SETTING

Two tertiary academic hospitals in Boston, MA.

PATIENTS

A total of 51,413 patients, aged ≥ 18 yrs, who received critical care between 1997 and 2007.

INTERVENTIONS

None.

MEASUREMENTS AND MAIN RESULTS

The exposure of interest was red cell distribution width as a predictor of mortality in the general population. The prevalence of increased red cell distribution width and its significance in the intensive care unit are unknown and categorized a priori in quintiles as ≤ 13.3%, 13.3% to 14.0%, 14.0% to 14.7%, 14.7% to 15.8%, and >15.8%. Logistic regression examined death by days 30, 90, and 365 postcritical care initiation, inhospital mortality, and bloodstream infection. Adjusted odds ratios were estimated by multivariable logistic regression models. Adjustment included age, sex, race, Deyo-Charlson index, coronary artery bypass grafting, myocardial infarction, congestive heart failure, hematocrit, white blood cell count, mean corpuscular volume, blood urea nitrogen, red blood cell transfusion, sepsis, and creatinine. Red cell distribution width was a particularly strong predictor of all-cause mortality 30 days after critical care initiation with a significant risk gradient across red cell distribution width quintiles after multivariable adjustment: red cell distribution width 13.3% to 14.0% (odds ratio [OR], 1.19; 95% confidence interval [CI], 1.08-1.30; p <.001); red cell distribution width 14.0% to 14.7% (OR, 1.28; 95% CI, 1.16-1.42; p <.001); red cell distribution width 14.7% to 15.8% (OR, 1.69; 95% CI, 1.52-1.86; p <.001); red cell distribution width >15.8% (OR, 2.61; 95% CI, 2.37-2.86; p <.001), all relative to patients with red cell distribution width ≤ 13.3%. Similar significant robust associations postmultivariable adjustments are seen with death by days 90 and 365 postcritical care initiation as well as inhospital mortality. In a subanalysis of patients with blood cultures drawn (n = 18,525), red cell distribution width at critical care initiation was associated with the risk of bloodstream infection and remained significant after multivariable adjustment. The adjusted risk of bloodstream infection was 1.40- and 1.44-fold higher in patients with red cell distribution width values in the 14.7% to 15.8% and >15.8% quintiles, respectively, compared with those with red cell distribution width ≤ 13.3%. Estimating the receiver operating characteristic area under the curve shows that red cell distribution width has moderate discriminative power for 30-day mortality (area under the curve = 0.67).

CONCLUSION

Red cell distribution width is a robust predictor of the risk of all-cause patient mortality and bloodstream infection in the critically ill. Red cell distribution width is commonly measured, inexpensive, and widely available and may reflect overall inflammation, oxidative stress, or arterial underfilling in the critically ill.

The accuracy of present-on-admission reporting in administrative data

OBJECTIVE

To test the accuracy of reporting present-on-admission (POA) and to assess whether POA reporting accuracy differs by hospital characteristics.

DATA SOURCES

We performed an audit of POA reporting of secondary diagnoses in 1,059 medical records from 48 California hospitals.

STUDY DESIGN

We used patient discharge data (PDD) to select records with secondary diagnoses that are powerful predictors of mortality and could potentially represent comorbidities or complications among patients who either had a primary procedure of a percutaneous transluminal coronary angioplasty or a primary diagnosis of acute myocardial infarction, community-acquired pneumonia, or congestive heart failure. We modeled the relationship between secondary diagnoses POA reporting accuracy (over-reporting and under-reporting) and hospital characteristics.

DATA COLLECTION

We created a gold standard from blind reabstraction of the medical records and compared the accuracy of the PDD against the gold standard.

PRINCIPAL FINDINGS

The PDD and gold standard agreed on POA reporting in 74.3 percent of records, with 13.7 percent over-reporting and 11.9 percent under-reporting. For-profit hospitals tended to overcode secondary diagnoses as present on admission (odds ratios [OR] 1.96; 95 percent confidence interval [CI] 1.11, 3.44), whereas teaching hospitals tended to undercode secondary diagnoses as present on admission (OR 2.61; 95 percent CI 1.36, 5.03).

CONCLUSIONS

POA reporting of secondary diagnoses is moderately accurate but varies by hospitals. Steps should be taken to improve POA reporting accuracy before using POA in hospital assessments tied to payments.

Association of low serum 25-hydroxyvitamin D levels and mortality in the critically ill

OBJECTIVE

We hypothesized that deficiency in 25-hydroxyvitamin D before hospital admission would be associated with all-cause mortality in the critically ill.

DESIGN

Multicenter observational study of patients treated in medical and surgical intensive care units.

SETTING

A total of 209 medical and surgical intensive care beds in two teaching hospitals in Boston, MA.

PATIENTS

A total of 2399 patients, age ≥ 18 yrs, in whom 25-hydroxyvitamin D was measured before hospitalization between 1998 and 2009.

INTERVENTIONS

None.

MEASUREMENTS AND MAIN RESULTS

Preadmission 25-hydroxyvitamin D was categorized as deficiency in 25-hydroxyvitamin D (≤ 15 ng/mL), insufficiency (16-29 ng/mL), and sufficiency (≥ 30 ng/mL). Logistic regression examined death by days 30, 90, and 365 post-intensive care unit admission, in-hospital mortality, and blood culture positivity. Adjusted odds ratios were estimated by multivariable logistic regression models. Preadmission 25-hydroxyvitamin D deficiency is predictive for short-term and long-term mortality. At 30 days following intensive care unit admission, patients with 25-hydroxyvitamin D deficiency have an odds ratio for mortality of 1.69 (95% confidence interval of 1.28-2.23, p < .0001) relative to patients with 25-hydroxyvitamin D sufficiency. 25-Hydroxyvitamin D deficiency remains a significant predictor of mortality at 30 days following intensive care unit admission following multivariable adjustment (adjusted odds ratio of 1.69, 95% confidence interval of 1.26-2.26, p < .0001). At 30 days following intensive care unit admission, patients with 25-hydroxyvitamin D insufficiency have an odds ratio of 1.32 (95% confidence interval of 1.02-1.72, p = .036) and an adjusted odds ratio of 1.36 (95% confidence interval of 1.03-1.79, p = .029) relative to patients with 25-hydroxyvitamin D sufficiency. Results were similar at 90 and 365 days following intensive care unit admission and for in-hospital mortality. In a subgroup analysis of patients who had blood cultures drawn (n = 1160), 25-hydroxyvitamin D deficiency was associated with increased risk of blood culture positivity. Patients with 25-hydroxyvitamin D insufficiency have an odds ratio for blood culture positivity of 1.64 (95% confidence interval of 1.05-2.55, p = .03) relative to patients with 25-hydroxyvitamin D sufficiency, which remains significant following multivariable adjustment (odds ratio of 1.58, 95% confidence interval of 1.01-2.49, p = .048).

