Administrative Diagnostic Codes Research Articles

Impact of Potential Case Misclassification by Administrative Diagnostic Codes on Outcome Assessment of Observational Study for People Who Inject Drugs.

Initiation of medications for opioid use disorder (MOUD) within the hospital setting may improve outcomes for people who inject drugs (PWID) hospitalized because of an infection. Many studies used International Classification of Diseases (ICD) codes to identify PWID, although these may be misclassified and thus, inaccurate. We hypothesized that bias from misclassification of PWID using ICD codes may impact analyses of MOUD outcomes. We analyzed a cohort of 36 868 cases of patients diagnosed with Staphylococcus aureus bacteremia at 124 US Veterans Health Administration hospitals between 2003 and 2014. To identify PWID, we implemented an ICD code-based algorithm and a natural language processing (NLP) algorithm for classification of admission notes. We analyzed outcomes of prescribing MOUD as an inpatient using both approaches. Our primary outcome was 365-day all-cause mortality. We fit mixed-effects Cox regression models with receipt or not of MOUD during the index hospitalization as the primary predictor and 365-day mortality as the outcome. NLP identified 2389 cases as PWID, whereas ICD codes identified 6804 cases as PWID. In the cohort identified by NLP, receipt of inpatient MOUD was associated with a protective effect on 365-day survival (adjusted hazard ratio, 0.48; 95% confidence interval, .29-.81; P < .01) compared with those not receiving MOUD. There was no significant effect of MOUD receipt in the cohort identified by ICD codes (adjusted hazard ratio, 1.00; 95% confidence interval, .77-1.30; P = .99). MOUD was protective of all-cause mortality when NLP was used to identify PWID, but not significant when ICD codes were used to identify the analytic subjects.

Minimizing misclassification bias with a model to identify acetabular fractures using health administrative data: A cohort study.

Acetabular fractures (AFs) are relatively uncommon thereby limiting their study. Analyses using population-based health administrative data can return erroneous results if case identification is inaccurate (‘misclassification bias’). This study measured the impact of an AF prediction model based exclusively on administrative data upon misclassification bias.We applied text analytical methods to all radiology reports over 11 years at a large, tertiary care teaching hospital to identify all AFs. Using clinically-based variable selection techniques, a logistic regression model was created.We identified 728 AFs in 438,098 hospitalizations (15.1 cases/10,000 admissions). The International Classification of Disease, 10th revision (ICD-10) code for AF (S32.4) missed almost half of cases and misclassified more than a quarter (sensitivity 51.2%, positive predictive value 73.0%). The AF model was very accurate (optimism adjusted R2 0.618, c-statistic 0.988, calibration slope 1.06). When model-based expected probabilities were used to determine AF status using bootstrap imputation methods, misclassification bias for AF prevalence and its association with other variables was much lower than with International Classification of Disease, 10th revision S32.4 (median [range] relative difference 1.0% [0%–9.0%] vs 18.0% [5.4%–75.0%]).Lone administrative database diagnostic codes are inadequate to create AF cohorts. The probability of AF can be accurately determined using health administrative data. This probability can be used in bootstrap imputation methods to importantly reduce misclassification bias.

Evaluating Differences in Opioid and Stimulant Use-associated Infectious Disease Hospitalizations in Florida, 2016-2017.

BackgroundThe opioid epidemic has led to increases in injection drug use (IDU)-associated infectious diseases; however, little is known about how more recent increases in stimulant use have affected the incidence and outcomes of hospitalizations for infections among people who inject drugs (PWID).MethodsAll hospitalizations of PWID for IDU-associated infections in Florida were identified using administrative diagnostic codes and were grouped by substance used (opioids, stimulants, or both) and site of infection. We evaluated the association between substance used and the outcomes: patient-directed discharge (PDD, or “against medical advice”) and in-hospital mortality.ResultsThere were 22 856 hospitalizations for infections among PWID. Opioid use was present in 73%, any stimulants in 43%, and stimulants-only in 27%. Skin and soft tissue infection was present in 50%, sepsis/bacteremia in 52%, osteomyelitis in 10%, and endocarditis in 10%. PWID using opioids/stimulants were youngest, most uninsured, and had the highest rates of endocarditis (16%) and hepatitis C (44%). Additionally, 25% of patients with opioid/stimulant use had PDD versus 12% for those using opioids-only. In adjusted models, opioid/stimulant use was associated with PDD compared to opioid-only use (aRR 1.28, 95% CI 1.17–1.40). Younger age and endocarditis were also associated with PDD. Compared to opioid-only use, stimulant-only use had higher risk of in-hospital mortality (aRR 1.26, 95% CI 1.03–1.46).ConclusionsWhile opioid use contributed to most IDU-associated infections, many hospitalizations also involved stimulants. Increasing access to harm reduction interventions could help prevent these infections, while further research on the acute management of stimulant use disorder-associated infections is needed.

