The quality of medical record documentation and External cause of fall injury coding in a tertiary teaching hospital

Transcription

1 The quality of medical record documentation and External cause of fall injury coding in a tertiary teaching hospital Janet Cunningham, Dianne Williamson, Kerin M. Robinson, Rhonda Carroll, Ross Buchanan and Lindsay Paul Abstract This paper reviews the documentation and coding of External causes of admitted fall cases in a major hospital. Intensive analysis of a random selection of 100 medical records included blind re-coding in the International Statistical Classification of Diseases and Related Health Problems, Tenth revision, Australian Modification (ICD-10-AM), Fifth Edition for External causes to ascertain whether: (i) the medical records contained sufficient information for assignment of specific External cause codes; and (ii) the most appropriate External cause codes were assigned per available documentation. Comparison of the hospital data with the state-wide Victorian Admitted Episodes Database (VAED) data on frequency of use of External cause codes revealed that the index hospital, a major trauma centre, treated comparatively more falls involving steps, stairs and ladders. The hospital sample reflected lower usage, than state-wide, of unspecified External cause codes and Other specified activity codes; otherwise, there was similarity in External cause coding. A comparison of researcher and hospital codes for the falls study sample revealed differences. The ambulance report was identified as the best source of External cause information; only 50% of hospital medical records contained sufficient information for specific code assignation for all three External cause codes, mechanism of injury, place of occurrence and activity at time of injury. Whilst all medical records contained mechanism of falls injury information, 16% contained insufficient details, indicating a deficiency in medical record documentation to underpin external cause coding. This was compounded by flaws in the ICD- 10-AM classification. Keywords (MeSH): ICD-10-AM; International Classification of Diseases; Medical Records; Falls, Accidental; Wounds and Injuries; Clinical Coding; Documentation. Supplementary Term: Health Information Management. Introduction The focus of this paper is the quality of the recording and coding of External cause information in hospitalised falls cases in an exemplar (i.e. index) hospital, specifically a large tertiary teaching hospital that is one of the state of Victoria s major trauma centres. This study was undertaken in conjunction with that described in a previous article in which the Victoriawide coded data for External cause of fall injuries were examined and compared with the corresponding Australia-wide data (Cunningham et al. 2013). Collecting, documenting and coding External cause information Injury resulting from falls, particularly in those over 65 years of age, is a major health and economic issue. Currently in Australia, the average hospital stay due to falls injuries for those in this age group is 15.5 days (Bradley 2013). Clinically, the recording and analysis of accurate and comprehensive data are essential to identify problems, prioritise interventions and assess change in the patient s condition. Hospital doctors and nurses, who may have little or no knowledge of the level of documentary detail necessary for coding of injuries, are responsible for eliciting External cause information from patients who present with injuries from falls. While knowledge of the mechanism of injury may help in patient care and treatment, the place of occurrence and activity at the time of the injury tend to be less likely to be documented. The National Centre for Classification in Health [NCCH] (2003) recommends specificity in medical record documentation of injuries. The (index) hospital used in the study has a policy of advising its doctors of the need to include External cause information in the Discharge Summary. In Victoria, expert Health Information Managers (or, in some hospitals, clinical coders) are responsible 6 HEALTH INFORMATION MANAGEMENT JOURNAL Vol 43 No ISSN (PRINT) ISSN (ONLINE)

2 for the assignment of ICD-10-AM codes based on the information documented in the medical record. The Clinical Coders Creed advises that the associated decision-making process is subjective and that expert coders have to rely on experience and, sometimes, common sense since rules and standards will never cover all cases (National Centre for Classification in Health 2008a). Maintaining integrity of the coding-drg processes and data Since the inception of casemix-based funding in 1993 the public hospital system in Victoria has relied upon comprehensive, accurate and timely clinical coding. At the administrative governance level internal clinical coding audits have been routinely conducted for many years in the state s public and private hospitals to measure and monitor