CONCLUSION

Deficiency of 25-hydroxyvitamin D before hospital admission is a significant predictor of short- and long-term all-cause patient mortality and blood culture positivity in a critically ill patient population.

Elevation of blood urea nitrogen is predictive of long-term mortality in critically ill patients independent of "normal" creatinine

OBJECTIVE

We hypothesized that elevated blood urea nitrogen can be associated with all-cause mortality independent of creatinine in a heterogeneous critically ill population.

DESIGN

Multicenter observational study of patients treated in medical and surgical intensive care units.

SETTING

Twenty intensive care units in two teaching hospitals in Boston, MA.

PATIENTS

A total of 26,288 patients, age ≥ 18 yrs, hospitalized between 1997 and 2007 with creatinine of 0.80-1.30 mg/dL.

INTERVENTIONS

None.

MEASUREMENTS

Blood urea nitrogen at intensive care unit admission was categorized as 10-20, 20-40, and >40 mg/dL. Logistic regression examined death at days 30, 90, and 365 after intensive care unit admission as well as in-hospital mortality. Adjusted odds ratios were estimated by multivariable logistic regression models.

MAIN RESULTS

Blood urea nitrogen at intensive care unit admission was predictive for short- and long-term mortality independent of creatinine. Thirty days following intensive care unit admission, patients with blood urea nitrogen of >40 mg/dL had an odds ratio for mortality of 5.12 (95% confidence interval, 4.30-6.09; p < .0001) relative to patients with blood urea nitrogen of 10-20 mg/dL. Blood urea nitrogen remained a significant predictor of mortality at 30 days after intensive care unit admission following multivariable adjustment for confounders; patients with blood urea nitrogen of >40 mg/dL had an odds ratio for mortality of 2.78 (95% confidence interval, 2.27-3.39; p < .0001) relative to patients with blood urea nitrogen of 10-20 mg/dL. Thirty days following intensive care unit admission, patients with blood urea nitrogen of 20-40 mg/dL had an odds ratio of 2.15 (95% confidence interval, 1.98-2.33; <.0001) and a multivariable odds ratio of 1.53 (95% confidence interval, 1.40-1.68; p < .0001) relative to patients with blood urea nitrogen of 10-20 mg/dL. Results were similar at 90 and 365 days following intensive care unit admission as well as for in-hospital mortality. A subanalysis of patients with blood cultures (n = 7,482) demonstrated that blood urea nitrogen at intensive care unit admission was associated with the risk of blood culture positivity.

CONCLUSION

Among critically ill patients with creatinine of 0.8-1.3 mg/dL, an elevated blood urea nitrogen was associated with increased mortality, independent of serum creatinine.

Do coder characteristics influence validity of ICD-10 hospital discharge data?

Background

Administrative data are widely used to study health systems and make important health policy decisions. Yet little is known about the influence of coder characteristics on administrative data validity in these studies. Our goal was to describe the relationship between several measures of validity in coded hospital discharge data and 1) coders' volume of coding (≥13,000 vs. <13,000 records), 2) coders' employment status (full- vs. part-time), and 3) hospital type.

Methods

This descriptive study examined 6 indicators of face validity in ICD-10 coded discharge records from 4 hospitals in Calgary, Canada between April 2002 and March 2007. Specifically, mean number of coded diagnoses, procedures, complications, Z-codes, and codes ending in 8 or 9 were compared by coding volume and employment status, as well as hospital type. The mean number of diagnoses was also compared across coder characteristics for 6 major conditions of varying complexity. Next, kappa statistics were computed to assess agreement between discharge data and linked chart data reabstracted by nursing chart reviewers. Kappas were compared across coder characteristics.

Results

422,618 discharge records were coded by 59 coders during the study period. The mean number of diagnoses per record decreased from 5.2 in 2002/2003 to 3.9 in 2006/2007, while the number of records coded annually increased from 69,613 to 102,842. Coders at the tertiary hospital coded the most diagnoses (5.0 compared with 3.9 and 3.8 at other sites). There was no variation by coder or site characteristics for any other face validity indicator. The mean number of diagnoses increased from 1.5 to 7.9 with increasing complexity of the major diagnosis, but did not vary with coder characteristics. Agreement (kappa) between coded data and chart review did not show any consistent pattern with respect to coder characteristics.

Conclusions

This large study suggests that coder characteristics do not influence the validity of hospital discharge data. Other jurisdictions might benefit from implementing similar employment programs to ours, e.g.: a requirement for a 2-year college training program, a single management structure across sites, and rotation of coders between sites. Limitations include few coder characteristics available for study due to privacy concerns.

An algorithm to identify incident myocardial infarction using Medicaid data

PURPOSE

Studies of non-steroidal anti-inflammatory drugs (NSAIDs) and cardiovascular events using administrative data require identification of incident acute myocardial infarctions (AMIs) and information on whether confounders differ by NSAID status.

METHODS

We identified patients with a first AMI hospitalization from Tennessee Medicaid files as those with primary ICD-9 discharge diagnosis 410.x and hospitalization stay of > 2 calendar days. Eligible persons were non-institutionalized, aged 50-84 years between 1999-2004, had continuous enrollment and no AMI, stroke, or non-cardiovascular serious medical illness in the prior year. Of 5524 patients with a potential first AMI, a systematic sample (n = 350) was selected for review. Using defined criteria, we classified events using chest pain history, EKG, and cardiac enzymes, and calculated the positive predictive value (PPV) for definite or probable AMI.