A253 ADMINISTRATIVE DATA CAN ACCURATELY IDENTIFY PATIENTS WITH PERIANAL CROHN’S DISEASE

Abstract Background Perianal fistulas (PAF) are a frequent complication of Crohn’s disease (CD) associated with substantial morbidity. Population-based studies with CD are lacking, in part due to the difficulty in identifying these patients from health administrative databases. Aims To determine if administrative claims and diagnostic codes can reliably identify patients with perianal fistulas in a cohort of patients with CD. Methods A retrospective cohort study was performed using data from The Ottawa Hospital (TOH), which was linked to The Institute for Clinical Evaluative Sciences (ICES) Ontario Crohn’s and Colitis Cohort (OCCC) data using Ontario Health Insurance Plan (OHIP) numbers. Patients admitted with CD from 1 Jan 2009 to 31 Dec 2016 were identified from TOH data warehouse using the ICD-10 code K50.x. Confirmation of CD diagnosis and determination of the presence or absence of PAF was achieved by a longitudinal, manual chart review. Patients with and without PAF were abstracted in a 1:2 ratio to serve as a reference gold standard for PAF status. ICES captures all publicly reimbursed diagnostic tests, interventional procedures and physician billing codes, including MRI pelvis utilization and surgical procedures associated with perianal fistulas in Ontario. Sixteen case definitions for PAF in ICES were specified a priori. Two by two contingency tables were constructed to assess the sensitivity, specificity, positive predictive value (PPV) and negative predicative value (NPV) of each case definition against the gold standard PAF status as determined by TOH data. Youden’s index and Kappa were used to select a case definition that best identified TOH PAF patients. Results: A total of 136 patients with active PAF and 351 without PAF were included in the linked analysis. There were a similar proportion of male patients with and without PAF (49% vs. 43%). Patients with PAF were slightly younger; 69% were aged 18–44 compared to 58% of patients without PAF. Sensitivity of the case definitions ranged from 0.44 to 0.98, and specificity from 0.45 to 1.00. A case definition that combined at least two of fistula diagnosis code, perianal surgical procedures associated with fistulas, and radiologic imaging codes of the pelvis, all within 2 years, had the best performance using Youden’s index and Kappa. It discriminated between patients with or without PAF with sensitivity of 0.80 and specificity of 0.92. Conclusions Using a cohort of CD patients from a single tertiary care center we derived a case definition that could accurately distinguish CD patients with and without PAF in a provincial health administrative database. Once validated this will allow for future population-based studies to assess trends in PAF. Funding Agencies Takeda Pharmaceuticals

Identifying incident Parkinson's disease using administrative diagnostic codes: a validation study

BackgroundAdministrative databases that capture diagnostic codes are increasingly being used worldwide for research because they can save time and reduce costs. However, assessing validity is necessary before defining diseases using only diagnostic codes in research applications. ObjectiveOur objective was to assess the validity of using diagnostic codes to identify incident Parkinson's disease (PD) cases in Olmsted County, Minnesota using an established standard for comparison (1976–2005). MethodsCases were identified solely using computer programs applied to administrative diagnostic code indexes from the Rochester Epidemiology Project (REP). Two codes >30 days apart or one code on the death certificate constituted PD. The standard was a clinical diagnosis by movement disorders specialists based on medical record review. Validity was assessed using positive predictive value (PPV) and sensitivity. Numbers of incident cases and incidence rates were compared between the two ascertainment methods by sex. ResultsThe codes only method over-counted the number of incident PD cases by 73% (804 versus 464), and this over-counting generally increased with calendar year. Sensitivity was 80% (95% CI [76%, 84%]) and PPV was 46% (95% CI [34%, 50%]). Disease status misclassification accounted for two-thirds of falsely identified cases, where individuals were found to not have PD (43%) or even parkinsonism (23%) after medical record review. The codes only method also over-estimated the incidence rate time trend for men and women by approximately two-fold. ConclusionIn our context, using administrative diagnostic codes only to identify incident PD cases is not recommended unless more accurate algorithms are developed.

Diagnosis and Antibiotic Management of Otitis Media and Otitis Externa in United States Veterans