the quality and integrity of the ICD-10-AM coding and Diagnosis-Related Group (DRG) allocation processes and the coded data (Cheng, Shepheard & Robinson 2005). In reality most internal coding audits are targeted to the diagnosis and procedure codes and DRG changes most likely to affect hospital revenues, that is, rather than coding of External cause of fall injuries, for instance. Rigorous, systematic external coding audits are also conducted regularly by Victoria s Department of Health and are accountable ultimately to the state s Auditor-General, in the case of public sector hospitals (Shepheard & Moore 2010). Similarly, external audits are conducted in private hospitals at the discretion of proprietors (Price & Robinson 2011). Another type of coding audit tool, Performance Indicators for Coding Quality, can be used to support the analysis of ICD-10-AM coded hospital data to identify problems which may require correction and to measure data accuracy against particular criteria 1. Quality of External cause of injury documentation and coding The coding of a fall injury requires assignment of codes for the diagnosis (body part and injury type), mechanism of injury, patient s activity at the time of the fall and the place of occurrence. The classification literature, however, tends more to describe general, than fall-related, injuries. Overseas research In Washington State, USA, LeMier, Cummings and West (2001) evaluated the accuracy of external cause 1 The indicators relevant to injury surveillance data include External cause code required but not present with Chapter 19 code and, introduced in 2008, Use of unspecified fall external code compared to use of all fall external cause codes (National Centre for Classification in Health 2008b). of injury codes in hospital discharge records. They compared assigned codes in computerised records for a stratified random sample of over 1,200 injury cases, with codes assigned by an expert coder reviewing the full, original medical record. The findings revealed overuse of non-specific codes. Hong, Walker and McKenzie (2009) examined the quality of hospital injury data in Vietnam to determine their value for injury surveillance. Based on a random sample of 205 child injury cases admitted to a large general hospital, the study involved the abstraction of information from medical records for ICD-10 coding by external coders from Hanoi and Australia; these codes were compared with the original hospital coding. At the time of the study, the ICD-10 Index had not been translated into Vietnamese. The researchers found poor documentation of External cause of injury in the medical records and no coding of external causes of injury by the hospital coders, who tended to use non-specific codes for injuries. The researchers recommended the introduction of a standard data collection form to capture injury related details required for coding. Australian research McKenzie et al. (2008) elicited the views of 402 Health Information Managers and clinical coders on lack of specificity in External cause coding, and sought suggestions for improvement. The respondents opined that poor documentation in the medical record had a major impact on External cause coding. Ambulance reports were considered the best, and Discharge Summaries and doctors notes the poorest sources of External cause information. The researchers suggested the use of a standard form to collect External cause information, for improved data quality. In their analysis of data from the Australian National Hospital Morbidity Database, Soo et al. (2009) explored usage of additional activity codes introduced in the ICD-10-AM, Third Edition. They found that although the majority of available activity codes were used, there was an over-reliance on nonspecific codes, with Unspecified activity accounting for around 50% of the activity codes assigned for injury admissions between 2001 and An analysis of injury mechanism revealed that falls accounted for 44% of Unspecified activity codes in Soo et al. also found high usage of non-specific sports activity codes such as Football unspecified and Rugby unspecified. They concluded that whilst there are sufficient activity codes, the majority of these are for sports, leisure and work. Falls are more common in people aged 65 years, yet there are few activity codes to describe activities typical of this age group (Bradley & HEALTH INFORMATION MANAGEMENT JOURNAL Vol 43 No ISSN (PRINT) ISSN (ONLINE) Cunningham

