Accuracy of the ICD-10 discharge diagnosis for syncope

Transcription

1 Europace (2013) 15, doi: /europace/eus359 CLINICAL RESEARCH Syncope and implantable loop recorders Accuracy of the ICD-10 discharge diagnosis for syncope Martin Huth Ruwald 1 *, Morten Lock Hansen 2, Morten Lamberts 1, Søren Lund Kristensen 1, Mads Wissenberg 1, Anne-Marie Schjerning Olsen 1, Stefan Bisgaard Christensen 1, Michael Vinther 1, Lars Køber 3, Christian Torp-Pedersen 1, Jim Hansen 1, and Gunnar Hilmar Gislason 1 1 Department of Cardiology, Copenhagen University Hospital, Gentofte, Denmark; 2 Department of Cardiology, Copenhagen University Hospital, Herlev, Denmark; and 3 Department of Cardiology, The Heart Centre, Rigshospitalet, Copenhagen University Hospital, Denmark Received 9 August 2012; accepted after revision 3 October 2012; online publish-ahead-of-print 4 November 2012 Aims Administrative discharge codes are widely used in epidemiology, but the specificity and sensitivity of this coding is unknown and must be validated. We assessed the validity of the discharge diagnosis of syncope in administrative registers and reviewed the etiology of syncope after workup.... Methods Two samples were investigated. One sample consisted of 5262 randomly selected medical patients. The other sample and results consisted of 750 patients admitted or seen in the emergency department (ED) for syncope (ICD-10: R55.9) in three hospitals in Denmark. All charts were reviewed for baseline characteristics and to confirm the presence/absence of syncope and to compare with the administrative coding. In a sample of 600 admitted patients 570 (95%) and of 150 patients from ED 140 (93%) had syncope representing the positive predictive values. Median age of the population was 69 years (IQR: +14). In the second sample of 5262 randomly selected medical patients, 75 (1.4%) had syncope, of which 47 were coded as R55.9 yielding a sensitivity of 62.7%, a negative predictive value of 99.5%, and a specificity of 99.9%.... Conclusion ED and hospital discharge diagnostic coding for syncope has a positive predictive value of 95% and a sensitivity of 63% Keywords Syncope Epidemiology Validity Introduction Syncope is a common condition associated with frequent hospitalizations or emergency department (ED) visits. 1 4 It is difficult to evaluate and is associated with a high mortality rate in selected subgroups of patients Hospital discharge diagnoses are frequently used to identify syncope subjects in epidemiological observational studies, but no validation studies have been carried out on the International Classification of Diseases (ICD), 1994, the 10th revision (ICD-10) discharge diagnosis. Syncope is in the ICD-10 coding system coded as R55.9 (syncope and collapse). Some patients with syncope and an underlying predisposing disorder may have the discharge diagnosis classified elsewhere such as patients experiencing syncope due to aortic stenosis, myocardial infarction, ventricular tachycardia, and similar disorders. Prospective syncope observational programs are extremely rare, primarily because of the high cost and therefore administrative registries have become a highly sought after source of data for disease observation, assessment of health resource consumption, and evaluation of outcomes. The potential benefits of administrative databases are their sizes and rich contents, particularly when combined with other administrative databases. Information on hospitalization and comorbidities from the Danish National Patient Register, where information on all hospital admissions in Denmark has been registered since 1978, is widely used in Danish epidemiological research. 12 At discharge, each * Corresponding author. Tel: , mruwald@hotmail.com Published on behalf of the European Society of Cardiology. All rights reserved. & The Author For permissions please journals.permissions@oup.com.

