Comparison of the performance of mental health, drug and alcohol comorbidities based on ICD-10-AM and medical records for predicting 12-month outcomes in trauma patients

Comparison of the performance of mental health, drug and alcohol comorbidities based on ICD-10-AM... Background: Many outcome studies capture the presence of mental health, drug and alcohol comorbidities from administrative datasets and medical records. How these sources compare as predictors of patient outcomes has not been determined. The purpose of the present study was to compare mental health, drug and alcohol comorbidities based on ICD-10-AM coding and medical record documentation for predicting longer-term outcomes in injured patients. Methods: A random sample of patients (n = 500) captured by the Victorian State Trauma Registry was selected for the study. Retrospective medical record reviews were conducted to collect data about documented mental health, drug and alcohol comorbidities while ICD-10-AM codes were obtained from routinely collected hospital data. Outcomes at 12-months post-injury were the Glasgow Outcome Scale – Extended (GOS-E), European Quality of Life Five Dimensions (EQ-5D-3L), and return to work. Linear and logistic regression models, adjusted for age and gender, using medical record derived comorbidity and ICD-10-AM were compared using measures of calibration (Hosmer-Lemeshow statistic) and discrimination (C-statistic and R ). Results: There was no demonstrable difference in predictive performance between the medical record and ICD- 10-AM models for predicting the GOS-E, EQ-5D-3L utility sore and EQ-5D-3L mobility, self-care, usual activities and pain/discomfort items. The area under the receiver operating characteristic (AUC) for models using medical record derived comorbidity (AUC 0.68, 95% CI: 0.63, 0.73) was higher than the model using ICD-10-AM data (AUC 0.62, 95% CI: 0.57, 0.67) for predicting the EQ-5D-3L anxiety/depression item. The discrimination of the model for predicting return to work was higher with inclusion of the medical record data (AUC 0.69, 95% CI: 0.63, 0.76) than the ICD-10-AM data (AUC 0.59, 95% CL: 0.52, 0.65). Conclusions: Mental health, drug and alcohol comorbidity information derived from medical record review was not clearly superior for predicting the majority of the outcomes assessed when compared to ICD-10-AM. While information available in medical records may be more comprehensive than in the ICD-10-AM, there appears to be little difference in the discriminative capacity of comorbidities coded in the two sources. Keywords: Trauma, Mental comorbidities, Post-injury outcomes, Prediction, Validation, Medical record coding * Correspondence: tu.nguyen@monash.edu Department of Epidemiology and Preventive Medicine, Monash University, Melbourne, Victoria, Australia Full list of author information is available at the end of the article © The Author(s). 2018 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Nguyen et al. BMC Health Services Research (2018) 18:408 Page 2 of 8 Background of Victoria [20]. The VSTR follows up all adult survivors Injuries accounted for 11% of the global burden of disease to discharge after their injury including function, health- in terms of Years Lived with Disability in 2015 [1]. related quality of life (HRQoL), return to work and post- Risk-adjustment analyses of long-term outcomes after in- discharge mortality. Standardised telephone interviews are jury have included compensable status, injury mechanism, conducted by trained interviewers [21]. A random sample injury severity, perceived level of social support, socioeco- of 500 patients registered on the VSTR were selected for nomic position and educational level [2–5]. Mental health, the study. Patients were eligible for inclusion in this study drug and alcohol comorbidities have been cited as factors if they met each of the following criteria: that impact on recovery and quality of life following injury [6, 7]. However, the importance of these comorbidities i. Aged 18 years or older; may be under-estimated in risk modelling of long-term ii. Definitively managed at one of the two adult major outcomes [8]. Data sources used in research studies to trauma services in Victoria; capture comorbid conditions are ostensibly linked to the iii. Survived to 12 months post-injury and 12-month research question, patient population and resources avail- follow-up completed. able. Consequently, poor comorbidity capture in different data sources has an impact on the resulting comorbidity Procedures profile of injured people and the ability to predict patient In Australia, hospital data are collected in the form of outcomes after injury. International Statistical Classification of Diseases, Tenth Although time- and labour-intensive to review, medical Revision, Australian Modification (ICD-10-AM) codes and records represent a comprehensive source of comorbidity apply to all public and private hospitals. In Victoria, each data, due to the detailed clinical information contained code contains a prefix with “P” representing principal diag- within the notes [9]. Alternatively, routinely collected ad- nosis requiring treatment during the stay, “A” representing ministrative data are commonly relied on in epidemiological additional diagnoses and “C” representing in-hospital research as they are relatively cost-effective and readily complications. These codes are assigned by the hospital available [10]. Findings from previous research suggest coders as part of routinely collected hospital data and that mental health, drug and alcohol comorbidities derived are used for activity-based funding purposes [22]. from administrative data has been shown to be a strong Data were extracted from the VSTR for patients in predictor of in-hospital mortality [11], hospital-acquired the sample including clinical and demographic data and infection [12] and longer hospital length of stay [13]. How- follow-up outcome data collected at 12-months post- ever, prediction of longer-term recovery outcomes using injury. This included patients’ compensation status by comorbidities based on administrative data have generated the no-fault, third party insurers for road traffic injury inconsistent results [14–16]. This could be explained by and work-related injury and the Charlson Comorbidity biases in administrative data, such as those associated with Index (CCI), a weighting used to assess comorbidity [23]. collecting data in order to account for different types of The ICD-10-AM “P” and “A” prefixed codes, which are patient episodes and services provided, in the case of provided to the VSTR by the hospitals, were also extracted. activity-based funding [17]. Retrospective review of the medical records of the sample Direct comparisons between administrative coded mental of patients was conducted, by the author of this paper health, drug and alcohol diagnoses with other sources in consultation with researchers with clinical and health of data in injured populations are scarce [18, 19]. The expertise, at the two adult major trauma services. completeness of administrative coding largely relies on To enable comparison with ICD-10-AM codes, all the quality of the medical record documentation but little mental health, drug and alcohol comorbidities described in is known about the impact of using the two alternative the medical record were captured based on the categories data sources on long-term outcome modelling. The aim of of mental and behavioural disorder (Chapter V) specified this study was to compare the contribution of two different by ICD-10-AM. If a mental health, drug or alcohol con- sources of mental health, drug and alcohol comorbidity dition was documented, based on clinical notation in data in predicting 12-month functional, return to work addition to substantiating evidence such as an alcohol and health-related quality of life outcomes in a major withdrawal scale or psychiatrist referral letter for on- trauma population. going treatment, this was recorded on a project data collection form in addition to the specified diagnosis. Methods Pre-existing mental health disorders were captured from Study design and inclusion criteria the medical record based on the following categories: The Victorian State Trauma Registry (VSTR) is a organic mental disorders, schizophrenia, mood disorders, population-based registry which captures information neurotic disorders, behavioural and personality disorders about virtually all major trauma patients across the state and other specified mental disorders. Nguyen et al. BMC Health Services Research (2018) 18:408 Page 3 of 8 A single binary indicator variable were generated to mobility, self-care, usual activities, pain/discomfort represent the presence of ICD-10-AM coded mental, drug and anxiety/depression. A linear regression model was and alcohol use disorders. This indicator variable included fitted for the continuous outcome EQ-5D-3L utility the major categories: alcohol use disorders (F10), drug score. Age was categorised for inclusion in the models use disorders (F11-F15, F16, F19) and mental disorders as it was not linearly related to the log odds of the out- (F00-F03, F07, F09, F20-F48, F60-F69, F90-F99, U792, comes of interest for the logistic models in its continuous U793). form. The following models were fitted for each the out- Outcome measures comes measured: The primary outcome measure recorded on the VSTR used to assess patients’ level of function is the Glasgow Outcome i) Age and gender; Scale-Extended (GOS-E). The GOS-E classifies patients ii) Medical record coded mental health, drug and into eight levels of function from death (GOS-E = 1), upper alcohol comorbidities, adjusted for age and gender severe disability (GOS-E = 4) to upper good recovery iii) ICD-10-AM coded mental health, drug and alcohol (GOS-E = 8), indicating a complete return to normal comorbidities, adjusted for age and gender activities of daily life [24]. The three-level European Quality of Life Five Dimensions (EQ-5D-3L) is used to Discrimination was evaluated using the C-statistic, or measure of health-related quality of life and consists of area under the receiver operating characteristic curve five dimensions (mobility, self-care, usual