3 Pointer 2009). Because of the lack of subcategories for the most commonly recorded activity codes in hospitalised falls cases involving those aged over 65 years, crucial information regarding the types of activities resulting in falls injuries in this age group is available only for the 1.4% of cases involving workplace or sports-related activities (Bradley 2013). McKenzie et al. (2009) conducted a retrospective medical record review to assess coder agreement in the assignment of ICD-10-AM External cause codes for injury-related hospitalisations in Australia. A random sample of over 4,300 injury separations from 2002 to 2004 was drawn from a stratified random sample of 50 public hospitals in four states. The researchers found 68% agreement for complete External cause coding, 75% for place of occurrence and 68% for activity codes. With regard to the latter, the coders were most likely to agree on Unspecified activity. In their review of 752 surgical inpatient separations in a Melbourne tertiary teaching hospital, Cheng et al. (2009) found that the major cause of coding error affecting casemix-based funding was lack of proper medical record documentation available at the time of coding. Chin et al. (2013) also reported the need for comprehensive documentation of principal diagnosis/es, co-morbidities and their complications to enable optimal DRG and WIES allocation in a Victorian hospital. It is evident from the findings reported in the companion paper (Cunningham et al. 2013) and the wider literature that key factors mitigating against accurate and reliable classification of hospitalised fall injuries are threefold: inadequate clinical documentation in the medical record describing the cause and location of the patient s injury; flaws in the classification, especially a paucity of activity codes to describe activities engaged in by members of the population aged 65 years; and coders overuse of non-specific codes, including non-specific sports activity codes and the general assignment of Unspecified activity codes. Aims of the study This study aimed to address the following questions via analysis of the content of, and coding from, medical records in a tertiary teaching hospital: Do hospital medical records contain sufficient information to support and justify the assignment of specific External cause codes? Are the most specific External cause codes being assigned to match documentation in the medical record? These aims were achieved in systematic, analytical reviews of medical records and coded data: to determine whether the index hospital s medical record documentation was sufficient to assign specific External cause codes to examine the use of ICD-10-AM External cause codes assigned in a random sample of admitted episodes of care for injury caused by a fall, and compare these at two levels: (i) researcher- and hospital-assigned codes for the sample; and subsequently (ii) the hospital s codes with those for the state to analyse the quality of injury-related information from the key source documents: Discharge Summary; Ambulance Report; Emergency Department notes; and Progress Notes in the medical record to determine whether the index hospital had allocated the most specific External cause codes, consistent with documentation in the medical record. Method Research process The hospital was chosen for study because of its complex case mix and status as a major trauma centre. Additionally, the researchers had an established professional relationship with the hospital s senior Health Information Managers, based on student placement and supervisory arrangements. Approval for the research was obtained from the hospital s Ethics Committee and the La Trobe University Faculty of Health Sciences Human Ethics Committee. A randomly selected sample, comprising 100 medical records of hospitalised falls cases, was obtained as follows. An extract was obtained from the hospital s database, comprising a set of complex data items relating to separations from 1 July 2007 to 30 June 2008 (Table 1). Table 1: Index hospital extract, data items and selection criteria DATA ITEM Unit record number Principal diagnosis Principal mechanism of injury code Principal activity code Principal place of occurrence code Age of patient Sex of patient SELECTION CRITERIA (Required to locate record in file) ICD-10-AM code = S00-T75 or T79 (Injury) ICD-10-AM code = W00-W19 (Fall) (Required to calculate 5-year age group) 8 HEALTH INFORMATION MANAGEMENT JOURNAL Vol 43 No ISSN (PRINT) ISSN (ONLINE)

4 The extract, supplied as an Excel spreadsheet, contained potential cases ordered by Unit Record (UR) Number, and included admission number for each patient, all admitted episodes including a fall and all diagnoses. The extract data were validated to ensure that only cases meeting the selection criteria were used; those in which the principal diagnosis did not meet the selection criteria, or the principal mechanism of injury was not a fall, were excluded from study. Readmissions for the same principal diagnosis and same External cause codes were rejected to prevent doublecounting of cases. A search of the hospital s Patient Administration System identified the medical records of potential cases held in the primary file, and produced a list of UR and admission numbers and patient names to enable record retrieval. Identification of a case was only possible via the UR Number for persons with authorised access