2 596 M.H. Ruwald et al. What s new The International Classification of Diseases (ICD), 1994, the 10th revision (ICD-10) discharge diagnosis of syncope, R55.9 has a positive predictive value of 95%. The ICD-10 discharge diagnosis of syncope, R55.9 has a sensitivity of 63% and excludes some with a more severe diagnosis. Validation of the administrative coding of syncope yields a tool for epidemiological surveillance. hospital admission is coded with one primary diagnosis and if appropriate one or more secondary diagnoses according to the ICD-10. Validation of ICD-coded discharge data from EDs and hospitals is one important step towards developing a valid and accurate case definition for a condition of interest for future observations. To our knowledge only one study has previously examined syncope discharge diagnosis. Sun et al. 4 validated the discharge diagnosis of ICD-9-CM code (comparable with ICD-10 diagnosis R55.9) in the USA, finding a positive predictive value of 92% of identifying patients with syncope or near syncope. We undertook a syncope validation and review study with a primary objective of determining the accuracy of syncope discharge coding in the setting of three Danish hospitals. Methods Two samples were investigated Sample 1 We retrospectively identified a cohort of subjects who underwent admission or was seen in an ED for medical reasons from 1 to 31 January 2008 in three university hospitals in the Capital Region of Copenhagen. A total of 4045 charts of admitted patients and 1255 ED patients were identified through the electronic patient management system, of which 38 charts could not be accessed yielding a total of 5262 reviewed charts. These charts were divided among the authors who all used the latest definition of syncope according to the European Society of Cardiology (ESC) as a screening tool for syncope in the recorded history of the actual cause of hospitalization or referral. This sample was used for calculation of the sensitivity, negative predictive value, and specificity. Inter-observer variation was done by blinding two reviewers and comparing their results from the review of 200 charts. Sample 2 To obtain a valid positive predictive value and to record baseline characteristics for the syncope patients we retrospectively identified a cohort of subjects who underwent a hospitalization for syncope according to the R55.9 diagnosis (primary or most responsible discharge diagnosis only) from 1 January 2007 to 31 December 2010 in the same three hospitals. A total of 1223 charts of admitted patients with syncope were identified through the electronic patient management system, 23 charts were insufficient for documentation or the chart could not be accessed by the reviewers. From this overall syncope population of 1200, we randomly selected 50% from each hospital of the total admitted patients for individual validation of their syncope, while a random selection of 50 patients per hospital for a total of 150 was used for ED validation. Next, we calculated the positive predictive value for the sample and analyzed the results within subgroups based on type of hospital and type of contact. For each patient, we determined if the hospital chart documentation satisfied the definition and diagnosis of syncope according to the ESC. 13 The current definition of syncope is a total loss of consciousness (T-LOC) characterized by a rapid onset, short duration, and spontaneous complete recovery. Those presenting a traumatic cause for T-LOC were excluded as well as epileptic seizures and psychogenic T-LOC, thus not fulfilling the criteria for syncope. The term addressed pre-syncopal or near-syncope characterized by lightheadedness, nausea, and weakness was not considered sufficient to fulfil the criteria. Hospitals chosen for validation Of the three hospitals, one is a major centre of cardiology with a specialized syncope unit and catheter laboratory for pacemaker implantations, coronary angiography, and cardiac ablations, while the other two are representatives of large volume hospitals with large open-referral EDs and designated departments of internal medicine and neurology. All charts from admitted patients were read independently, reviewing all notes from the physician history and physical examination, physicians notes during rounds and at discharge. On the basis of this information and the presence or absence of syncope each individual was assigned to each chart. The reviews were divided among the reviewers (mean value 501 