Abstract
Background
Coding of patients' diagnosis and surgical procedures is subject to error levels of up to 40% with consequences on distribution of resources and financial recompense. Our aim was to explore and address reasons behind coding errors of shoulder diagnosis and surgical procedures and to evaluate a potential solution.
Methods
A retrospective review of 100 patients who had undergone surgery was carried out. Coding errors were identified and the reasons explored. A coding proforma was designed to address these errors and was prospectively evaluated for 100 patients. The financial implications were also considered.
Results
Retrospective analysis revealed the correct primary diagnosis was assigned in 54 patients (54%) had an entirely correct diagnosis, and only 7 (7%) patients had a correct procedure code assigned. Coders identified indistinct clinical notes and poor clarity of procedure codes as reasons for errors. The proforma was significantly more likely to assign the correct diagnosis (odds ratio 18.2, p < 0.0001) and the correct procedure code (odds ratio 310.0, p < 0.0001). Using the proforma resulted in a £28,562 increase in revenue for the 100 patients evaluated relative to the income generated from the coding department.
Conclusion
High error levels for coding are due to misinterpretation of notes and ambiguity of procedure codes. This can be addressed by allowing surgeons to assign the diagnosis and procedure using a simplified list that is passed directly to coding.
Introduction
The Department of Health outlined plans to introduce a new system of financing secondary care institutions in April 2002. 1 This led to the Payment by Results (PbR) scheme being implemented throughout England and Wales in 2005. 2 This payment scheme reimburses hospitals for care they provide according to a fixed case-mix tariff. This marks a fundamental change in the way hospitals are financed, with primary care trusts being stakeholders and purchasing services from secondary care. The tariff is derived from the average cost of providing a specific treatment. This cost is enhanced with increased age and comorbidity of the patient. It is also dependent upon region. Codes are used to record primary diagnosis, comorbidity and complications using the International Classification of Diseases 10 (ICD-10). The surgical procedure is coded using Office of Population Censuses and Surveys (OPCS-4) code(s) that are used to generate a Health Care Resource Group (HRG) code.
Inaccurate coding leads to the assignment of an incorrect tariff with loss of revenue to the hospital. 3 Error rate for arthroscopic shoulder procedures of up to 81% have been reported. 4 One source of error is using the discharge summary alone to code the procedure, which would result in significant loss of income for the hospital. 5 Coding is also used to analyse clinical performance at local and national levels and to produce league tables. 6 The British Elbow and Shoulder Society (BESS) recognised the high rate of inaccurate coding declaring the tariff ‘Not fit for purpose’. 7 The inaccuracy of the coding is partly attributable to the 1200 plus OPCS-4 codes that currently exist. In an effort to simplify this BESS proposed 31 shoulder codes.
We conducted an audit to explore and address the reasons behind errors in the coding of shoulder pathology and surgical procedures, and to evaluate a potential solution. The financial implications were considered.
Methods
The study institution is a large teaching hospital with a catchment population of approximately 780,000 people. 8
Retrospective review of coding errors
A retrospective review of 100 patients consecutive patients who underwent arthroscopic shoulder surgery by the senior author was carried out. We compared the codes assigned by two independent investigators analysing clinical and operative notes with the ICD-10 and OPCS-4 codes provided by the coding department. Coding errors were identified and recorded. These errors were discussed with our colleagues within the coding department and the reasons for these errors were explored. Their thoughts and recommendations were taken into account.
Implementation and evaluation of coding proforma
We designed a coding proforma to address the reasons for coding errors (Figure 1). This was prospectively evaluated for 100 consecutive patients who underwent arthroscopic shoulder surgery. The surgeon completed the proforma at the time of the operation. This recorded the diagnosis using ICD-10 code and the surgical procedure using a simplified list as proposed by British Elbow and Shoulder Society) and the current OPCS-4 code(s). The coding department supplied the ICD-10 and OPCS-4 codes that they had recorded for the same group of patients. Both the proforma and the coding departments results were compared to the diagnosis and operation that was recorded in the operation note recorded immediately after the surgery by the surgeon.
Shoulder coding proforma.
Financial implications
A costing analysis was carried out using the National Schedule for Reference Costs. Patient co-morbidities and the Market Forces Factors (cost of living factors dependent on the geographical region of the hospital) were not accounted for as these are influenced by region and case-mix. We therefore present baseline costs. Total remuneration for the procedure recorded on the proforma was compared to the coding department calculations using the HRG tariff for the assigned OPCS-4 code(s). A further cost analysis was performed comparing the actual cost of arthroscopic subacromial decompression with arthroscopic rotator cuff repair. Recently published expenditure 9 for these procedures was used to compare the income generated for the proforma and by hospital coding.