to the hospital s Health Information Service and computer system. All potentially identifying data were restricted to the premises, specifically within the Health Information Service. The verified sample data from the hospital database were transferred into an Access database. Queries were run to provide counts of mechanism of injury (fall), place of occurrence, and activity codes. Results from the index hospital were compared with state-wide VAED data on frequency of use of External cause codes to determine whether coding in the index hospital was representative of the state as a whole. Systematic sampling was used because it is a convenient sampling method that usually produces a representative sample of the population (Polgar & Thomas 2000). A sample of the hospital s medical records was obtained by selecting every third case on the data extract and identified as being in the main file, until a sample of 100 was achieved. The extract was ordered by UR Number rather than admission date, to minimise potential bias due to any seasonal variation in falls or associated with specific trauma or coding staff. Medical records for the 100 selected cases were retrieved and the following details entered in the worksheet: UR Number, principal diagnosis, and hospital-assigned External cause codes. External cause information was also recorded by source, under the headings: Discharge Summary; Ambulance Report; Emergency Department notes; and Progress Notes. Data items extracted from the medical records were entered into a custom-designed Excel worksheet, trialled initially on a pilot sample of five medical records. The hospital s Manager of Coding Services independently replicated the pilot process to validate the research data extraction technique. The UR Number was then added to the worksheet to enable cross-referencing with the hospital data extract. Modified worksheets were printed to enable researcher-blinding during the re-coding process. The cases were re-coded from the worksheets and External cause codes assigned, based on abstraction from the source documents. These codes were compared with those assigned by the hospital s Health Information Managers and clinical coder to determine level of agreement and comparative specificity of the two code-sets. For each case it was noted whether or not information about the mechanism of injury, the place of occurrence and the activity at the time of injury was present in the source documents. 2 Data analysis Data from the hospital were analysed to determine: counts and percentages of ICD-10-AM codes assigned for: (i) mechanism of injury; (ii) place of occurrence; (iii) activity; (iv) Unspecified activity U73.9 degree of agreement in External cause codes as allocated by the hospital and, subsequently, via researcher blind re-coding: complete agreement; agreement at the 4 th digit level; agreement at the 3 rd digit level; complete disagreement more specific External cause codes as allocated by researcher completeness of documentation within the source documents for mechanism of injury, place of occurrence and activity information. 3 Results from the hospital analysis were then compared with those from the VAED analysis to determine whether the falls coding practice in the index hospital was in general representative of that in the state s hospitals. Results Comparing the index hospital with the state Mechanism of injury coding There were 63,243 separations from the index hospital in the study period, 1 July 2007 to 30 June 2008 (Victorian Government Department of Human Services 2008). Injuries due to a fall accounted for 3.1% (n = 1,963) of these separations, 50.4% comprising females (n = 989) and 49.6% males (n = 974). The hospital recorded higher percentages of falls involving steps and ladders (16.7%) and falls from buildings and scaffolding (5.5%) than those for hospitals Victoria-wide (10.1% and 2.6%, respectively) (Table 2). 2 Discharge Summary, Ambulance Report, Emergency Department notes and Progress Notes. 3 Discharge Summary, Ambulance Report, Emergency Department notes and Progress Notes. HEALTH INFORMATION MANAGEMENT JOURNAL Vol 43 No ISSN (PRINT) ISSN (ONLINE) Cunningham

5 Table 2: Mechanism of injury for hospitalised falls, Victoria and index hospital MECHANISM OF INJURY VICTORIA % INDEX HOSPITAL % Slips, trips and stumbles 9, % % Falls involving skates, skateboards etc % % Fall involving collision with person % % Falls involving furniture 2, % % Falls involving playground equipment 1, % 5 0.3% Falls involving steps, stairs, ladders 3, % % Fall from, through out of building, scaffolding % % Other fall from one level to another 1, % % Other fall on same level 5, % % Unspecified fall 9, % % Total 38, % 1, % Table 3: Place of occurrence for hospitalised falls, Victoria and index hospital PLACE OF OCCURRENCE VICTORIA % INDEX HOSPITAL % Home 13, % % Residential institution 5, % % School, other institution and public administration area 2, % % Sports