activities, pain/ (AUC), which ranges from zero to one, with a value of discomfort, and anxiety/depression), with each item having 0.5 representing a model with no predictive power and a three possible responses: no problems, some problems and value of 1 indicating a model with perfect predictive power. extreme problems. The EQ-5D-3L item responses are used A value between 0.7 and 0.8 is considered acceptable to calculate a single utility score using age and gender spe- predictive performance, while values greater than 0.8 cific social preference weights [25]. Algorithms were used demonstrate excellent predictive performance [28]. To to convert EQ-5D-3L responses into health utility scores investigate model fit, likelihood ratio (LR) tests were [range: 0 (equivalent to death) to 1 (equivalent to full used. A higher LR statistic and significant p-value supports health)]. The original preference weights for the the hypothesis that including the comorbidity data was a EQ-5D-3L from the UK population were applied as significant improvement in model fit over the model of age these are most commonly used [25]. If patients worked and gender alone [29]. Hosmer-Lemeshow (H-L) statistics or studied prior to injury, they were asked whether they were used to compare model calibration of logistic regres- had returned to work or study. sion models [28]. A linear regression model was fitted for To establish whether the capacity to predict patient the EQ-5D-3L utility scores outcome and the coefficient of outcomes at 12-months post-injury differed between determination (R ) of the models used to assess pre- medical record review of drug, alcohol and mental health dictive performance. A p-value < 0.05 was considered comorbidities and those coded in ICD-10-AM, comparison significant. All analyses were conducted using Stata 13.0 of model performance was undertaken. A ‘good recovery’ (StataCorp, College Station, TX). wasdefined as aGOS-E scoreof7–8and a score less than seven represented ongoing functional limitations. The EQ-5D-3L item responses were dichotomised as Results no problems versus some or severe problems. The clinical and demographic characteristics of the study sample are shown in Table 1. There was an even Analysis distribution of patients among the age groups and most In order to detect a statistically significant difference be- patients were male. The median (IQR) injury severity tween the area under the Receiver Operating Characteristic score was 17 (14–22), and these figures are representative (AUC) of 0.7 and 0.8 with a 5% one-sided type I error and of major trauma patients in Victoria [30]. Most patients 80% power, 158 patients was the smallest sample size did not have a Charlson Comorbidity Index condition required for each hospital [26]. We over-sampled patients recorded and did not claim compensation for their beyond this minimum sample size and accessed the medical injuries (Table 1). records of 250 patients from each hospital, in order to Based on the GOS-E measure, 41% (n =205) of patients account for issues with availability of paper medical records in the sample reported a good recovery at and missing volumes which are common in medical record 12-months post-injury. According to the H-L statistic, review studies [27]. Binary logistic regression models the models predicting functional recovery demonstrated were fitted for the categorical outcomes functional acceptable calibration. Inclusion of the mental health, drug recovery, return to work, and the EQ-5D-3L items of and alcohol comorbidity data, from either the medical Nguyen et al. BMC Health Services Research (2018) 18:408 Page 4 of 8 Table 1 Demographic and clinical characteristics of patients The EQ-5D-3L was complete for 98% of patients at within the sample (n = 500) 12 months. The mean (SD) EQ-5D-3L utility score in Variable Total n (%) the sample was 0.75 (0.29). The prevalence of problems was highest for the mobility and anxiety or depression Age (years) items (Table 3). Overall, the prevalence of problems was 18–24 63 (12.6) lowest for the self-care item. 25–34 92 (18.4) The models predicting problems on each of the dimen- 35–44 98 (19.6) sions of the EQ-5D-3L showed less than acceptable 45–54 91 (18.2) discrimination (Table 4). The capacity to discriminate 55–64 72 (14.4) between patients with and without problems in the mobility, self-care, usual activities, pain/discomfort ≥ 65 84 (16.8) items and utility score was similar for the ICD-10-AM Gender and medical record models except for patients with Male 374 (74.8) anxiety/depression problems. For predicting the EQ-5D-3L Female 126 (25.2) utility score, the highest R was observed for the medical Charlson Comorbidity Index weighting record model. Both models were an improvement over the 0 351 (70.2) age and gender model, but addition of the ICD-10-AM coded conditions was only a marginal improvement 1 115 (23.0) over age and gender (Table 4). Calibration was acceptable ≥ 2 34 (6.8) according to the H-L statistic for predicting all of the Compensation status EQ-5D-3L items. Compensable 180 (36.1) Non-compensable 318 (63.9) Discussion Compensation status not available for n =2 Comorbidity information available in different data sources have their own advantages and limitations, which can record or ICD-10-AM, improved the model fit but none of affect the validity of studies relying on such data [31]. It the models showed acceptable discrimination (Table 2) has been acknowledged that routinely collected adminis- (Additional file 1). trative data sub-optimally represent a patient’s condition Three hundred and thirty eight (68%) patients reported and the totality of their comorbid illnesses, particularly for that they were working or studying prior to injury. Of mental health, drug and alcohol comorbidities which are those who were working/studying prior to injury, 71% historically difficult to characterise, classify and diagnose (n = 240) had returned to work or study at 12-months [32]. Previous analyses of administrative coding in routinely post-injury. The models predicting return to work all collected hospital data compared with medical records have showed less than acceptable discrimination but demon- revealed the prevalence of comorbid conditions are fre- strated acceptable calibration (Table 2). The addition of quently under-estimated [33–35]. Observable discrepancies the medical record data, but not the ICD-10-AM coded between the data sources exist [36], and are driven largely data, improved discrimination over age and gender. by coding rules, activity-based funding, variability in Table 2 Discrimination and calibration statistics for predicting GOS-E good recovery and return to work Good recovery (GOS-E 7–8) Model N AUC (95% CI) H-L statistic (p value) LR-test (p-value) Age and gender 500 0.57 (0.52, 0.63) 2.3 (0.89) – Age, gender and medical record data 500 0.62 (0.57, 0.67) 7.1 (0.52) 11.7 (< 0.01) Age, gender, ICD-10-AM 500 0.58 (0.53, 0.63) 3.3 (0.91) 0.7 (0.42) Return to work N AUC (95% CI) H-L statistic (p value) LR-test (p-value) Age and gender 348 0.55 (0.49, 0.62) 4.3 (0.75) – Age, gender and medical record data 348 0.69 (0.63, 0.76) 7.5 (0.48) 38.1 (< 0.01) Age, gender, ICD-10-AM 348 0.59 (0.52, 0.65) 3.6 (0.89) 5.9 (0.01) Compared to age and gender model Nguyen et al. BMC Health Services Research (2018) 18:408 Page 5 of 8 Table 3 EQ-5D-3L responses at 12-months post-injury the review due to missing documentation or illegible handwriting. As a result, mental health, drug and alcohol EQ-5D-3L dimension n (%) comorbidities obtained from the medical record may not Mobility be reliably predictive of long-term function. Further re- No problems 337 (68.4) search is needed to compare the predictive performance Some or severe problems 156 (31.6) of medical record data with self-report and clinical inter- Self-care view to assess these comorbidities for long-term outcome No problems 398 (80.8) modelling. Some or severe problems 95 (19.3) Mental health, drug and alcohol comorbidity data derived from medical records have been used previously for risk- Usual activities adjustment in studies of return to work [42, 44]. Mental No problems 263 (53.5) health, drug and alcohol comorbidities coded in ICD- Some or severe problems 229 (46.5) 10-AM may not be as reliable or be able to perform to the Pain or discomfort same capacity as medical record abstracted conditions for No problems 278 (56.4) predicting return to work. This is difficult to assess, as few Some or severe problems 215 (43.6) existing studies have evaluated the association between ICD-10 coded mental health comorbidities and returning Anxiety or depression to work after injury. A previous study of major trauma No problems 316 (64.1) patients using ICD-10-AM coded data reported with no Some or severe problems 177 (35.9) significant association with returning to work [15], but this Missing data (n = 7) for mobility, self-care, pain/discomfort may have been attributed to limited statistical power. and anxiety/depression Missing data (n = 8) for usual activities Whilethemedical record mayprovideahigh degreeof detail about mental health factors that are associated with coder interpretation and inconsistent or unclear charting lower odds of returning to work following severe injury, [17, 37]. We sought to compare the statistical performance this information may be incomplete when coded in the of two commonly used data sources: medical record docu- ICD-10-AM. mentation and ICD-10-AM administrative data. Mental Survivors of traumatic injuries often experience deficits health, drug and alcohol comorbidities were shown to in perceived health status, including pain or discomfort, be important for predicting long-term outcomes, as the and difficulty with usual activities and mobility [45]. Aside addition of the comorbidity data fit the data better than from anxiety or depression, there were no differences in age and gender alone. Overall, the addition of comorbidi- the discriminative ability of the ICD-10-AM and medical ties obtained from medical record review did not pro- record data for predicting the remaining EQ-5D-3L vide improvement in discrimination over age and gender, dimensions or utility score. Neither the medical record which was