charts per reviewer). Statistics Patient demographics (age, sex, pharmacotherapy, and comorbidity) were analysed for all patients identified as having syncope based on the administrative coding in the database. Student s t-tests and chi-square tests were used to assess differences between patients included in the chart review. The validity of the coding was described as positive predictive value defined as the proportion of patients that actually had the condition by a positive history of syncope in the chart with an administrative coding of syncope. The sensitivity was calculated in the first sample by dividing the total amount of patients with syncope recorded as R559 with the total amount of syncope as found in the review. All data management and analyses were conducted using SAS Version 9.2 (SAS Institute, Inc., Cary, NC, USA), and findings with P, 0.05 were considered statistically significant. Ethics The study was approved by the Danish Data Protection Agency (ref , int. ref: GEH ). Ethical approval in retrospective register-based analysis is not compulsory in Denmark. Results A total of 750 charts from patients with syncope (R559) were reviewed systematically: 150 ED charts and 600 inpatient charts, revealing 140 and 570 cases fulfilling syncope criteria, respectively. The calculated positive predictive value of the administrative coding was high with a positive predictive value of 93% in an ED and 95% in admitted patients. The numbers of patient charts per hospital and patient baseline characteristics of comorbidities and pharmacotherapy are presented in Table 1. The median age of the

3 Accuracy of the ICD-10 discharge diagnosis for syncope 597 Table 1 Baseline characteristics Characteristics Number (%) Number (%) Hospital 1 Number (%) Number (%) Hospital 3 Total Internal medicine and Hospital 2 Internal medicine, neurology departments Cardiology cardiology and neurology departments... Number of patients 570 (100) 171 (30) 158 (27.7) 241 (42.3) Men 297 (52) 82 (48) 89 (56) 126 (52) Age years (IQR) 68.5 (53 81) 65 (40 80) 68 (52 80) 70 (57 82) Previous syncope 130 (23) 39 (23) 39 (25) 52 (22) Prodromal symptoms 249 (44) 87 (51) 50 (32) 112 (46) Comorbidities Ischaemic heart disease 126 (22) 37 (22) 33 (21) 56 (23) Peripheral vascular disease 78 (14) 30 (18) 26 (16) 22 (9) Previous myocardial infarction 87 (15) 23 (13) 23 (15) 41 (17) Systemic hypertension 288 (51) 83 (49) 87 (55) 118 (49) Previous or current atrial fibrillation 77 (14) 19 (11) 27 (17) 31 (13) Other arrhythmias 37 (6) 11 (6) 13 (8) 13 (5) Previous stroke 72 (13) 21 (12) 11 (7) 40 (17) Congestive heart failure 61 (11) 17 (10) 20 (13) 24 (10) Previous or ongoing cancer 49 (9) 18 (11) 8 (5) 23 (10) Chronic obstructive pulmonary disorder 46 (8) 16 (9) 10 (6) 20 (8) Diabetes 56 (10) 18 (11) 14 (9) 24 (10) Cardiac pacemaker or ICD unit 21 (4) 6 (4) 4 (3) 11 (5) Epilepsy 7 (1) 2 (1) 3 (2) 2 (1) Alcoholism 18 (3) 10 (6) 1 (1) 7(3) Dementia 29 (5) 9 (5) 4 (3) 16 (7) Depression 25 (4) 2 (1) 1 (1) 22 (9) Previous PCI or CABG 61 (11) 17 (10) 18 (11) 26 (11) Previous or current smoker 223 (39) 78 (46) 56 (35) 89 (37) Alcohol intake above recommended level 124 (22) 57 (33) 27 (17) 40 (17) Pharmacotherapy Statins 148 (26) 39 (23) 45 (28) 64 (27) Aspirin 188 (33) 43 (25) 60 (38) 85 (35) Platelet inhibitors 57 (10) 19 (11) 11 (7) 27 (11) Beta-blockers 140 (25) 34 (20) 40 (25) 66 (27) ACEI/ARB 210 (37) 62 (36) 56 (35) 92 (38) Digoxin 21 (4) 4 (2) 6 (4) 11 (5) Nitrates 34 (6) 12 (7) 9 (6) 13 (5) Calcium channel blockers 85 (15) 26 (15) 28 (18) 31 (13) Spironolactone 21 (4) 5 (3) 4 (3) 12 (5) Thiazide 111 (19) 24 (14) 36 (23) 51 (21) Loop diuretics 85 (15) 22 (13) 25 (16) 38(16) VKA 35 (6) 9 (5) 10 (6) 16 (7) Class Ic antiarrhythmic drugs 1 (0) 0 (0) 0 (0) 1 (0) Class III antiarrhythmic drugs 13 (2) 5 (3) 6 (4) 2 (1) Bronchial dilators 42 (7) 15 (9) 8 (5) 19 (8) Anxiolytics 56 (10) 20 (12) 13 (8) 23 (10) Antipsychotics 27 (5) 10 (6) 6 (4) 11 (5) Antidepressants 76 (13 22 (13) 14 (9) 40 (17) Glucose-lowering drugs 51 (9) 15 (9) 10 (6) 26 (11) Non-steroidal anti-inflammatory drugs 40 (7) 18 (11) 9 (6) 13 (5) Morphine a 37 (7) 14 (8) 5 (3) 18 (7) ICD, implantable cardioverter defibrillator; PCI, percutaneous coronary intervention; CABG, coronary artery bypass graft; ACEI, angiotensin-converting enzyme inhibitor; ARB, angiotensin receptor blocker; VKA, vitamin K antagonists. a Morphine use was appropriate in all patients taking morphine. Age is given in median and interquartile range (IQR).