and athletics areas 1, % % Street and highway 1, % % Trade and service areas 1, % % Industrial, construction % % Farm % 5 0.3% Other specified place 1, % % Unspecified place of occurrence 11, % % Total 38, % 1, % Table 4: Activity at time of injury for hospitalised falls, index hospital and Victoria ACTIVITY VICTORIA % INDEX HOSPITAL % While engaged in sports 3, % % While engaged in leisure activity % % While working for income 1, % % Other work (e.g. domestic duties, learning activities) 1, % % While resting, sleeping, eating, etc. 3, % % Other specified activity 5, % % Unspecified activity 22, % 1, % Total 38, % 1, % Table 5: Age and gender for hospitalised falls: index hospital and study sample INDEX HOSPITAL 100 CASES AGE GROUP FEMALE MALE PERSONS FEMALE MALE PERSONS ( 4.9%) 101 (10.4%) 150 ( 7.6%) (24.9%) 449 (46.1%) 695 (35.4%) 17 (28.8%) 22 (53.7%) (70.2%) 424 (43.5%) 1,118 (57.0%) 42 (71.2%) 19 (46.3%) 61 Total 989 (50.4%) 974 (49.6%) 1,963 (100%) 59 (59.0%) 41 (41.0%) HEALTH INFORMATION MANAGEMENT JOURNAL Vol 43 No ISSN (PRINT) ISSN (ONLINE)

6 Substantially fewer Unspecified falls were recorded by the index hospital (21.0%) than by all Victorian hospitals, collectively (25.6%), amounting to a difference of 18%. The hospital does not normally treat paediatric cases, thus accounting for comparatively fewer falls involving playground equipment. Place of occurrence coding There were similarities in the coding of place of occurrence, as shown in Table 3. The index hospital recorded street and highway as place of occurrence more than twice as often as all hospitals Victoria-wide (9.2% and 4.2%, respectively). Unspecified place of occurrence was assigned substantially less frequently by the index hospital (20.9%) than state-wide (29.0%). Activity coding The pattern of activity coding was similar for state and hospital (Table 4). The index hospital recorded Unspecified activity less frequently, and Other specified more frequently, than the state s hospitals collectively. Analysis of the study sample The study sample of 100 cases, comprising 59 female participants and 41 males (Table 5) was then compared, by gender and age group, with all hospitalised falls for the index hospital during the study period. Mechanism of injury: comparison of hospital and study sample Slips, trips and stumbles were the leading cause of fall in hospital-coded (35 cases) and researcher-(re-) coded (30) episodes (Table 6). Researcher assignment of Unspecified fall (n = 16) was less frequent than by hospital staff (n = 20). Table 6: Mechanism of injury: comparison of index hospital and study sample FIRST EXTERNAL CAUSE HOSPITAL RESEARCHER Slips, trips and stumbles Falls involving furniture 5 6 Falls involving steps, stairs, ladders Fall from, through, out of building 5 5 Other fall on same level Unspecified fall Total Place of occurrence: comparison of hospital and study sample Coding of place of occurrence showed some variations; however, there was concordance in hospital- and researcher-assignation of Unspecified place of occurrence (Table 7). Table 7: Place of occurrence: comparison of hospital and study sample PLACE OF OCCURRENCE HOSPITAL RESEARCHER Home Aged care facility Street and highway Other specified place of occurrence Unspecified place of occurrence Total Activity: comparison of hospital and study sample Activity coding (Table 8) showed the greatest differences between hospital- and researcher-(re-)coded cases, with sport and leisure assigned in seven cases by the former and in 19 by the latter. Researcher assignment of Unspecified activity (31 cases) was less than by the hospital staff (44). Table 8: Activity at time of injury: comparison of hospital and study sample ACTIVITY HOSPITAL RESEARCHER Sport and leisure 7 18 While working for income and other work While resting, sleeping, eating etc 7 14 Other specified activity Unspecified activity Total Level of agreement in External cause coding There was complete researcher and hospital staff code agreement in 31 of the 100 cases. All place of occurrence codes concurred at the 3-character level (all start with Y92). Figure 1 shows complete agreement for coding of 71 mechanism of injury codes, 81 place of occurrence codes and 56 activity codes. More specific, researcher-assigned codes There were more researcher- than hospital-assigned specific codes for mechanism of injury with slips, HEALTH INFORMATION MANAGEMENT JOURNAL Vol 43 No ISSN (PRINT) ISSN (ONLINE) Cunningham