Acute otitis media (AOM) and otitis media with effusion (OME) occur primarily in children, whereas acute otitis externa (AOE) occurs with similar frequency in children and adults. Data on the incidence and management of otitis in adults are limited. This study characterizes the incidence, antibiotic management, and outcomes for adults with otitis diagnoses. A retrospective cohort of ambulatory adult veterans who presented with acute respiratory tract infection (ARI) diagnoses at 6 VA Medical Centers during 2014-2018 was created. Then, a subcohort of patients with acute otitis diagnoses was developed. Patient visits were categorized with administrative diagnostic codes for ARI (eg, sinusitis, pharyngitis) and otitis (OME, AOM, and AOE). Incidence rates for each diagnosis were calculated. Proportions of otitis visits with antibiotic prescribing, complications, and specialty referral were summarized. Of 46 634 ARI visits, 3898 (8%) included an otitis diagnosis: OME (22%), AOM (44%), AOE (31%), and multiple otitis diagnoses (3%). Incidence rates were otitis media 4.0 (95% confidence interval [CI], 3.9-4.2) and AOE 2.0 (95% CI, 1.9-2.1) diagnoses per 1000 patient-years. By comparison, the incidence rates for pharyngitis (8.4; 95% CI, 8.2-8.6) and sinusitis (15.2; 95% CI, 14.9-15.5) were higher. Systemic antibiotics were prescribed in 75%, 63%, and 21% of AOM, OME, and AOE visits, respectively. Complications for otitis visits were low irrespective of antibiotic treatment. Administrative data indicated that otitis media diagnoses in adults were half as common as acute pharyngitis, and the majority received antibiotic treatment, which may be inappropriate. Prospective studies verifying diagnostic accuracy and antibiotic appropriateness are warranted.

Sensitivity of Administrative Coding in Identifying Inpatient Acute Strokes Complicating Procedures or Other Diseases in UK Hospitals.

BackgroundAdministrative hospital diagnostic coding data are increasingly used in “big data” research and to assess complication rates after surgery or acute medical conditions. Acute stroke is a common complication of several procedures/conditions, such as carotid interventions, but data are lacking on the sensitivity of administrative coding in identifying acute stroke during inpatient stay.Methods and ResultsUsing all acute strokes ascertained in a population‐based cohort (2002–2017) as the reference, we determined the sensitivity of hospital administrative diagnostic codes (International Classification of Diseases, Tenth Revision; ICD‐10) for identifying acute strokes that occurred during hospital admission for other reasons, stratified by coding strategies, study periods, and stroke severity (National Institutes of Health Stroke Score</≥5). Of 3011 acute strokes, 198 (6.6%) occurred during hospital admissions for procedures/other diseases, including 122 (61.6%) major strokes. Using stroke‐specific codes (ICD‐10=I60–I61 and I63–I64) in the primary diagnostic position, 66 of the 198 cases were correctly identified (sensitivity for any stroke, 33.3%; 95% CI, 27.1–40.2; minor stroke, 30.3%; 95% CI, 21.0–41.5; major stroke, 35.2%; 95% CI, 27.2–44.2), with no improvement of sensitivity over time (P trend=0.54). Sensitivity was lower during admissions for surgery/procedures than for other acute medical admissions (n/% 17/23.3% versus 49/39.2%; P=0.02). Sensitivity improved to 60.6% (53.6–67.2) for all and 61.6% (50.0–72.1) for surgery/procedures if other diagnostic positions were used, and to 65.2% (58.2–71.5) and 68.5% (56.9–78.1) respectively if combined with use of all possible nonspecific stroke‐related codes (ie, adding ICD‐10=I62 and I65–I68).ConclusionsLow sensitivity of administrative coding in identifying acute strokes that occurred during admission does not support its use alone for audit of complication rates of procedures or hospitalization for other reasons.

Post-traumatic stress disorder and serum cytokine and chemokine concentrations in patients with rheumatoid arthritis✰

ObjectiveAlthough post-traumatic stress disorder (PTSD) is identified as a risk factor in the development of rheumatoid arthritis (RA), associations of PTSD with disease progression are less clear. To explore whether PTSD might influence disease-related measures of systemic inflammation in RA, we compared serum cytokine/chemokine (cytokine) concentrations in RA patients with and without PTSD. MethodsParticipants were U.S. Veterans with RA and were categorized as having PTSD, other forms of depression/anxiety, or neither based on administrative diagnostic codes. Multiplex cytokines were measured using banked serum. Associations of PTSD with cytokine parameters (including a weighted cytokine score) were assessed using multivariable regression, stratified by anti-CCP status and adjusted for age, sex, race, and smoking status. ResultsAmong 1,460 RA subjects with mean (SD) age of 64 (11) years and disease duration of 11 (11) years, 91% were male, 77% anti-CCP positive, and 80% ever smokers. Of these, 11.6% had PTSD, 23.7% other depression/anxiety, and 64.7% had neither. PTSD, but not depression/anxiety, was associated with a higher cytokine score and number of high-concentration analytes in adjusted models, though this was limited to anti-CCP positive subjects. PTSD was associated with heightened expression of several individual cytokines including IL-1β, IL-2, IL-4, IL-5, IL-6, IL-7, IL-10, IL-12, IL-17, IFN-γ, GM-CSF, MCP-1, and TNF-α. ConclusionAnti-CCP positive RA patients with PTSD have higher serum cytokine concentrations than those without PTSD, demonstrating that systemic inflammation characteristic of RA is heightened in the context of this relatively common psychiatric comorbidity.

Incident Atrial Fibrillation and the Risk of Stroke in Adults with Chronic Kidney Disease: The Stockholm CREAtinine Measurements (SCREAM) Project.

