Does Structured Audit and Feedback Improve the Accuracy of Residents CPT E&M Coding of Clinic Visits?

Transcription

1 648 October 2010 Family Medicine Does Structured Audit and Feedback Improve the Accuracy of Residents CPT E&M Coding of Clinic Visits? Kelly S. Skelly, MD; George R. Bergus, MD, MA Background and Objectives: Family physicians frequently err when applying Current Procedural Terminology (CPT) evaluation and management (E&M) codes to their office visits, but there are few published prospective studies on educational interventions to improve coding. Methods: Over a 6-year intervention period, 429 resident patient notes from return clinic visits were recoded by a faculty member with coding expertise. Feedback on coding accuracy and annual educational coding workshops were provided to the residents. Coding accuracy was calculated by subtracting residents code from that of the faculty. Coding accuracy was analyzed cross-sectionally using all available data and longitudinally for 14 residents with data from all 3 years of the residency. Results: Analysis of codings by 68 residents found that residents undercoded their clinic visits by 0.49 levels of service. Higher training year of the resident was associated with more accurate coding. Improvement over time was also found with the longitudinal analysis. However, comparison of 23 residents coding from before the first feedback and didactic session to codings after starting feedback suggests that these improvements were not due to the intervention. Conclusions: Residents improved in coding accuracy over time, but our educational intervention may not have been responsible for the improvement. (Fam Med 2010;42(9): ) Family physicians frequently err when applying Current Procedural Terminology (CPT) evaluation and management (E&M) codes to their office visits. An observation study of 138 family physicians found that only about half of the CPT E&M codes generated by medical record review concorded with the actual billing code assigned by the physicians. 1 In a recent study in which a randomly selected group of family physicians evaluated six hypothetical progress notes developed to represent different levels of service, physicians agreed with expert coders only about half of the time. Typically, the physicians undercoded the visits. 2 Clinic visits are focused on patient care and frequently leave little time for coding and billing. However, without correct coding, the physician is at risk for decreased revenue and investigations resulting in discipline or loss of future participation in Medicare and other programs. With appropriate documentation and coding, physicians get the payment they deserve for their work, and they are more likely to have a From the Department of Family Medicine, University of Iowa. sustainable practice. Multiple mechanisms have been implemented to attempt to improve the accuracy of coding. Because of the importance of accurate coding of physician services, the Residency Review Committee (RRC) requirements for family medicine residencies state that programs must provide training for residents in documentation and coding. This requirement does not stipulate how the training must be undertaken. Through a literature search, no previous study investigating the impact of an educational intervention alone on CPT E&M coding by residents was found. There is a lack of published research on educational interventions to improve CPT E&M coding by family medicine residents. Therefore, the purpose of this study was to determine whether an intervention auditing resident coding performance with feedback on accuracy combined and educational coding workshops would result in improved coding accuracy. Methods After University of Iowa Institutional Review Board approval for human subjects, a multifaceted educational intervention s impact on family medicine residents outpatient coding over a 6-year period ( )

2 Original Article Vol. 42, No was evaluated. Once yearly, all residents (n=20 28 per year) attended a 1-hour didactic interactive conference on CPT E&M coding, taught by a faculty member with expertise in coding. During this session, the residents were taught the basic rules and concepts related to CPT E&M coding and to reinforce this content, during the workshops, residents were asked to code several family medicine clinical notes copied from the medical records of active patients with the aid of a standardized coding tool. During this interactive session, a faculty then provided the correct coding for these clinical notes and used the coding tool to explain why. Additionally, each resident s coding accuracy was assessed several times each year by reviewing two clinic notes of each resident, pulled at random, from their first day of clinic of the month selected. This re-coding of the residents notes was done by the same faculty member with expertise in coding who led the coding workshops and used the same standardized coding tool used in the workshops. Based on