4 598 M.H. Ruwald et al. Table 2 Validity of the principal discharge diagnosis of syncope R55.9 All medical patients... Syncope positive Syncope negative... Test outcome R559 True positive (TP) ¼ 47 False positive (FP) ¼ 2 Positive predictive value ¼TP/(TP + FP) ¼47/(47 + 1) ¼95.9% Other diagnosis False negative (FN) ¼ 28 True negative (TN) ¼ 5185 Negative predictive value ¼TN/(FN + TN) ¼5185/( ) 99.5% Sensitivity Specificity ¼TP/(TP + FN) ¼TN/(FP + TN) ¼47/( ) ¼5185/( ) 62.7% ¼99.9% A more precise positive predictive value was found by reviewing 600 charts from admitted patients only with R559 diagnosis, revealing 570 true syncope ¼ 570/600 ¼ 95.0%. The positive predictive value of syncope in an ED was found to be 93.3 (140 out of 150). Table 3 Other discharge codes used where syncope was one of the predominant symptoms ICD-10 code Diagnosis Number (contributing percentage)... I495 Sick sinus node syndrome 4 (8.5%) Z033 Observation on suspicion of neurological disease 3 (6.4%) I214 Non-ST-elevation acute myocardial infarction with no Q-wave 3 (6.4%) I649, I951, J189, K290, Z039, R298, I455, I489, I350, I422, J969, Z034, I459, I210, K553, I509, T430 Cerebral infarction, orthostatic hypotension, pneumonia, acute bleeding gastritis, symptom from the nervous system, observation on suspicion of disease, unspecified heart block, atrial fibrillation, aortic stenosis, hypertrophic cardiomyopathy, insufficient respiration, observation on suspicion of myocardial infarction, Adam-stokes syndrome, anterior myocardial infarction with Q-wave, unspecified heart failure, poisoning with tricyclic antidepressives patients was 68.5 years [interquartile range (IQR): years, and 48% of them were female. Of the 570 patients; a history of hypertension was present in 51%, ischaemic heart disease in 22%, atrial fibrillation in 14%, congestive heart failure in 11%, and diabetes in 10%. No major differences between the three hospitals were seen in terms of comorbidities or pharmacotherapy. The sensitivity was calculated through validation of the first sample revealing 75 (1.4%) cases of syncope of which 47 (62.7%) were coded as R559. The negative predictive value and the specificity were as high as 99%. The terms are presented in Table 2. Inter-observer variation was compared in two blinded reviewers and no disagreement was found. Roughly one-third (28 patients) of patients with syncope as a primary or significantly contributing symptom are thus coded as other discharge diagnoses. The discharge diagnoses most often used was I49.5 (8.5%) sick sinus node syndrome, Z03.3 (6.4%) observation on suspicion of neurologic disease, and I21.4 (6.4%) non-st-elevation acute myocardial infarction with no Q-wave. Other codes used as discharge diagnoses covering the remaining cases of syncope are presented in Table 3. Discussion 1 (2.1%) The objective of this study was to determine whether the administrative coding of syncope could accurately be used to identify patients with syncope. The major finding of the study was a very high validity and positive predictive value (95%) of the principal discharge diagnosis of syncope R55.9. Another important finding is that 62.7% of cases with syncope are covered by the discharge diagnosis of R55.9, whereas the remaining part of syncope is covered by a wide range of discharge diagnosis, mostly cardiologic, such as third-degree atrioventricular block, myocardial infarction, and some observation codes. To our knowledge, this study is the first to systematically validate administrative data against medical chart data for the identification of syncope, either admitted or seen in an ED. A high positive predictive value suggests that the proposed coding can be used to identify patients with syncope in administrative databases with a high level of accuracy, introducing the possibility of epidemiological surveillance, whereas one-third of patients suffering syncope are not included in the R55.9 diagnosis limiting the use of R55.9 when investigating