Patients with CKD have a high risk of atrial fibrillation. Both CKD and atrial fibrillation are associated with higher risk of stroke and death. However, the effect of incident atrial fibrillation on stroke risk among patients with CKD is unknown. Our study included adults with CKD (eGFR<60 ml/min per 1.73 m2) without previously documented atrial fibrillation who had been in contact with health care in Stockholm, Sweden during 2006-2011. Incident atrial fibrillation was identified by administrative diagnostic codes in outpatient or inpatient care and treated as a time-updated exposure in the analysis of stroke and death risk. Stroke events and deaths were ascertained from regional and national registers with complete coverage. Covariates included demographics, comorbidities, therapeutic procedures, and medications. Multivariable Cox regression analysis and competing risk analysis (accounting for death) were used to estimate the association between incident atrial fibrillation and stroke. Among 116,184 adults with CKD, 13,412 (12%) developed clinically recognized atrial fibrillation during a mean follow-up of 3.9 years (interquartile range, 2.3-5.7 years). Incidence of atrial fibrillation increased across lower eGFR strata: from 29.4 to 46.3 atrial fibrillations per 1000 person-years in subjects with eGFR=45-60 and <30 ml/min per 1.73 m2, respectively; 1388 (53.8 per 1000 person-years) cases of stroke and 5592 (205.1 per 1000 person-years) deaths occurred after incident atrial fibrillation compared with 6850 (16.6 per 1000 person-years) cases of stroke and 28,613 (67.5 per 1000 person-years) deaths during periods without atrial fibrillation. After adjustment, incident atrial fibrillation was associated with higher risk of stroke (hazard ratio, 2.00; 95% confidence interval, 1.88 to 2.14) and death (hazard ratio, 1.76; 95% confidence interval, 1.71 to 1.82). This was attributed to both ischemic stroke (hazard ratio, 2.11; 95% confidence interval, 1.96 to 2.28) and intracranial bleeds (hazard ratio, 1.64; 95% confidence interval, 1.42 to 1.90). Stroke risk was similar across all eGFR strata. In competing risk analyses accounting for death, the association between incident atrial fibrillation and stroke was attenuated but remained higher (subhazard ratio, 1.49; 95% confidence interval, 1.39 to 1.60). Patients with CKD who develop atrial fibrillation are at higher risk of stroke and death.

Assessing the accuracy of using diagnostic codes from administrative data to infer antidepressant treatment indications: a validation study.

PurposeTo assess the accuracy of using diagnostic codes from administrative data to infer treatment indications for antidepressants prescribed in primary care.MethodsValidation study of administrative diagnostic codes for 13 plausible indications for antidepressants compared with physician‐documented treatment indications from an indication‐based electronic prescribing system in Quebec, Canada. The analysis included all antidepressant prescriptions written by primary care physicians between January 1, 2003 and December 31, 2012 using the electronic prescribing system. Patient prescribed antidepressants were linked to physician claims and hospitalization data to obtain all diagnoses recorded in the past year.ResultsDiagnostic codes had poor sensitivity for all treatment indications, ranging from a high of only 31.2% (95% CI, 26.8%‐35.9%) for anxiety/stress disorders to as low as 1.3% (95% CI, 0.0%‐5.2%) for sexual dysfunction. Sensitivity was notably worse among older patients and patients with more chronic comorbidities. Physician claims data were a better source of diagnostic codes for antidepressant treatment indications than hospitalization data.ConclusionsAdministrative diagnostic codes are poor proxies for antidepressant treatment indications. Future work should determine whether the use of other variables in administrative data besides diagnostic codes can improve the ability to predict antidepressant treatment indications.

Analyses of the Complexity of Patients Undergoing Attended Polysomnography in the Era of Home Sleep Apnea Tests.

Health care complexity includes dimensions of patient comorbidity and the level of services needed to meet patient demands. Home sleep apnea tests (HSAT) are increasingly used to test medically uncomplicated patients suspected of having moderate to severe obstructive sleep apnea (OSA). Patients with significant comorbidities or other sleep disorders are not candidates for HSAT and require attended in-center polysomnography. We hypothesized that this trend would result in increasingly complex patients being studied in sleep centers. Our study had two parts. To ascertain trends in sleep patient comorbidity, we used administrative diagnostic codes from patients undergoing polysomnography at the Mayo Clinic Center for Sleep Medicine from 2005 to June 2015 to calculate the Charlson and the Elixhauser comorbidity indices. We measured the level of services provided in two ways: (1) in a subset of patients from the past 2 months of 2015, we evaluated correlation of these morbidity indices with an internally developed Polysomnogram Clinical Index (PSGCI) rating anticipated patient care needs from 0 to 3 and (2) we measured the sleep study complexity based on polysomnography protocol design. In 43,780 patients studied from 2005 to June 2015, the Charlson index increased from a mean of 1.38 to 1.88 (3.1% per year, P < .001) and the mean Elixhauser index increased from 2.61 to 3.35 (2.5% per year, P < .001). Both comorbidity indices were significantly higher at the highest (Level 3) level of the PSGCI (P < .001), and sleep study complexity increased over time. The complexity of patients undergoing attended polysomnography has increased by 28% to 36% over the past decade as measured by validated comorbidity indices, and these indices correlate with the complexity of rendered care during polysomnography. These findings have implications for increasing requirements for staffing, monitoring capabilities, and facility design of future sleep centers. A commentary on this article appears in this issue on page 499.