7 trips and stumbles researcher-assigned for five cases originally coded Unspecified fall by the hospital. Fall in or into bath-tub or shower was researcher-assigned for three cases hospital-assigned as Fall on same level from slipping. Home, aged care facilities, road and highway were researcher-assigned as place of occurrence for seven cases assigned Unspecified place of occurrence by the hospital. Specific researcher-assigned activity codes seen most often in place of a non-specific hospitalassigned code were: U56.2 Walking (7); U73.1 While engaged in other type of work (7); and U73.2 While resting, sleeping, eating (7). Completeness of documentation The Ambulance Report, when present, was the best source of external cause information for mechanism of injury, place of occurrence and activity (Table 9). The Emergency Department notes were very reliable sources of mechanism of injury information, excepting one case involving a 93-year old patient with no recollection of how she was injured. The Discharge Summary was the poorest source for all elements. Details of the mechanism of injury were present for all 100 cases in the study sample. Discussion Level of agreement in External cause coding, study sample The complete agreement on External cause codes in 31% of the study sample represents a substantially lower statistic than the 68% reported by McKenzie et al. (2009); however the current study was restricted to fall injuries, and was relatively small compared with 4,300 cases in the national study. The current study involved thorough searches of both the Index and the Tabular volume of ICD-10-AM for specific codes. Searching the Tabular directly to locate codes is not normally accepted practice in hospital coding environments. This method of re-coding differed from previous studies (LeMier et al. 2001; McKenzie et al. 2009; Hong et al. 2009) that assessed coding accuracy. Agreement between researcher- and hospitalallocated codes was found to be highest for place of occurrence coding; it was lowest for activity coding, perhaps because this element relies more on the coder s interpretation of the information. Many common activities cannot be subsumed under one broad category: shopping, for instance, may be a leisure activity for some people and for others a chore. This demonstrates the need for clear guidance from the ICD-10-AM Index to facilitate consistent activity coding. Differences in coding may also be attribut- Percentage agreement at each code level 100% 90% 80% 70% 60% 50% 40% 30% Complete disagreement Agreement at 3rd digit but disagreement at 4th Agreement at 4th digit but disagreement at 5th Complete agreement 20% 10% 0% Fall code Place of occurrence Activity Figure 1: Percentage agreement, at each code level, for hospital and study sample 12 HEALTH INFORMATION MANAGEMENT JOURNAL Vol 43 No ISSN (PRINT) ISSN (ONLINE)

8 able to conflicting documentary descriptions of the injury event; for example, it was observed that the clinical documentation for one case incorporated three different versions for mechanism of injury, place of occurrence and activity. Are the most specific codes assigned? Hospital staff did not always assign the most specific External cause codes to match documentation in the medical record, compared to the researcher-allocated codes for mechanism of injury, place of occurrence and activity. Codes assigned to the study sample were only counted as more specific when the hospital-assigned code was either Unspecified or Other specified, except for W18.2 Fall in or into bath-tub or shower. Since most falls in the shower are likely to be due to slipping it is probable that a coder will search the Index looking for slip. There is no Exclusion Note at W01.0 Fall on same level from slipping for slip in shower and thus it is likely that falls in showers and baths are under-counted. The code W18.2 was assigned six times in the index hospital and only 115 times Victoria-wide in Four of the more specific place of occurrence codes assigned to the study sample were researcher-assigned for the home or aged care facilities, with extra location information found variously in Emergency Department notes, Ambulance Report and Progress Notes. Descriptions of walking, out for a walk and walking back from the shops were frequently provided in the medical records yet the hospital coders did not assign U56.2. Notably, this code was assigned only 99 times Victoria-wide and nine times in the hospital, for hospitalised falls in the study period. It is probably under-used because it is listed under sport, and coders may not necessarily consider walking to be a sport or leisure activity. Descriptions of walking and running were coded as Other specified activity four times by hospital coders, suggesting difficulty in finding the codes. Some cases lacked detail of the injury event as the respective medical records described dementia or poor recollection by the patient. Is medical record detail sufficient for coding? Only 50% of the medical records in the study sample contained sufficient information to assign specific codes for all three External cause codes, mechanism of injury, place of occurrence and activity at time of injury. Whilst all cases contained mechanism of injury information to identify a fall, 16% contained insufficient detail to assign a specific code. The level of detail varied considerably between medical records. When present, the Ambulance Report was the best