documentation available, each clinical note received a CPT E&M code by the faculty that was compared with the code assigned by the resident at the time of the clinical encounter. The clinical notes and faculty coding were returned to the residents as formative feedback several times each year in a random fashion. The faculty coder s accuracy was assessed by having a hospital compliance and coding specialist who is a certified professional coder separately recode 5% or 22 of the notes randomly selected from the 429 notes. Complete agreement was present in 18 of 22 re-coded notes, and on all four of the discordant notes, the two codings were within 1 point most commonly as in instead of During the 6-year study period in which we used the audit and feedback intervention, we assessed the change of our residents performances over time. To assess whether the educational intervention resulted in improved coding accuracy, resident performances were assessed both cross-sectionally and longitudinally. Data Analysis The clinical encounters of only return patients that were coded using CPT E&M codes visits were used for this analysis. For each encounter, an error score was calculated by subtracting the faculty level of coding from the residents level. Thus, error scores could range from +4 to -4 with negative scores representing undercoding on the part of the resident. The error score data were then analyzed cross-sectionally using a one-way ANOVA with the resident s year of training as the factor variable and accuracy of CPT E&M coding as the response variable. Newman-Keuls Multiple-Comparison Test was used as the analysis of variance technique. Data were also analyzed longitudinally using reported measures ANOVA for the group of residents for whom there are error score data for all 3 years of their participation in the training program. Residents did not have data from all 3 years if they left the residency program before completion, if they were in their second or third year of training at the time of the initiation of the coding intervention, or if they were not in their third year when we analyzed these data. Because this was an educational intervention within a residency training program, there was no prospective control group. To assess the impact of the education intervention on residents coding accuracy, we used a baseline coding as the residents codings analyzed prior to their first didactic session on coding. This was used as a baseline control group. These residents were in all 3 years of training and had not received any formal feedback on their coding. A one-way ANOVA was undertaken with coding error as the dependent variable and resident year of training and before-and-after intervention status of the coding as the independent variables. Analyses were undertaken using NCSS 2007, and a P value less than.05 was considered statistically significant. Results During the 6-year study period, 496 family medicine resident codings were analyzed for all residents. Forty-one visits were coded as new visits, and three were coded as preventive medicine visits. These codings were excluded from the analysis. An additional 23 codings were completed by residents before the first feedback and education session and were removed to be used as baseline controls. This left 429 returning patient visits that were reviewed for accuracy of coding. A total of 150 of these visits were completed by first-year residents, 147 were with second-year residents, and 132 were with third-year residents. The controls and intervention notes had a similar distribution of the 3 years of the residency (P=.66) A mean of 6.3 visits per resident were reviewed by the faculty member, with a range of 1 to 18 encounters per resident. The median number of visits reviewed was 4.5. These visits were provided by a total of 68 different family medicine residents. The CPT code was the most frequent in this dataset and made up nearly 75% of the visits. Further details of the coding can be found in Table 1. Cross-sectional Analysis The mean coding error score for these notes suggested that the residents systematically under-coded their visits by 0.49 levels of service (SD=0.65). The distribution of the error scores is shown in Figure 1. Residents errors were associated to the actual level of service (P<.001) as they tended to over-code simple visits and under-code more complex visits (Table 1). Cross-sectional analysis found that the training year of a resident was strongly associated with the mean