5 Accuracy of the ICD-10 discharge diagnosis for syncope 599 syncope as an outcome in epidemiological studies or in drugrelated adverse effects. However, reliable data can be extracted from this administrative coding regarding hospitalization costs due to syncope, in-hospital days, and estimations in general. The general evaluation of a patient with syncope involves a myriad of diagnostic tests, but the annual cost of syncope-related admissions is very hard to calculate as no administrative coding is specifically designed to syncope. It is, however, estimated that the annual costs of syncope-related admissions in the USA exceeds $2 billion. 14 Syncope accounted for 1.4% of all medically hospitalized patients which is comparable with other retrospective studies and to our recent study on nationwide incidence. 1,2,4,15,16 Validity Our findings support the notion that population-based hospital discharge administrative data can be used to identify a syncope cohort accurately. That is the only way appropriate adjustment for the prevalence and hospital incidence of syncope could be made. Syncope ICD-10 coding by physicians for ED visits and admission is accurate when looking at other mimics of syncope, such as pre-syncope, dizziness, vertigo, alcohol intoxication, exhaustion, epilepsy, transient ischaemic attack, hypoglyacemia, and convulsions. It should be noted that accordingly only 5% of patients are actually non-syncopal T-LOC and not true syncopes. Likewise it should be noted that 37% of all patients with syncope are missed in this diagnostic coding and thus are represented by other ICD-10 diagnoses. Accuracy of syncope coded in administrative data is high across all hospital settings. We showed that accuracy of syncope coding by physicians is high, whether the encounter (patient visit) was an ED visit or a hospital admission, and whether it was at a centre with a syncope clinic or a non-specialized unit. This probably reflects that physicians in these settings use the same coding guidelines throughout the health region and we would expect similar findings in other hospitals across the country. The finding corresponds with the lone validation study of administrative coding of syncope 4 from the USA. A high validity has been found in other countries of other conditions such as epilepsy, amyotrophic lateral sclerosis, and stroke, but other validation studies have proved that this cannot be extrapolated to all diseases or symptoms. 20,21 Based on our findings, registries can be used to identify patients with a general diagnosis of syncope accurately including a broad mixture of assumed aetiologies, but general inferences from coding cannot be made to identify any specific cause. Other discharge diagnosis codes used whenever syncope was a predominant symptom As suspected earlier this retrospective analysis R55.9 did not cover all cases of syncope. The findings that one-third of the patients received other discharge diagnoses were actually lower than that anticipated. Most of the other discharge diagnosis used were diagnoses with well-established aetiological and pathophysiological coherence to syncope as a symptom covering a wide spectrum of specific cardiac diseases but also less-specific observational coding. In other words we do not find the ICD-10 particularly helpful in terms of describing any established aetiological explanation to syncope. However, the possibility of using Adam stokes syndrome and orthostatic hypotension remain applicable in the ICD-10 system. Thus, the optimal description of aetiological considerations in syncope in the ICD-10 system is unfortunately at the moment to use either syncope R55.9 as the primary diagnosis and a secondary diagnosis that most accurately covers the cause of the syncope, such as aortic stenosis, myocardial infarction, and the like, or consequently use R55.9 as a secondary diagnosis whenever syncope is a predominant symptom. Limitations First, we assessed only ED and inpatient databases. A majority of syncope cases are treated in outpatient and general practitioner (GP) settings and generalizing our findings to outpatient or GP databases should not be done. Therefore, data from those individuals who do