RESULTS

337 of 350 (96.3%) charts were abstracted and 307 (91.1%), 6 (1.8%), and 24 (7.1%) events were categorized as definite, probable, and no AMI, respectively. PPV for any definite or probable AMI was 92.8% (95% CI 89.6-95.2); for an AMI without an event in the past year 91.7% (95% CI 88.3-94.2), and for an incident AMI was 72.7% (95% CI 67.7-77.2). Age-adjusted prevalence of current smoking (46.4% vs. 39.1%, p = 0.35) and aspirin use (36.9% vs. 35.9%, p = 0.90) was similar among NSAID users and non-users

CONCLUSIONS

ICD-9 code 410.x had high predictive value for identifying AMI. Among those with AMI, smoking and aspirin use was similar in NSAID exposure groups, suggesting these factors will not confound the relationship between NSAIDs and cardiovascular outcomes.

Pay for performance: quality- and value-based reimbursement

There is urgent need to reform health care reimbursement models, including physician compensation, to address high health care costs, despite numerous quality initiatives. Pay for performance (P4P) is a model that attempts to align financial incentives with better outcomes and value rather than the current system of rewarding volume and intensity of care delivered. P4P has been implemented in other countries besides the United States and is perhaps most advanced in the United Kingdom. Measurement for P4P is evolving, as are the types of incentives; neither is perfect at this time. For P4P to succeed, all health care stakeholders will need to collaborate.

Assessing validity of ICD-9-CM and ICD-10 administrative data in recording clinical conditions in a unique dually coded database

OBJECTIVE

The goal of this study was to assess the validity of the International Classification of Disease, 10th Version (ICD-10) administrative hospital discharge data and to determine whether there were improvements in the validity of coding for clinical conditions compared with ICD-9 Clinical Modification (ICD-9-CM) data.

METHODS

We reviewed 4,008 randomly selected charts for patients admitted from January 1 to June 30, 2003 at four teaching hospitals in Alberta, Canada to determine the presence or absence of 32 clinical conditions and to assess the agreement between ICD-10 data and chart data. We then re-coded the same charts using ICD-9-CM and determined the agreement between the ICD-9-CM data and chart data for recording those same conditions. The accuracy of ICD-10 data relative to chart data was compared with the accuracy of ICD-9-CM data relative to chart data.

RESULTS

Sensitivity values ranged from 9.3 to 83.1 percent for ICD-9-CM and from 12.7 to 80.8 percent for ICD-10 data. Positive predictive values ranged from 23.1 to 100 percent for ICD-9-CM and from 32.0 to 100 percent for ICD-10 data. Specificity and negative predictive values were consistently high for both ICD-9-CM and ICD-10 databases. Of the 32 conditions assessed, ICD-10 data had significantly higher sensitivity for one condition and lower sensitivity for seven conditions relative to ICD-9-CM data. The two databases had similar sensitivity values for the remaining 24 conditions.

CONCLUSIONS

The validity of ICD-9-CM and ICD-10 administrative data in recording clinical conditions was generally similar though validity differed between coding versions for some conditions. The implementation of ICD-10 coding has not significantly improved the quality of administrative data relative to ICD-9-CM. Future assessments like this one are needed because the validity of ICD-10 data may get better as coders gain experience with the new coding system.

Accuracy of data collected by surgical residents

PURPOSE

Clinician records are the primary information source for assessing the quality of facial injury care, billing, risk management, planning of health services, and health-system management and reporting. Inaccuracies obscure outcomes assessment and affect the planning of health services. We sought to determine the accuracy of the clinician collected data by comparing them to similar information elicited by professional interviewers.

MATERIALS AND METHODS

We abstracted admissions data from the medical records of 185 patients treated for orofacial injury between January 2005 and January 2007. Clinician data on sociodemographics and substance use were compared with similar information elicited by trained research staff as part of a prospective study.

RESULTS

The accuracy of the clinician data sets varied considerably depending on the variable. Concordance with the interviewer data sets was highest for age (paired t test P = .09), gender (kappa = 1), and ethnicity (kappa = .84) but dropped off considerably for marital status (kappa = .22) and alcohol (kappa = .18) and drug use (kappa = .16). The missing data per variable ranged from 4.5% (gender) to 46.9% (employment and education).

CONCLUSIONS

Although more research is needed to evaluate the cause of inaccuracies and the relative contributions of patient, provider, and system level effects, it seems that significant inaccuracies in administrative data are common. In particular, patient information collected by surgical residents under-reports substance use behaviors. Interventions aimed at identifying the sources and correcting these errors are necessary.

Feasibility of determining myocardial infarction type from medical record review

BACKGROUND

Hospital discharge data are used extensively in health research. Given the clinical differences between ST segment elevation myocardial infarction (STEMI) and non-ST segment elevation myocardial infarction (NSTEMI), it is important that these entities be distinguishable in a medical record. The authors sought to determine the extent to which the type of MI is recorded in medical records, as well as the consistency of this designation within individual records.

METHODS

Records of all MI patients admitted to a tertiary care centre in Canada from April 1, 2000, to March 31, 2001, were reviewed. Documentation and consistency of the use of the terms STEMI (Q wave, ST elevation or transmural MI) or NSTEMI (non-Q wave, subendocardial or nontransmural MI) were assessed in the admission history, progress notes, coronary care unit summary and discharge summary sections of each record.

RESULTS

Missing data were common; each chart section mentioned MI type in fewer than one-half of charts. When information was combined, it was possible to determine the type of MI in 81.1% of cases. MI type was consistently described as STEMI in 48.7% of cases, and as NSTEMI in 32.4%. Of concern, MI type was discrepant across sections in 10.5% of cases and missing entirely in 8.4% of cases.

CONCLUSIONS

The designation of MI cases as STEMI or NSTEMI is both incomplete and inconsistent in hospital records. This has implications for health services research conducted retrospectively using medical record data, because it is difficult to comprehensively study processes and outcomes of MI care if the type cannot be retrospectively determined.