3 650 October 2010 Family Medicine Table 1 Errors in CPT Coding by Residents of the 429 Return Patient Visits Included in This Study True CPT Coding Number of Visits % of Total Visits Visits Coded Correctly (%) Mean Coding Error % 2 (50%) % 67 (74.4%) % 132 (41.6%) % 3 (16.7%) Totals % CPT Current Procedural Terminology True CPT coding was determined by the reviewing faculty. Figure 1 Coding Error Scores for the CPT E&M Codes Assigned by Family Medicine Residents to 429 Patient Encounters in the Family Medicine Model Office coding error scores (P=.02). Third-year residents had the lowest mean error score (-0.37) followed by the second-year residents (-0.50), and first-year residents had the largest mean error score (-0.59). Post hoc comparison confirms that there is a significant difference between the mean coding error scores of first- and third-year residents. Longitudinal Analysis There was a subset of 14 residents with CPT E&M coding data from 191 visits over their 3 years of the residency program. This analysis, by repeated measures ANOVA, found significant variation in the coding accuracy by residents (P=.001). More importantly, we found a significant association between year of training and the coding accuracy of residents (P<.001). The mean deviation of first-year residents was -0.59, and this improved to in the third year of training. These mean deviation scores were significantly different. Coding error is calculated by subtracting the last digit of residents visit coding from that of the reviewing faculty coder. A negative error indicates that the visit was undercoded by the resident. Before and After Intervention While residents improved over time in their coding accuracy, a comparison of resident performance before and after the intervention suggest that the observed improvement in accuracy was not the result of the intervention. The coding accuracy for the 23 clinic visits before the intervention were used for this comparison. Postgraduate year (PGY) of the residents training was clearly associated with coding accuracy (P=.01) in both the pre- or postintervention codings by residents (P=.97). A graph showing coding accuracy as a function of PGY of training and whether the accuracy was assessed before or after the intervention is shown in Figure 2.

4 Original Article Vol. 42, No Figure 2 Plot of Mean (With SE) of Coding Errors for the Residents Return Patients Visits as a Function of Resident Year of Training and Whether the Coding was Completed Before or After the Education Intervention SE standard error Discussion Accurate CPT E&M coding is an essential skill for physicians. We found that residents routinely undercoded their patient visits in a family medicine clinic. This finding suggests that residency programs should introduce interventions to address this performance problem. Indeed, the Family Medicine RRC mandates that all training program will provide residents with feedback on their coding. Goals for teaching correct CPT coding include improvement in accuracy, increased interest in documenting correctly, and continued improvement in accuracy over time. Our program designed to improve resident coding performance utilized regular faculty review of residents coding combined with delayed formative feedback to individual residents and formal didactics about coding for all residents. Audit and feedback interventions are widely used in clinical medicine, and research suggests that these interventions can improve professional practices although these interventions are not always effective. We did find that residents significantly improved as they advanced in their training. However, our intervention may not be responsible for the improvement in coding accuracy. Although we only have a small number of resident codings from before starting our coding didactics and the faculty feedback, we find a similar coding accuracy after our intervention and a similar pattern of improvement as residents advanced in their training. All the reasons for their improvement over time are not clear but could include resident educational exposure within the act of patient care with preceptor input. It is not accurate to conclude, based on our findings, that residents do not need any structured instruction on coding. While our residents might have improved independently of our intervention, we still find that they systematically under-coded in their final year of residency. Lastly, we do not know the coding proficiency of our residents as they leave our program and enter practice. While we did not demonstrate the positive impact of our intervention of audit, feedback, and educational workshops on CPT E&M accuracy, others have reported success. 3-5 While these studies found improvement in coding accuracy, compared to our study, resident performance was monitored over shorter periods of time than we did and the interventions were generally more intense with more frequent educational or feedback sessions. These studies also used a before and after intervention design to assess the effect of the intervention. In considering interventions to improve coding accuracy, our data suggest that undercoding of complex (99214 and 99215) visits should receive considerable attention. This is where we found most of the coding errors. Although residents accurately coded the great majority of their visits, most of the undercoded complex visits were coded by residents as Thus, we would recommend that if resources were limited that faculty should preferentially focus their reviews on the visits that their residents code as This research has several limitations that should be considered when attempting to generalize our findings. The first is that these data are from one residency program. In addition, the number of pre-intervention codings is small, being only about 5% of the data. Having a larger control group would provide greater confidence in our conclusion that the intervention was not effective. Ideally, a control group of residents whose coding practices could be followed throughout