not seek medical attention or who are only seen in outpatient clinics or offices are not captured. For syncope, this results in a slight bias to more severe cases of syncope because patients with milder/less severe symptoms may not seek medical attention. Secondly, our gold standard relied solely on chart documentation. However, when admitted to an ED or in a hospital, patients tend to have more extensive workup. Therefore, the charts most likely reflected the true diagnoses. Conclusion In conclusion, ICD-10 coding for the identification of those with syncope who visit an ED or who are admitted to a hospital is highly specific. To identify a cohort of true syncope patients, the ICD-10 coding R55.9 can be used with a positive predictive value of 95% and a sensitivity of 63%. Conflict of interest: none declared. Funding This study was supported by an unrestricted grant from the Danish Heart Association (12-04-R90-A ) and the Lundbeck Foundation (R108-A10415). References 1. Sarasin FP, Louis-Simonet M, Carballo D, Slama S, Rajeswaran A, Metzger JT et al. Prospective evaluation of patients with syncope: a population-based study. Am J Med 2001;111: Quinn J, McDermott D, Kramer N, Yeh C, Kohn MA, Stiell I et al. Death after emergency department visits for syncope: how common and can it be predicted? Ann Emerg Med 2008;51: Day SC, Cook EF, Funkenstein H, Goldman L. Evaluation and outcome of emergency room patients with transient loss of consciousness. Am J Med 1982;73: Sun BC, Derose SF, Liang LJ, Gabayan GZ, Hoffman JR, Moore AA et al. Predictors of 30-day serious events in older patients with syncope. Ann Emerg Med 2009;54: e Soteriades ES, Evans JC, Larson MG, Chen MH, Chen L, Benjamin EJ et al. Incidence and prognosis of syncope. N Engl J Med 2002;347: Alshekhlee A, Shen WK, Mackall J, Chelimsky TC. Incidence and mortality rates of syncope in the United States. Am J Med 2009;122:181 8.

6 600 M.H. Ruwald et al. 7. Lipsitz LA, Wei JY, Rowe JW. Syncope in an elderly, institutionalised population: prevalence, incidence, and associated risk. Q J Med 1985;55: Getchell WS, Larsen GC, Morris CD, McAnulty JH. Epidemiology of syncope in hospitalized patients. J Gen Intern Med 1999;14: Kapoor WN. Evaluation and outcome of patients with syncope. Medicine (Baltimore) 1990;69: Kapoor WN, Hanusa BH. Is syncope a risk factor for poor outcomes? Comparison of patients with and without syncope. Am J Med 1996;100: Kapoor W, Snustad D, Peterson J, Wieand HS, Cha R, Karpf M. Syncope in the elderly. Am J Med 1986;80: Andersen TF, Madsen M, Jorgensen J, Mellemkjoer L, Olsen JH. The Danish national hospital register. A valuable source of data for modern health sciences. Dan Med Bull 1999;46: Moya A, Sutton R, Ammirati F, Blanc JJ, Brignole M, Dahm JB et al. Guidelines for the diagnosis and management of syncope (version 2009). Eur Heart J 2009;30: Sun BC, Emond JA, Camargo CA Jr. Direct medical costs of syncope-related hospitalizations in the United States. Am J Cardiol 2005;95: Sun BC, Emond JA, Camargo CA Jr. Characteristics and admission patterns of patients presenting with syncope to U.S. Emergency Departments, Acad Emerg Med 2004;11: Ruwald MH, Hansen ML, Lamberts M, Hansen CM, Hojgaard MV, Kober L et al. The relation between age, sex, comorbidity, and pharmacotherapy and the risk of syncope: a Danish nationwide study. Europace Kokotailo RA, Hill MD. Coding of stroke and stroke risk factors using international classification of diseases, revisions 9 and 10. Stroke 2005;36: Jette N, Reid AY, Quan H, Hill MD, Wiebe S. How accurate is ICD coding for epilepsy? Epilepsia 2010;51: Stickler DE, Royer JA, Hardin JW. Validity of hospital discharge data for identifying cases of amyotrophic lateral sclerosis. Muscle Nerve 2011;44: Pedersen M, Klarlund M, Jacobsen S, Svendsen AJ, Frisch M. Validity of rheumatoid arthritis diagnoses in the Danish National Patient Registry. Eur J Epidemiol 2004;19: Nielsen EH, Lindholm J, Laurberg P. Use of combined search criteria improved validity of rare disease (craniopharyngioma) diagnosis in a national registry. J Clin Epidemiol 2011;64: