BRIEF OBSERVATION

Capturing the Diagnosis: An Internal Medicine EducationProgram to Improve DocumentationBrad Spellberg, MD,a,b Darrell Harrington, MD,a,b Susan Black, RNP,c Darryl Sue, MD,b,d William Stringer, MD,b,d

Mallory Witt, MDb,eaDivision of General Internal Medicine, Los Angeles Biomedical Research Institute at Harbor-UCLA Medical Center, Torrance, Calif;bDavid Geffen School of Medicine at UCLA, Los Angeles, Calif; cDepartment of Quality Improvement, Harbor-UCLA Medical Center, LosAngeles, Calif; dDivision of Pulmonary and Critical Care Medicine, Los Angeles Biomedical Research Institute at Harbor-UCLA MedicalCenter, Los Angeles, Calif; eDivision of HIV Medicine, Los Angeles Biomedical Research Institute at Harbor-UCLA Medical Center, LosAngeles, Calif.

Funding: NonConflict of InAuthorship: A

writing this manusRequests for re

West Carson St, LE-mail address

0002-9343/$ -seehttp://dx.doi.org/1

ABSTRACT

BACKGROUND: Specific and accurate documentation of patient diagnoses and comorbidities in the medicalrecord is critical to drive quality improvement and to ensure accuracy of publicly reported data. Unfor-tunately, inpatient documentation is taught to internal medicine trainees and practitioners sporadically, if atall. At Harbor-UCLA Medical Center, a public, tertiary care, academic medical center, we implemented aneducational program to enhance documentation of diagnoses and comorbidities by internal medicineresident and attending physicians.METHODS: The program consisted of a series of lectures and the creation of a pocket card. These weredesigned to guide providers in accurate documentation of common diagnoses that group to different levelsof disease severity, achieved by capturing Centers for Medicare and Medicaid Services complication codesand major complication codes. We started the educational program in January 2010 and used a pre-postdesign to compare outcomes. The program’s impact on complication codes and major complication codescapture rates, mortality index, and case mix index was evaluated using the University Health Consortiumdatabase.RESULTS: The median quarterly complication codes and major complication codes capture rate for inpa-tients on the internal medicine service was 42% before the intervention versus 48% after (P ¼ .003).Observed mortality did not change but expected mortality increased, resulting in a 30% decline in medianquarterly mortality index (P ¼ .001). The median quarterly case mix index increased from 1.27 to 1.36(P ¼ .004).CONCLUSIONS: Thus, implementation of an internal medicine documentation curriculum improved accu-racy in documenting diagnoses and comorbidities, resulting in improved capture of complication codes.� 2013 Elsevier Inc. All rights reserved. � The American Journal of Medicine (2013) 126, 739-743

KEYWORDS: Case mix index; Centers for Medicare and Medicaid Services; Coding; Complication code; Docu-

et

cpi

f0

mentation; Graduate medical education; Mortality index; Resident education

Specific and accurate documentation in the medical recordof patient diagnoses and comorbidities is critical to drivequality improvement and to ensure accuracy of publiclyreported data. Since the 1983 introduction by the Centers for

.erest: None.ll authors had access to the data and played a role inript.rints should be addressed to Brad Spellberg, MD, 1124u Vaccine Center, Torrance, CA 90502.: bspellberg@labiomed.org

ront matter � 2013 Elsevier Inc. All rights reserved..1016/j.amjmed.2012.11.035

Medicare and Medicaid Services (CMS) of the InpatientProspective Payer System, medical documentation of patientdiagnoses and comorbidities has grown increasinglycomplex.1,2 Unfortunately, although training programs aresupposed to consider “systems-based practice” as part oftheir curricula, there have been only limited attempts toevaluate teaching of documentation in internal medicineresidency training.3,4 Thus, trainees may find themselves ata disadvantage when they enter practice and are suddenlyexpected to have competency in specific and accuratedocumentation of diagnoses and comorbidities.