Results
Retrospective review of coding errors
The correct primary diagnosis was assigned in 78 patients. However, a secondary diagnosis was not coded in 24 patients, so only 54 patients (54%) had a truly correct diagnosis. A correct or partially correct procedure code (W84.8 = therapeutic endoscopy, n = 94) was assigned for 98 patients, but 91 patients had additional procedure(s) that had not been coded. This included 30 patients undergoing an arthroscopic rotator cuff repair that was not recorded (T79.1 = plastic repair of rotator cuff). No patient undergoing subacromial decompression had an additional code assigned other than the W84.8. No specific code is assigned to this procedure within the OPCS-4 cipher. Hence, only 7 (7%) patients had a truly correct procedure code assigned.
The coding department described three areas that caused difficulty when coding and could be a source for coding errors: complex OCPS-4 coding system, identifying the diagnosis and operative procedure from the notes, and not being able to retrieve case notes.
Coding proforma
The proforma assigned the correct diagnosis for 97 patients, and failed in two patients to assign a correct procedure code. Hospital coding correctly coded the primary diagnosis in 82 patients. A secondary diagnosis was omitted by coders in 18 patients, resulting in 64 patients (64%) having an entirely correct procedure code(s) assigned. A correct or partially correct procedure code (W84.8 = therapeutic endoscopy, n = 90) was identified in 97 patients; however, 83 patients had additional non-coded procedure(s). Hence, only 14 patients (14%) had a truly correct procedure code assigned. The proforma was significantly more likely to assign the correct diagnosis (odds ratio 18.2, p < 0.0001 Fishers exact test) and the correct procedure code (odds ratio 310.0, p < 0.0001 Fishers exact test).
All except two procedures were covered by the simplified list of procedures.
Financial implications
Using the HRG tariff, the inaccuracy of coding in our series resulted in a £28,562 loss of income to the trust for the 100 patients that were prospectively complied. This was calculated by subtracting the gross reimbursement according to the codes assigned by the coders, from gross reimbursement according to the codes assigned by the surgeon using the HRG tariffs. We believe that the actual cost of this inaccurate coding is far greater due to the inaccurate HRG tariff codes currently available. Using the actual cost of arthroscopic subacromial decompression (£1307) and arthroscopic rotator cuff repair (£2672) the true loss of income would be £55,322, if the HRG codes reflected the actual cost.
Discussion
The current system of data collection and coding varies between NHS hospitals. Trusts are audited on the accuracy of data that they collect. Previous audits have demonstrated coding error rates of up to 40% of patients. 10 One source of coding error is that of inadequate or illegible documentation within case notes. 11 Errors arise from erroneous interpretation of case notes. The provisional preoperative diagnosis and planned procedure may differ from the intraoperative diagnosis and actual procedure preformed. This may again differ from that recorded on the discharge summary which has been shown to result in a 15% to 20% loss of income if they are used to assign the HRG code. 5 However, Jameson and Reid 12 conducted a postal questionnaire of all NHS hospital trusts in England and Wales and found that only 55% of coders had access to patient notes.
The administrative costs of PbR have been estimated at £180,000 per hospital trust.13,14 We have shown that current data collection is inaccurate and results in an incorrect representation of patient information. Our simple proforma which can be adapted by other orthopaedic specialties allows accurate data to be gathered without relying on third party interpretation of patient notes with clear communication between clinicians, coders and the financial department.
We report a 97.5% (195/200) partial accuracy rate for the assignment of the correct procedure code by our coding department. This falls to 13.0% (26/200) if a complete code is used as the minimum standard. Other authors have demonstrated similar results in shoulder surgery. 4 Harshavardhana et al. 15 reported a coding accuracy for the primary procedure as 91%, falling to 71% for a complete description of surgery. More recently Astle et al. 16 reported an overall accuracy rate of 45%. They identified deficiencies in the clinical knowledge amongst coders to be a source of error. Our proforma eliminates this error. Astle et al. also raised issues with the OPCS-4 coding and suggested the introduction of a standardised coding system for common shoulder procedures. The simplified list of procedures put forward by BESS, covering the majority of shoulder procedures (98%), provides a solution.
There is scant literature regarding coding accuracy from countries other than the UK. This may be due to the fact that incorrectly coding procedures may be classed as fraud in some medical systems upon billing the patients and insurance companies. 17 The American system of coding does seem more complicated than our own, with fines and civil penalties if miscoding is identified. An audit by the American Academy of Orthopaedic Surgeons committee on coding revealed non-physicians to under code (assign a procedure with less value) in fear of penalties for over coding.17 They also demonstrated the most accurate coding to be achieved when a physician assigns the code(s). We have also found the surgeon to be the most reliable coder, and hence should be involved in the assignment of codes to ensure their accuracy.
We have shown that a simple proforma results in accurate communication of data between clinician and coder. Application of the correct procedure code will ultimately reflect true clinical practice thereby acknowledging the performance of the surgical unit and reap the appropriate financial reward.
Footnotes
Declaration of conflicting interests
None declared.
Funding
This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.