source of information for mechanism of injury, place of occurrence and activity, and contained more External cause information than Emergency Department notes, Progress Notes or the Discharge Summary. Consistent with McKenzie et al. s (2008) findings, the Discharge Summary was found to be the poorest source of information. The Emergency Department record and Discharge Summary form in our sample lacked prompts on collection of injury surveillance data. Whilst the index hospital advises doctors to include External cause of injury information in the Discharge Summary, the relevant details have to be documented in the medical record upon presentation and admission so that subsequently the junior doctor responsible for writing the Discharge Summary can find and incorporate them. Strengths and limitations of the research Strengths The index hospital s complex casemix and trauma centre status lent credibility to its capacity and expertise in the treatment and coding of fall injuries. Furthermore, the multi-level analysis (hospital falls coding compared with the state-wide coding, hospitaland researcher-assigned code comparison within the Table 9: Completeness of documentation for External cause elements TYPE OF DOCUMENTATION PAPERWORK PRESENT MECHANISM OF INJURY PLACE OF OCCURRENCE ACTIVITY Ambulance report (100%) 61 (90%) 47 (69%) Emergency Department notes (99%) 49 (50%) 54 (55%) Progress notes (77%) 31 (37%) 31 (37%) Discharge summary (87%) 23 (32%) 18 (25%) Cases with element present: Hospital coder Researcher HEALTH INFORMATION MANAGEMENT JOURNAL Vol 43 No ISSN (PRINT) ISSN (ONLINE) Cunningham

9 sample and researcher analysis of fall injury documentation supporting hospital- and researcher-coding) were strengths of the research design. Limitations Differences between researcher- and hospital-assigned codes in the sample may be driven by the highpressure environment of Victorian hospital coding: (i) Under the casemix-based funding model, the hospital s revenue is dependent on the quality of the coding and DRG allocation, and External cause, Activity and Mechanism of Fall do not change the DRG and subsequently the amount of money allocated to the hospital; and (ii) Coding must be completed within tight timeframes to meet government-imposed reporting deadlines. The researcher (re)coding was not subject to these constraints; however, whilst there was researcher access to the paper medical records, the lack of access to online reports, correspondence and the trauma registry may have contributed to code disagreement. As the research focus was specifically on identifying falls injury documentation, and (re)coding, some difference between researcher and hospital outcomes might be expected in light of these limitations. Conclusion The main differences in coding practice between the state as a whole and the index hospital related to the hospital s greater coding specificity evidence in the pattern of lower usage of unspecified External cause codes and Other specified activity codes. This is possibly attributable to several factors: the high standard of coding at the index hospital, which was undertaken primarily by Health Information Managers; the Health Information Service s well-developed coding Professional Development Program; the hospital s rigorous internal coding audit and quality management program; and, probably, relatively greater experience and expertise of the hospital s staff in injury coding, reflective of its status as a teaching hospital and major trauma centre. The quality of medical record documentation directly affects the quality of clinical coding. This is evidenced in our findings that half of the medical records in the study sample contained insufficient information for specific code assignation for all three External cause codes, mechanism of injury, place of occurrence and activity at time of injury and, whilst all cases contained mechanism of injury information to identify a fall, 16% had insufficient detail for specific code assignation. Flaws in the classification (e.g. guidance from the Index, and insufficient activity codes) are also likely to contribute to coding variations, a situation that warrants further investigation. Finally, our finding of lack of relevant detail in the Discharge Summary should alert all Health Information Managers and Clinical Coders to abstract comprehensively from all sources including the entire medical record. References Bradley, C. (2013). Hospitalisations due to falls by older people, Australia Injury research and statistics series number 70. Cat. no. INJCAT 146. Canberra, Australian Institute of Health & Welfare. Available at: aspx?id= (accessed 5 May 2013). Bradley, C. and Pointer, S. (2009). Hospitalisations due to falls by older people, Australia Injury research and statistics series number 50. Cat. No. INJCAT 122. Canberra, ACT: Australian Institute of Health & Welfare. Summary available at: publication-detail/?id= Cheng, D., Shepheard, J. and