5 652 October 2010 Family Medicine training without our formal teaching and ongoing feedback would more clearly demonstrate whether our intervention is definitely having an impact. However, the RRC requirements make this difficult when ongoing feedback is required in some form. While our education and feedback intervention might not have been sufficiently intense to be successful, there are non-education-based interventions we could have considered that have been reported to enhance coding accuracy. Implementing clinician-coder double reading improves accuracy and increases revenue. 6 Although there is additional cost associated with this approach, captured revenue can make this cost-effective. Others have offered different coding tools to enhance coding accuracy. 7,8 There are a small number of studies on the impact of having residents use note templates to improve CPT coding These job aids have been found to have a positive effect on coding accuracy, but it is unknown whether they would continue to increase coding accuracy over long time periods. The EMR also has been shown to have potential as a way to increase coding accuracy and improve charge capture. 12 Our project differed from these studies because we used a coding aid for our educational workshops but did not attempt to have residents use the instrument while working in the clinical setting. Conclusions Over a 6-year study period we found significant evidence that residents improved in their CPT E&M coding of return patient visits as they progressed in the residency program. This improvement occurred concurrently with an educational intervention involving interactive coding workshops and delayed formative feedback to residents on their coding performance in a family medicine clinic. Our data also suggest the improvement in coding accuracy we observed might not be the result of our workshops and feedback to residents. Coding accuracy is an important outcome of a residency program, and development of effective educational interventions that can be implemented for long periods of time should be pursued using rigorous research methodologies. Acknowledgments: The authors thank Rozanne Murphy, CPC (certified professional coder), for her work on this project. Corresponding Author: Address correspondence to Dr Skelly, University of Iowa, Department of Family Medicine, 200 Hawkins Drive, Iowa City, IA Fax: kelly-skelly@uiowa.edu. Re f e r e n c e s 1. Kikano GE, Goodwin MA, Stange KC. Evaluation and management services. A comparison of medical record documentation with actual billing in a community family practice. Arch Fam Med 2000;9: King MS, Sharp L, Lipsky MS. Accuracy of CPT evaluation and management coding by family physicians. J Am Board Fam Pract 2001;14(3): As-Sanie S, Zolnoun D, Wechter ME, Lamvu G, Tu F, Steege J. Teaching residents coding and documentation: effectiveness of a problem-oriented approach. Am J Obstet Gynecol 2005;193: , Carter KA, Dawson BC, Brewer K, Lawson L. RVU ready? Preparing emergency medicine resident physicians in documentation for an incentive-based work environment. Acad Emerg Med 2009;16: Jones K, Lebron RA, Mangram A. Practice management education during surgical residency. Am J Surg 2008;196: Nouraei SA, O Hanlon S, Butler CR, et al. A multidisciplinary audit of clinical coding accuracy in otolaryngology: financial, managerial and clinical governance considerations under payment-by-results. Clin Otolaryng 2009;34: Sosa RE, Cascardo D. Between a rock and a hard place with E/M coding: dilemmas of compliance and practice financial viability. J Med Pract Manage 2000;16(1): Staiger TO, Chew LD, Helenius I. A case-based approach to outpatient evaluation and management service coding. Postgrad Med 2008;120(4): Sprtel SJ, Zlabek JA. Does the use of standardized history and physical forms improve billable income and resident physician awareness of billing codes? South Med J 2005;98(5): Grogan EL, Speroff T, Deppen SA, et al. Improving documentation of patient acuity level using a progress note template. J Am Coll Surg 2004;199(3): Mulvehill S, Schneider G, Cullen CM, Roaten S, Foster B, Porter A. Template-guided versus undirected written medical documentation: a prospective, randomized trial in a family medicine residency clinic. J Am Board Fam Pract 2005;18(6): Silfen E. Documentation and coding of ED patient encounters: an evaluation of the accuracy of an electronic medical record. Am J Emerg Med 2006;24(6):