740 The American Journal of Medicine, Vol 126, No 8, August 2013

The impact of research and quality interventions designedto improve processes of care can only be accurately measuredif the processes are in control, which requires accuratemeasurement of the diagnoses of the patients in the system.Thus, publicly reported quality metrics, such as case mixindex and mortality index, vary depending on the adequacy

CLINICAL SIGNIFICANCE

� A documentation-enhancement pro-gram based on education of internalmedicine resident physicians, supportedby a pocket tool, resulted in improvedcapture of complication codes, highercase mix index, and lower mortalityindex.

� The program required minimal up-frontinvestment.

� Such programs are likely to be ofincreasing importance given the currentnational emphasis on improved qualityof care and publicly reported data.

of documentation of patients’diagnoses. Case mix index isa measure of the average diseaseseverity of patients taken care ofby healthcare providers and dependson what diagnoses the patientsactually have and how accuratelythose diagnoses are documented bythe provider. Medicare reimburse-ment is linearly related to casemix index; a doubling of casemix index results in a doublingof hospital income. In addition,mortality index allows compari-sons of the risk of patients dying athospitals that have different patientpopulations with different severityof illnesses. Hospitals’ mortalityindices can be compared by patientsand payors to determine which

hospitals provide care resulting in lower risks of death.However, these metrics can be fairly compared across facil-ities and systems only if each facility and system is using thesame language to describe its disease severity—absent stan-dardization of documentation, apples are compared withoranges. Thus, to support quality improvement efforts andhealthcare outcomes research, it is necessary that providers betrained to accurately document their patients’ diagnoses.Accurate documentation of diagnosis also is critical tosupport CMS reimbursement and public reporting of thequality of care and risk of dying at individual hospitals.Finally, it is critical to train providers not to overdocument, orfalsely enhance disease severity, which results in fraudulent,excessive hospital billing.

Because of these concerns, we sought to develop a prac-tical method to teach internal medicine residents andattending physicians documentation skills with a goal ofmore accurately capturing the severity of disease for patientscared for at an urban public teaching hospital. We hypoth-esized that formal training in documentation would havea positive effect on several measures, including case mixindex and mortality index.

MATERIALS AND METHODS

Study DesignWe used a pre-post study design to evaluate the effect ofa documentation enhancement tool as part of an educationalprogram for internal medicine residents. The programfocused on accurate documentation of diagnoses and capture

of CMS complication codes and major complication codes.Mortality index (defined as the observed/expected mortality)and case mix index (defined as the sum of the relative weightsof the diagnostic related groups of patients cared for onthe internal medicine ward service divided by the number ofadmissions over a period of time) were the metrics used

for comparison.

Development ofa Documentation TeachingToolTo focus on the highest impactdiseases, we generated a list of themost common diagnoses encoun-tered on the internal medicineinpatient service. The diagnoseswere cross-referenced to CMScomplication codes and majorcomplication codes, which areused in the Medicare Severitydiagnosis-related group system toenhance capture of disease severity.The 10 most frequent diagnosesthat related to complication codes/major complication codes were

included on a documentation pocket card. In addition, aneleventh set of diagnoses was based on information providedby medical record coders, who cannot enter documentationcodes based on symbols or unacceptable abbreviations (eg,YK for hypokalemia). The pocket card instructed residentsand attending physicians to avoid use of arrows or othersymbols for diagnoses. The handout was designed withcolored fonts and highlights to describe explicitly what wordswere required to enable proper International Classification ofDiseases, Ninth Revision, Clinical Modification (ICD-9-CM)codes to be applied on the basis of individual diagnoses(Supplemental Figure, online).

Deployment of the Teaching ToolIn January of 2010, laminated documentation pocketcards were given to each internal medicine resident andattending physician at the Harbor-UCLA Medical Center.We instructed residents to carry the pocket cards and to referto them when writing history and physical examinations,daily progress notes, and discharge summaries. Didacticlectures on use of the pocket card were given by the authors(BS, MW, SB) to housestaff and to new, incoming interns inJuly of 2010 and July of 2011, as well as to attendingphysicians at faculty meetings.

Analysis of ResultsQueries were run using the University Hospital Consortium(UHC) database and search engine to gather data on theDepartment of Medicine’s case mix index, mortality, and

Spellberg et al Medical Record Documentation 741

complication codes/major complication codes capture for 2years before and 2 years after implementation of the docu-mentation enhancement program. Cumulative data beforeand after the intervention were compared by the nonpara-metric Mann-Whitney U test or chi-square test, with Pvalues �0.05 considered significant. To analyze trends, datawere plotted using control charts, with any point above theupper control limit, or 2 of 3 points between the 95%confidence interval limit and the upper control limit,considered a statistically significant change from the base-line process.5 Control and confidence interval limits werecalculated on the basis of values before the intervention todetect special cause variation from the baseline process.5

Figure 1 Quarterly complication code and major complica-tion code capture rate before and after the intervention. Quar-terly complication code and major complication code capturerates for patients on the Harbor-UCLA Medical Center internalmedicine services (solid lines, closed circles) are comparedwith the average capture rates for patients cared for on internalmedicine services across all other hospitals in the UHC data-base (dashed lines, open circles). Data are shown from 2008 tosecond quarter 2012. CC ¼ complication code; HUMC ¼Harbor-UCLA Medical Center; MCC ¼ major complicationcode; UHC ¼ University Hospital Consortium.

RESULTS

Complication Code CaptureComplication codes and major complication codes capturerates (ie, the percentage of patients admitted for whom coderscaptured complication codes or major complication codes)were compared before and after the intervention. Baselinedata from UHC were available starting from 2008. Themedian (range) quarterly complication codes and majorcomplication codes capture rate was 42% (39%-45%) beforethe intervention versus 48% (46%-50%) after the intervention(P¼ .003). The median (range) quarterly complication codescapture rate increased from 18% (17%-20%) to 23% (20%-26%) (P ¼ .003). The major complication codes capture ratedid not significantly increase (median, 23% to 24%).

As a comparative control for local complication codes/major complication codes capture rates, the UHC databasewas queried for average complication codes and majorcomplication codes capture rates for patients on InternalMedicine services at all other hospitals in the database from2008 through the second quarter of 2012. The annualcomplication codes and major complication codes capturerates were plotted, comparing Harbor-UCLA MedicalCenter internal medicine cases and the internal medicinecases at the other hospitals. After the intervention, thecomplication codes and major complication codes capturerate gaps narrowed between Harbor-UCLA and the averageof other UHC hospitals (Figure 1).

Mortality IndexObserved mortality rates by quarter fluctuated between 2.3%and 3.6% throughout the entire period (Figure 2). Themedian (range) quarterly observed mortality rate pre- versuspost-intervention was 2.8% (2.3%-3.6%) versus 2.7%(2.3%-3.4%) (P ¼ .9), respectively.

In contrast, the expected mortality rate significantlyincreased during the first quarter of the intervention(Figure 2). Pre-intervention, the median (range) quarterlyexpected mortality rate was 2.3% (1.8%-2.7%), whereaspost-intervention it was 3.3% (2.6-3.8%) (P ¼ .002). Asa result, the mortality index decreased significantly during thefirst quarter of the intervention (Figure 2). Pre-intervention,

the median (range) quarterly mortality index was 1.2 (1-1.5),whereas post-intervention it was 0.9 (0.7-1) (P ¼ .001).

Case Mix IndexBaseline case mix index for the Department of Medicinewas analyzed from the fourth quarter of 2007 (when the newMedicare Severity diagnostic-related group system wasintroduced by the CMS) through the fourth quarter of 2009.During that time, the median quarterly case mix index forpatients admitted to the internal medicine inpatient servicewas 1.27 (range, 1.24-1.34) (Figure 3A). After introductionof the documentation enhancement pocket card in the firstquarter of 2010, the median quarterly case mix indexincreased to 1.36 (range, 1.32-1.5) (P ¼ .004 by Mann-Whitney U test compared with pre-intervention case mixindex). Control charts demonstrated that the shift in process

Figure 2 The impact of the documentationeducation on mortality data. Observed mortalityrates, expected mortality rates, and mortality index(calculated by dividing observed/expected) forHarbor-UCLA Medical Center from the UHCdatabase are plotted on control charts. Arrows markthe first quarter of the intervention. Only commoncause variation is seen for observed mortality, withno significant changes before or after the inter-vention. However, expected mortality significantlyincreased during the first quarter of the interven-tion, because the data crossed the upper controllimit (marked by asterisk). The reduction inmortality index became significant by the secondquarter of the intervention (�2 of 3 datapointsbetween the �95% bound and the lower controllimit [*]). LCL ¼ lower control limit; UCL ¼ uppercontrol limit.

Figure 3 The documentation educational programincreased quarterly case mix index for patientsadmitted on the internal medicine inpatient service.(A), Quarterly case mix index from the UHC databasefor patients on the internal medicine service startingfrom the introduction of the Medicare Severity diag-nostic-related group system (fourth quarter calendaryear 2007) plotted on a control chart. The educationalprogram initiated in the first quarter of 2010 (arrow).To evaluate for a shift in the process, the controllimits are set on the basis of the variance of thebaseline data (ie, pre-2010). By the second quarter of2010, the case mix index had shifted upward abovethe upper control limit, indicating special causevariation (*). (B), Data subsequent to the interventionare plotted on a control chart using limits based on thevariance after the intervention to determine whetherthe process remained within control (ie, was main-tained) after the intervention was introduced. All datapoints remain with the control limits. LCL ¼ lowercontrol limit; þ95% ¼ upper bound of the 95%confidence interval; �95% ¼ lower bound of the 95%confidence interval; UCL ¼ upper control limit.

742 The American Journal of Medicine, Vol 126, No 8, August 2013

became significant during the second quarter of 2010.Subsequent values fluctuated around a new mean, but whenplotted with new control limits based on post-interventionvariance, all data remained within the control limits (ie, notsignificantly different) (Figure 3B).

DISCUSSIONSpecific documentation of diagnoses and comorbidities isincreasingly important to support quality initiatives andimprove accuracy of publicly reported data. The importanceof training providers to accurately document to supportquality metrics and healthcare outcomes research is under-scored by the improvement we found in our departmentalmortality index. Before the intervention, our median quar-terly departmental mortality index was 1.2, indicating that20% more patients died than were expected to on the basis ofthe diagnoses they had. However, with no significant change

Spellberg et al Medical Record Documentation 743

in observed mortality, and no change in the overall patientpopulation cared for, improving the quality of our docu-mentation resulted in a decline in the median quarterlymortality index to 0.9 during the post-interventionperiod, and to 0.8 during the most recent 12-month period,indicating that 20% fewer patients died than would havebeen predicted. Our mortality index was inaccurate beforethe intervention because we were failing to document manyof the diagnoses our patients had, and this inaccuracyconfounded cross-facility or cross-system comparisons ofour hospital’s outcomes. Likewise, Rosenthal et al6,7 previ-ously described marked variations in facilities’ mortalityindex and length of stay when compared across healthcaresystems, raising concerns about the ability to make “apples-to-apples” comparisons of such quality data. The impact ofresearch and quality interventions designed to improveprocesses of care can only be accurately measured if theprocesses are in control, which requires accurate measure-ment of the diagnoses of the patients in the system. Thus, it isimperative that training programs prepare housestaff forpractice, and help educate attending physicians and otherproviders, by teaching them how to document to the level ofspecificity required to support accurate coding.

We developed a novel documentation training programto help housestaff and attending physicians capture accuratedescriptors of patient diagnoses and comorbidities, withexplicit warnings not to overstate or embellish documenta-tion to increase coding artificially or falsely. It is critical notto overdocument (sometimes referred to as “upcoding”),which creates potential for fraud. Rather, the goal is tocapture accurately all of the diagnoses patients have, withoutembellishment, which is essential for proper reimbursementand to support quality improvement and healthcare researchefforts. By providing a pocket information tool, supple-mented by repeated lectures and reminders to residents, weachieved a substantial increase in the capture of complica-tion codes, a lower mortality index, and a higher depart-ment-wide case mix index. Our results are similar to thoseachieved in a surgical training program that created anelectronic template for progress notes that supportedenhanced diagnostic capture,8 underscoring the potential forelectronic medical records to similarly support enhance-ments in diagnostic capture and increases in case mix indexand decreases in mortality index.

The improvements in complication code capture tookplace without any other changes in coding or documentationpractices. The majority of documentation at our hospital wasin the form of paper charting, because our medical systemhas not yet undergone conversion to electronic medicalrecord. The improvements were sustained through the2.5-year period of follow-up. There were minimal up-frontdirect costs of the intervention, aside from the costs ofprinting the pocket cards and the time spent giving lecturesand meeting with the housestaff. These data indicate thata simple pocket tool supplemented by an ongoing educa-tional program can result in substantive longer-termenhancements in documentation even without large up-front

costs. Further supplementation by documentation enhance-ment software would likely magnify the effect.

Study LimitationsThe primary limitations of the study include its single-centerdesign and focus on internal medicine residents. Neverthe-less, because the hospital under study was a public teachinghospital with limited resources to support documentationenhancement programs, it is likely that other hospitals withadditional resources would be able to achieve similar oreven better results. We also cannot exclude the possibilitythat other changes to practice or coder’s behavior led tothe increases in case mix index and expected mortalityseen after the documentation enhancement intervention.However, we are unaware of any other specific changes tolocal provider or coder practice that could account for thechanges in case mix index, complication codes/majorcomplication codes capture rate, and expected mortality.Finally, it was not possible to determine whether theprimary drivers of the improvement in case mix index andmortality index were residents, attending physicians, orboth. Residents provide the majority of the documentationin our system; however, attending physicians have theability to amend and add addendums to notes. It is likelythat both contributed to the improvements seen.

CONCLUSIONSA documentation-enhancement program based on educationof internal medicine resident and attending physicians,supported by a pocket tool, resulted in improved capture ofcomplication codes, higher case mix index, and lowermortality index. Such programs are likely to be of increasingimportance given the current national emphasis on improvedquality of care and publicly reported data, and the loomingconversion from an ICD-9-CMebased documentationsystem to the more complex ICD-10 system.

References1. Iglehart JK. Medicare begins prospective payment of hospitals. N Engl J

Med. 1983;308:1428-1432.2. Steinwald B, Dummit LA. Hospital case-mix change: sicker patients or

DRG creep? Health Aff (Millwood). 1989;8:35-47.3. Miller DD, Magill G. Evolving internal medicine faculty and resident

comprehension of and attitudes toward a departmental documentationcompliance program (1999-2002). Am J Med. 2003;114:250-256.

4. Oxentenko AS, West CP, Popkave C, et al. Time spent on clinicaldocumentation: a survey of internal medicine residents and programdirectors. Arch Intern Med. 2010;170:377-380.

5. Carey RG, Lloyd RC. Measuring Quality Improvement in Healthcare.Milwaukee, WI: Quality Press; 2001.

6. Rosenthal GE, Shah A, Way LE, Harper DL. Variations in standardizedhospital mortality rates for six common medical diagnoses: implicationsfor profiling hospital quality. Med Care. 1998;36:955-964.

7. Rosenthal GE, Harper DL, Quinn LM, Cooper GS. Severity-adjustedmortality and length of stay in teaching and nonteaching hospitals.Results of a regional study. JAMA. 1997;278:485-490.

8. Grogan EL, Speroff T, Deppen SA, et al. Improving documentation ofpatient acuity level using a progress note template. J Am Coll Surg.2004;199:468-475.

Supplemental Figure Documentation enhancement tri-fold pocket card.

AIDS ¼ acquired immunodeficiency syndrome; BP ¼ blood pressure; CHF ¼ congestive heart failure; CKD ¼ chronic kidney disease; COPD ¼ chronicobstructive pulmonary disease; DVT ¼ deep venous thrombosis; EKG ¼ electrocardiogram; GERD ¼ gastroesophageal reflux disease; GFR ¼ glomerularfiltration rate; GI ¼ gastrointestinal; GNR ¼ gram negative rod; HR ¼ heart rate; HTN ¼ hypertension; MIBI ¼ myocardial perfusion imaging; NSTEMI ¼non-ST elevation myocardial infarction; RR ¼ respiratory rate; SIRS ¼ systemic inflammatory response syndrome; STEMI ¼ ST elevation myocardialinfarction; UTI ¼ urinary tract infection; WBC ¼ white blood cell.

743.e1 The American Journal of Medicine, Vol 126, No 8, August 2013