Robinson, K. (2005). The role of coding auditing in health information classification in Victoria, Australia. Proceedings of the First Asia-Pacific Medical Record Congress in Conjunction with the Chinese Medical Record Association 14 th National Convention, Beijing, China, September. Beijing, China: CHMRA: 1-3 (Chinese); 4-9 (English). Cheng, P., Gilchrist, A., Robinson, K.M. and Paul, L. (2009). The risk and consequences of clinical miscoding due to inadequate medical documentation: a case study of the impact on health services funding. Health Information Management Journal 38(1): Chin, N., Perera, P., Roberts, A. and Nagappan, R. (2013). Review of medical discharge summaries and medical documentation in a metropolitan hospital: impact on Diagnosis Related Groups (DRGs) and Weighted Inlier Equivalent Separation (WIES). Internal Medicine Journal DOI: /imj (Published online 24 January). [Available at: doi/ /imj.12084/abstract (accessed 5 May 2013). Details of hard copy publication not available at time of writing]. Cunningham, J., Williamson, D., Robinson, K.M. and Paul, L. (2013). A comparison of state and national Australian data on external cause of injury due to falls. Health Information Management Journal 42(3): Available: Cunningham Hong, T. T., Walker, S. M. and McKenzie, K. (2009). The quality of injury data from hospital records in Vietnam. Health Information Management Journal 38(1): LeMier, M., Cummings, P. and West, T.A. (2001). Accuracy of external cause of injury codes reported in Washington State hospital discharge records. Injury Prevention 7(4): HEALTH INFORMATION MANAGEMENT JOURNAL Vol 43 No ISSN (PRINT) ISSN (ONLINE)

10 McKenzie, K., Enraght-Moony, E., Harding, L., Walker, S., Waller, G. and Chen, L. (2008). Coding external causes of injuries: problems and solutions. Accident Analysis and Prevention 40 (2008): McKenzie, K., Enraght-Moony, E.L., Waller, G., Walker, S.M., Harrison, J.E. and McClure, R.J. (2009). Causes of injuries resulting in hospitalisation in Australia: Assessing coder agreement on external causes. Injury Prevention 2009 (15): National Centre for Classification in Health (2008a). The International Statistical Classification of Diseases and Related Health Problems, Tenth Revision, Australian Modification 6 th ed. (ICD-10-AM). Lidcombe, NSW: NCCH. National Centre for Classification in Health (2008b). Performance Indicators for Coding Quality (PICQ) 2008: User guide. Lidcombe, NSW: NCCH. National Centre for Classification in Health (2003). The good clinical documentation guide. Lidcombe, NSW: NCCH. Polgar, S. and Thomas, S.A. (2000). Introduction to research in the health sciences. 4 th ed. Edinburgh, Churchill Livingstone. Price, E. and Robinson, K. (2011). The coding masterpiece: a framework for the formal pathways and processes of health classification. Health Information Management Journal 40(1): Shepheard, J. and Moore, V. (2010). Victorian state-wide audit program: methodology, results and significance for activity-based funding. Paper presented at the 2010 National Conference of the Health Information Management Association of Australia, Sydney, New South Wales, October. Soo, I.H-Y., Lam, M.K., Rust, J. and Madden, R. (2009). Do we have enough information? How ICD-10-AM Activity codes measure up. Health Information Management Journal 38(1): Victorian Government, Department of Human Services (2008). Your hospitals: a report on Victoria s public hospitals July 2007 to June Melbourne, Vic Department of Human Services. Available at: yourhosps1008.pdf (accessed 30 March 2013). Janet Cunningham*, BHlthInfoMangt(Hons) Health Information Advisor Admitted Emergency and Elective Data Department of Health Melbourne, Victoria, 3000 AUSTRALIA Janet.Cunningham@health.vic.gov.au Dianne Williamson, BAppSc(MRA), GradDipErg Senior Lecturer Department of Health Information Management School of Public Health & Human Biosciences Faculty of Health Sciences La Trobe University Bundoora VIC 3086 AUSTRALIA D.Williamson@latrobe.edu.au Kerin M. Robinson, BHA, BAppSc(MRA), MHP, CHIM Head, Department of Health Information Management School of Public Health & Human Biosciences Faculty of Health Sciences La Trobe University Bundoora VIC 3086 AUSTRALIA K.Robinson@latrobe.edu.au Rhonda Carroll, RMRL Director Information and Performance Melbourne Health Victoria Rhonda.Carroll@mh.org.au Ross Buchanan, BMedRecAdmin(Hons), MHSc(HlthAdmin) Director, Health Information Services Information Development Division Alfred Health Melbourne VIC r.buchanan@alfred.org.au Corresponding author: Lindsay Paul, BSc, GradDipCommHlth, PhD Adjunct Senior Lecturer School of Public Health & Human Biosciences Faculty of Health Sciences La Trobe University Bundoora VIC 3086 AUSTRALIA Lindsay1645@bigpond.com * Research undertaken during Honours studies, Department of Health Information Management, La Trobe University. HEALTH INFORMATION MANAGEMENT JOURNAL Vol 43 No ISSN (PRINT) ISSN (ONLINE) Cunningham