Validation of an algorithm to identify children with biopsy-proven celiac disease from within health administrative data: An assessment of health services utilization patterns in Ontario, Canada.

ImportanceCeliac disease (CD) is a common pediatric illness, and awareness of gluten-related disorders including CD is growing. Health administrative data represents a unique opportunity to conduct population-based surveillance of this chronic condition and assess the impact of caring for children with CD on the health system.ObjectiveThe objective of the study was to validate an algorithm based on health administrative data diagnostic codes to accurately identify children with biopsy-proven CD. We also evaluated trends over time in the use of health services related to CD by children in Ontario, Canada.Study design and settingWe conducted a retrospective cohort study and validation study of population-based health administrative data in Ontario, Canada. All cases of biopsy-proven CD diagnosed 2005–2011 in Ottawa were identified through chart review from a large pediatric health care center, and linked to the Ontario health administrative data to serve as positive reference standard. All other children living within Ottawa served as the negative reference standard. Case-identifying algorithms based on outpatient physician visits with associated ICD-9 code for CD plus endoscopy billing code were constructed and tested. Sensitivity, specificity, PPV and NPV were tested for each algorithm (with 95% CI). Poisson regression, adjusting for sex and age at diagnosis, was used to explore the trend in outpatient visits associated with a CD diagnostic code from 1995–2011.ResultsThe best algorithm to identify CD consisted of an endoscopy billing claim follow by 1 or more adult or pediatric gastroenterologist encounters after the endoscopic procedure. The sensitivity, specificity, PPV, and NPV for the algorithm were: 70.4% (95% CI 61.1–78.4%), >99.9% (95% CI >99.9->99.9%), 53.3% (95% CI 45.1–61.4%) and >99.9% (95% CI >99.9->99.9%) respectively. It identified 1289 suspected CD cases from Ontario-wide administrative data. There was a 9% annual increase in the use of this combination of CD-associated diagnostic codes in physician billing data (RR 1.09, 95% CI 1.07–1.10, P<0.001).ConclusionsWith its current structure and variables Ontario health administrative data is not suitable in identifying incident pediatric CD cases. The tested algorithms suffer from poor sensitivity and/or poor PPV, which increase the risk of case misclassification that could lead to biased estimation of CD incidence rate. This study reinforced the importance of validating the codes used to identify cohorts or outcomes when conducting research using health administrative data.

The spectrum of rheumatic in-patient diagnoses at a pediatric hospital in Kenya

BackgroundPediatric rheumatic diseases are chronic illnesses that can cause considerable disease burden to children and their families. There is limited epidemiologic data on these diseases in East Africa. The aim of this study was to assess the spectrum of pediatric rheumatic diagnoses in an in-patient setting and determine the accuracy of ICD-10 codes in identifying these conditions.MethodsMedical records from Gertrude’s Children’s Hospital in Kenya were reviewed for patients diagnosed with “diseases of the musculoskeletal system and connective tissue” as per ICD-10 diagnostic codes assigned at discharge between January and December 2011. Cases were classified as “rheumatic” or “non-rheumatic”. Accuracy of the assigned ICD-10 code was ascertained. Death records were reviewed. Longitudinal follow-up of “rheumatic” cases was done by chart review up to March 2014.ResultsTwenty six patients were classified as having a “rheumatic” condition accounting for 0.32% of patients admitted. Of these, 11 (42.3%) had an acute inflammatory arthropathy, 6 (23.1%) had septic arthritis, 4 (15.4%) had Kawasaki disease, 2 (7.7%) had pyomyositis, and there was one case each of septic bursitis, rheumatic fever, and a non-specific soft tissue disorder. No cases of juvenile idiopathic arthritis (JIA) were identified. One case of systemic lupus erythematosus was documented by death records. The agreement between the treating physician’s discharge diagnosis and medical records ICD-10 code assignment was good (Kappa: 0.769). On follow-up, one child had recurrent knee swelling that was suspicious for JIA.ConclusionsPediatric rheumatic conditions represented 0.32% of admissions at a pediatric hospital in Kenya. Acute inflammatory arthropathies, septic arthritis and Kawasaki disease were the most frequent in-patient rheumatic diagnoses. Chronic pediatric rheumatic diseases were rare amongst this in-patient population. Despite limitations associated with the use of administrative diagnostic codes, they can be a first step in evaluating the spectrum of pediatric rheumatic conditions in Kenya and other countries in East Africa.

Target Serum Urate: Do Gout Patients Know Their Goal?

To examine gout patients' knowledge of their condition, including the central role of achieving and maintaining the serum urate (SU) goal with the use of urate-lowering therapy (ULT). This study of 612 gout patients was conducted at a Veterans Affairs medical center. Gout patients were included based on administrative diagnostic codes and receipt of at least 1 allopurinol prescription over a 1-year period. Questionnaires were mailed to patients and linked to medical records data. The questionnaire included gout-specific knowledge questions, the Patient Activation Measure, and self-reported health outcomes. Knowledge was assessed descriptively. Multivariable logistic regression was used to determine predictors of SU goal knowledge. Associations of knowledge with health outcomes were examined in exploratory analyses. The questionnaire had a 62% response rate. Only 14% of patients knew their SU goal, while the majority answered correctly for the other 5 gout-specific knowledge questions. In adjusted analyses, having a rheumatologist as initial prescriber (odds ratio [OR] 3.0 [95% confidence interval (95% CI) 1.4-6.2]) and knowing all of the other 5 gout-specific knowledge questions (OR 2.1 [95% CI 1.3-3.4]) were associated with greater odds of knowing the SU goal. SU goal knowledge was associated with self-reported global health status, but not with self-reported health-related quality of life or gout-specific health status. There is a knowledge deficit regarding the SU treatment goal among gout patients receiving ULT, despite generally high levels of other gout-specific knowledge. SU goal information may be an important and underutilized concept among providers treating gout patients.

Development of an algorithm to identify metastatic prostate cancer in electronic medical records using natural language processing.

164 Background: Prostate cancer patients who develop metastases are a difficult population to identify through administrative diagnostic codes, due to their protracted time to metastases, limited survival and the inconsistent use of specific codes. As a result, research that is needed to inform the delivery of high-quality care in this setting is limited. Therefore, the goal of this study was to develop an algorithm, which utilizes EMR data to identify men who progress to metastatic prostate cancer after diagnosis using natural language processing (NLP). Methods: An electronic algorithm was developed to search unstructured text using NLP to identify progression to metastases among men with a diagnosis of prostate cancer between 1992 and 2010 in a large, diverse cohort of men who were part of an ongoing study focused on prostate cancer mortality. A training set of 449 men who were diagnosed as early stage prostate cancer was used for development. Pathology, radiology and clinic notes were searched from diagnosis until death or loss to follow-up. Pathology reports were searched for mention of adenocarcinoma in the metastatic lesion, radiology reports were searched for abnormal findings consistent with metastases, and clinic notes were searched for mentions of increasing pain or narcotic use related to metastases. Each NLP component was validated against manual review of the corresponding records. Results: Of the 449 men in the training set, 40 (8.9%) were found to have metastatic prostate cancer. The majority of cases had evidence of metastases in their clinic notes (98%). Radiology reports identified 18% of cases, and pathology reports identified 5%. Of the 40 cases identified, 25% did not have a corresponding ICD-9 codes for metastatic cancer. However, 7.5% used ADT, 37.5% had increasing oncology visits and 22.5% had rapidly rising PSA levels. Conclusions: Our results suggest that NLP can be used to identify men with metastatic prostate cancer in the EMR more accurately than diagnosis codes alone. The automated identification of patients with metastatic cancer facilitates quality of care research in this setting to ensure the delivery of appropriate and high-quality care.

PRM72 - The impact of two different case definitions of major congenital malformations identified from Quebec (Canada) administrative databases on their association with asthma controller medications taken during pregnancy

Accurate identification of major malformation cases from administrative databases is crucial for perinatal epidemiology. In Quebec, most of major malformations are detected in hospital, however administrative databases capture diagnoses data from both hospitals and other medical facilities. We aimed to compare the prevalence of major congenital malformations identified from administrative databases diagnostic codes (ICD-9 and ICD-10) with 2 definitions: diagnoses made in hospital only vs. diagnoses made in hospital and other medical facilities. We further evaluated the impact of using each definition to quantify the association between maternal use of asthma controller medications and the prevalence of major malformations at birth and during the first year of life. A cohort of pregnancies from asthmatic women between 1990 and 2010 was formed through the linkage of administrative databases from Quebec. We calculated the prevalence of major malformations identified in the 1st year of life of the newborn using the 2 case definitions. We also calculated the crude odds ratio (OR) of major malformations associated with maternal use of inhaled corticosteroids (ICS) and long-acting beta2-agonists (LABA) in the 1st trimester of pregnancy. From 30655 pregnancies, 2090 (6.8%) major malformations were identified with hospital diagnostic codes only vs. 2748 (9.0%) with hospital and other medical facilities diagnostic codes. The OR of major malformations associated with ICS was 1.1 (95%CI: 1.0-1.2) with the first and 1.1 (95%CI: 1.0-1.2) with the second case definition. Corresponding results were 1.3 (95%CI: 1.0-1.6) and 1.1 (95%CI: 0.9-1.4) for LABA exposure. The case definition of congenital malformations had a considerable impact on the prevalence of major congenital malformations, but less impact on the associations we examined. The percentage of false positive cases when using all medical facilities diagnostic codes in the case definition is unknown and should be the objective for future research.