comparable with the ICD-10-AM coded comor- documentation nor the ICD-10-AM data were able to bidities, for most of the outcomes assessed. differentiate between groups at risk of ongoing problems Mental health, drug and alcohol comorbidities are with mobility, self-care, usual activities and pain/discom- associated with delays in recovery and disability after fort with any great accuracy. An explanation for this injury, but the majority of the evidence predicting these finding may be that self-reported measures of comorbidity outcomes have relied on self-report or clinical interview may be a better predictor of the EQ-5D-3L outcomes to assess comorbidity [38–40]. Pre-injury mental health whereas indirect comorbidity measures may not be diagnoses derived from the medical record has been optimal for predicting these health status outcomes associated with poorer function in previous studies [41, 42], [46]. This suggests the importance of other factors not whereas ascertaining pre-existing mental health history explained by mental health, drug and alcohol comor- based on administrative data coding has yielded differing bidities including personal and environmental factors results [16, 43]. Despite the lower prevalence of comorbid- for predicting HRQoL following injury. Other factors ities coded in ICD-10-AM than in the medical record, the that have been shown to be important in the literature findings of this study indicate the capacity to predict func- include compensable status, education, and social tional recovery was largely comparable with the medical circumstances [47–49]. record. This suggests that key mental health, drug and The strengths of the study were the use of long-term alcohol comorbidities, which are predictors and have outcomes data at 12-months post-injury and the random significant impacts on long-term function after injury, are sample of a population-based cohort, which was repre- not comprehensively captured in the medical record at sentative of major trauma patients in Victoria. Although the outset. Notably, some comorbidities within the med- patients from each hospital were over-sampled in order to ical record could have been inadvertently excluded from mitigate potential issues with missing charts, only seven Nguyen et al. BMC Health Services Research (2018) 18:408 Page 6 of 8 Table 4 Discrimination and calibration statistics for predicting EQ-5D-3L dimensions and utility score EQ-5D Mobility Model N AUC (95% CI) H-L statistic (p-value) LR-test (p-value) Age and gender 493 0.56 (0.51, 0.62) 0.6 (0.99) – Age, gender and medical record data 493 0.58 (0.53, 0.63) 2.3 (0.97) 4.3 (0.04) Age, gender, ICD-10-AM 493 0.56 (0.51, 0.62) 3.7 (0.81) 0.2 (0.64) EQ-5D Self-care N AUC (95% CI) H-L statistic (p-value) LR-test (p-value) Age and gender 493 0.58 (0.51, 0.64) 3.6 (0.73) – Age, gender and medical record data 493 0.60 (0.53, 0.66) 4.5 (0.80) 4.6 (0.03) Age, gender, ICD-10-AM 493 0.57 (0.51, 0.64) 6.8 (0.45) 0.4 (0.54) EQ-5D Usual activities N AUC (95% CI) H-L statistic (p-value) LR-test (p-value) Age and gender 492 0.57 (0.52, 0.62) 1.8 (0.94) – Age, gender and medical record data 492 0.59 (0.54, 0.64) 5.1 (0.75) 6.0 (0.01) Age, gender, ICD-10-AM 492 0.57 (0.52, 0.62) 6.2 (0.63) 0.1 (0.72) EQ-5D Pain or discomfort N AUC (95% CI) H-L statistic (p-value) LR-test (p-value) Age and gender 493 0.58 (0.53, 0.63) 3.2 (0.78) – Age, gender and medical record data 493 0.58 (0.53, 0.63) 8.2 (0.41) 2.7 (0.10) Age, gender, ICD-10-AM 493 0.58 (0.53, 0.63) 5.6 (0.58) 0.0 (0.92) EQ-5D Anxiety or depression N AUC (95% CI) H-L statistic (p-value) LR-test (p-value) Age and gender 493 0.55 (0.50, 0.60) 1.7 (0.94) – Age, gender and medical record data 493 0.68 (0.63, 0.73) 9.9 (0.28) 42.1 (< 0.01) Age, gender, ICD-10-AM 493 0.62 (0.57, 0.67) 6.7 (0.57) 10.9 (0.01) EQ-5D-3L utility score NR LR-test (p-value) Age and gender 492 0.01 – Age, gender and medical record data 492 0.06 22.1 (< 0.01) Age, gender, ICD-10-AM 492 0.02 4.0 (0.04) Compared to age and gender model patient records (3%) of the total sample could not be lo- choose the best predictors of the outcome [50], and cated, and these patients were replaced from the VSTR. A the results cannot be interpreted for this purpose with- limitation of the study includes the extent to which these out careful selection and further examination of other findings may be extrapolated to other patients. This re- variables. search relied on ICD-10-AM recorded mental health, drug and alcohol diagnoses which were originally collected pri- Conclusion marily for activity-based funding. The findings may not be While the prevalence of mental health, drug and alcohol applicable to other jurisdictions, which vary in coding comorbidities is lower in ICD-10-AM than medical records, practices and funding systems. This aim of the study was data derived from medical record review was not clearly su- not to develop the best prediction model for long-term perior for predicting the majority of the outcomes assessed. outcomes. The selection of variables to identify the There was no demonstrable difference in discriminative ‘optimal’ risk prediction model is based on trying to capacity between the medical record and ICD-10-AM Nguyen et al. BMC Health Services Research (2018) 18:408 Page 7 of 8 models for predicting the GOS-E, EQ-5D-3L utility sore intellectual content. BJG oversaw the project, contributed to the study design, data interpretation, and drafting and revising of the manuscript. All and EQ-5D-3L mobility, self-care, usual activities and pain/ authors read and approved the final version of the manuscript. discomfort items. Further research is needed to com- pare the contribution of medical record documented Ethics approval and consent to participate This study was approved by the Alfred Hospital Human Research Ethics mental health, drug and alcohol comorbidities with alter- Committee (Project Number: 275/16), the Royal Melbourne Hospital (Project native data sources. Number: 2016.179) and the Monash University Human Research Ethics Committee (Project Number: 0290). Access to the VSTR data for this study was granted by the Victorian State Trauma Outcomes Registry and Additional file Monitoring (VSTORM) Group after receiving ethical approvals. Identifiable information extracted from the VSTR was only used to identify the correct Additional file 1: Figure S1. Receiver operating characteristic curves for medical record for the sample of patients at the hospitals. Thereafter, all the models incorporating medical record and ICD-10 data predicting a good analyses were performed in non-identifiable format. recovery (GOS-E 7–8), Figure S2. Receiver operating characteristic curves for the models incorporating medical record and ICD-10 data predicting return Competing interests to work, Figure S3. Receiver operating characteristic curves for models The authors declare that they have no competing interests. incorporating medical record and ICD-10 data predicting EQ-5D-3L problems with ongoing mobility, Figure S4. Receiver operating characteristic curves for models incorporating medical record and ICD-10 data predicting Publisher’sNote EQ-5D-3L problems with ongoing self-care, Figure S5. Receiver operating Springer Nature remains neutral with regard to jurisdictional claims in characteristic curves for models incorporating medical record and published maps and institutional affiliations. ICD-10 data predicting EQ-5D-3L ongoing problems with usual activities, Figure S6. Receiver operating characteristic curves for models incorporating Author details medical record and ICD-10 data predicting EQ-5D-3L problems with on- Department of Epidemiology and Preventive Medicine, Monash University, going pain or discomfort, Figure S7. Receiver operating characteristic Melbourne, Victoria, Australia. Emergency and Trauma Centre, The Alfred curves for models incorporating medical record and ICD-10 data predicting Hospital, Melbourne, Victoria, Australia. Trauma Service, Royal Melbourne EQ-5D-3L problems with ongoing anxiety or depression (DOCX 11452 kb) Hospital, Melbourne, Victoria, Australia. Farr Institute, Swansea University Medical School, Swansea University, Swansea, UK. Abbreviations ACS: Australian Coding Standards; AUC: Area under the curve; EQ-5D: European Received: 14 December 2017 Accepted: 29 May 2018 Quality of Life Five Dimensions; H-L: Hosmer-Lemeshow; HRQoL: Health-related quality of life; ICD-10-AM: International Classification of Diseases, Tenth Revision, Australian Modification; ROC: Receiver operating characteristic; VSTR: Victorian References State Trauma Registry 1. GBD 2015 Disease and Injury Incidence and Prevalence Collaborators. Global, regional, and national incidence, prevalence, and years lived with Acknowledgements disability for 310 diseases and injuries, 1990–2015: a systematic analysis for The authors would like to acknowledge The Alfred and Royal Melbourne the Global Burden of Disease Study 2015. Lancet. 2016;388:1545–602. Hospitals and all project-associated investigators: Susan McLellan, Jane Ford, 2. Vles WJ, Steyerberg EW, Essink-Bot ML, van Beeck EF, Meeuwis JD, Leenen Melissa Hart, Adrian Buzgau, Tani Thomas, Mimi Morgan. The Victorian State LP. Prevalence and determinants of disabilities and return to work after Trauma Outcomes Registry and Monitoring (VSTORM) Group is thanked for major trauma. J Trauma. 2005;58:126–35. the provision of data. The authors acknowledge the data collectors and 3. Holbrook TL, Hoyt DB. The impact of major trauma: quality-of-life outcomes participating hospitals of the VSTR and members of the VSTR Steering are worse in women than in men, independent of mechanism and injury Committee. The authors would also like to thank Mary Lou Greer and Carol severity. J Trauma. 2004;56:284–90. Roberts who facilitated the study. 4. Langley J, Davie G, Wilson S, Lilley R, Ameratunga S, Wyeth E, Derrett S. Difficulties in functioning 1 year after injury: the role of preinjury Funding sociodemographic and health characteristics, health care and injury-related The Victorian State Trauma Registry (VSTR) is a Department of Health and factors. Arch Phys Med Rehabil. 2013;94:1277–86. Human Services, State Government of Victoria and Transport Accident 5. Michaels AJ, Michaels CE, Smith JS, Moon CH, Peterson C, Long WB. Commission funded project. Tu Nguyen was supported by an Australian Outcome from injury: general health, work status, and satisfaction 12 Government Research Training Program Scholarship during the preparation months after trauma. J Trauma Acute Care Surg. 2000;48:841–50. of this manuscript. Belinda Gabbe was supported by an Australian Research 6. Wan JJ, Morabito DJ, Khaw L, Knudson MM, Dicker RA. Mental illness as an Council Future Fellowship (FT170100048) during the preparation of this independent risk factor for unintentional injury and injury recidivism. J manuscript. The funding sources had no role in the design or conduct of Trauma. 2006;61:1299–304. the study. 7. Dicker RA, Mah J, Lopez D, Tran C, Reidy R, Moore M, Kreniske P, Crane I, Knudson MM, Li M, et al. Screening for mental illness in a trauma center: Availability of data and materials rooting out a risk factor for unintentional injury. J Trauma. 2011;70:1337–44. The data that support the findings of this study are available from the 8. Archer KR, Heins SE, Abraham CM, Obremskey WT, Wegener ST, Castillo RC. Victorian State Trauma Outcomes Registry and Monitoring Group (VSTORM) Clinical significance of pain at hospital discharge following traumatic but restrictions apply to the availability of these data, which were used orthopedic injury: general health, depression, and PTSD outcomes at 1 year. under license for the current study, and so are not publicly available. Data Clin J Pain. 2016;32:196–202. are however available from the authors upon reasonable request and with 9. Gregory KE, Radovinsky L. Research strategies that result in optimal data permission of VSTORM. collection from the patient medical record. Appl Nurs Res. 2012;25:108–16. 10. Gavrielov-Yusim N, Friger M. Use of administrative medical databases in Authors’ contributions population-based research. J Epidemiol Community Health. 2014;68:283–7. TQN contributed to the study design, collected and analysed the data and 11. Pine M, Norusis M, Jones B, Rosenthal GE. Predictions of hospital mortality drafted the manuscript. SCB contributed to the study design, data rates: a comparison of data sources. Ann Intern Med. 1997;126:347–54. interpretation, and drafting and revising of the manuscript. PMS contributed 12. Dickinson CM, Karlin DR, Nunez HR, Cao SA, Heffernan DS, Monaghan SF, to statistical analysis, interpretation of the data and revision of the Kheirbek T, Adams CA, Stephen AH. Do patients with pre-existing manuscript for important intellectual content. PAC and RJ contributed to psychiatric illness have an increased risk of infection after injury? Surg Infect. interpretation of the data and revision of the manuscript for important 2017;18:545. Nguyen et al. BMC Health Services Research (2018) 18:408 Page 8 of 8 13. Falsgraf E, Inaba K, de Roulet A, Johnson M, Benjamin E, Lam L, Matsushima 37. Cheng P, Gilchrist A, Robinson KM, Paul L. The risk and consequences of K, Strumwasser A, Demetriades D. Outcomes following traumatic injury in clinical miscoding due to inadequate medical documentation: a case study patients with preexisting psychiatric illness. J Trauma Acute Care Surg. 2017; of the impact on health services funding. HIM J. 2009;38:35–46. 83:882. 38. Hibbard MR, Ashman TA, Spielman LA, Chun D, Charatz HJ, Melvin S. 14. Davydow DS, Zatzick DF, Rivara FP, Jurkovich GJ, Wang J, Roy-Byrne PP, Relationship between depression and psychosocial functioning after Katon WJ, Hough CL, Kross EK, Fan MY, et al. Predictors of posttraumatic traumatic brain injury. Arch Phys Med Rehabil. 2004;85:S43–53. stress disorder and return to usual major activity in traumatically injured 39. O’Dell KM, Hannay HJ, Biney FO, Robertson CS, Tian TS. The effect of blood intensive care unit survivors. Gen Hosp Psychiatry. 2009;31:428–35. alcohol level and pre-injury chronic alcohol use on outcome from severe traumatic brain injury in Hispanics, Anglo-Caucasians, and African 15. Nguyen TQ, Simpson PM, Braaf SC, Gabbe BJ. Mortality, functional and Americans. J Head Trauma Rehabil. 2012;27:361–9. return to work outcomes of major trauma patients injured from deliberate 40. Schweininger S, Forbes D, Creamer M, McFarlane AC, Silove D, Bryant RA, self-harm. Injury. 2017;48:184–94. O'Donnell ML. The temporal relationship between mental health and 16. Cameron CM, Purdie DM, Kliewer EV, McClure RJ. Mental health: a cause or disability after injury. Depress Anxiety. 2015;32:64–71. consequence of injury? A population-based matched cohort study. BMC 41. Lankester BJA, Garneti N, Gargan MF, Bannister GC. Factors predicting Public Health. 2006;6:114. outcome after whiplash injury in subjects pursuing litigation. Eur Spine J. 17. Steinbusch PJ, Oostenbrink JB, Zuurbier JJ, Schaepkens FJ. The risk of 2006;15:902–7. upcoding in casemix systems: a comparative study. Health Policy. 2007;81: 42. MacMillan PJ, Hart RP, Martelli MF, Zasler ND. Pre-injury status and 289–99. adaptation following traumatic brain injury. Brain Inj. 2002;16:41–9. 18. Kim HM, Smith EG, Stano CM, Ganoczy D, Zivin K, Walters H, Valenstein M. 43. Chubb HL, Bisson JI. The development of psychological reactions in a group Validation of key behaviourally based mental health diagnoses in of individuals with pre-existing and enduring psychiatric disorder following administrative data: suicide attempt, alcohol abuse, illicit drug abuse and a fatal coach crash. J Ment Health. 1999;8:203–12. tobacco use. BMC Health Serv Res. 2012;12:18. 44. Brych SB, Engrav LH, Rivara FP, Ptacek JT, Lezotte DC, Esselman PC, 19. Fowles JB, Fowler EJ, Craft C. Validation of claims diagnoses and self- Kowalske KJ, Gibran NS. Time off work and return to work rates after burns: reported conditions compared with medical records for selected chronic systematic review of the literature and a large two-center series. J Burn Care diseases. J Ambul Care Manage. 1998;21:24–34. Rehabil. 2001;22:401–5. 20. Cameron PA, Finch CF, Gabbe BJ, Collins LJ, Smith KL, McNeil JJ. Developing 45. Black JA, Herbison GP, Lyons RA, Polinder S, Derrett S. Recovery after injury: Australia’s first statewide trauma registry: what are the lessons? ANZ J Surg. an individual patient data meta-analysis of general health status using the 2004;74:424–8. EQ-5D. J Trauma. 2011;71:1003–10. 21. Gabbe BJ, Cameron PA, Hannaford AP, Sutherland AM, McNeil JJ. Routine 46. Tan Z, Liang Y, Liu S, Cao W, Tu H, Guo L, Xu Y. Health-related quality of life follow up of major trauma patients from trauma registries: what are the as measured with EQ-5D among populations with and without specific outcomes? J Trauma. 2006;61:1393–9. chronic conditions: a population-based survey in Shaanxi Province, China. 22. National Centre for Classification in Health. International Statistical PLoS One. 2013;8:e65958. Classification of Diseases and Related Health Problems, Tenth Revision, 47. Hu XB, Feng Z, Fan YC, Xiong ZY, Huang QW. Health-related quality-of-life Australian Modification (ICD-10-AM) Australian Coding Standards. after traumatic brain injury: a 2-year follow-up study in Wuhan, China. Brain Darlinghurst: University of Sydney; 1998. Inj. 2012;26:183–7. 23. Charlson ME, Charlson RE, Peterson JC, Marinopoulos SS, Briggs WM, 48. Leduc BE, Lepage Y. Health-related quality of life after spinal cord injury. Hollenberg JP. The Charlson comorbidity index is adapted to predict costs Disabil Rehabil. 2002;24:196–202. of chronic disease in primary care patients. J Clin Epidemiol. 2008;61:1234–40. 49. Polinder S, Haagsma JA, van Klaveren D, Steyerberg EW, van Beeck EF. 24. Jennett B, Bond M. Assessment of outcome after severe brain damage. Health-related quality of life after TBI: a systematic review of study design, Lancet. 1975;1:480–4. instruments, measurement properties, and outcome. Popul Health Metr. 25. Dolan P. Modeling valuations for EuroQol health states. Med Care. 1997;35: 2015;13:4. 1095–108. 50. Steyerberg EW, Vergouwe Y. Towards better clinical prediction models: 26. Hanley JA, McNeil BJ. The meaning and use of the area under a receiver seven steps for development and an ABCD for validation. Eur Heart J. 2014; operating characteristic (ROC) curve. Radiology. 1982;143:29–36. 35:1925–31. 27. Badcock D, Kelly A-M, Kerr D, Reade T. The quality of medical record review studies in the international emergency medicine literature. Ann Emerg Med. 2005;45:444–7. 28. Hosmer DW, Lemeshow S. Applied logistic regression. 2nd ed. New York: Wiley; 2000. 29. Steyerberg EW, Eijkemans MJ, Harrell FE Jr, Habbema JD. Prognostic modelling with logistic regression analysis: a comparison of selection and estimation methods in small data sets. Stat Med. 2000;19:1059–79. 30. Gabbe BJ, Harrison JE, Lyons RA, Edwards ER, Cameron PA, Victorian Orthopaedic Trauma Outcomes R. Comparison of measures of comorbidity for predicting disability 12-months post-injury. BMC Health Serv Res. 2013;13:30. 31. Bender M, Smith TC. Using administrative mental health indicators in heart failure outcomes research: comparison of clinical records and international classification of disease coding. J Card Fail. 2016;22:56–60. 32. McGorry PD. The next stage for diagnosis: validity through utility. World Psychiatry. 2013;12:213–5. 33. Henderson T, Shepheard J, Sundararajan V. Quality of diagnosis and procedure coding in ICD-10 administrative data. Med Care. 2006;44:1011–9. 34. Humphries KH, Rankin JM, Carere RG, Buller CE, Kiely FM, Spinelli JJ. Co- morbidity data in outcomes research: are clinical data derived from administrative databases a reliable alternative to chart review? J Clin Epidemiol. 2000;53:343–9. 35. Malenka DJ, McLerran D, Roos N, Fisher ES, Wennberg JE. Using administrative data to describe casemix: a comparison with the medical record. J Clin Epidemiol. 1994;47:1027–32. 36. O'Malley KJ, Cook KF, Price MD, Wildes KR, Hurdle JF, Ashton CM. Measuring diagnoses: ICD code accuracy. Health Serv Res. 2005;40:1620–39. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png BMC Health Services Research Springer Journals

Comparison of the performance of mental health, drug and alcohol comorbidities based on ICD-10-AM and medical records for predicting 12-month outcomes in trauma patients

Free
8 pages

Loading next page...
 
/lp/springer_journal/comparison-of-the-performance-of-mental-health-drug-and-alcohol-K4Epf2MpIX
Publisher
Springer Journals
Copyright
Copyright © 2018 by The Author(s).
Subject
Medicine & Public Health; Public Health; Health Administration; Health Informatics; Nursing Research
eISSN
1472-6963
D.O.I.
10.1186/s12913-018-3248-x
Publisher site
See Article on Publisher Site

Abstract

Background: Many outcome studies capture the presence of mental health, drug and alcohol comorbidities from administrative datasets and medical records. How these sources compare as predictors of patient outcomes has not been determined. The purpose of the present study was to compare mental health, drug and alcohol comorbidities based on ICD-10-AM coding and medical record documentation for predicting longer-term outcomes in injured patients. Methods: A random sample of patients (n = 500) captured by the Victorian State Trauma Registry was selected for the study. Retrospective medical record reviews were conducted to collect data about documented mental health, drug and alcohol comorbidities while ICD-10-AM codes were obtained from routinely collected hospital data. Outcomes at 12-months post-injury were the Glasgow Outcome Scale – Extended (GOS-E), European Quality of Life Five Dimensions (EQ-5D-3L), and return to work. Linear and logistic regression models, adjusted for age and gender, using medical record derived comorbidity and ICD-10-AM were compared using measures of calibration (Hosmer-Lemeshow statistic) and discrimination (C-statistic and R ). Results: There was no demonstrable difference in predictive performance between the medical record and ICD- 10-AM models for predicting the GOS-E, EQ-5D-3L utility sore and EQ-5D-3L mobility, self-care, usual activities and pain/discomfort items. The area under the receiver operating characteristic (AUC) for models using medical record derived comorbidity (AUC 0.68, 95% CI: 0.63, 0.73) was higher than the model using ICD-10-AM data (AUC 0.62, 95% CI: 0.57, 0.67) for predicting the EQ-5D-3L anxiety/depression item. The discrimination of the model for predicting return to work was higher with inclusion of the medical record data (AUC 0.69, 95% CI: 0.63, 0.76) than the ICD-10-AM data (AUC 0.59, 95% CL: 0.52, 0.65). Conclusions: Mental health, drug and alcohol comorbidity information derived from medical record review was not clearly superior for predicting the majority of the outcomes assessed when compared to ICD-10-AM. While information available in medical records may be more comprehensive than in the ICD-10-AM, there appears to be little difference in the discriminative capacity of comorbidities coded in the two sources. Keywords: Trauma, Mental comorbidities, Post-injury outcomes, Prediction, Validation, Medical record coding * Correspondence: tu.nguyen@monash.edu Department of Epidemiology and Preventive Medicine, Monash University, Melbourne, Victoria, Australia Full list of author information is available at the end of the article © The Author(s). 2018 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Nguyen et al. BMC Health Services Research (2018) 18:408 Page 2 of 8 Background of Victoria [20]. The VSTR follows up all adult survivors Injuries accounted for 11% of the global burden of disease to discharge after their injury including function, health- in terms of Years Lived with Disability in 2015 [1]. related quality of life (HRQoL), return to work and post- Risk-adjustment analyses of long-term outcomes after in- discharge mortality. Standardised telephone interviews are jury have included compensable status, injury mechanism, conducted by trained interviewers [21]. A random sample injury severity, perceived level of social support, socioeco- of 500 patients registered on the VSTR were selected for nomic position and educational level [2–5]. Mental health, the study. Patients were eligible for inclusion in this study drug and alcohol comorbidities have been cited as factors if they met each of the following criteria: that impact on recovery and quality of life following injury [6, 7]. However, the importance of these comorbidities i. Aged 18 years or older; may be under-estimated in risk modelling of long-term ii. Definitively managed at one of the two adult major outcomes [8]. Data sources used in research studies to trauma services in Victoria; capture comorbid conditions are ostensibly linked to the iii. Survived to 12 months post-injury and 12-month research question, patient population and resources avail- follow-up completed. able. Consequently, poor comorbidity capture in different data sources has an impact on the resulting comorbidity Procedures profile of injured people and the ability to predict patient In Australia, hospital data are collected in the form of outcomes after injury. International Statistical Classification of Diseases, Tenth Although time- and labour-intensive to review, medical Revision, Australian Modification (ICD-10-AM) codes and records represent a comprehensive source of comorbidity apply to all public and private hospitals. In Victoria, each data, due to the detailed clinical information contained code contains a prefix with “P” representing principal diag- within the notes [9]. Alternatively, routinely collected ad- nosis requiring treatment during the stay, “A” representing ministrative data are commonly relied on in epidemiological additional diagnoses and “C” representing in-hospital research as they are relatively cost-effective and readily complications. These codes are assigned by the hospital available [10]. Findings from previous research suggest coders as part of routinely collected hospital data and that mental health, drug and alcohol comorbidities derived are used for activity-based funding purposes [22]. from administrative data has been shown to be a strong Data were extracted from the VSTR for patients in predictor of in-hospital mortality [11], hospital-acquired the sample including clinical and demographic data and infection [12] and longer hospital length of stay [13]. How- follow-up outcome data collected at 12-months post- ever, prediction of longer-term recovery outcomes using injury. This included patients’ compensation status by comorbidities based on administrative data have generated the no-fault, third party insurers for road traffic injury inconsistent results [14–16]. This could be explained by and work-related injury and the Charlson Comorbidity biases in administrative data, such as those associated with Index (CCI), a weighting used to assess comorbidity [23]. collecting data in order to account for different types of The ICD-10-AM “P” and “A” prefixed codes, which are patient episodes and services provided, in the case of provided to the VSTR by the hospitals, were also extracted. activity-based funding [17]. Retrospective review of the medical records of the sample Direct comparisons between administrative coded mental of patients was conducted, by the author of this paper health, drug and alcohol diagnoses with other sources in consultation with researchers with clinical and health of data in injured populations are scarce [18, 19]. The expertise, at the two adult major trauma services. completeness of administrative coding largely relies on To enable comparison with ICD-10-AM codes, all the quality of the medical record documentation but little mental health, drug and alcohol comorbidities described in is known about the impact of using the two alternative the medical record were captured based on the categories data sources on long-term outcome modelling. The aim of of mental and behavioural disorder (Chapter V) specified this study was to compare the contribution of two different by ICD-10-AM. If a mental health, drug or alcohol con- sources of mental health, drug and alcohol comorbidity dition was documented, based on clinical notation in data in predicting 12-month functional, return to work addition to substantiating evidence such as an alcohol and health-related quality of life outcomes in a major withdrawal scale or psychiatrist referral letter for on- trauma population. going treatment, this was recorded on a project data collection form in addition to the specified diagnosis. Methods Pre-existing mental health disorders were captured from Study design and inclusion criteria the medical record based on the following categories: The Victorian State Trauma Registry (VSTR) is a organic mental disorders, schizophrenia, mood disorders, population-based registry which captures information neurotic disorders, behavioural and personality disorders about virtually all major trauma patients across the state and other specified mental disorders. Nguyen et al. BMC Health Services Research (2018) 18:408 Page 3 of 8 A single binary indicator variable were generated to mobility, self-care, usual activities, pain/discomfort represent the presence of ICD-10-AM coded mental, drug and anxiety/depression. A linear regression model was and alcohol use disorders. This indicator variable included fitted for the continuous outcome EQ-5D-3L utility the major categories: alcohol use disorders (F10), drug score. Age was categorised for inclusion in the models use disorders (F11-F15, F16, F19) and mental disorders as it was not linearly related to the log odds of the out- (F00-F03, F07, F09, F20-F48, F60-F69, F90-F99, U792, comes of interest for the logistic models in its continuous U793). form. The following models were fitted for each the out- Outcome measures comes measured: The primary outcome measure recorded on the VSTR used to assess patients’ level of function is the Glasgow Outcome i) Age and gender; Scale-Extended (GOS-E). The GOS-E classifies patients ii) Medical record coded mental health, drug and into eight levels of function from death (GOS-E = 1), upper alcohol comorbidities, adjusted for age and gender severe disability (GOS-E = 4) to upper good recovery iii) ICD-10-AM coded mental health, drug and alcohol (GOS-E = 8), indicating a complete return to normal comorbidities, adjusted for age and gender activities of daily life [24]. The three-level European Quality of Life Five Dimensions (EQ-5D-3L) is used to Discrimination was evaluated using the C-statistic, or measure of health-related quality of life and consists of area under the receiver operating characteristic curve five dimensions (mobility, self-care, usual activities, pain/ (AUC), which ranges from zero to one, with a value of discomfort, and anxiety/depression), with each item having 0.5 representing a model with no predictive power and a three possible responses: no problems, some problems and value of 1 indicating a model with perfect predictive power. extreme problems. The EQ-5D-3L item responses are used A value between 0.7 and 0.8 is considered acceptable to calculate a single utility score using age and gender spe- predictive performance, while values greater than 0.8 cific social preference weights [25]. Algorithms were used demonstrate excellent predictive performance [28]. To to convert EQ-5D-3L responses into health utility scores investigate model fit, likelihood ratio (LR) tests were [range: 0 (equivalent to death) to 1 (equivalent to full used. A higher LR statistic and significant p-value supports health)]. The original preference weights for the the hypothesis that including the comorbidity data was a EQ-5D-3L from the UK population were applied as significant improvement in model fit over the model of age these are most commonly used [25]. If patients worked and gender alone [29]. Hosmer-Lemeshow (H-L) statistics or studied prior to injury, they were asked whether they were used to compare model calibration of logistic regres- had returned to work or study. sion models [28]. A linear regression model was fitted for To establish whether the capacity to predict patient the EQ-5D-3L utility scores outcome and the coefficient of outcomes at 12-months post-injury differed between determination (R ) of the models used to assess pre- medical record review of drug, alcohol and mental health dictive performance. A p-value < 0.05 was considered comorbidities and those coded in ICD-10-AM, comparison significant. All analyses were conducted using Stata 13.0 of model performance was undertaken. A ‘good recovery’ (StataCorp, College Station, TX). wasdefined as aGOS-E scoreof7–8and a score less than seven represented ongoing functional limitations. The EQ-5D-3L item responses were dichotomised as Results no problems versus some or severe problems. The clinical and demographic characteristics of the study sample are shown in Table 1. There was an even Analysis distribution of patients among the age groups and most In order to detect a statistically significant difference be- patients were male. The median (IQR) injury severity tween the area under the Receiver Operating Characteristic score was 17 (14–22), and these figures are representative (AUC) of 0.7 and 0.8 with a 5% one-sided type I error and of major trauma patients in Victoria [30]. Most patients 80% power, 158 patients was the smallest sample size did not have a Charlson Comorbidity Index condition required for each hospital [26]. We over-sampled patients recorded and did not claim compensation for their beyond this minimum sample size and accessed the medical injuries (Table 1). records of 250 patients from each hospital, in order to Based on the GOS-E measure, 41% (n =205) of patients account for issues with availability of paper medical records in the sample reported a good recovery at and missing volumes which are common in medical record 12-months post-injury. According to the H-L statistic, review studies [27]. Binary logistic regression models the models predicting functional recovery demonstrated were fitted for the categorical outcomes functional acceptable calibration. Inclusion of the mental health, drug recovery, return to work, and the EQ-5D-3L items of and alcohol comorbidity data, from either the medical Nguyen et al. BMC Health Services Research (2018) 18:408 Page 4 of 8 Table 1 Demographic and clinical characteristics of patients The EQ-5D-3L was complete for 98% of patients at within the sample (n = 500) 12 months. The mean (SD) EQ-5D-3L utility score in Variable Total n (%) the sample was 0.75 (0.29). The prevalence of problems was highest for the mobility and anxiety or depression Age (years) items (Table 3). Overall, the prevalence of problems was 18–24 63 (12.6) lowest for the self-care item. 25–34 92 (18.4) The models predicting problems on each of the dimen- 35–44 98 (19.6) sions of the EQ-5D-3L showed less than acceptable 45–54 91 (18.2) discrimination (Table 4). The capacity to discriminate 55–64 72 (14.4) between patients with and without problems in the mobility, self-care, usual activities, pain/discomfort ≥ 65 84 (16.8) items and utility score was similar for the ICD-10-AM Gender and medical record models except for patients with Male 374 (74.8) anxiety/depression problems. For predicting the EQ-5D-3L Female 126 (25.2) utility score, the highest R was observed for the medical Charlson Comorbidity Index weighting record model. Both models were an improvement over the 0 351 (70.2) age and gender model, but addition of the ICD-10-AM coded conditions was only a marginal improvement 1 115 (23.0) over age and gender (Table 4). Calibration was acceptable ≥ 2 34 (6.8) according to the H-L statistic for predicting all of the Compensation status EQ-5D-3L items. Compensable 180 (36.1) Non-compensable 318 (63.9) Discussion Compensation status not available for n =2 Comorbidity information available in different data sources have their own advantages and limitations, which can record or ICD-10-AM, improved the model fit but none of affect the validity of studies relying on such data [31]. It the models showed acceptable discrimination (Table 2) has been acknowledged that routinely collected adminis- (Additional file 1). trative data sub-optimally represent a patient’s condition Three hundred and thirty eight (68%) patients reported and the totality of their comorbid illnesses, particularly for that they were working or studying prior to injury. Of mental health, drug and alcohol comorbidities which are those who were working/studying prior to injury, 71% historically difficult to characterise, classify and diagnose (n = 240) had returned to work or study at 12-months [32]. Previous analyses of administrative coding in routinely post-injury. The models predicting return to work all collected hospital data compared with medical records have showed less than acceptable discrimination but demon- revealed the prevalence of comorbid conditions are fre- strated acceptable calibration (Table 2). The addition of quently under-estimated [33–35]. Observable discrepancies the medical record data, but not the ICD-10-AM coded between the data sources exist [36], and are driven largely data, improved discrimination over age and gender. by coding rules, activity-based funding, variability in Table 2 Discrimination and calibration statistics for predicting GOS-E good recovery and return to work Good recovery (GOS-E 7–8) Model N AUC (95% CI) H-L statistic (p value) LR-test (p-value) Age and gender 500 0.57 (0.52, 0.63) 2.3 (0.89) – Age, gender and medical record data 500 0.62 (0.57, 0.67) 7.1 (0.52) 11.7 (< 0.01) Age, gender, ICD-10-AM 500 0.58 (0.53, 0.63) 3.3 (0.91) 0.7 (0.42) Return to work N AUC (95% CI) H-L statistic (p value) LR-test (p-value) Age and gender 348 0.55 (0.49, 0.62) 4.3 (0.75) – Age, gender and medical record data 348 0.69 (0.63, 0.76) 7.5 (0.48) 38.1 (< 0.01) Age, gender, ICD-10-AM 348 0.59 (0.52, 0.65) 3.6 (0.89) 5.9 (0.01) Compared to age and gender model Nguyen et al. BMC Health Services Research (2018) 18:408 Page 5 of 8 Table 3 EQ-5D-3L responses at 12-months post-injury the review due to missing documentation or illegible handwriting. As a result, mental health, drug and alcohol EQ-5D-3L dimension n (%) comorbidities obtained from the medical record may not Mobility be reliably predictive of long-term function. Further re- No problems 337 (68.4) search is needed to compare the predictive performance Some or severe problems 156 (31.6) of medical record data with self-report and clinical inter- Self-care view to assess these comorbidities for long-term outcome No problems 398 (80.8) modelling. Some or severe problems 95 (19.3) Mental health, drug and alcohol comorbidity data derived from medical records have been used previously for risk- Usual activities adjustment in studies of return to work [42, 44]. Mental No problems 263 (53.5) health, drug and alcohol comorbidities coded in ICD- Some or severe problems 229 (46.5) 10-AM may not be as reliable or be able to perform to the Pain or discomfort same capacity as medical record abstracted conditions for No problems 278 (56.4) predicting return to work. This is difficult to assess, as few Some or severe problems 215 (43.6) existing studies have evaluated the association between ICD-10 coded mental health comorbidities and returning Anxiety or depression to work after injury. A previous study of major trauma No problems 316 (64.1) patients using ICD-10-AM coded data reported with no Some or severe problems 177 (35.9) significant association with returning to work [15], but this Missing data (n = 7) for mobility, self-care, pain/discomfort may have been attributed to limited statistical power. and anxiety/depression Missing data (n = 8) for usual activities Whilethemedical record mayprovideahigh degreeof detail about mental health factors that are associated with coder interpretation and inconsistent or unclear charting lower odds of returning to work following severe injury, [17, 37]. We sought to compare the statistical performance this information may be incomplete when coded in the of two commonly used data sources: medical record docu- ICD-10-AM. mentation and ICD-10-AM administrative data. Mental Survivors of traumatic injuries often experience deficits health, drug and alcohol comorbidities were shown to in perceived health status, including pain or discomfort, be important for predicting long-term outcomes, as the and difficulty with usual activities and mobility [45]. Aside addition of the comorbidity data fit the data better than from anxiety or depression, there were no differences in age and gender alone. Overall, the addition of comorbidi- the discriminative ability of the ICD-10-AM and medical ties obtained from medical record review did not pro- record data for predicting the remaining EQ-5D-3L vide improvement in discrimination over age and gender, dimensions or utility score. Neither the medical record which was comparable with the ICD-10-AM coded comor- documentation nor the ICD-10-AM data were able to bidities, for most of the outcomes assessed. differentiate between groups at risk of ongoing problems Mental health, drug and alcohol comorbidities are with mobility, self-care, usual activities and pain/discom- associated with delays in recovery and disability after fort with any great accuracy. An explanation for this injury, but the majority of the evidence predicting these finding may be that self-reported measures of comorbidity outcomes have relied on self-report or clinical interview may be a better predictor of the EQ-5D-3L outcomes to assess comorbidity [38–40]. Pre-injury mental health whereas indirect comorbidity measures may not be diagnoses derived from the medical record has been optimal for predicting these health status outcomes associated with poorer function in previous studies [41, 42], [46]. This suggests the importance of other factors not whereas ascertaining pre-existing mental health history explained by mental health, drug and alcohol comor- based on administrative data coding has yielded differing bidities including personal and environmental factors results [16, 43]. Despite the lower prevalence of comorbid- for predicting HRQoL following injury. Other factors ities coded in ICD-10-AM than in the medical record, the that have been shown to be important in the literature findings of this study indicate the capacity to predict func- include compensable status, education, and social tional recovery was largely comparable with the medical circumstances [47–49]. record. This suggests that key mental health, drug and The strengths of the study were the use of long-term alcohol comorbidities, which are predictors and have outcomes data at 12-months post-injury and the random significant impacts on long-term function after injury, are sample of a population-based cohort, which was repre- not comprehensively captured in the medical record at sentative of major trauma patients in Victoria. Although the outset. Notably, some comorbidities within the med- patients from each hospital were over-sampled in order to ical record could have been inadvertently excluded from mitigate potential issues with missing charts, only seven Nguyen et al. BMC Health Services Research (2018) 18:408 Page 6 of 8 Table 4 Discrimination and calibration statistics for predicting EQ-5D-3L dimensions and utility score EQ-5D Mobility Model N AUC (95% CI) H-L statistic (p-value) LR-test (p-value) Age and gender 493 0.56 (0.51, 0.62) 0.6 (0.99) – Age, gender and medical record data 493 0.58 (0.53, 0.63) 2.3 (0.97) 4.3 (0.04) Age, gender, ICD-10-AM 493 0.56 (0.51, 0.62) 3.7 (0.81) 0.2 (0.64) EQ-5D Self-care N AUC (95% CI) H-L statistic (p-value) LR-test (p-value) Age and gender 493 0.58 (0.51, 0.64) 3.6 (0.73) – Age, gender and medical record data 493 0.60 (0.53, 0.66) 4.5 (0.80) 4.6 (0.03) Age, gender, ICD-10-AM 493 0.57 (0.51, 0.64) 6.8 (0.45) 0.4 (0.54) EQ-5D Usual activities N AUC (95% CI) H-L statistic (p-value) LR-test (p-value) Age and gender 492 0.57 (0.52, 0.62) 1.8 (0.94) – Age, gender and medical record data 492 0.59 (0.54, 0.64) 5.1 (0.75) 6.0 (0.01) Age, gender, ICD-10-AM 492 0.57 (0.52, 0.62) 6.2 (0.63) 0.1 (0.72) EQ-5D Pain or discomfort N AUC (95% CI) H-L statistic (p-value) LR-test (p-value) Age and gender 493 0.58 (0.53, 0.63) 3.2 (0.78) – Age, gender and medical record data 493 0.58 (0.53, 0.63) 8.2 (0.41) 2.7 (0.10) Age, gender, ICD-10-AM 493 0.58 (0.53, 0.63) 5.6 (0.58) 0.0 (0.92) EQ-5D Anxiety or depression N AUC (95% CI) H-L statistic (p-value) LR-test (p-value) Age and gender 493 0.55 (0.50, 0.60) 1.7 (0.94) – Age, gender and medical record data 493 0.68 (0.63, 0.73) 9.9 (0.28) 42.1 (< 0.01) Age, gender, ICD-10-AM 493 0.62 (0.57, 0.67) 6.7 (0.57) 10.9 (0.01) EQ-5D-3L utility score NR LR-test (p-value) Age and gender 492 0.01 – Age, gender and medical record data 492 0.06 22.1 (< 0.01) Age, gender, ICD-10-AM 492 0.02 4.0 (0.04) Compared to age and gender model patient records (3%) of the total sample could not be lo- choose the best predictors of the outcome [50], and cated, and these patients were replaced from the VSTR. A the results cannot be interpreted for this purpose with- limitation of the study includes the extent to which these out careful selection and further examination of other findings may be extrapolated to other patients. This re- variables. search relied on ICD-10-AM recorded mental health, drug and alcohol diagnoses which were originally collected pri- Conclusion marily for activity-based funding. The findings may not be While the prevalence of mental health, drug and alcohol applicable to other jurisdictions, which vary in coding comorbidities is lower in ICD-10-AM than medical records, practices and funding systems. This aim of the study was data derived from medical record review was not clearly su- not to develop the best prediction model for long-term perior for predicting the majority of the outcomes assessed. outcomes. The selection of variables to identify the There was no demonstrable difference in discriminative ‘optimal’ risk prediction model is based on trying to capacity between the medical record and ICD-10-AM Nguyen et al. BMC Health Services Research (2018) 18:408 Page 7 of 8 models for predicting the GOS-E, EQ-5D-3L utility sore intellectual content. BJG oversaw the project, contributed to the study design, data interpretation, and drafting and revising of the manuscript. All and EQ-5D-3L mobility, self-care, usual activities and pain/ authors read and approved the final version of the manuscript. discomfort items. Further research is needed to com- pare the contribution of medical record documented Ethics approval and consent to participate This study was approved by the Alfred Hospital Human Research Ethics mental health, drug and alcohol comorbidities with alter- Committee (Project Number: 275/16), the Royal Melbourne Hospital (Project native data sources. Number: 2016.179) and the Monash University Human Research Ethics Committee (Project Number: 0290). Access to the VSTR data for this study was granted by the Victorian State Trauma Outcomes Registry and Additional file Monitoring (VSTORM) Group after receiving ethical approvals. Identifiable information extracted from the VSTR was only used to identify the correct Additional file 1: Figure S1. Receiver operating characteristic curves for medical record for the sample of patients at the hospitals. Thereafter, all the models incorporating medical record and ICD-10 data predicting a good analyses were performed in non-identifiable format. recovery (GOS-E 7–8), Figure S2. Receiver operating characteristic curves for the models incorporating medical record and ICD-10 data predicting return Competing interests to work, Figure S3. Receiver operating characteristic curves for models The authors declare that they have no competing interests. incorporating medical record and ICD-10 data predicting EQ-5D-3L problems with ongoing mobility, Figure S4. Receiver operating characteristic curves for models incorporating medical record and ICD-10 data predicting Publisher’sNote EQ-5D-3L problems with ongoing self-care, Figure S5. Receiver operating Springer Nature remains neutral with regard to jurisdictional claims in characteristic curves for models incorporating medical record and published maps and institutional affiliations. ICD-10 data predicting EQ-5D-3L ongoing problems with usual activities, Figure S6. Receiver operating characteristic curves for models incorporating Author details medical record and ICD-10 data predicting EQ-5D-3L problems with on- Department of Epidemiology and Preventive Medicine, Monash University, going pain or discomfort, Figure S7. Receiver operating characteristic Melbourne, Victoria, Australia. Emergency and Trauma Centre, The Alfred curves for models incorporating medical record and ICD-10 data predicting Hospital, Melbourne, Victoria, Australia. Trauma Service, Royal Melbourne EQ-5D-3L problems with ongoing anxiety or depression (DOCX 11452 kb) Hospital, Melbourne, Victoria, Australia. Farr Institute, Swansea University Medical School, Swansea University, Swansea, UK. Abbreviations ACS: Australian Coding Standards; AUC: Area under the curve; EQ-5D: European Received: 14 December 2017 Accepted: 29 May 2018 Quality of Life Five Dimensions; H-L: Hosmer-Lemeshow; HRQoL: Health-related quality of life; ICD-10-AM: International Classification of Diseases, Tenth Revision, Australian Modification; ROC: Receiver operating characteristic; VSTR: Victorian References State Trauma Registry 1. GBD 2015 Disease and Injury Incidence and Prevalence Collaborators. Global, regional, and national incidence, prevalence, and years lived with Acknowledgements disability for 310 diseases and injuries, 1990–2015: a systematic analysis for The authors would like to acknowledge The Alfred and Royal Melbourne the Global Burden of Disease Study 2015. Lancet. 2016;388:1545–602. Hospitals and all project-associated investigators: Susan McLellan, Jane Ford, 2. Vles WJ, Steyerberg EW, Essink-Bot ML, van Beeck EF, Meeuwis JD, Leenen Melissa Hart, Adrian Buzgau, Tani Thomas, Mimi Morgan. The Victorian State LP. Prevalence and determinants of disabilities and return to work after Trauma Outcomes Registry and Monitoring (VSTORM) Group is thanked for major trauma. J Trauma. 2005;58:126–35. the provision of data. The authors acknowledge the data collectors and 3. Holbrook TL, Hoyt DB. The impact of major trauma: quality-of-life outcomes participating hospitals of the VSTR and members of the VSTR Steering are worse in women than in men, independent of mechanism and injury Committee. The authors would also like to thank Mary Lou Greer and Carol severity. J Trauma. 2004;56:284–90. Roberts who facilitated the study. 4. Langley J, Davie G, Wilson S, Lilley R, Ameratunga S, Wyeth E, Derrett S. Difficulties in functioning 1 year after injury: the role of preinjury Funding sociodemographic and health characteristics, health care and injury-related The Victorian State Trauma Registry (VSTR) is a Department of Health and factors. Arch Phys Med Rehabil. 2013;94:1277–86. Human Services, State Government of Victoria and Transport Accident 5. Michaels AJ, Michaels CE, Smith JS, Moon CH, Peterson C, Long WB. Commission funded project. Tu Nguyen was supported by an Australian Outcome from injury: general health, work status, and satisfaction 12 Government Research Training Program Scholarship during the preparation months after trauma. J Trauma Acute Care Surg. 2000;48:841–50. of this manuscript. Belinda Gabbe was supported by an Australian Research 6. Wan JJ, Morabito DJ, Khaw L, Knudson MM, Dicker RA. Mental illness as an Council Future Fellowship (FT170100048) during the preparation of this independent risk factor for unintentional injury and injury recidivism. J manuscript. The funding sources had no role in the design or conduct of Trauma. 2006;61:1299–304. the study. 7. Dicker RA, Mah J, Lopez D, Tran C, Reidy R, Moore M, Kreniske P, Crane I, Knudson MM, Li M, et al. Screening for mental illness in a trauma center: Availability of data and materials rooting out a risk factor for unintentional injury. J Trauma. 2011;70:1337–44. The data that support the findings of this study are available from the 8. Archer KR, Heins SE, Abraham CM, Obremskey WT, Wegener ST, Castillo RC. Victorian State Trauma Outcomes Registry and Monitoring Group (VSTORM) Clinical significance of pain at hospital discharge following traumatic but restrictions apply to the availability of these data, which were used orthopedic injury: general health, depression, and PTSD outcomes at 1 year. under license for the current study, and so are not publicly available. Data Clin J Pain. 2016;32:196–202. are however available from the authors upon reasonable request and with 9. Gregory KE, Radovinsky L. Research strategies that result in optimal data permission of VSTORM. collection from the patient medical record. Appl Nurs Res. 2012;25:108–16. 10. Gavrielov-Yusim N, Friger M. Use of administrative medical databases in Authors’ contributions population-based research. J Epidemiol Community Health. 2014;68:283–7. TQN contributed to the study design, collected and analysed the data and 11. Pine M, Norusis M, Jones B, Rosenthal GE. Predictions of hospital mortality drafted the manuscript. SCB contributed to the study design, data rates: a comparison of data sources. Ann Intern Med. 1997;126:347–54. interpretation, and drafting and revising of the manuscript. PMS contributed 12. Dickinson CM, Karlin DR, Nunez HR, Cao SA, Heffernan DS, Monaghan SF, to statistical analysis, interpretation of the data and revision of the Kheirbek T, Adams CA, Stephen AH. Do patients with pre-existing manuscript for important intellectual content. PAC and RJ contributed to psychiatric illness have an increased risk of infection after injury? Surg Infect. interpretation of the data and revision of the manuscript for important 2017;18:545. Nguyen et al. BMC Health Services Research (2018) 18:408 Page 8 of 8 13. Falsgraf E, Inaba K, de Roulet A, Johnson M, Benjamin E, Lam L, Matsushima 37. Cheng P, Gilchrist A, Robinson KM, Paul L. The risk and consequences of K, Strumwasser A, Demetriades D. Outcomes following traumatic injury in clinical miscoding due to inadequate medical documentation: a case study patients with preexisting psychiatric illness. J Trauma Acute Care Surg. 2017; of the impact on health services funding. HIM J. 2009;38:35–46. 83:882. 38. Hibbard MR, Ashman TA, Spielman LA, Chun D, Charatz HJ, Melvin S. 14. Davydow DS, Zatzick DF, Rivara FP, Jurkovich GJ, Wang J, Roy-Byrne PP, Relationship between depression and psychosocial functioning after Katon WJ, Hough CL, Kross EK, Fan MY, et al. Predictors of posttraumatic traumatic brain injury. Arch Phys Med Rehabil. 2004;85:S43–53. stress disorder and return to usual major activity in traumatically injured 39. O’Dell KM, Hannay HJ, Biney FO, Robertson CS, Tian TS. The effect of blood intensive care unit survivors. Gen Hosp Psychiatry. 2009;31:428–35. alcohol level and pre-injury chronic alcohol use on outcome from severe traumatic brain injury in Hispanics, Anglo-Caucasians, and African 15. Nguyen TQ, Simpson PM, Braaf SC, Gabbe BJ. Mortality, functional and Americans. J Head Trauma Rehabil. 2012;27:361–9. return to work outcomes of major trauma patients injured from deliberate 40. Schweininger S, Forbes D, Creamer M, McFarlane AC, Silove D, Bryant RA, self-harm. Injury. 2017;48:184–94. O'Donnell ML. The temporal relationship between mental health and 16. Cameron CM, Purdie DM, Kliewer EV, McClure RJ. Mental health: a cause or disability after injury. Depress Anxiety. 2015;32:64–71. consequence of injury? A population-based matched cohort study. BMC 41. Lankester BJA, Garneti N, Gargan MF, Bannister GC. Factors predicting Public Health. 2006;6:114. outcome after whiplash injury in subjects pursuing litigation. Eur Spine J. 17. Steinbusch PJ, Oostenbrink JB, Zuurbier JJ, Schaepkens FJ. The risk of 2006;15:902–7. upcoding in casemix systems: a comparative study. Health Policy. 2007;81: 42. MacMillan PJ, Hart RP, Martelli MF, Zasler ND. Pre-injury status and 289–99. adaptation following traumatic brain injury. Brain Inj. 2002;16:41–9. 18. Kim HM, Smith EG, Stano CM, Ganoczy D, Zivin K, Walters H, Valenstein M. 43. Chubb HL, Bisson JI. The development of psychological reactions in a group Validation of key behaviourally based mental health diagnoses in of individuals with pre-existing and enduring psychiatric disorder following administrative data: suicide attempt, alcohol abuse, illicit drug abuse and a fatal coach crash. J Ment Health. 1999;8:203–12. tobacco use. BMC Health Serv Res. 2012;12:18. 44. Brych SB, Engrav LH, Rivara FP, Ptacek JT, Lezotte DC, Esselman PC, 19. Fowles JB, Fowler EJ, Craft C. Validation of claims diagnoses and self- Kowalske KJ, Gibran NS. Time off work and return to work rates after burns: reported conditions compared with medical records for selected chronic systematic review of the literature and a large two-center series. J Burn Care diseases. J Ambul Care Manage. 1998;21:24–34. Rehabil. 2001;22:401–5. 20. Cameron PA, Finch CF, Gabbe BJ, Collins LJ, Smith KL, McNeil JJ. Developing 45. Black JA, Herbison GP, Lyons RA, Polinder S, Derrett S. Recovery after injury: Australia’s first statewide trauma registry: what are the lessons? ANZ J Surg. an individual patient data meta-analysis of general health status using the 2004;74:424–8. EQ-5D. J Trauma. 2011;71:1003–10. 21. Gabbe BJ, Cameron PA, Hannaford AP, Sutherland AM, McNeil JJ. Routine 46. Tan Z, Liang Y, Liu S, Cao W, Tu H, Guo L, Xu Y. Health-related quality of life follow up of major trauma patients from trauma registries: what are the as measured with EQ-5D among populations with and without specific outcomes? J Trauma. 2006;61:1393–9. chronic conditions: a population-based survey in Shaanxi Province, China. 22. National Centre for Classification in Health. International Statistical PLoS One. 2013;8:e65958. Classification of Diseases and Related Health Problems, Tenth Revision, 47. Hu XB, Feng Z, Fan YC, Xiong ZY, Huang QW. Health-related quality-of-life Australian Modification (ICD-10-AM) Australian Coding Standards. after traumatic brain injury: a 2-year follow-up study in Wuhan, China. Brain Darlinghurst: University of Sydney; 1998. Inj. 2012;26:183–7. 23. Charlson ME, Charlson RE, Peterson JC, Marinopoulos SS, Briggs WM, 48. Leduc BE, Lepage Y. Health-related quality of life after spinal cord injury. Hollenberg JP. The Charlson comorbidity index is adapted to predict costs Disabil Rehabil. 2002;24:196–202. of chronic disease in primary care patients. J Clin Epidemiol. 2008;61:1234–40. 49. Polinder S, Haagsma JA, van Klaveren D, Steyerberg EW, van Beeck EF. 24. Jennett B, Bond M. Assessment of outcome after severe brain damage. Health-related quality of life after TBI: a systematic review of study design, Lancet. 1975;1:480–4. instruments, measurement properties, and outcome. Popul Health Metr. 25. Dolan P. Modeling valuations for EuroQol health states. Med Care. 1997;35: 2015;13:4. 1095–108. 50. Steyerberg EW, Vergouwe Y. Towards better clinical prediction models: 26. Hanley JA, McNeil BJ. The meaning and use of the area under a receiver seven steps for development and an ABCD for validation. Eur Heart J. 2014; operating characteristic (ROC) curve. Radiology. 1982;143:29–36. 35:1925–31. 27. Badcock D, Kelly A-M, Kerr D, Reade T. The quality of medical record review studies in the international emergency medicine literature. Ann Emerg Med. 2005;45:444–7. 28. Hosmer DW, Lemeshow S. Applied logistic regression. 2nd ed. New York: Wiley; 2000. 29. Steyerberg EW, Eijkemans MJ, Harrell FE Jr, Habbema JD. Prognostic modelling with logistic regression analysis: a comparison of selection and estimation methods in small data sets. Stat Med. 2000;19:1059–79. 30. Gabbe BJ, Harrison JE, Lyons RA, Edwards ER, Cameron PA, Victorian Orthopaedic Trauma Outcomes R. Comparison of measures of comorbidity for predicting disability 12-months post-injury. BMC Health Serv Res. 2013;13:30. 31. Bender M, Smith TC. Using administrative mental health indicators in heart failure outcomes research: comparison of clinical records and international classification of disease coding. J Card Fail. 2016;22:56–60. 32. McGorry PD. The next stage for diagnosis: validity through utility. World Psychiatry. 2013;12:213–5. 33. Henderson T, Shepheard J, Sundararajan V. Quality of diagnosis and procedure coding in ICD-10 administrative data. Med Care. 2006;44:1011–9. 34. Humphries KH, Rankin JM, Carere RG, Buller CE, Kiely FM, Spinelli JJ. Co- morbidity data in outcomes research: are clinical data derived from administrative databases a reliable alternative to chart review? J Clin Epidemiol. 2000;53:343–9. 35. Malenka DJ, McLerran D, Roos N, Fisher ES, Wennberg JE. Using administrative data to describe casemix: a comparison with the medical record. J Clin Epidemiol. 1994;47:1027–32. 36. O'Malley KJ, Cook KF, Price MD, Wildes KR, Hurdle JF, Ashton CM. Measuring diagnoses: ICD code accuracy. Health Serv Res. 2005;40:1620–39.

Journal

BMC Health Services ResearchSpringer Journals

Published: Jun 5, 2018

References

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off