Comparison of methodologies for calculating quality measures based on administrative data versus clinical data from an electronic health record system: implications for performance measures

New reimbursement policies and pay-for-performance programs to reward providers for producing better outcomes are proliferating. Although electronic health record (EHR) systems could provide essential clinical data upon which to base quality measures, most metrics in use were derived from administrative claims data. We compared commonly used quality measures calculated from administrative data to those derived from clinical data in an EHR based on a random sample of 125 charts of Medicare patients with diabetes. Using standard definitions based on administrative data (which require two visits with an encounter diagnosis of diabetes during the measurement period), only 75% of diabetics determined by manually reviewing the EHR (the gold standard) were identified. In contrast, 97% of diabetics were identified using coded information in the EHR. The discrepancies in identified patients resulted in statistically significant differences in the quality measures for frequency of HbA1c testing, control of blood pressure, frequency of testing for urine protein, and frequency of eye exams for diabetic patients. New development of standardized quality measures should shift from claims-based measures to clinically based measures that can be derived from coded information in an EHR. Using data from EHRs will also leverage their clinical content without adding burden to the care process.

Validity of International Classification of Diseases, Ninth Revision, Clinical Modification Codes for Acute Renal Failure

Administrative and claims databases may be useful for the study of acute renal failure (ARF) and ARF that requires dialysis (ARF-D), but the validity of the corresponding diagnosis and procedure codes is unknown. The performance characteristics of International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes for ARF were assessed against serum creatinine-based definitions of ARF in 97,705 adult discharges from three Boston hospitals in 2004. For ARF-D, ICD-9-CM codes were compared with review of medical records in 150 patients with ARF-D and 150 control patients. As compared with a diagnostic standard of a 100% change in serum creatinine, ICD-9-CM codes for ARF had a sensitivity of 35.4%, specificity of 97.7%, positive predictive value of 47.9%, and negative predictive value of 96.1%. As compared with review of medical records, ICD-9-CM codes for ARF-D had positive predictive value of 94.0% and negative predictive value of 90.0%. It is concluded that administrative databases may be a powerful tool for the study of ARF, although the low sensitivity of ARF codes is an important caveat. The excellent performance characteristics of ICD-9-CM codes for ARF-D suggest that administrative data sets may be particularly well suited for research endeavors that involve patients with ARF-D.

Peripartum myocardial ischemia: a review of Canadian deliveries from 1970 to 1998

OBJECTIVE

The purpose of this study was to evaluate the incidence of peripartum myocardial ischemia in Canada.

STUDY DESIGN

We identified the cohort of women who were admitted to Canadian hospitals for delivery between 1970 and 1998 to calculate the incidence rate and to evaluate potential risk factors.

RESULTS

One hundred fourteen of 10,032,375 women delivered in Canadian hospitals between 1970 and 1998 had peripartum myocardial ischemia recorded as a discharge diagnosis. The overall crude incidence rate was 1.1 (95% confidence interval 0.93, 1.37) women with peripartum myocardial ischemia per 100,000 women delivering per year as noted in the Canadian Hospital Morbidity database. Rates did not increase over time but increased with maternal age. Identified risk factors were diabetes mellitus, hyperlipidemia, and chronic heart disease. The case fatality rate among women with the disease was 1.8%.

CONCLUSION

The incidence of peripartum myocardial ischemia did not increase between 1970 and 1998 in Canada, despite an aging cohort with more prevalent medical comorbidities. Maternal mortality from this event is lower than previously described.

Triage and diagnosis of chest pain in rural hospitals: implementation of the ACI-TIPI in the High Plains Research Network

PURPOSE

The Acute Cardiac Ischemia Time-Insensitive Predictive Instrument (ACI-TIPI) has been shown to improve diagnostic accuracy of acute cardiac ischemia (ACI) and decrease coronary care unit admissions in urban emergency departments. The purpose of this study was to determine the impact of the ACI-TIPI on triage and diagnosis of patients with chest pain in rural hospitals.

METHODS

We undertook a controlled trial of the impact ACI-TIPI use in the High Plains Research Network (HPRN). Main outcome measures were the triage of patients in emergency departments (admission, transfer, or discharge home) and diagnostic accuracy.

RESULTS

There were 1,861 patients seen during a 10-month period. Forty-five percent of all patients complaining of chest pain were discharged home from the emergency department. Eight percent were transferred from the emergency department, and another 10% were transferred later after admission. Among patients with acute myocardial infarction or unstable angina, 22.2% were transferred directly from the emergency department and only 3% were discharged home when ACI-TIPI was available, compared with 18.7% transferred and 5.2% discharged home when not available (P = .4). Diagnostic accuracy was high and not statistically different with the addition of the ACI-TIPI score (86.8% ACI-TIPI off vs 89.0% ACI-TIPI on, P = .15),

CONCLUSIONS

Physicians in the HPRN provided appropriate diagnosis and triage to patients with chest pain. Routine addition of the ACI-TIPI score did not improve diagnostic accuracy or significantly change triage. Further research on ACI-TIPI in rural hospitals is necessary before recommending routine use of the ACI-TIPI.

Depressive symptoms have an independent, gradient risk for coronary heart disease incidence in a random, population-based sample

PURPOSE

Depression is a risk factor for incident coronary heart disease (CHD), and predicts poor prognosis for patients post-myocardial infarction (MI). Few population-based, prospective studies have tested a gradient risk for depressive symptoms on CHD incidence.

METHODS

The sample (n=1302) was derived from the Nova Scotia Health Survey-1995 (NSHS95), an age- and sex-stratified, random, population-based health survey. All subjects were 45 years or older, free of overt CHD at baseline, and completed the Center for Epidemiological Studies-Depression (CES-D) scale. Covariates included age, sex, body mass index, physical activity level, family history of premature CHD, diastolic blood pressure, lipids, smoking, alcohol use, diabetes, and education level. For the 4 years following NSHS95, MI-related hospitalizations (ICD-9-CM code 410) and CHD-related deaths (ICD-9-CM codes 410-414) were extracted from the provincial, universal healthcare registry.

RESULTS

Fifty-two participants experienced a CHD event. A one standard-deviation increase in CES-D score was associated with a 1.32 hazard risk (confidence interval, 1.01-1.71) of CHD events, controlling for established CHD risk factors.

CONCLUSIONS

An independent, gradient association between depression and incident CHD was detected in a population-based sample with complete 4-year CHD data. This evidence supports the value of investigating mechanisms linking depression and CHD.

Pursuing integration of performance measures into electronic medical records: beta-adrenergic receptor antagonist medications

OBJECTIVE

Electronic medical records seldom integrate performance indicators into daily operations. Assessing quality indicators traditionally requires resource intensive chart reviews of small samples. We sought to use an electronic medical record to assess use of beta-adrenergic antagonist medications (beta-blockers) following myocardial infarction, to compare a standardized manual assessment with assessment using electronic medical records, and to discuss potential for future integration of performance indicators into electronic records.

DESIGN

Cross-sectional data analysis.

SETTING

An urban academic medical center.

PARTICIPANTS

US Medicare beneficiaries 65 years of age or older, admitted to hospital with myocardial infarction between 1995 and 1999.

MEASUREMENTS AND MAIN RESULTS

Manual chart review was compared with a computer driven assessment of electronic records. Administration of beta-blockers and cases excluded from use of beta-blockers were measured, based on Medicare criteria. Among 4490 older adults, 391 (4%) of 9018 hospital admissions contained codes for myocardial infarction. In 323 (83%) of the 391 hospital admissions, criteria for excluding beta-blockers were met; 235 (60%) were excluded due to heart failure. Of 68 hospital admissions for myocardial infarction that did not meet exclusion criteria, physicians prescribed beta-blockers in 49 (72%) on admission and 42 (62%) at discharge. Compared with manual chart review, electronic review had a sensitivity of 83-100% and led to fewer false negative findings.

CONCLUSIONS

An electronic medical records system can be used instead of chart review to measure use of beta-blockers after myocardial infarction. This should lead to integration of real time automated performance measurement into electronic medical records.

Comparison of a commercially available clinical information system with other methods of measuring critical care outcomes data

PURPOSE

To compare the quality of data recorded by a commercially available clinical information system (CIS) to other commonly used methods for obtaining large amounts of patient data.

MATERIALS AND METHODS

Five sets of clinical patient data were chosen as a cross-section of all the data collected by a CIS in our intensive care unit (ICU): 1) Length of stay in the ICU, 2) Vital signs, 3) Days of mechanical ventilation, 4) medications, and 5) diagnoses. Data generated by our ICU CIS was compared with other parallel data sets commonly used to obtain the same data for clinical research.

RESULTS

When compared with our CIS, the hospital database recorded a length of stay at least 1 day longer than the actual length of stay 53% of the time. A search of 139,387 sets of vital signs showed less than 0.1% rate of suspected artifact. When compared to direct observation, our CIS correctly recorded days of mechanical ventilation in 23 of 26 patients (88%). Two other data sets, medical diagnoses and medications given showed significant differences with other commonly used databases of the same information collected outside the ICU (billing codes and pharmacy records respectively

CONCLUSIONS

Compared to other commonly used data sources for clinical research, a commercially available CIS is an acceptable source of ICU patient data.

Validity of procedure codes in International Classification of Diseases, 9th revision, clinical modification administrative data

BACKGROUND

Administrative hospital discharge data are widely used to assess quality of care in patients undergoing certain procedures. However, little is known about the validity of administrative coding of procedure data. We conducted a detailed chart review to evaluate the accuracy and completeness of information on procedures in administrative data.

METHODS

We randomly selected 1200 hospital separations in the period April 1, 1996, to March 31, 1997, from administrative discharge data of 3 acute adult hospitals in Calgary, Alberta, Canada. Each separation record in administrative data contains up to 10 procedure coding fields. The corresponding medical charts were reviewed for recording presence or absence of procedures. We then determined sensitivity to quantify the accuracy of coding presence of procedures in administrative data when these are present in the chart data (criterion standard).

RESULTS

The agreement between the 2 databases varied greatly across 35 procedures studied. The sensitivity ranged from 0% to 94%. Of 6 major procedures studied, validity of coding was generally good, with 5 procedures having coding sensitivity of 69% and over and only 1 (lysis of peritoneal adhesion) with a low sensitivity of 41%. In contrast, many minor procedures had low sensitivities. Of 29 minor procedures studied, sensitivity was lower than 50% for 15 procedures, between 50% and 79% for 10, and 80% and over for 4.

CONCLUSION

Validity of information on procedures in administrative discharge data appears to be related to type of procedures. Major procedures that are usually performed in operating rooms are reasonably well-coded. Meanwhile, minor procedures that are routinely performed on wards or in radiology departments are generally undercoded.

Accuracy of Medicare claims-based diagnosis of acute myocardial infarction: estimating positive predictive value on the basis of review of hospital records

BACKGROUND

Many cardiovascular epidemiologic studies rely on diagnosis codes in health care claims databases. Despite important changes in the care and diagnosis of acute myocardial infarction (AMI), the validity of hospital discharge diagnosis codes for AMI in the US Medicare system has not been recently examined. Our objective was to examine the accuracy of International Classification of Diseases--ninth revision--Clinical Modifications (ICD-9-CM) discharge diagnosis codes and diagnosis-related groups (DRG) codes for AMI in a Medicare claims database.

METHODS

We sampled hospitalization episodes from Medicare beneficiaries in Pennsylvania during 1999, 2000, or both. We used Medicare data to identify patients with hospitalizations containing indicators of AMI (ICD-9-CM diagnosis codes 410.X0 and 410.X1 or DRG codes 121, 122, and 123). Hospital records for these episodes were reviewed by trained abstractors using World Health Organization criteria for diagnosing AMI. We then calculated the positive predictive value of Medicare claims-based definitions of AMI.

RESULTS

Of 2200 hospitalization episodes with Medicare diagnosis codes suggestive of AMI, 2022 hospital records (91.9%) were obtained. The positive predictive value for a primary Medicare claims-based definition was 94.1% (95% CI, 93.0%-95.2%). Positive predictive values for alternative claims-based definitions ranged slightly, with the definition including DRG codes and length-of-stay restrictions yielding the highest positive predictive value, 95.4% (95% CI, 94.3%-96.4%). Subjects with a history of myocardial infarction had a significantly lower positive predictive value than subjects without a history of myocardial infarction (88.1% vs 94.6%, P <.001).

CONCLUSIONS

In this study, we observed high positive predictive values for a Medicare claims-based diagnosis of AMI and a diagnosis based on structured hospital record review.

Do regulations limiting residents' work hours affect patient mortality?

OBJECTIVE

To conduct a statewide analysis of the effect of New York's regulations, limiting internal medicine and family practice residents' work hours, on patient mortality.

DESIGN

Retrospective study of inpatient discharge files for 1988 (before the regulations) and 1991 (after the regulations).

SETTING AND PATIENTS

Adult patients discharged from New York teaching hospitals (170214) and nonteaching hospitals (143,455) with a principal diagnosis of congestive heart failure, acute myocardial infarction, or pneumonia, for the years 1988 and 1991 (periods before and after Code 405 regulations went into law). Patients from nonteaching hospitals served as controls.

MEASUREMENT

In-hospital mortality.

RESULTS

Combined unadjusted mortality for congestive heart failure, acute myocardial infarction, and pneumonia patients declined between 1988 and 1991 in both teaching (14.1% to 13.0%; P =.0001) and nonteaching hospitals (14.0% to 12.5%; P =.0001). Adjusted mortality also declined between 1988 and 1991 in both teaching (odds ratio [OR], death 1991/1988, 0.868; 95% confidence interval [CI], 0.843 to 0.894; P =.0001) and nonteaching hospitals (OR, death 1991/1988, 0.853; 95% CI, 0.826 to 0.881; P =.0001). This beneficial trend toward lower mortality over time was nearly identical between teaching and nonteaching hospitals (P =.4348).

CONCLUSION

New York's mandated limitations on residents' work hours do not appear to have positively or negatively affected in-hospital mortality from congestive heart failure, acute myocardial infarction, or pneumonia in teaching hospitals.

Elevated serum creatinine is associated with 1-year mortality after acute myocardial infarction

BACKGROUND

Cardiovascular mortality is high in individuals with end-stage renal disease. However, less is known about the prognostic importance of moderate renal insufficiency in patients with acute myocardial infarction.

METHODS

We studied all patients with acute myocardial infarction admitted through the emergency department to an urban, academic hospital over 1 year. Patients were classified as having elevated (>133 micromol/L [1.5 mg/dL]) or normal (< or =133 micromol/L) serum creatinine at presentation.

RESULTS

Of 483 patients, 22% had elevated creatinine and 78% had normal creatinine. By 1 year, 46% of patients with elevated creatinine and 15% of patients with normal creatinine had died (P <.001). The unadjusted hazard ratio for 1-year mortality was increased in patients with elevated creatinine compared with those with normal creatinine (hazard ratio 3.85, 95% CI 2.61-5.67). After adjustment for baseline characteristics and treatment, the multivariable-adjusted hazard ratio for 1-year mortality remained increased in patients with elevated creatinine compared with those with normal creatinine (hazard ratio 2.40, 95% CI 1.55-3.72). There was an important modification of the prognostic value of creatinine by the presence of congestive heart failure at presentation (P value for interaction =.04). The adjusted hazard ratio for 1-year death associated with elevated creatinine compared with normal creatinine was 3.89 (95% CI 1.87-8.07) in patients without congestive heart failure and 1.92 (95% CI 1.10-3.36) in patients with congestive heart failure.

CONCLUSIONS

Elevated serum creatinine at presentation is associated with 1-year mortality after acute myocardial infarction. Further study is needed to optimize treatment after myocardial infarction in this high-risk group.

Effect of statin (HMG-Co-A-Reductase Inhibitor) use on 1-year mortality and hospitalization rates in older patients with cardiovascular disease living in nursing homes

OBJECTIVES

To quantify the effect of statins on 1-year mortality, hospitalizations, and decline in physical function among patients with cardiovascular disease (CVD) aged 65 and older living in nursing homes.

DESIGN

Retrospective cohort study.

SETTING

All Medicare/Medicaid certified nursing homes (N = 1,492) in Maine, New York, Mississippi, and South Dakota.

PARTICIPANTS

We identified 51,559 older patients with CVD from a population database that merged sociodemographic data and functional, clinical, and drug treatments from more than 300,000 newly admitted nursing home residents from 1992 to 1997. Statin users (n = 1,313) were matched with nonusers (n = 1,313) in the same facilities.

MEASUREMENTS

All-cause mortality, hospitalization, combined endpoint of mortality or hospitalization, and decline in physical function were determined at 1 year, and survival analysis was performed.

RESULTS

Prevalence of statin use in this frail older cohort with CVD was 2.6%. Statin use varied by age, gender, comorbid condition, medication use, and cognitive and physical function. One-year mortality was 229/1,000 person-years in the statin group and 404/1,000 person-years in the nonusers, with an adjusted hazard rate ratio (HRR) of 0.69, 95% confidence interval (CI) = 0.58-0.81. The estimated number needed to treat was seven (95% CI = 5-13). This association with improved all-cause mortality was evident for women and men and for age groups 75 to 84, and 85 and older.

CONCLUSION

Statin therapy is associated with improved clinical outcomes, including reduction in 1-year all-cause mortality, and the combined endpoint of death or hospitalization in a frail older population with CVD. Some caution should be taken in interpreting these results because potential bias from residual confounding could affect these results.

A multicenter study of the coding accuracy of hospital discharge administrative data for patients admitted to cardiac care units in Ontario

BACKGROUND

Cardiac health services researchers frequently use cohorts derived from administrative hospital discharge abstract data to study the outcomes and treatment of coronary artery disease. However, relatively limited data exist on the accuracy of the coding of cardiac diagnoses in discharge abstract data. The goal of this study was to examine the accuracy of the coding of acute myocardial infarction and other cardiac diagnoses in the Canadian Institute of Health Information hospital discharge abstracts.

METHODS

Patients admitted to 58 cardiac care units (CCUs) in Ontario that participated in the Fastrak II Acute Coronary Syndromes registry were linked to CIHI hospital discharge abstracts. The most responsible diagnosis at hospital discharge in the administrative data was compared with the CCU discharge diagnosis in the clinical registry.

RESULTS

A total of 58,816 CCU patients were linked to hospital discharge abstract data. The specificity, sensitivity, and positive predictive value of a most responsible diagnosis of acute myocardial infarction were 92.8%, 88.8%, and 88.5%, respectively. The specificity of CIHI diagnosis codes for arrhythmia, congestive heart failure, unstable angina, and chest pain not yet diagnosed were all at least 93.9%. However, the sensitivity of these CIHI diagnosis codes was no greater than 60.7%. Furthermore, the positive predictive values were no larger than 80.8%.

CONCLUSION

Myocardial infarction is generally accurately coded in Ontario hospital discharge abstract data. However, other cardiac diagnoses are less reliably coded in discharge abstract data.

Validity of information on comorbidity derived rom ICD-9-CCM administrative data

BACKGROUND

The comorbidity variables that constitute the Charlson index are widely used in health care research using administrative data. However, little is known about the validity of administrative data in these comorbidities. The agreement between administrative hospital discharge data and chart data for the recording of information on comorbidity was evaluated. The predictive ability of comorbidity information in the two data sets for predicting in-hospital mortality was also compared.

METHODS

One thousand two hundred administrative hospital discharge records were randomly selected in the region of Calgary, Alberta, Canada in 1996 and used a published coding algorithm to define the 17 comorbidities that constitute the Charlson index. Corresponding patient charts for the selected records were reviewed as the "criterion standard" against which validity of the administrative data were judged.

RESULTS

Compared with the chart data, administrative data had a lower prevalence in 10 comorbidities, a higher prevalence in 3 and a similar prevalence in 4. The kappa values ranged from a high of 0.87 to a low of 0.34; agreement was therefore near perfect for one variable, substantial for six, moderate for nine, and only fair for one variable. For the Charlson index score ranging from 0 to 5 to 6 or higher, agreement was moderate to substantial (kappa = 0.56, weighted kappa = 0.71). When 16 Charlson comorbidities from administrative data were used to predict in-hospital mortality, 10 comorbidities and the index scores defined using administrative data yielded odds ratios that were similar to those derived from chart data. The remaining six comorbidities yielded odds ratios that were quite different from those derived from chart data.

CONCLUSIONS

Administrative data generally agree with patient chart data for recording of comorbidities although comorbidities tend to be under-reported in administrative data. The ability to predict in-hospital mortality is less reliable for some of the individual comorbidities than it is for the summarized Charlson index scores in administrative data.

Racial differences in the use of cardiac procedures in patients with acute myocardial infarction

BACKGROUND

Although numerous studies have shown that Black patients are less likely than White patients to undergo cardiac procedures, most of these studies did not consider clinical and demographic factors that could account for observed racial differences.

OBJECTIVE

To determine if there are racial differences in the use of coronary angiography and revascularization procedures in patients with acute myocardial infarction, while controlling for multiple potentially important demographic and clinical variables.

METHODS

In this retrospective cohort study, data were obtained from medical records of 642 consecutive Black and White patients with acute myocardial infarction at a regional cardiac referral center in southern New England.

RESULTS

Blacks were significantly less likely than Whites to undergo angiography (p =.004; adjusted odds ratio =.36; 95% confidence interval =.18 -.72) and revascularization procedures (p =.006; adjusted odds ratio =.21; 95% confidence interval =.07 -.64). In the subgroup admitted directly to the hospital (n = 465), rather than transferred in from outlying hospitals, there were no racial differences in the use of angiography, but Blacks were significantly less likely to undergo revascularization procedures (p =.004; adjusted odds ratio =.18; 95% confidence interval =.06 -.58).

CONCLUSIONS

In patients hospitalized with acute myocardial infarction, there are substantial racial differences in the use of angiography and revascularization procedures that cannot be explained by clinical or demographic factors.

Occurrence of secondary ischemic events among persons with atherosclerotic vascular disease

BACKGROUND AND PURPOSE

Few data exist for large managed care populations on the occurrence of subsequent acute ischemic events in persons with established atherosclerotic vascular disease. We estimated the occurrence of secondary stroke, acute myocardial infarction (AMI), and vascular deaths among 2 large, managed care samples.

METHODS

With the use of International Classification of Diseases, Ninth Revision, Clinical Modification codes, patients aged > or =40 years and with stroke, AMI, or peripheral arterial disease (PAD) were identified from administrative data of UnitedHealthcare plans during 1995-1998. Stroke, AMI, and PAD cohorts were identified within a commercial insurance sample and a Medicare sample. Cumulative occurrences of subsequent stroke, AMI, or vascular death were estimated by survival analysis.

RESULTS

In the stroke commercial cohort (n=1631; mean age, 62.1 years), cumulative occurrence of subsequent events was 4.2%, 6.5%, 9.8%, and 11.8% at 0.5, 1, 2, and 3 years, respectively; cumulative secondary event occurrence in the AMI commercial cohort (n=6458; mean age, 56.0 years) was 3.5%, 4.8%, 7.3%, and 8.5% and in the PAD commercial cohort (n=5813; mean age, 59.2 years) was 1.5%, 2.8%, 4.8%, and 6.5%, respectively. Cumulative secondary event occurrences were even higher in stroke (n=1518; mean age, 79.5 years), AMI (n=2197; mean age, 76.2 years), and PAD (n=5033; mean age, 76.6 years) cohorts of the Medicare sample: 18.1%, 17.0%, and 8.7%, respectively, at 3 years. More than 75% of each stroke cohort's secondary events were strokes; more than 75% of each AMI cohort's secondary events were AMIs. Of the PAD cohorts' secondary events, 27% to 39% were strokes, 48% to 57% were AMIs, and 13% to 16% were vascular deaths.

CONCLUSIONS

Among these managed care enrollees with existing atherosclerotic vascular disease, subsequent ischemic events represent a significant symptomatic disease burden. Given these findings, it is very important to determine whether secondary prevention strategies are being effectively used to manage patients with diagnosed atherosclerosis.

Factoring socioeconomic status into cardiac performance profiling for hospitals: does it matter?

BACKGROUND

Critics of "scorecard medicine" often highlight the incompleteness of risk-adjustment methods used when accounting for baseline patient differences. Although socioeconomic status is a highly important determinant of adverse outcome for patients admitted to the hospital with acute myocardial infarction, it has not been used in most risk-adjustment models for cardiovascular report cards.

OBJECTIVES

To determine the incremental impact of socioeconomic status adjustments on age, sex, and illness severity for hospital-specific 30-day mortality rates after acute myocardial infarction.

METHODS

The authors compared the absolute and relative hospital-specific 30-day acute myocardial infarction mortality rates in 169 hospitals throughout Ontario between April 1, 1994 and March 31, 1997. Patient socioeconomic status was characterized by median neighborhood income using postal codes and 1996 Canadian census data. They examined two risk-adjustment models: the first adjusted for age, sex, and illness severity (standard), whereas the second adjusted for age, sex, illness severity, and median neighborhood income level (socioeconomic status).

RESULTS

There was an extremely strong correlation between 'standard' and 'socioeconomic status' risk-adjusted mortality rates (r = 0.99). Absolute differences in 30-day risk-adjusted mortality rates between the socioeconomic status and standard risk-adjustment models were small (median, 0.1%; 25th-75th percentile, 0.1-0.2). The agreement in the quintile rankings of hospitals between the socioeconomic status and standard risk-adjustment models was high (weighted kappa = 0.93).

CONCLUSION

Despite its importance as a determinant of patient outcomes, the effect of socioeconomic status on hospital-specific mortality rates over and above standard risk-adjustment methods for acute myocardial infarction hospital profiling in Ontario was negligible.

Time trends in the occurrence and outcome of acute myocardial infarction and coronary heart disease death between 1986 and 1996 (a New Jersey statewide study)

Most reports of the decrease in age-adjusted coronary heart disease (CHD) are based on databases with upper age cut-offs that exclude approximately half of the events. We report changes in rates of acute myocardial infarction (AMI) and of out-of-hospital coronary death between 1986 and 1996 among New Jersey residents > or =15 years old. Data on patients discharged with the diagnosis of AMI from nonfederal acute care hospitals in the state (n = 270,091) and all records in the New Jersey death registration files with CHD (n = 172,175) listed as the cause of death from 1986 to 1996 (total study n = 442,266) were analyzed. The rate of hospitalized AMI cases in the state remained essentially unchanged during these 11 years, whereas in-hospital and 30-day case fatality among all age groups and both sexes declined. Age-adjusted CHD rates showed a decrease in fatal events, a smaller decrease in total events, and a slight increase in nonfatal events. The proportion of fatal CHD events occurring out-of-hospital decreased especially among men. The median age at occurrence of events increased by 1 year. Despite a decrease in CHD mortality, the rate of nonfatal events increased, especially among persons > or =75 years old. Thus, the decrease in age-adjusted CHD mortality is not all due to treatment and true prevention of CHD, but the disease simply occurs at an older age.

Quality of the information contained in the minimum basic data set: results from an evaluation in eight hospitals

To assess the quality of the information included in the minimum basic data set (MBDS) of the eight public hospitals of the Murcia region in order to ascertain what should be improved to be valid and reliable. An external encoder performed a recoding of a random sample of hospital discharges, using the patients hospital records and comparing afterwards the information obtained with the one reflected in the MBDS databases. Quality was assessed using 12 criteria. The reviewed discharges sample consisted at least of 96 cases per hospital (Type I error = 0.05, Type II = 0.10, for the most unfavorable case). A total of 796 cases were reviewed. The MBDS disagreement percentages with the patient record data were higher for the clinical data, with 41.6% for the main diagnosis and 33.5% for the main surgical procedure, being in both cases higher in those hospitals that had used to codify just the discharge record with regard to those that did so with the complete patient record. The variation rate in the diagnosis-related group (DRG) assignment was of 29.6%, and there was a decrease in the case-mix index of 1.07397 when reviewing with the patient record to 1.05555 in the MBDS. Within the administrative data, the highest disagreement rate was for the physician that signs the discharge (60.5%) and the patient's address (31.6%). In many of these assessed aspects there are significant differences between hospitals. A reliability problem was identified in the collected data, which mainly affects the clinical variables. It is therefore advisable to carefully assess the use of this information (both the MBDS directly as well as its grouping through the use of patient classification systems), and the indicators derived from it as its quality is not guaranteed. Systematic assessment and quality control of the MBDS production is advised.

Administrative data for public health surveillance and planning

Electronically available administrative data are increasingly used by public health researchers and planners. The validity of the data source has been established, and its strengths and weaknesses relative to data abstracted from medical records and obtained via survey are documented. Administrative data are available from a variety of state, federal, and private sources and can, in many cases, be combined. As a tool for planning and surveillance, administrative data show great promise: They contain consistent elements, are available in a timely manner, and provide information about large numbers of individuals. Because they are available in an electronic format, they are relatively inexpensive to obtain and use. In the United States, however, there is no administrative data set covering the entire population. Although Medicare provides health care for an estimated 96% of the elderly, age 65 years and older, there is no comparable source for those under 65.

Impact of double counting and transfer bias on estimated rates and outcomes of acute myocardial infarction

CONTEXT

Ischemic heart disease is the leading cause of death in the United States. Recent studies report inconsistent findings on the changes in the incidence of hospitalizations for ischemic heart disease. These reports have relied primarily on hospital discharge data. Preliminary data suggest that a significant percentage of patients suffering acute myocardial infarction (MI) in rural communities are transferred to urban centers for care. Patients transferred to a second hospital may be counted twice for one episode of ischemic heart disease.

OBJECTIVE

To describe the impact of double counting and transfer bias on the estimation of incidence rates and outcomes of ischemic heart disease, specifically acute MI, in the United States.

DESIGN

Analysis of state hospital discharge data from Kansas, Colorado (State Inpatient Database [SID]), Nebraska, Arizona, New Jersey, Michigan, Pennsylvania, and Illinois (SID) for the years 1995 to 1997. A matching algorithm was developed for hospital discharges to determine patients counted twice for one episode of ischemic heart disease. Validation of our matching algorithm.

PATIENTS

Patients reported to have suffered ischemic heart disease (ICD9 codes 410-414, 786.5).

MAIN OUTCOME MEASURES

Number of patients counted twice for one episode of acute MI.

RESULTS

It is estimated that double count rates range from 10% to 15% for all states and increased over the 3 years. Moderate sized rural counties had the highest estimated double count rates at 15% to 20% with a few counties having estimated double count rates a high as 35% to 50%. Older patients and females were less likely to be double counted (P <0.05).

CONCLUSIONS

Double counting patients has resulted in a significant overestimation in the incidence rate for hospitalization for acute MI. Correction of this double counting reveals a significantly lower incidence rate and a higher in-hospital mortality rate for acute MI. Transferred patients differ significantly from nontransferred patients, introducing significant bias into MI outcome studies. Double counting and transfer bias should be considered when conducting and interpreting research on ischemic heart disease, particularly in rural regions.