Proportion of hospital readmissions deemed avoidable: a systematic review

Readmissions to hospital are increasingly being used as an indicator of quality of care. However, this approach is valid only when we know what proportion of readmissions are avoidable. We conducted a systematic review of studies that measured the proportion of readmissions deemed avoidable. We examined how such readmissions were measured and estimated their prevalence. We searched the MEDLINE and EMBASE databases to identify all studies published from 1966 to July 2010 that reviewed hospital readmissions and that specified how many were classified as avoidable. Our search strategy identified 34 studies. Three of the studies used combinations of administrative diagnostic codes to determine whether readmissions were avoidable. Criteria used in the remaining studies were subjective. Most of the studies were conducted at single teaching hospitals, did not consider information from the community or treating physicians, and used only one reviewer to decide whether readmissions were avoidable. The median proportion of readmissions deemed avoidable was 27.1% but varied from 5% to 79%. Three study-level factors (teaching status of hospital, whether all diagnoses or only some were considered, and length of follow-up) were significantly associated with the proportion of admissions deemed to be avoidable and explained some, but not all, of the heterogeneity between the studies. All but three of the studies used subjective criteria to determine whether readmissions were avoidable. Study methods had notable deficits and varied extensively, as did the proportion of readmissions deemed avoidable. The true proportion of hospital readmissions that are potentially avoidable remains unclear.

Administrative Data Sets are Inaccurate for Assessing Functional Outcomes After Radical Prostatectomy

A recent report examined rates of urinary incontinence and erectile dysfunction following radical prostatectomy by evaluating administrative claims data. However, the validity of this approach for reporting functional outcomes has not been established. Therefore, we determined the prognostic value of administrative claims data for reporting urinary incontinence and erectile dysfunction after radical prostatectomy. We identified 562 patients who underwent radical prostatectomy from 2004 to 2007 and were followed at our institution with self-reported standardized survey data available at least 1 year after surgery. Urinary incontinence was assessed by self-reported pad use and the urinary function domain of the Expanded Prostate Cancer Index. Erectile dysfunction was assessed with the International Index of Erectile Function. These results were then compared with administrative claims data using ICD-9 and Hospital International Classification of Diseases Adapted codes for urinary incontinence and erectile dysfunction. Administrative claims data demonstrated a poor correlation with patient self-reported questionnaire data. The administrative identification of erectile dysfunction was associated with a sensitivity of 0.598 and a specificity of 0.591. Poor correlation was also illustrated by the low kappa correlation coefficient of 0.184. Similarly urinary incontinence was poorly correlated with self-reported pad use and the urinary function domain of the Expanded Prostate Cancer Index (correlation coefficient 0.195). Administrative claims data correlate poorly with validated questionnaire data when assessing functional outcomes after radical prostatectomy such as urinary incontinence and erectile dysfunction. Therefore, outcomes data generated using this approach may not reflect the development or severity of such complications.

Pelvic Inflammatory Disease

Pelvic inflammatory disease (PID) is an inflammatory process elicited by the migration of pathogenic microorganisms from the lower to upper genital tract. Although PID is known to increase the risk of tubal factor infertility, many other aspects of this disease remain less well defined. For example, while PID is often caused by Chlamydia trachomatis or Neisseria gonorrhoeae infection, frequently, neither of these bacterial pathogens is isolated from the upper genital tract of women with PID. This etiologic ambiguity also creates uncertainty regarding the decision to include antibiotics effective against genital mycoplasmas and anaerobic vaginal flora in PID treatment. Moreover, since clinical signs and laboratory measurements do not precisely identify all PID cases, and as the accuracy with which imaging modalities identify upper genital tract inflammation is not firmly established, PID remains a diagnostic challenge. It is clear, however, that better understanding of disease pathogenesis, diagnosis, and treatment is needed to improve the care provided women with PID, and this issue of Infectious Diseases in Obstetrics and Gynecology was constructed to deliver specific focus on these topics. Three papers in this issue examine the role of C. trachomatis in PID pathogenesis. The first examines findings from C. trachomatis infection control programs that have altered our understanding of the host immune response to chlamydial infection, and considers implications of these findings for prophylactic vaccine development. The second paper concisely reviews how non-human primate models of chlamydia infection have improved our understanding of PID pathogenesis, while the third explores a possible association between C. trachomatis-specific humoral immunity and genital tract inflammation. The fourth PID pathogenesis-focused paper is a case report that reminds readers of the link between Actinomyces israelii and this disease among women using intrauterine devices, while the fifth reviews evidence supporting Mycoplasma genitalium as a cause of PID. The five remaining papers in this issue address PID diagnosis and treatment. The first diagnosis-related paper defines a practical approach for the identification of women with PID, while the second tests an algorithm for PID case identification in epidemiological research using administrative diagnostic codes rather than the more unwieldy medical record review. The third describes an investigation among a cohort of women with high prevalence of HIV-1 that found that the conventional markers of histologic endometritis (i.e., neutrophils and plasma cells) are, at least in this population, unreliable surrogate markers for laparoscopically confirmed salpingitis. The fourth paper reviewed existing literature regarding serologic diagnosis of C. trachomatis infection in order to construct an algorithm for chlamydial serologic antibody testing in clinical work-up of the infertile couple. We close this issue with a comprehensive review of the antimicrobial therapies currently available for PID treatment. Thomas L. Cherpes Peter A. Rice Richard L. Sweet

Anemia in Chronic Heart Failure

Anemia occurs commonly in patients with chronic heart failure (CHF) and has been proposed as a novel therapeutic target in this patient population. The present review will summarize the current knowledge of the prevalence, causative factors, and pathophysiological correlates of anemia in CHF. Potential risks and benefits of therapy with erythropoietic agents for the treatment of anemia in CHF will also be discussed. Prevalence of Anemia in CHF Estimates of the prevalence of anemia in patients with CHF and low ejection fraction range widely from 4% to 61% (median 18%).1–15 Variability in estimated prevalence is partly attributable to use of inconsistent definitions of anemia in individual reports. The World Health Organization definition of anemia (hemoglobin concentration 13.0 g/dL in men and 12.0 g/dL in women) takes into account known gender differences in distribution of hemoglobin values, 16 whereas the National Kidney Foundation defines anemia as hemoglobin 12 g/dL in men and postmenopausal women. 17 These standard definitions of anemia are not based on wellestablished physiological or population norms. Published reports in CHF populations have used these and other study-specific definitions of anemia (including other arbitrary or statistically defined hemoglobin and hematocrit categories and administrative diagnostic codes from hospital records). Despite these inconsistencies in the definition of anemia cases, most studies indicate that the prevalence of anemia is increased in CHF populations with comorbid kidney disease, advanced age, and more severe symptoms (range, 30% to 61%) when compared with less symptomatic ambulatory populations (range, 4% to 23%). In patients with CHF and preserved ejection fraction, the few published reports indicate that anemia is also highly prevalent in this group.18–20 Underlying Cause of Anemia in CHF Anemia occurs when there is a deficiency in new erythrocyte production relative to the rate of removal of aged erythrocytes. Erythropoietin, a 30.4-kDa glycoprotein growth factor produced primarily by kidney, is the key component of the homeostatic system for regulation of red blood cell mass and tissue oxygen delivery.21–24 Erythropoietin prevents the programmed cell death of erythrocyte progenitor cells and thereby stimulates their proliferation, maturation, and terminal differentiation. 23 Any abnormality that reduces renal secretion of or bone marrow response to erythropoietin may result in anemia. Iron deficiency is present in 30% of anemic patients with CHF, so the majority of observed anemia is normocytic, often classified as anemia of chronic disease. Clinical characteristics commonly associated with increased risk of anemia in CHF populations are listed in the Table. Although risk factors for anemia identified in cross-sectional studies do not provide evidence of a causal link, these observations suggest that several distinct mechanisms may commonly contribute to anemia in patients with CHF. Several of the most important potential causal pathways will be discussed briefly below and are summarized in Figure 1. Chronic kidney disease is a common comorbidity in patients with CHF and is a strong independent predictor of increased risk of anemia in several studies. In chronic kidney disease populations without heart failure, moderate to severe kidney disease (defined as glomerular filtration rate [GFR] 60 mL/min) is associated with diminished erythropoietin production and a progressive decrease in hemoglobin values that is linearly related to reduction in GFR.25 The estimated prevalence of at least moderate chronic kidney disease (defined as GFR 60 mL/min) in CHF populations is 20% to 40%. 4,9,26,27 Anemia is frequently associated with decreased body mass index in published reports, a finding that suggests that patients with cachexia are at greater risk for anemia. Serum levels of proinflammatory cytokines are increased in cachectic patients with CHF and may contribute to development of anemia by several mechanisms. Proinflammatory cytokines including tumor necrosis factor- (TNF-), interleukin-1, and interleukin-6 have been shown to disrupt multiple aspects of erythropoiesis, including reduction of renal erythropoietin secretion, suppression of erythropoietin activity in red blood cell precursors in the bone marrow level, and reduction of bioavailability of iron stores for hemoglobin synthesis.28–31 Proinflammatory cytokines also increase levels of the liverderived peptide hormone, hepcidin.32,33 Hepcidin interacts with ferroportin and other iron transport proteins in the enterocyte to inhibit gut iron absorption and thereby reduces iron bioavailability for hemoglobin synthesis. 32,33 In a mouse